diff --git a/file751.txt b/file751.txt new file mode 100644 index 0000000000000000000000000000000000000000..71d5ddc8f617e370fe4f67866959dde9d4c7f14a --- /dev/null +++ b/file751.txt @@ -0,0 +1,60 @@ + + + + +IntroductionStudies have been conducted to investigate the cueing fidelity requirements for roll-lateral interactions but little has been done for the pitch-longitudinal interactions'T3.From a psychophysics point of view, there is little perceptual difference as provided by a human's angular sensors4.In Zacharias's' assessment of the pilot's motion sensor models, the mathematical treatment of the roll and pitch angular rate sensing characteristics are also similar.However, in most motion-based simulators, the pitch and longitudinal visual cues are typically poor as compared to roll-lateral visual cues.Vertical field-of-view (FOV) is usually more constrained so that the horizon can disappear taking away the pilot's pitch attitude awareness.Also little ground is seen in front or below the pilot providing little if any longitudinal position awareness.Depth cues are also typically poor, thus damaging the pilot's longitudinal awareness.Furthermore, from a tactile consideration, the pilot's contacts with the seat and restraint system in the pitch-longitudinal axes are different from the roll-lateral axes.Taking these differences into account, the cueing fidelity requirements for pitch-longitudinal interactions need to be examined in their own right.This investigation was undertaken to provide some preliminary insights into the motion requirements for combined pitch-longitudinal motion.In flight simulation, pitch and longitudinal motion need to be treated together.For a coordinated maneuver, longitudinal platform motion provides an acceleration to counteract the "leans" that would result from only pitching the cockpit.Typically, the longitudinal motion used for this purpose does not eliminate the leans completely but is conventionally attenuated through a washout filter to maintain the simulator within its physical travel limit.This compromise introduces an erroneous specific force6 depending on the magnitude of the pitch attitude.This investigation examined the effects of this erroneous cue and compared the results to the roll-lateral requirements.Experiment Description Aircraft Model.Force Characteristics.and The Task A two DOF helicopter model with a pitch rate command and a fully coordinated pitch-longitudinal response about hover is given by equation 1,(1)The pitch acceleration due to pitch rate (or the pitch damping stability derivative), M,, was set at -4.0 l/set and the pitch control power, Mslon, was 0.67 rad/sec21in.The rotational center was set at the pilot's abdomen.The pitch rate response to the stick was developed to have the satisfactory handling qualities according to the Reference 7 specification.The longitudinal stick force feel characteristics was set at 2 lb/in with a 0.5 lb breakout force. +The TaskThe task was a 20 ft dash-and-stop task performed at a constant altitude of 23 ft as shown in Figure 1 along the center of a runway.The piloted task started with a 20 ft translation towards the desired hover position situated in front of the helicopter's initial hover position, followed by 20 seconds of stationkeeping.The stationkeeping point was denoted by a series of four walls located 100 ft to the right and 100 ft in front and positioned at an angle of 45 degrees.At the desired stationkeeping point, the pilot should see the walls "edge on" out the right window.At the proper position, a building in the background appeared in the middle of the two inner walls.The walls were spaced such that the distance between the two inner walls translated to a distance of *5 ft with respect to the desired stationkeeping point in the axis of travel.Likewise, the distance between the two outer walls translated to a distance of + 10 ft with respect to the desired stationkeeping point in the axis of travel.A bob-up target was placed in the center of the runway and various red cones were placed on the runway visible through the right chin window for additional visual cueing.,Pilots were instructed to complete the dash-and-stop task in one smooth maneuver within the performance standards shown in Table 1.A sum-of-sines, Table 2, was added to the pilot's stick input to simulate wind gusts.Four experienced test pilots participated in the test.Each pilot practiced with randomly selected motion configurations at the beginning of each session.During the data collection, each pilot evaluated the motion configurations in a random order.The pilots were asked to fly each motion configuration three times and to evaluate the third repetition.Pilots were asked to give handling qualities ratings (HQRs)* and motion fidelity scale (MFS) ratings as shown in Table 3.A questionnaire was also given to elicit comments from the pilots regarding control strategy and motion fidelity. +Test FacilitiesThe NASA Ames Vertical Motion Simulator (VMS) was used for this investigation, which is shown in Figure 2.The cockpit was oriented such that the longitudinal axis of the simulator cab traveled along the beam which has a usable travel of 40 ft.The large longitudinal travel of the VMS allowed for the 20 ft dashand-stop task without any motion cue attenuation.The motion system responses were collected using white noise with a Gaussian distribution and checked with frequency response technique developed for system identification called CIFERog.For this investigation, each motion axis had an equivalent time delay of 60 msec.The visual FOV of the helicopter cab is shown in Figure 4.The visual system was an Evans and Sutherland ESIG 4530 image generator.The out-the-window scene was presented on four monitors with a collimation/beam-splitter system.The visual system time delay was measured at 60 msec, which matches the equivalent time delay in the pitch and longitudinal motion axes. +Test ConfigurationsTwo second-order washout filters, Figure 3, were developed for this investigation.A pitch washout filter generated the pitch motion commands, and a coordinated longitudinal washout filter provided the longitudinal motion to counteract the leans due to pitch.Two pitch washout filter configurations and eleven longitudinal washout filter configurations, shown in Tables 3 and4 respectively and presented in Figure 5, were selected to investigate the interaction between pitch motion and longitudinal motion.Each configuration consisted of the gain and filter frequency for the respective washout filter.The configurations were selected to represent low, medium, and high-motion fidelity as defined in Reference 10. +ResultsFour pilots took part in this test.Their average MFS ratings and the average HQR are shown in Figures 6 and7 respectively.All pilots rated the fixed-base case as low fidelity.In general, the amount of longitudinal platform motion increases as both the longitudinal filter's predicted fidelity and the pitch filter's predicted fidelity increases, i.e., increasing motion gain and decreasing phase distortion.However, as the amount of required longitudinal motion increases any parasitic differences in the dynamics between the pitch and longitudinal motion become exacerbated and potentially objectionable.Pilots commented on the sharpness of the motion cues due to the high gains.This is evident in the full-motion case (Al, Tl) in that the pilots felt the motion cues to be objectionable and rated it as low fidelity.As pitch motion is attenuated (A3, Tl), the longitudinal platform motion decreased and artifacts such as above were reduced significantly which resulted in an improved motion fidelity rating.In both the full and the attenuated pitch motion cases, the motion cues became objectionable when the longitudinal phase distortion was 80 degrees at 1 rad/sec (co% = 0.9). Figure 8 shows the average MFS as a function of co,.As expected, the MFS decreases as longitudinal phase distortion is increased regardless of pitch motion.This is consistent with the roll-lateral motion requirements.In both the full and the attenuated pitch motion cases, it is clear that pilots felt the motion cues to be objectionable when little coordinated motion was provided, i.e., cases where the longitudinal motion gain was at or below 0.2.This is consistent with Schroeder2 and Chung' in their roll-lateral motion fidelity requirement investigations.Figure 9 shows the average MFS 'as a function of G,.For the attenuated pitch motion case (A3), MFS ratings generally decrease as longitudinal motion gain decreases.Again this is consistent with roll-lateral motion requirements.For the full pitch motion case (Al), there is no clear trend due to the low rating for the (Al, T3) case where it is expected to be rated medium fidelity.Undesired motion artifacts might have played a role in receiving the low fidelity rating while the longitudinal motion gain is still considerably high at 0.8.One observation is made from Figure 6 when comparing the average MFS between the full pitch motion (Al) and the attenuated pitch motion (A3).For the cases that represent the medium and high fidelity region, i.e., cases T2, T3, T5, T6, the average MFS ratings for the attenuated pitch motion case are noticeably improved when compared with the full pitch motion case.This improvement may be contributed to less pitch motion resulting in less "leans".Therefore any objectionable situations would be expected to occur less frequently, and consequently result in the noticeable improvement in MFS ratings.For both the full and attenuated pitch motion cases, the averaged HQRconsistently followed the averaged motion fidelity ratings, i.e., HQR improves when MFS improves.Concludine Remarks 1) Large phase distortion in longitudinal motion due to the pitch motion is detrimental to the motion fidelity.Pilots found it to be objectionable.This is consistent with the roll-lateral motion requirements.2) Ratings generally improved as the longitudinal motion gain increased until the undesired motion artifacts became noticeable.3) Additional testing is recommended to further investigate the pitch-longitudinal motion cueing fidelity requirements. +References(c)l999 American Institute of Aeronautics & Astronautic;or.pubkedwith permikion of aithor(s) and/or author(s)' sponsoring organization. +*Jex, H.R., Jewell, W.F., Magdaleno, R.E., and Junker, A.M.: "Effects of Various Lateral-Beam Washouts on Pilot Tracking and Opionio in the Lamar Simulator," AFFDL-TR-79-3134, pp.244-266.2Schroeder, J.A. and Chung, W.Y.: "Effects of Roll and Lateral Flight Simulation Motion Gains on a Sidestep Task," American Helicopter Society's 53rd Annual Forum, April 1997.3Chung, W.Y., Robinson, D.J., Wong, J., and Tran, D..L.: "Motion Cue Models for Pilot Vehicle Analysis," AMRL TR-78-2, 1978.6Grant, P.R. and Reid, L.D.: "Motion Washout Filter Tuning: Rules and Requirements," AIAA Flight Simulation Technologies Conference, August 1995.7Aeronautical Design Standard, Handling Qualities +Figure 5 .FullFigure 9 .59Figure 1.The longitudinal dash-and-stop task +Table 1 .1Task performance standardsDesiredAdequateLongitudinaltranslation7 set11 setcompletion timeStationkeepingposition+I-5 ft+I-10 fttoleranceTable 2. External disturbanceFrequency (rad/sec)0.280.490.801.502.674.638.50Amplitude (in)0.002 0.006 0.014 0.032 0.054 0.068 0.06 +Table 3 .3Motion fidelity scaleDescriptionScore +Table 4 .4Angular motion washout gains and filter frequencies configurationsPitch MotionMotion GainWashout Filter@ 1 rad/secConfigurationsG,Frequency, os (radkec)GainPhase distortion(degree)Al1.00.000110A30.60.50.5843 +Table 5 .5Translational washout gains and filter frequencies configurationsLongitudinalMotion GainWashout FilterMotionGXFrequency, co,Configurations + (c)l999 American Institute of Aeronautics & Astronautics or published with permission of author(s) and/or author(s)' sponsoring organization. + + + + + + + + + Frequency-Response Method for Rotorcraft System Identification: Flight Applications to BO-105 Coupled Rotor/Fuselage Dynamics + + JBOsinacori + + + GECooper + + + RPHarper + + + MBJr + + + MGCauffman + + + + The Use of Pilot Requirements for a Helicopter Flight Research Simulation Rating in the Evaluation of Aircraft Handling Qualities + St. Louis, pgs + + American Institute of Aeronautics & Astronautics or published with permission of author( + July 1992. July 1994. September 1977. April 1969 + 37 + + + + s) &d/or author(s)' sponsoring organization. Requirements for Military Rotorcraft, ADS-33D + American Institute of Aeronautics & Astronautics or published with permission of author(s) &d/or author(s)' sponsoring organization. Requirements for Military Rotorcraft, ADS-33D, St. Louis, pgs. 3-17, July 1992. MO, July 1994. 'OSinacori, J. B.: "The Determination of Some *Cooper, G. E., and Harper, R. P., Jr.: "The Use of Pilot Requirements for a Helicopter Flight Research Simulation Rating in the Evaluation of Aircraft Handling Qualities," Facility," NASA CR 152066, September 1977. NASA TN D-5153, April 1969. gTischler, M. B., Cauffman, M.G.: "Frequency-Response Method for Rotorcraft System Identification: Flight Applications to BO-105 Coupled Rotor/Fuselage Dynamics," Journal of the American Helicopter Society, Vol. 37, No. 3, + + + + + + diff --git a/file752.txt b/file752.txt new file mode 100644 index 0000000000000000000000000000000000000000..c6600acdb870c12e6dfc123ce8b640eeb2d224d3 --- /dev/null +++ b/file752.txt @@ -0,0 +1,352 @@ + + + + +I. IntroductionASA and the FAA have been involved in extensive efforts to develop advanced concepts, technologies, and procedures for the Next Generation Air Transportation System (NextGen). 1 The objective of these research efforts has been to improve the capacity, efficiency, and safety in the next-generation National Airspace System (NAS).Improvements come in the form of more accurate and autonomous onboard navigational capabilities based upon the Global Positioning System, more accurate surveillance capabilities such as Automatic Dependent Surveillance-Broadcast, advanced communication capabilities such as Data Communications, improved vehicle designs, and improved air traffic operations realized through advanced automation systems.A significant portion of the NextGen research is aimed at (i) developing ground-based automation systems to assist controllers in strategic planning operations, (ii) developing controller decision support tools to tactically separate and space the traffic, and (iii) developing flight-deck automation to assist pilots in accomplishing airborne merging and spacing operations. 2 A concept for future high-density terminal air traffic operations that has been proposed by the Airspace Super Density Operations (ASDO) researchers at NASA. 3 The concept includes five core componentstwo strategic planning elements in the form of Extended Terminal Area Routing and Precision Scheduling, as well as three tactical control elements in the form of Merging and Spacing, Tactical Separation, and Off-Nominal Recovery.Successful implementation of the Precision Scheduling component requires the following questions to be answered:1. What is the range of feasible flight times for an aircraft to transit between two points along its flight path (e.g., from top-of-descent to the meter fix and from the meter fix to the runway threshold)? 2. What is the accuracy with which an aircraft can meet a scheduled time-of-arrival (i.e., the absolute timing error)?3. What is the accuracy with which an aircraft can maintain self-separation with respect to a leading aircraft (i.e., the relative timing error)?From the foregoing list of key questions, the range of feasible flight times depends upon the following characteristics of the arrival operation: Aircraft performance characteristics  Cruise and descent speeds selected by the aircraft's onboard flight management system  Terminal area route geometry  Observed atmospheric conditions such as temperature and winds aloft Similarly, the time-of-arrival accuracy and self-separation performance depend upon the following:  Uncertainty associated with the predicted atmospheric conditions  Advisories from air traffic controllers, both manual as well as those assisted by automation tools  Current-day and next-generation aircraft navigation capabilities Not surprisingly, the prediction of atmospheric conditions and its associated uncertainty (most importantly the prediction of winds aloft) is a key driver in the successful implementation of Precision Scheduling.Therefore, the current work investigates the time-of-arrival uncertainty which results from the uncertainty associated with wind forecast errors.This paper specifically focuses on Phoenix Sky Harbor International Airport (PHX) arrival operations.The Rapid Update Cycle (RUC) forecast products generated by the National Oceanic and Atmospheric Administration (NOAA) is used to define the predicted winds.Aircraft Communication Addressing and Reporting System (ACARS) reports from participating aircraft are used to model the wind forecast errors.The data is further described in Section II.The PHX terminal airspace used for analysis is described in Section III.Definitions of wind magnitude and wind uncertainty metrics based upon time-of-arrival estimates are provided in Section IV.Finally, the data analysis results are presented in Section V, and some initial conclusions are discussed in Section VI. +II. Data SourcesThe wind error is defined as the deviation between the actual (truth) wind vectors and the predicted (forecasted) wind vectors.Therefore, the statistical properties of a large number of observed deviations are used to model the probability distribution of the wind uncertainty.Sample truth wind sets are generated from these statistics and used to investigate the wind error's impact on time-of-arrival.This requires the following data: (i) truth wind data and (ii) forecast wind data.In this research, wind reports obtained from equipped aircraft are used as the truth data and wind predictions obtained from NOAA are used as the forecast data.These data are described in more detail in this section. +A. Truth Wind ReportsMany commercial aircraft operating today are equipped with sensors that can provide real-time weather observations (primarily winds and temperatures) via radio downlinks.The Meteorological Assimilation Data Ingest System's (MADIS) 4 automated aircraft dataset provides ACARS 5 data obtained from many U.S., European and Asian airlines.Each participating aircraft provides position and wind information at approximately one-minute intervals.Since this data is obtained from commercial aircraft flying through the airspace, the ACARS data are not available for arbitrary locations and times.It is only available for those spatial locations and times that the aircraft actually was present in.Moreover, not all aircraft report this data.However, a large amount of historical data is still available to characterize the statistics of the wind errors.Figure 1 shows a sample of ACARS wind report locations from aircraft operating in the Phoenix terminal area.PHX is located at the origin of this plot.Each small black arrow represents a single ACARS report received during between January 1, 2011 and January 15, 2011.The arrival and departure routes are clearly marked by a higher density of ACARS reports. +B. Forecast Wind DataThe NOAA provides predictions of wind and atmospheric conditions for the entire United States.These forecasts are obtained through a weather product referred to as RUC 6 .RUC is an operational weather prediction system covering North America that updates on an hourly basis.It consists of a numerical forecast model and an analysis/assimilation system to initialize that model.RUC provides 1-18 hour forecasts, updated hourly using either 40-, 20-, or 13-km horizontal resolutions and 50 vertical levels.This study uses the 40-km RUC 1-hour forecast (RUC-40) previously used by air traffic management applications.It provides the predicted North and East components of the wind velocity.Unlike the ACARS data, RUC data is available over a fixed grid of spatial locations.A bilinear interpolation scheme is used to compute the wind predictions for spatial locations in between the RUC grid points. +C. Potential Truth Wind DataThe ACARS wind reports represent individual wind conditions.They are not suitable for a comprehensive analysis of wind error on time-of-arrival accuracy, since they are spatial and temporally sparse.Reference 7 describes an approach to construct a spatio-temporally correlated model of wind error from a set of truth wind reports and forecast wind data.Application of this technique to a particular wind forecast (e.g., the RUC-40 1-hour forecast for March 4, 2005 at 06Z) is used to create a set of potential truth wind data that could have been present.This set of potential winds are used to investigate the range of times-of-arrival that would have been observed.Throughout the remainder of the paper, spatio-temporally correlated truth winds refer to these potential set of truth winds corresponding to a particular wind forecast based upon the historically observed wind errors. +D. Limitations and ApplicabilityThis study compares the RUC-40 wind predictions and ACARS wind reports from 2011 in order to remain consistent with previous analyses performed for a series of ongoing NASA human-in-the-loop simulations of PHX arrival operations. 8,9These simulations used traffic and wind scenarios constructed from 2011 data.The wind errors at PHX are assumed to exhibit similar year-to-year seasonal variations.Additional work is required to determine the extent to which the wind errors at PHX are similar to those present at other locations around the NAS.In March 2012, NOAA replaced the RUC forecast model with the newer and more accurate Rapid Refresh (RAP) 10 forecast model.Although not part of the results presented here, later comparison of RAP wind predictions to the ACARS wind reports shows markedly reduced wind errors for altitudes above FL200.A discussion of this improvement is provided in subsequent sections.Finally, although the current paper focuses on wind modeling, the same approach is applicable to other atmospheric data such as temperature and pressure.In that context, it is worth noting that both ACARS and RUC provide temperature and pressure data in addition to wind components. +III. Arrival Route Modeling +IV. Definition of Wind MetricsThis section defines the wind magnitude and wind uncertainty metrics analyzed in this study.These metrics are based upon the ETA along each route from the en route transition fix to the runway threshold.Therefore, it is important to first identify the methodology used to compute the ETA. +A. ETA ComputationA piece-wise linear indicated airspeed (IAS) profile is developed from the following points along each of the routes shown in Figure 2 and Figure 3  140 KIAS at 0 nmi (typical final approach speed) The locations are expressed as distances to travel along the route (i.e., path distance).This speed profile was formulated as a reasonable approximation of the published speed constraints and standard operating procedures at PHX.They are similar to the observed speed profiles at other airports.In between the prescribed distances, the airspeeds are linearly interpolated.It should be noted that the full time-of-arrival error will be a combination of the error due to the differences between the modeled airspeed profile and the actual profile as well as the errors due to the predicted atmospheric conditions versus the actual conditions.The following approach is then used to compute a set of ETAs along each route:  Convert the IAS profile to a true airspeed (TAS) profile at the prescribed path distances  Interpolate for the TAS as a function of path length between the prescribed path distances  Discretize the horizontal path at 3 nmi intervals  Calculate the winds at the 3 nmi intervals for three sets of wind data: o Zero wind o RUC-40 forecast wind o Set of sample random spatio-temporally correlated wind values  For each scenario, compute the ground speed at the 3 nmi intervals by combining the prescribed true airspeed profile and the associated wind field  Compute transit time over each 3 nmi segment using the computed ground speed The ETA corresponding to the zero-wind scenario is represented as ; the ETA corresponding to the RUC-40 forecast wind is represented as .The 5 th and 95 th percentile ETAs for the set of sample spatio-temporally correlated winds are represented as and , respectively. +B. Wind Magnitude MetricThe Wind Magnitude Metric (WMM) characterizes the nominal strength of the wind in terms of its impact on the ETA, irrespective of the accuracy of the forecast.It is expected that winds affect the ETA; the stronger the wind magnitude the bigger the difference between the and .However, the ETA difference will be different for each route.Winds aloft will make flights on some routes travel faster and flights on the opposite routes travel slower.For this study, a scalar metric that encompasses all routes is sought.The following definition satisfies the above requirements and can be applied to an arbitrary number of routes: (1) where, is a particular route, and is the total number of routes.The absolute value of the numerator prevents cancellation of wind effects on opposite routes; the normalization by the denominator treats variations along shorter and longer routes appropriately; and the average over the number of routes prevents the expression from assuming very large values for large numbers of routes.The WMM can be interpreted as the average percentage variation of ETA with respect to the zero wind ETA for a given set of routes.Though the WMM is computed using a specific wind forecast product, namely the RUC-40 1-hour forecast, it is expected to be largely invariant to the particular wind forecast product, since all wind forecast products are intended to reflect similar large scale changes in atmospheric conditions. +C. Wind Uncertainty MetricsThe Wind Uncertainty Metric (WUM) characterizes the accuracy of the wind forecast products in terms of ETA variability.Again, a scalar metric that encompasses all routes is sought.Two different metrics are evaluated in the current research.The WUMs are based on the statistics of the ETA distributions obtained from a set of spatiotemporally correlated truth winds generated by 5000 Monte-Carlo simulation runs conducted for all routes and each RUC-40 1-hour wind forecast.The ETA variation in each Monte-Carlo simulation is used as a measure of the wind uncertainty.The variation is characterized using two different measures: (i) the 90-percentile interquartile range, and (ii) the standard deviation.The former is suited for all statistical distributions, whereas the latter is best suited for normal distributions.The WUM definitions are given below:(2)(3) This wind uncertainty metric specifically focuses on the impact of the wind errors on the corresponding ETA variation, i.e., ETA error.An alternative formulation of the wind uncertainty metric could investigate the impact of the wind errors on the corresponding inter-arrival error (i.e., the differential ETA error between two successive flights).This impact may be the subject of future analyses. +V. ResultsResults obtained from analyzing the RUC-40 1-hour wind forecasts and ACARS wind reports in the PHX terminal area for the entire year of 2011 are presented in this section.As described in Reference 7, the wind error statistics are determined using a 15-day moving average in order to capture the seasonal variation of the atmospheric conditions and their associated errors. +A. Variation of Wind Uncertainty with AltitudeFigure 4 shows the variation of the North and East wind errors observed in the PHX terminal area as a function of altitude.These wind errors are the difference between the ACARS wind reports and the corresponding RUC-40 1-hour wind forecasts for the same time, location, and altitude.These results are consistent with other studies that have found similar wind error magnitudes. 11,12Overall, there is no statistical difference between the North and East wind errors.For all altitudes, the mean wind error (not shown) is found to be effectively zero (i.e., the RUC-40 1hour forecast is unbiased for sufficiently large data sets).For altitudes below FL200, the standard deviations of the North and East wind errors are approximately 5-10 knots.Meanwhile, for altitudes above FL200, the standard deviations of the North and East wind errors increase rapidly to 15-20 knots. +Figure 4. Variation of RUC-40 Forecast Wind Errors with AltitudeAs previously mentioned, NOAA replaced the RUC forecast model with the newer and more accurate RAP forecast model in March 2012.Subsequent limited comparison of RAP wind predictions to the ACARS wind reports shows markedly reduced North and East wind errors for altitudes above FL200.This cursory examination was performed to understand how the RUC analyses would translate to the current wind forecast products available.While not presented in this paper, those results showed that the standard deviations of the wind errors of the RAP wind predictions above FL200 do not exhibit the same significant increase with altitude as shown in Figure 4.As a result of this improvement, the impact of the wind errors on ETA accuracy is expected to be lessened for the initial portion of the arrival route. +B. Spatio-Temporal Correlation in Wind ErrorsAt a given altitude, the standard deviations of North and East wind errors at two spatially and temporally separated locations and instances of time, are found to be correlated.From a physical perspective, this is expected since the wind error at a particular physical location will change gradually over time due to changes in the prevailing winds.Furthermore, the wind errors at nearby locations will be similar due to the large-scale nature of atmospheric winds and the overall sparseness of the RUC-40 grid.Consideration of this spatio-temporal correlation is necessary to properly simulate wind errors during high-fidelity air traffic simulations.Without both temporal and spatial correlation, an aircraft flying along a given path could potentially experience unrealistic changes of the predicted wind such as a strong headwind immediately followed by a strong tailwind.Modeling the proper amount of correlation is important since too little correlation underestimates the effect of the wind error on ETA accuracy but too much correlation overestimates the effect of the wind error on ETA accuracy.An example of the correlation coefficient variation with respect to relative distance and time is shown in Figure 5.The correlation coefficient variation was calculated for each RUC-40 forecast using an N-day moving average, which in the present study was selected as 15 days.A strong correlation with respect to time is found at the same physical location (relative distance = 0 nmi).The correlation with respect to physical location at the same time (relative time = 0 minutes) is found to decrease exponentially.An interesting feature is the appearance of a 'ridge' of strong correlation along a line through the relative distance and relative time space.A possible cause is the availability of ACARS wind reports along specific arrival routes in the terminal area, instead of across a larger generic area.Wake vortices from a leading aircraft along an arrival route can affect the local wind pattern encountered by a following aircraft along the same route.This phenomenon requires further analysis which is beyond the scope of this paper.Generally, the wind errors are found to exhibit low correlation when spatially separated by more than 10 nmi.Similarly, the wind errors are found to exhibit low correlation when temporally separated by more than 5 minutes.Both of these scalesdistance and timeare significantly shorter than the typical air traffic management arrival planning horizons which are approximately 30-60 minutes and 150-250 nmi.As a result, the effect of wind error over sufficiently long routes will be attenuated.Both scales are also significantly shorter than the grid size and update period of the available RUC and RAP wind forecast products.The experimental High Resolution Rapid Refresh (HRRR) model with a grid size of 3km has the potential to provide a grid size and update period more similar to the typical arrival planning horizons. 13Thus, use of the HRRR wind forecast product in future highfidelity air traffic simulations might allow the effects of the wind errors spatial correlation to be simulated. +C. Seasonal Variation of Wind ErrorsFigure 6 and Figure 7 show the seasonal variation of North and East wind errors as a function of the day of the year and altitude.The plots show different wind error behavior for altitudes above and below 10,000 feet.For altitudes above 10,000 feet, there is a significant seasonal variation.In this altitude range, the wind errors at PHX are smallest during the summer months of July through September and largest during the winter months of January through April.Conversely, there is little seasonal variation of the wind forecast errors for altitudes below 10,000 feet.In order to evaluate the effect of the wind and wind error on ETAs, a set of 5000 sample wind profiles were created by applying the observed spatio-temporally correlated wind errors corresponding to each RUC-40 1-hour forecast wind file as illustrated in Figure 4 (See Section IId).ETAs were then calculated based on that set of 5000 sample wind profiles, the RUC-40 1-hour forecast winds, and zero winds for each route.Figure 8 shows the results of the simulation runs conducted using the RUC-40 1-hour forecast for January 20, 2011 at 1300Z along the EAGUL5 arrival route from the GUP en route transition fix to PHX Runway 08.The difference between the blue line and the red line is a measure of the wind magnitude's effect on ETA.The spread of the cyan colored histogram is a measure of the wind error's effect on ETA.Subsequent determination of the seasonal variations of the WMM and WUM are generated using results like those shown in Figure 8 for all RUC-40 1-hour forecasts and all arrival routes. +E. Seasonal Variation of the Wind Magnitude MetricFigure 9 shows the seasonal variation of the WMM (as defined by Eq. ( 1)).This WMM can be interpreted as a statistical measure of the average ETA change due to wind aloft.The time series clearly shows both high frequency variations (i.e., changes over a period of a few days) and low frequency variations (i.e., changes over a period of months or the entire season).The results indicate that the months of July and August are the time period when the winds have the lowest impact on ETA.During these summer months, the ETA change due to the winds is less than 5% of the flight time from the en route transition to the runway threshold.During the other months of the year, the ETA change due to winds increases to 10-15% with some periods spiking to 20-25%.Furthermore, during these summer months, the wind effect is more consistent and the high frequency variations are attenuated. +F. Seasonal Variation of Wind Uncertainty MetricFigure 10 shows the seasonal variation of the WUM based on the standard deviation metric (as defined by Eq. ( 3)).The WUM results are very similar to the WMM results discussed earlier.The time series clearly shows both high frequency and low frequency variations.The results again indicate that the months of July and August are the time period when the wind errors have the lowest impact on ETA variability.During these summer months, the estimated standard deviation of the ETA due to the wind error is approximately 1% of the flight time.During the other months of the year, the estimated standard deviation of the ETA due to wind error increases to approximately 2% with some periods spiking as high as 2.5-3%.Furthermore, during these summer months, the wind error effect is more consistent and the high frequency variations are attenuated.These results suggests that the ETA uncertainty induced by wind error will be less than 1 minute for an arrival phase of flight generally lasting 20-30 minutes.Figure 11 shows the seasonal variation of the WUM based on the 90-percentile interquartile range metric (as defined by Eq. ( 2)).The results behave nearly identical to the seasonal variation of the WUM based on the standard deviation.The magnitude is appropriately larger since the 90-percentile interquartile range is naturally larger than the standard deviation; for normal distributions, it is approximately 3.3 times larger. +VI. ConclusionsThe current paper investigates the effect of wind uncertainty resulting from forecast errors on the expected times-of-arrival.The study specifically focuses on the PHX terminal area, NOAA's RUC-40 1-hour wind forecast product, and the year of 2011.It can be concluded from the results of this study that the wind forecast errors are: (i) dependent on the altitude, (ii) spatio-temporally correlated, and (iii) exhibit modest seasonal variation.The wind magnitudes generally cause ETA variations less than 10% when compared to ETAs with zero wind.However, the wind magnitudes can cause ETA variations as high as 25% in limited cases.The wind uncertainty resulting from forecast errors generally cause ETA variations less than 7% when compared to the ETA with zero wind errors.Under certain conditions, the wind uncertainty resulting from forecast errors can reach 10%.The study serves as a formal basis for understanding the effect of wind uncertainty on the trajectory prediction of NextGen air traffic management concepts.Figure 1 .1Figure 1.Sample of ACARS Wind Reports from Aircraft in the Phoenix Terminal Area +Figure 2 and2Figure 2 and Figure 3 show the East flow and West arrival routes respectively in the Phoenix terminal area used in this study.Each of the arrival routes is shown in a different color.Table 1 lists the combinations of the Area Navigation (RNAV) Standard Terminal Arrival Routes (STAR), en route transition fixes, and arrival runways.The en route transitions were chosen to represent the longest route from each direction.Runways 08 and 26 were used to represent the other parallel runways at PHX -Runways 07L/07R in East Flow airport configurations and 25L/25R in West Flow airport configurations, respectively. +Figure 2 .2Figure 2. PHX East Flow Configuration Figure 3. PHX West Flow Configuration +Figure 5 .5Figure 5. Spatio-Temporal Correlation of North and East Wind Errors +Figure 6 .Figure 7 .Figure 8 .678Figure 6.Seasonal Variation of North Wind Errors at PHX +Figure 9 .9Figure 9. Seasonal Variation of the Wind Magnitude Metric +Figure 10 .Figure 11 .1011Figure 10.Seasonal Variation of Wind Uncertainty Based on Standard Deviation Metric + + + +Table 1 . STARs, Fixes, and Runway Combinations1RNAV STARTRANSITION FIXARRIVAL RUNWAYGEELA6BLH08GEELA6BLH26MAIER5BLD08MAIER5BLD26KOOLY4SSO08KOOLY4SSO26EAGUL5GUP08EAGUL5GUP26 + Downloaded by NASA AMES RESEARCH CENTER on June 20, 2014 | http://arc.aiaa.org| DOI: 10.2514/6.2014-2863 + + + + +AcknowledgmentsThe study was performed under the current research sponsored by a NASA Contract No. NNA13AA07C, Task Order CTO413 to Optimal Synthesis Inc. + + + + + + + + + The Next-Generation Air Transportation System's Joint Planning Environment: A Decision Support System + + EdgarWaggoner + + + ScottGoldsmith + + + JoshElliot + + 10.2514/6.2009-7011 + + + 9th AIAA Aviation Technology, Integration, and Operations Conference (ATIO) + + American Institute of Aeronautics and Astronautics + 30 September 2010 + + + Joint Planning and Development Office + Joint Planning and Development Office, "Concept of Operations for the Next Generation Air Transportation System, Version 3.2,", 30 September 2010. + + + + + Practice for Application of Federal Aviation Administration (FAA) Federal Aviation Regulations Part 21 Requirements to Unmanned Aircraft Systems (UAS) + 10.1520/f2505-07 + + + Federal Aviation Administration + Washington, DC + + ASTM International + Feb 10, 2009 + + + "FAA's NextGen Implementation Plan," Federal Aviation Administration, Washington, DC, Revised Feb 10, 2009 + + + + + A Concept for Robust, High Density Terminal Air Traffic Operations + + DougIsaacson + + + JohnRobinson + + + HarrySwenson + + + DallasDenery + + 10.2514/6.2010-9292 + + + 10th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Fort Worth, TX + + American Institute of Aeronautics and Astronautics + September 2010 + + + + Isaacson, D. R., Swenson, H. N., and Robinson III, J. E., "A Concept for Robust, High Density Terminal Air Traffic Operations," 10 th AIAA Aviation Technology, Integration, and Operations Conference (ATIO), Fort Worth, TX, 13-15 September 2010. + + + + + Program MADIS + 10.1016/b978-0-408-01396-3.50016-3 + + + + Kinetic Energy Storage + + Elsevier + 27 February 2013 + + + + MADIS website: http://madis.noaa.gov/, accessed on 27 February 2013. + + + + + ARINC communications addressing and reporting system /ACARS/ - The data link that got implemented and why + + NSteele, Jr. + + 10.2514/6.1977-1522 + + + + Computers in Aerospace Conference + + American Institute of Aeronautics and Astronautics + 27 February 2013 + + + Aircraft Communications Addressing and Reporting System (ACARS): http://www.arinc.com/products/voice_data_comm/acars.html, accessed on 27 February 2013. + + + + + + MDTandale + + + SSVaddi + + + PSengupta + + + SLin + + Spatio-Temporally Correlated Wind Uncertainty Model for Simulation of Terminal Airspace Operations + Los Angeles, CA + + August 2013 + + + + 13 th Aviation Technology, Integration, and Operations (ATIO) Conference + Tandale, M. D., Vaddi, S. S., Sengupta, P., and Lin, S., "Spatio-Temporally Correlated Wind Uncertainty Model for Simulation of Terminal Airspace Operations," 13 th Aviation Technology, Integration, and Operations (ATIO) Conference, Los Angeles, CA, 12-14 August 2013. + + + + + Evaluation of the Terminal Sequencing and Spacing system for Performance-Based Navigation arrivals + + JaneThipphavong + + + JaewooJung + + + HarrySwenson + + + LynneMartin + + + MelodyLin + + + JimmyNguyen + + 10.1109/dasc.2013.6712503 + + + 2013 IEEE/AIAA 32nd Digital Avionics Systems Conference (DASC) + + IEEE + June 2013 + + + + 11 th USA/Europe Air Traffic Management Research and Development Seminar + + + Thipphavong, J., Swenson, H., Jung, J., Witzberger, K., Lin, M., Nguyen, J., Martin, L., Downs, M., and Smith, T., "Evaluation of the Controller-Managed Spacing Tools, Flight-deck Interval Management, and Terminal Area Metering Capabilities for the ATM Technology Demonstration #1," 11 th USA/Europe Air Traffic Management Research and Development Seminar, 10-13 June 2013. + + + + + Initial Investigations of Controller Tools and Procedures for Schedule-Based Arrival Operations with Mixed Flight-Deck Interval Management Equipage + + ToddCallantine + + + ChristopherCabrall + + + MichaelKupfer + + + JoeyMercer + + + ThomasPrevot + + 10.2514/6.2012-5673 + + + 12th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference and 14th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference + + American Institute of Aeronautics and Astronautics + June 2013 + + + + 11 th USA/ + Callantine, T., Kupfer, M., Martin, L., and Prevot, T., "Simulations of Continuous Descent Operations with Arrival Management Automation and Mixed Flight-Deck Interval Management Equipage," 11 th USA/Europe Air Traffic Management Research and Development Seminar, 10-13 June 2013. + + + + + NOAA Paleoclimatology + + RapidRefresh + + 10.1108/rr-06-2013-0143 + + + + Reference Reviews + 0950-4125 + + 27 + 7 + + 27 February 2013 + Emerald + + + Rapid Refresh (RAP): http://rapidrefresh.noaa.gov/, accessed on 27 February 2013. + + + + + Accuracy of RUC-1 and RUC-2 Wind and Aircraft Trajectory Forecasts by Comparison with ACARS Observations + + BarryESchwartz + + + StanleyGBenjamin + + + StevenMGreen + + + MatthewRJardin + + 10.1175/1520-0434(2000)015<0313:aorarw>2.0.co;2 + + + Weather and Forecasting + Wea. Forecasting + 0882-8156 + 1520-0434 + + 15 + 3 + + + American Meteorological Society + + + Schwartz, B. E., Benjamin, S. G., Green, S. M., and Jardin, M. R., "Accuracy of RUC-1 and RUC-2 Wind and Aircraft Trajectory Forecasts By Comparisons with ACARS Observations," Weather and Forecasting, Vol. 15, No. + + + + + Wind analysis in aviation applications + + CWynnyk + + 10.1109/dasc.2012.6383045 + + + 2012 IEEE/AIAA 31st Digital Avionics Systems Conference (DASC) + + IEEE + 14 January 2013 + + + PowerPoint presentation + Wynnyk, C., "Meteorological Accuracy for Aviation Applications," MITRE Center for Advanced Aviation System Development Case No. 13-0246 (PowerPoint presentation), 14 January 2013. + + + + + Evaluation of high-resolution rapid refresh (HRRR) forecasts for extreme precipitation + + HaowenYue + 0000-0002-4068-8810 + + + MekonnenGebremichael + + 10.1088/2515-7620/ab9002 + + + + Environmental Research Communications + Environ. Res. Commun. + 2515-7620 + + 2 + 6 + 065004 + 5 May 2014 + IOP Publishing + + + High-Resolution Rapid Refresh (HRRR): http://ruc.noaa.gov/hrrr/, accessed on 5 May 2014. + + + + + + diff --git a/file753.txt b/file753.txt new file mode 100644 index 0000000000000000000000000000000000000000..c2d79f79c68562fbcd83c03cc98bb061854d7b06 --- /dev/null +++ b/file753.txt @@ -0,0 +1,433 @@ + + + + +American Institute of Aeronautics and Astronautics National Air Space (NAS), the commercial market will drive the rapid diversification of UAS technologies and missions to serve new market sectors in the United States and abroad. 7For example, numerous commercial UAS and sUAS applications have been proposed, with a wide range of vehicles and concept of operations (ConOps) such as lighter than air dirigibles inspecting photovoltaic arrays 8 , multi-hundred pound helicopters applying pesticides to farms 9 , infrastructure inspection and monitoring with multi-rotor and fixed-wing sUAS 10 , and aerial package delivery with UAS. 11n light of the anticipated increase in commercial UAS flights, in particular sUAS operations at low altitudes in airspace not currently managed by the FAA, NASA developed the UAS Traffic Management (UTM) project to research prototype technologies, such as dynamic geo-fencing and congestion management, to "enable safe, efficient low-altitude operations." 12The UTM prototype also applies a set of operational constraints, such as a maximum altitude flight ceiling and visual line of sight requirement, in traffic management.Previous works on UAS market value have been conducted without consideration of a traffic management perspective.This paper estimates the value of commercial sUAS markets supported by the UTM prototype technologies in the near future, and assesses the impact of operational constraints on the market value.This paper is organized as follows.Section II describes methods used to classify commercial sUAS market sectors that could be supported by the UTM by year 2020, and presents metrics to quantify market value of each sector.Section III discusses sensitivity of the market value to changes in the operational constraints.Section IV concludes the paper with a summary of key findings. +II. Analysis Methods +A. Commercial UAS Application Identification and ClassificationThe first step of the analysis involved characterizing the commercial UAS industry as broadly as possible by identifying the range of proposed vehicles and missions.Vehicle type, performance and cost data were collected for about 400 vehicles by examining UAS literature and manufacturer provided information.Additionally, 135 unique representative UAS applications were identified that, while not a complete listing, provided a representative sampling of the UAS applications landscape.The ConOps for each of the 135 applications was roughly defined, and applications with similar mission architectures and technologies were grouped into broad mission categories.This process resulted in the definition of five fundamental UAS mission categories that represent the basic ConOps of currently proposed commercial UAS activities.The five UAS mission categories proposed below more succinctly capture the anticipated commercial market operations than a 10 mission category classification introduced by Ref. 13 and adopted by Ref. 5.1) Aerial Photography, Sensing and Surveillance: Aerial photography, sensing and surveillance missions span from property patrol to infrastructure management to real estate appraisal.UAS executing these missions collect imagery and data with a variety of electromagnetic or acoustic sensors and equipment.2) Communications: UAS are proposed to serve as strategic communications relays for underserved regions.While a majority of these vehicles will be large, high altitude, long endurance UAS and are beyond the current scope of UTM, this mission type may find some applications with small vehicles at low altitudes in dynamically congested or highly remote areas, such as for rapidly configurable search and rescue communication or cellular networks.3) Transportation, Delivery and Interaction: The capacity of a UAS to potentially transport and deliver goods and services is highly anticipated.Aerial crop dusting within Visual Line of Sight (VLOS) by UAS has been a reality in Japan for over two decades, and the aerial delivery of products is a major anticipated development.Missions of this type have been met with a large degree of public attention due to the involvement and development efforts of large companies such as Amazon, Google and FedEx.This mission type also presents promise to more cost effectively move goods in areas with seasonally available or unavailable ground transportation infrastructure, such as in Alaska.4) Atmospheric and Earth Science: UAS provide unique capabilities to conduct air quality testing, weather exploration and other atmospheric science missions at low cost.Although similar to mission category 1, these missions often require specialized equipment and direct interaction of the UAS with the phenomenon under study.5) Audiovisual Presence: UAS may have marketable applications that simply rely upon their physical presence, perhaps enhanced through visual and acoustic means.These include aerial advertising, tour guiding, or visual indication for search and rescue.American Institute of Aeronautics and AstronauticsThe Appendix contains a table displaying the 135 applications organized by the fundamental UAS mission category of their ConOps.Since the UTM prototype consists of four Technology Capability Levels (TCLs) of increasing complexity, these applications are further identified with requisite enabling UTM TCL level through color-coding.Following is an excerpt from Ref. 12, describing TCLs 1 to 4.UTM TCL 1: Concluded field testing in August 2015/ongoing testing at FAA site.Addressed rural UAS operations for agriculture, firefighting and infrastructure monitoring.In this TCL, the UAS ground pilot reserved the airspace and adjusted the flight plan if notified of a conflict.UTM TCL 2: Tests in October 2016 to address beyond-visual line-of-sight operations in sparsely populated areas, and provide flight procedures and traffic rules for longer-range applications.UTM TCL 3: Tests in January 2018 to include cooperative and uncooperative UAS tracking capabilities to ensure collective safety of manned and unmanned operations over moderately populated areas.UTM TCL 4: Test dates to be determined.Would involve UAS operations in higher-density urban areas for tasks such as news gathering and package delivery, and large-scale contingency mitigation. +B. 2020 Leading Applications Down-SelectionThis step of the market analysis identified the commercial sUAS applications that are likely to achieve moderate to large market penetration by 2020.UTM TCL 1 and 2 flight demonstrations are expected to occur in the fall months of 2015 and 2016, respectively.UTM TCL 3 is expected in January 2018 and TCL 4 dates are to be determined.Based upon these dates, and considering the time between the demonstration, transfer of results to the FAA, and the resultant introduction of new UAS operations in the industry, an assumption was made for this study that only UAS applications supported by the results of UTM TCLs 1 and 2 would be implemented on sufficient scale in the market by 2020 to warrant investigation at this time.Under this assumption, 74 commercial applications were identified from the initial list of 135 to be supported by UTM TCLs 1 or 2. From a review of these 74 potential applications, 16 representational market sectors were selected to form the economic impact market analysis.The down-selection of the applications from 74 to 16 was necessary to remove small, economically inconsequential applications (such as archeology mapping), combine applications with similar ConOps (maritime scouting and maritime search and rescue), and best utilize the limited study resources on the most significant market sectors.Each sector was developed with the following traits: +•The sector represented one or more sUAS applications supported by UTM TCLs 1 or 2 +•The sector represented a substantial market penetration potential by 2020 •The sector had the potential for significant economic impact, or the potential to provide a valuable ancillary benefit such as the protection of life and property Table I on the next page presents the 16 market sectors considered by this economic impact market analysis and provides a brief description of each.Each of these sectors will be defined and reviewed in depth through the remainder of this analysis.It should be noted that some sectors, such as dam and bridge inspection, represent a joint civil and public (government) UAS market.For these sectors, the market analysis shall be completed under the assumption that the public entities contract UAS services from commercial companies.Under this assumption the economic value of such sectors is captured in the civil market rather than the public market. +C. Sector Economic Impact Market Analysis MethodologyAn economic impact market analysis was completed for each of the 16 commercial sUAS sectors identified.While the exact approach for the market analysis varied depending upon unique factors of each sector, the general approach is outlined in Figure 2. The approach is best described as an iterative data collection process to develop three modules for each sector: the sector ConOps, the sector market size, and the general business case for a provider operating a sUAS business in that sector.Three types of data sources were utilized to collect information and validate assumptions: subject matter expert (SME) interviews, primary source data collection, and multi-source estimation from secondary sources.The final steps of the market analysis involved selecting a representative sUAS vehicle to meet the needs of the specific ConOps and business case.Once a representative vehicle had been selected, the economic impact of the sUAS sector was estimated along a variety of market metrics.Each of the steps in this methodology is described in detail through this section.American Institute of Aeronautics and Astronautics +Sector ConOps DefinitionA thorough understanding of the ConOps for each application in a sector is essential to characterize the airspace and technology needs required for UTM.Care was taken to identify how the application was previously achieved (use of manned aircraft, ground vehicles, climbers, not conducted, etc.).The vehicle requirements were defined through parameters such as speed, operating altitude, endurance, range and payload.Finally, the airspace required by the ConOps was defined through metrics such as the underlying population density, UAS and manned aircraft operations density, and altitude ceilings. +Sector Size CharacterizationThe second step of the market analysis was to determine the size of the sector.The sector size was evaluated in this research using the number of single operator service providers the sector could support as a proxy.The process of determining this characteristic varied widely from sector to sector, however the basic approach involved determining the total sUAS service opportunity (number of missions demanded), applying market adoption (penetration) estimates, and then dividing the remaining service opportunity by the projected The economic impact of each market sector resulted from the characterization of proposed or existing business operations in the sector, and an understanding of their potential 2020 market penetration.American Institute of Aeronautics and Astronautics capabilities of a single operator service provider (operating a single vehicle at a time).This method resulted in defining the total number of operators the market sector could feasibly support.For example, the total sUAS service opportunity for power line inspection may be represented by the mileage of power lines inspected annually in the country (forecasted to 2020).This total sUAS service opportunity for power line inspection may be reduced by market adoption assumptions: only transmission lines will be inspected by sUAS (excludes distribution lines), and only transmission lines located in rural areas will be inspected (excludes lines in suburban and urban areas).The market adoption rate, typically estimated by a SME, is then applied and represents the percent of the service potential that is projected to be fulfilled by sUAS service providers in 2020.Three different adoption estimations were evaluated in this analysis due to the uncertain nature of this estimate: base, medium and high market adoption.Finally, the capabilities of a single operator service provider are based upon the operational capabilities of the representative sUAS chosen for the sector and, where available, interviews with current service providers in the market.This process for defining sector size is dependent upon a few key assumptions central to this analysis.First of all, it should be reiterated that the sector size characterization typically assumes that all sUAS services are provided by an independent, single operator service provider.This assumption suggests that a farmer or oil company hires a service provider to complete their sUAS mission rather than purchasing their own sUAS to internally meet the need.While this assumption does not accurately represent the mix of service models that will exist in the actual market, it is useful to assess the total value of the industry, albeit not the distribution of who captures this value.The independent sUAS service provider model provides an accurate assessment of the total value of the market opportunity.However, the model likely underestimates the number of operators and vehicles that will actually exist because the model promotes an implicit assumption of 100% efficiency of sUAS service delivery.Actual service delivery efficiency is limited by logistics, non-uniform service demand, and other market factors.Finally, a service provider is assumed in this analysis to consist of a single sUAS operating outfit (pilot, observer, other necessary personnel for the mission) using a single vehicle.This assumption is intended to represent the current regulations that require pilot-in-the-loop flight and prohibits multiple UAS operating under one pilot.The metric is therefore useful to capture total employment potential for the sector.However, the single vehicle per service provider assumption likely overestimates the number of companies that would exist to meet the sector need a business will likely employ more than one pilot and sUAS system.Changes in future regulations may also allow a single pilot to operate multiple vehicles simultaneously.As a result of the limitations of the independent sUAS service provider model outline above, a second service model was also explored.For some industries an internal service provider model was assumed where a client invests in in-house sUAS operating capabilities and assets.This assumption was applied to sectors such as maritime security and surveillance where an independent service provider model was unlikely to meet the mission needs.The use of an internal service provider model likely results in an overestimation of the number of sUAS in operation and an underestimation of the market value.The actual split between an internal or independent sUAS service model will depend heavily upon the cost of UAS insurance, the cost of the vehicles, and the regulatory requirements for pilot certification. +Sector Business Case DefinitionThe third step of the market analysis was to outline the business case for a hypothetical sUAS operator in the sector.The purpose of this process is to understand the financial operating characteristics of the sUAS application and project the total sector value metrics, based on the number of service providers previously estimated in the sector size characterization.SME engagement, financial records, market studies, operator service fees, and news articles were utilized to estimate a variety of cost information including sUAS service fees, current manned mission costs, estimated efficiency improvements, production downtime reduction and safety improvement value, among others.Where necessary, differences between sUAS operations costs and status quo operations costs were elicited for both an internal provider model and an independent service provider model.It should be noted that all financial figures in this analysis are presented in 2015 equivalent dollars.Additionally, it was assumed that except in the case of bridge, dam, and trestle inspections, all industry sectors are independent and an operator does not use their sUAS for operations in two sectors (i.e. they exclusively serve one sector). +Representative UAS SelectionThrough a review of the collected type data for the 400+ vehicles, six general categories of sUAS were defined based on size, capabilities, and costs.A representative sUAS was defined from the average characteristics of the systems in the group.The ConOps and business case for the sector was examined to select the appropriate representative sUAS based on the sector needs.Table II displays the six categories of sUAS considered in this American Institute of Aeronautics and Astronautics analysis and displays their key cost parameters.The operating cost presented is an estimate of the average hourly operating cost for the vehicle in an external provider service model considering vehicle fuel costs, maintenance, travel to site, insurance, and pilot wage.Note that the operating cost does not include an amortization of acquisition cost.The acquisition cost represents average costs of the platform, software, and basic sensor packages. +Characterizing Market Value through Various MetricsThe final step of the market analysis is to assess the market value of the sector.Numerous metrics exist to measure market value.The goal of this analysis is to provide a holistic impression of the commercial UAS industry.Therefore, a variety of market metrics are defined for each sector that display the economic impact on different constituencies including the UAS manufacturers, service providers, clients (service purchasers), and employees/operators.The seven metrics chosen for this analysis are described below.Each metric provides unique insight, but each also is subject to specific assumptions and uncertainties that must be recognized.Market Size Metrics -The first two metrics concern the potential revenue opportunity available to sUAS service providers.The external service provider model likely overestimates both metrics while the internal provider model assumption likely leads to an underestimate.Metrics three and four characterize the size of the industry that may be supported by the sector, and their associated capital investment costs, respectively.The serviceable addressable market and UAS capital investment may be considered direct economic impact metrics. +1) Total Addressable Market (TAM)-The TAM represents the potential revenue opportunity for a sUAS service provider if the full market share were to be met by the UAS industry.2) Serviceable Addressable Market (SAM) -The SAM represents the potential revenue opportunity for a sUAS service provider considering limitations such as market penetration, service scalability, and non-UAS competition, among others.The SAM and TAM may best be understood through Figure 3.The blue rectangle represents the total sUAS market, across all sectors.The orange circle, or sector TAM, is the total value of operations in a specific sector if the entire sector need could be fulfilled through sUAS services.For example, the orange circle may be thought of in the power line inspection industry as the case where every mile of power line is inspected per regulation by sUAS services.The green circle represents the sector SAM, or the total market value that is actually anticipated to be captured by sUAS services in the year 2020.The green circle is smaller than the orange circle because sUAS will not likely reach 100% market penetration by 2020, if ever.There are situations where companies may desire to use traditional techniques or where sUAS are not practical.The sector TAM is the entire value of the UAS market that could potentially be captured assuming 100% market penetration, while the SAM is the subset that is forecast to be captured.American Institute of Aeronautics and Astronautics3) UAS Application Industry Size -For this analysis the size of the sUAS industry for a specific application is presented as the number of single operator service providers needed to meet the market service demands.Each service provider is assumed to provide 250 full days of operation per year.In implementation, it is likely there will be more service providers than predicted in this study as markets do not operate at 100% efficiency due to factors such as competition, weather, and logistics.As the market matures it is also likely that single operator service providers will coalesce into multi-operator service provider companies; this dynamic is not captured by the current analysis.4) UAS Capital Investment -The capital investment metric represents the potential sales of sUAS necessary to meet the demands of the service providers.The capital investment of the service providers is assumed to be the cost of their sUAS and basic command and control equipment.Specialized sensor package needs, company overhead, transportation vehicles, advanced ground stations and other likely capital costs that a sUAS service provider may encounter are not included in this initial analysis.Additional Direct Economic Impact Metrics -These economic impact metrics characterize capital savings directly accrued by a client as a result of UAS adoption. +5)Operations Savings Potential -The operations savings potential metric represents the historical operations costs to complete the mission through conventional methods less the sUAS serviceable addressable market (costs to complete the service with sUAS) for the segment of the market utilizing sUAS services.This metric therefore captures the operations savings a company may realize through use of sUAS services compared to the status quo service. +6)Revenue Recovery Potential -The revenue recovery potential represents client value that may be realized as a result of new or enhanced capabilities provided by the sUAS application.Potential revenue recovery may typically be considered as "production recovery" or as "failure avoidance."Production recovery represents the client value of reduced downtime for inspection or repair.Failure avoidance represents customer value due to the reduction of unplanned repair and damage payments.External Economic Impact Metric -Potential value realized by entities as a result of the sUAS operations, but not considered as an economic activity.7) Safety Improvement Potential -This metric seeks to characterize the value of safety improvements resulting from sUAS operation.This metric is typically presented in this research as a discussion of the safety improvements relative to current operations possible through sUAS utilization.It should be noted that multiple other metrics exist to characterize market size, particularly indirect and induced economic impact metrics.For example, AUVSI included in Ref. 14 multiple forms of indirect impacts and induced impacts that characterize the downstream multiplier effect of money infused into an economy (i.e."reexpenditures").This multiplier effect may at times be quite large.This analysis chose not to consider such indirect or induced impacts (except safety improvement) due to the difficulty of estimating these factors and the inflationary effect they have on market prediction by considering impacts outside of the direct UAS industry. +III. ResultsSector by sector economic impact market analyses were completed according to the methodology outlined in the previous section.Sensitivity studies within some sectors also provide insight into how study assumptions and UTM operational constraints impact the projected value of specific sUAS sectors.In order to review the potential overall influence of the UTM project on the UAS industry, the individual sectors are compared to one another and aggregate economic impact information is presented.The six direct impact economic metrics enable a detailed comparison among sectors in terms of where value is generated, which stakeholders capture the value generated, and which sectors are more sensitive to certain changes in ConOps or UTM TCL operational constraints.The discussion presents a variety of figures drawn from the economic impact market analysis.Key takeaways, as well as limitations and assumptions, are discussed for each figure.The goal of presenting the data through this means, rather than providing a simple overall dollar comparison of the sector values, is to promote a holistic understanding of the market landscape in the reader and communicate market trends and study limitations, rather than providing a direct numeric sector ranking.American Institute of Aeronautics and Astronautics +A. Full Market Commercial sUAS Economic Impact Findings for 2020Figure 4 displays the aggregate direct economic impact metrics for each of the commercial sUAS sectors supported by UTM TCLs 1 and 2. The "base," or lowest market adoption assumptions were utilized to create Figure 4; therefore the market projections may be thought of as conservative estimations given the assumptions of this analysis.The direct economic impact metrics include the serviceable addressable market (total revenue opportunity captured by sUAS service providers), capital investment (value of sector sUAS), operations savings (expenses saved by converting existing services to sUAS services) and revenue recovery (revenue gained as a result of increased production or reduced failure due to sUAS services).It is immediately recognizable from Figure 4 that the "cropland precision agriculture" sector is forecasted to be the largest commercial sUAS market by a significant margin.The "construction" and "heavy industry inspection" sectors are the next two largest sectors, but each is a mere fraction of the potential total economic impact of cropland precision agriculture.Beyond providing an overall picture of the commercial sUAS industry market projections in 2020, significant structural elements of the industry may be identified in Figure 4. First of all, it is apparent that a significant proportion of the direct economic impact of the three largest sectors is categorized as revenue recovery.This suggests that commercial sUAS services will provide cropland precision agriculture, heavy industry inspection and construction with a means to significantly increase their revenue over what was previously possible.Upon further investigation, the drivers of this revenue recovery are apparent.Unmanned aircraft systems in cropland precision agriculture have been estimated to provide a crop yield increase of up to 7%. 15 Resultantly, because the value of US crop production is $210 billion dollars annually, a 7% increase in revenue (even with a rather low adoption rate of roughly 6% of the market) results in the nearly one billion dollars of annual revenue recovery to the agricultural industry shown in Figure 4; this number grows rapidly with higher predicted market penetration assumptions.The heavy industry sector experiences similarly large revenue recovery due to the ability of facilities to continue the operation of some mission critical infrastructure during sUAS visual inspection, a feat that was not possible with hazardous manned inspections.Finally, the construction sector predicts sUAS will improve on-time project completion and identify building errors early.This capability, even if only affecting 0.5% of project revenue for a small percentage of the market, produces significant value in the one trillion dollar industry.A second takeaway from Figure 4 is that the other significant proportion of the precision agriculture value results from operations savings to the farmers.Industry advocacy groups suggest that precision agriculture by sUAS may reduce chemical use by up to 40%. 15The US agricultural chemical and fertilizer industry is a $30 billion dollar M American Institute of Aeronautics and Astronautics industry.Therefore a moderate reduction in chemical usage even in a small number of farms (low market penetration by 2020) results in significant overall economic impact.Figure 4 supports the reader to understand a few key takeaways and limitations of the current market analysis.First, cropland precision agriculture in particular may reflect the influence of technology and market optimism.For many years, precision agriculture has been identified by multiple market studies as the leading commercial application for UAS technologies. 14As a result, this application has had an unequal share of media coverage and service provider investment.It may therefore be possible that the estimations for sUAS impact, especially the projected increases in crop yield and reduction in chemical use, may be overstated.However, even if both of these parameters are reduced by 50%, the economic impact of the precision agriculture industry remains nearly as large as all the other sectors combined.Therefore, this analysis supports previous findings by other researchers and firms that precision agriculture represents one of the largest commercial sUAS market for the young industry.While Figure 4 supported a high level analysis of the economic impact structure of the three largest commercial UAS sectors, it did not provide the resolution to support the same analysis for the remainder of the sectors that exhibit smaller potential economic impacts.Therefore, Figure 5 was developed to display the economic impact composition of each sector based on the four direct impact economic metrics.While the sectors have dramatically different total values, this comparison enables a study of the structural differences between the sectors in terms of where value is derived.From Figure 5, it may be seen that the different sectors create economic impact through a wide mix of mechanisms.Some sectors, such as "maritime surveillance and scouting" and "search and rescue" create economic value nearly completely through capital investment in UAS and associated technologies.These two sectors are characterized by a large number of operators with a low rate of deployment but a high desired rate of readiness.Sectors such as "construction," "heavy industry inspection," and "solar array inspection" produce a majority of their economic impact by recovering revenue for the client.Yet another group of sectors, including "rail inspection," "communications tower inspection" and "aerial application," derive a majority of their economic impact from the serviceable addressable market, or the value captured by sUAS service providers.The differences between the commercial sUAS sectors in terms of where value is created also influences the distribution of stakeholder benefits.For example, those sectors that have a high percentage of their economic impact categorized as capital investment will most significantly benefit the sUAS manufactures, while potentially deterring Figure 5. Direct economic impacts structure analysis for 2020 commercial sUAS market.Industry composite displaying the contribution fractions for each of the four direction economic impact metrics to each sector.American Institute of Aeronautics and Astronautics sUAS service providers from entering the market unless there is a large SAM to capture.Sectors with large revenue recovery and operations savings potential will benefit the sUAS clients most significantly.It should be noted that Figure 5 presents the total economic impact of all the sectors as a normalized value.This may lead to confusion that an industry such as "solar array inspection" has a greater total revenue recovery than cropland precision agriculture, for example.However, Figure 5 is only presenting the composition percentages of the four impact metrics within each industry.Therefore, in this case, "cropland precision agriculture" delivers far more revenue recovery to farmers than "solar array inspection" returns to utility companies because the total market size of the former is much larger than that of the latter. +B. Commercial sUAS Market Impact Findings for UTM TCLs 1 and 2The remainder of the discussion shall focus on the serviceable addressable market and capital investment metrics only.While the revenue recovery and operations savings metrics represent a majority of the potential economic impacts of sUAS up to 2020, they represent value delivered to the client and not necessarily value captured within the sUAS industry (by manufactures or operators).Figure 4 and Figure 5 displayed the industries that represent the greatest overall economic impact.The interpretation of these figures may provide insight into UTM TCL operational constraints that will best support these high impact industries.However, the remainder of this analysis will focus on how sUAS ConOps limitations and operational constraints may influence the SAM and capital investment metrics, which act as a proxy for the internal revenue value of the sUAS industry.Figure 6 displays the relative contributions of all sectors to the total SAM supported by UTM TCL 1. From the findings of UTM TCL 1, these operations are assumed to be constrained to line of sight flight in geofenced airspace, by a single sUAS, weighing less than 55 lbs, without people or significant property underneath.From Figure 6 it is evident that cropland precision agriculture accounts for 50% of the total market, and agriculture applications in general accounts for 84% of the total.These percentages indicate that UTM TCL 1 produced technologies and capabilities that enable wide scale sUAS adoption in the agricultural industry and capture much of the potential value therein.Agriculture applications account for a majority of the UTM TCL 1 supported market as they are not generally constrained by the line of sight or population overflight constraints; however, as will be shown in the sensitivity studies in sub-section C, the 55 lb UAS weight limitation is binding on aerial application.Many of the other sectors are not supported to operate sUAS to any degree, or to only small degree due to the operational constraints of UTM TCL 1. Please note that the economic impacts presented are in 2015 dollars, calculated with the service provider model market (versus an internal provider market) assumption, and under the base market adoption assumption.7 displays the increase in total SAM and the shift in relative sector value contribution when sectors supported by UTM TCL 2 are included.The beyond visual line of sight capability supported by UTM TCL 2 is key in enabling operations such as railroad inspection and oil & gas pipeline inspection, each representing significant market wedges.Additionally, the UTM TCL 2 allowance to fly over personnel involved in the sUAS operation enables applications in the construction sector.Although farming applications do not realize significant economic gains between UTM TCLs 1 and 2, they continue to represent over 50% of the potential market. +C. Commercial sUAS Market Impact Finding Sensitivity Studies 1. Service Provider Model SensitivityAs discussed previously, a significant assumption in this study is that an independent service provider model will be used across all industries.This model assumes a client hires an independent sUAS service provider to complete the application.This model is viable for many industries such as solar panel inspection and bridge, trestle, and dam inspection where the utilization for a single client may be low and likely does not warrant the capital costs of the sUAS, training, insurance and other expenses of operation.The model less accurately represents industries such as rail and pipeline inspection where industries currently have a mix of internal and external providers.Finally, the independent service provider model is unreasonable for sectors such as maritime surveillance and security where it would not be economically viable to put a service provider on every seagoing vessel.In order to explore the sensitivity of the provider model assumption on the projected 2020 commercial sUAS market, an alternative internal service provider model was applied to industries that were likely to conduct sUAS operations without a 3 rd party service provider.Figure 8 displays the change in the SAM for the "mixed service provider market" where the three agriculture sectors, rail inspection, dam, trestle and bridge inspection, search and rescue, maritime surveillance and scouting, and construction sectors use the internal service provider model while the other sectors continue to use the independent provider model.The impact of the mixed service provider market is to generally reduce the SAM of the sectors that adopt the internal service provide model.This occurs because the internal operators do not charge a profit margin in their hourly sUAS operation fee.This means the SAM represents the estimated cost for labor, travel, insurance, energy and other non-revenue costs.The result is to significantly reduce the total commercial sUAS SAM from $284 million to $169 million.Much of this service provider revenue potential loss occurs in the farming sectors.Under the mixed provider model market assumption, oil and gas pipeline inspection emerges as the new leading serviceable addressable market for sUAS application by UTM TCLs 1 and 2. Remember, however, that this sector is not possible through UTM TCL 1 alone due to the visual line of sight limitation.While a change of market model from independent to internal service providers lowers the SAM of the farming sectors, it does quite the opposite in terms of capital investment.The independent service provider model simulates a highly efficient sUAS market where vehicles and pilots operate at 100% capacity across the sector (an overestimate when considering market distribution, logistical and temporal efficiency, competition, etc).The high efficiency assumption results in an underestimation of the capital investment necessary to meet the sector needs.Therefore, as Figure 9 and Figure 10 display, the mixed service provider market dramatically increases the capital investment in those industries that adopt an internal provider model.The total capital investment of the industry increases from $89 million to $2.8 billion.Most significantly, the farming sectors expand from accounting for 52% of the capital investment in the industry to 95%.This is a result of the large number of potential sUAS purchasing farms and their low utilization of the vehicles.A farm may only use a sUAS for 4 to 40 days of operations a year on average (pasture and range management may be much higher) resultantly paying a large cost per usage in order to have the vehicles on hand and readily deployable.It is not an unreasonable market assumption for farms to purchase their own vehicles as farmers may make operations decisions from day to day based on weather conditions, a factor that would make scheduling an independent sUAS service provider difficult.The discussion above suggests that the commercial sUAS market is highly sensitive to the service provider model that is assumed.In practice, the commercial sUAS market will likely contain sectors exhibiting both types of models, as well as mixed models.The driving factor that determines which model will prevail in a sector is likely to be regulation.While manufacturers and technologists may reduce the costs of advanced sUAS and increase performance so most operators can easily fly the vehicles, the FAA training requirements for pilots may prevent the widespread internal provider model and operational use of these The time and expense to certify as a private pilot (currently required to fly a UAS commercially) will dissuade most farmers and other potential operators from purchasing their own systems, thus promoting the independent service provider model. +Market Adoption Rate SensitivityA second significant assumption made through the market analysis concerns the unknown market adoption rate for commercial sUAS in each sector by 2020.Emerging technologies typically follow an "s-curve" adoption rate with initial usage among a small number of early-adopters.This is followed by a rapid period of growth in the mainstream market and concludes with an asymptotic approach to the steady-state adoption level as late-adopters begin to use the product or service.The slope of the s-curve (rate of market adoption) is dependent upon factors unique to the particular technology and industry. 16onfounding the situation further for commercial sUAS adoption is regulatory uncertainty.Figure 11 displays the proposed influence of regulatory uncertainty on sUAS market adoption.The blue curve represents what a traditional s-curve market adoption for commercial sUAS may have looked like.However, entities such as AUVSI have proposed that sUAS market adoption has been stunted by regulatory uncertainty and restrictive FAA policies for commercial sUAS operations.The red curve in Figure 11 visualizes the impact of such regulatory uncertainty through a lesser slope and delayed inflection of the technology adoption curve.It is anticipated that upon the establishment of permanent, favorable sUAS policies by the FAA, the industry will rapidly expand as manufactures and operators previously constrained by regulations compete for market segments. 14igure 11 also reveals two potentially interesting impacts of regulation uncertainty on the commercial sUAS industry.First of all, the area between the two curves (designated as region 1) may be interpreted as proportional to the total economic value either lost or delayed due to the influence of uncertain regulations.In an unimpeded market (blue curve), the market would have adopted sUAS technologies more quickly and derived the resulting commercial value from their use.However, in the case of sUAS, uncertain regulations and the temporary measures have stunted the market and delayed the market realization of sUAS potential value.The second interesting impact of regulatory uncertainly occurs during the late stages of market adoption and is indicative of a "path dependency" of technology adoption (designated as region 2 in Figure 11).As can be seen, two potential paths are represented for the stunted sUAS technology adoption curve: one that exceeds the natural curve adoption asymptote and one that achieves a lower adoption percentage.In actuality, there are an unlimited number of potential paths for the regulatory delayed curve resulting in a variety of final adoption states.While regulatory uncertainty may initially delay the adoption of sUAS technologies and stunt the market, this condition does not necessarily hobble the long-term market.If delayed rule-making by the FAA enables more advanced technologies to be developed to ensure safer sUAS operations, then perhaps incidents may be avoided that could have hypothetically crippled the natural market.Alternatively, some industries may find alternative solutions to sUAS while the technology is limited by uncertain regulation meaning these markets could potentially be unrecoverable.The concept of path dependency in market adoption is common in economics and the social sciences.This paper does not comment further on what the long-term impact of uncertain regulations in the commercial sUAS market may be, but this should be considered as an area for future efforts.The influence of regulatory uncertainty increases the uncertainty in the estimate of market adoption rate.Subject matter expertise was primarily used to develop these estimations for this study.Sensitivity of market adoption rate were conducted in each sector by calculating the economic impact with a base, medium and high adoption rate.A comparison of the forecasted markets reveals that the market adoption percentage will significantly influence the size of the commercial sUAS industry in 2020.As more accurate forecast models are developed, they should be implemented to refine the assumptions of this analysis.Furthermore, if final FAA regulations are developed before 2020 and lead to a step-change in the market adoption (rapid adoption by a significant portion of industry) the 2020 economic impact may reflect the higher market penetration assumptions of this analysis. +Technology Capability Level ConOps Constraint SensitivitiesIn addition to impacting the overall technology market adoption rate for commercial sUAS as shown above, regulatory constraints and UTM TCL operational constraints may also significantly influence the economic impact of some sectors.The dependence of three sectors on operations constraints for maximum altitude, maximum takeoff weight and population density restrictions for operation over people are presented as examples of how these sensitivity studies may inform the requirements for UTM or policy officials.The first operation constraint sensitivity study explores the variability in the communications tower inspection sector as a result of the current 500 ft commercial sUAS operation ceiling proposed by the FAA. 17Communications tower inspection involves the at-height imaging by UAS of structures ranging from under 100 ft to over 2000 ft in height.Figure 12 suggests that the increase of the sUAS ceiling above 500 ft to at least 2000 ft, while keeping the other conditions of UTM TCL 1 constant, increases the total direct economic impact of the sector by roughly 22%.The increase is a result of the additional structures that a sUAS service provider will be able to inspect due to the increase in operating altitude.Furthermore, current communications tower inspections are completed by climbers exposing these individuals to significant risk.Expanding the sUAS inspection services to additional facilities may therefore also provide potentially substantial safety benefits.The second operating constraint sensitivity study explores the variability in economic impact of the aerial application sector as a result of the sUAS 55 lb maximum takeoff weight operation restriction. 3Aerial application involves low altitude flight over croplands and the spraying of chemicals and other substances.Small UAS have been shown as capable of fulfilling the spraying needs of small farms and specialty crop spraying operations. 10However, current manned aerial application vehicles can carry thousands of pounds of payload.These missions, which represent the major share of the current aerial application sector, would not be possible under a 55 lb sUAS operating limit.Figure 13 displays that the relaxation of the maximum takeoff weight restriction for an aerial applicator sUAS up to 300 lbs may result in an increase in economic The increase is dramatic due to the significantly larger market sUAS will be able to serve, and the cost savings the vehicles are anticipated to provide compared to current manned aerial operations.There was no economic impact due to revenue recovery from aerial application as the missions are assumed to currently be completed by manned aircraft or ground based systems.The final operation constraint sensitivity study explores the variability of the economic impact of the communications tower inspection sector as a result of the population density overflight limitations of the UTM TCLs.Technology Capability Levels 1 and 2 do not support operations over people or significant property.These UTM TCLs will support communications tower inspections only in remote areas.Technology Capability Level 3 is proposed to support operations over moderately populated areas such as in a suburban setting.Finally, UTM TCL 4 is proposed to support operations over densely populated areas such as in an urban setting.Figure 14 indicates a significant dependence of economic impact on the population overflight standard supported by the UTM TCLs.The sensitivity of economic impact to population overflight allowance varies between sectors depending upon what percent of the missions are located in populated areas, however nearly all sectors have greater market potential with flight over suburban and urban areas. +IV. ConclusionThe commercial UAS economic impact market analysis presented in this study projects the state of the industry in the year 2020 through the consideration 16 leading market sectors.The 16 sectors were derived from 74 commercial sUAS applications that were predicted to be supported with the findings from the NASA UAS Traffic Management (UTM) project.The analysis determined the economic impact of each sector through the application of seven market metrics.Three market adoption assumptions and two service provider models were also considered to capture uncertainty in the rate of UAS services market penetration and type of service model, respectively.This work is unique in that it adopted a traffic management perspective to assess the potential economic value of new UAS technologies and services with respect to UTM Technology Capability Level (TCL) phased operational constraints.The market analysis found that sUAS application for precision agriculture created the largest economic impact by a significant degree compared to any other sector.The bulk of the precision agriculture sector value was supported by UTM TCL 1 results that enable line of sight flight over sparsely populated areas.The development of beyond visual line of sight capabilities in UTM TCL 2 was shown to be a key enabling technology in numerous other business sectors including pipeline and railroad inspection, construction, and maritime applications.Technology Capability Level 2 findings were shown to result in a serviceable addressable market (SAM) expansion of over 30% for commercial sUAS.A variety of sensitivity studies were completed to characterize the variability of select UAS sectors' economic value to key UTM operational constraints.A relaxation of the initial 500 ft maximum operating altitude constraint to 2000 ft provided a moderate market expansion for the communications tower inspection industry.While such a ceiling is not reasonable for all airspaces, this study suggests that UTM capabilities to relax the altitude constraint in the vicinity of large communication towers may provide increased economic value and personnel safety.Furthermore, the proposed maximum takeoff weight constraint of 55 lbs was found to have a significant market impact for the aerial application sector, but have little to no effect in the other sectors.If later UTM TCLs supported the use of agricultural UAS of weights up to 300 lbs, this analysis suggests an up to 450% market expansion may occur for UAS aerial application.Finally, the UTM TCL population density overflight constraints were found to have a large impact in the market forecast for a majority of the sectors.The support of UAS operations over suburban and urban regions though UTM TCLs 3 and 4 is projected to increase the market impact communications tower inspection sector by up to 600% compared to UTM TCLs 1 and 2 rural flight area limitations.Similar market expansions are expected in many other UAS sectors.Figure 3 .3Figure 3. Visual representation of the Serviceable Addressable Market (SAM) and Total AddressableMarket (SAM) for a UAS sector.The sector TAM is the entire value of the UAS market that could potentially be captured assuming 100% market penetration, while the SAM is the subset that is forecast to be captured. +Figure 4 .4Figure 4. Direct economic impacts of 2020 commercial sUAS market.Industry composite displaying projected direct market impacts of commercial applications enabled by UTM technologies in 2020. +Figure 6 .6Figure 6.UTM Technology Capability Level 1 enabled sector contributions to 2020 sUAS SAM with base market penetration and independent service provider model assumptions.The base, or lowest projected market penetration estimates were utilized.The independent service provider model assumes sUAS are owned and operated by 3 rd party contractors.Only sectors enabled by UTM TCL 1 ConOps and technologies are considered. +FigureFigure7displays the increase in total SAM and the shift in relative sector value contribution when sectors supported by UTM TCL 2 are included.The beyond visual line of sight capability supported by UTM TCL 2 is key in enabling operations such as railroad inspection and oil & gas pipeline inspection, each representing significant market wedges.Additionally, the UTM TCL 2 allowance to fly over personnel involved in the sUAS operation enables applications in the construction sector.Although farming applications do not realize significant economic gains between UTM TCLs 1 and 2, they continue to represent over 50% of the potential market. +Figure 7 .7Figure 7. UTM Technology Capability Level 2 enabled sector contributions to 2020 sUAS SAM with base market penetration and independent service provider model assumptions.The base, or lowest projected market penetration estimates were utilized.The independent service provider model assumes sUAS are owned and operated by 3 rd party contractors.Only sectors enabled by UTM TCLs 1 or 2 ConOps and technologies are considered. +Figure 8 .8Figure 8. UTM Technology Capability Level 2 enabled sector contributions to 2020 sUAS SAM with base market penetration and mixed service provider market assumptions.The base, or lowest projected market penetration estimates were utilized.The mixed service provider model assumes some sectors contact sUAS services, while other sectors internally provide their own sUAS services.Sectors enabled by UTM TCLs 1 or 2 ConOps and technologies are considered. +Figure 9 .9Figure 9. Technology Capability Level 2 enabled sector contributions to total sUAS capital investment with base market penetration and independent service provider model assumptions. +Figure 10 .10Figure 10.Technology Capability Level 2 enabled sector contributions to 2020 sUAS capital investment with base market penetration and mixed service provider market assumptions. +Figure 11 .11Figure 11.The influence of regulatory uncertainty on sUAS market adoption.The delay of FAA sUAS rulemaking may have slowed market adoption. +Figure 12 .12Figure 12.Communications tower inspection operating altitude constraint sensitivity study.A 500ft operational ceiling reduces potential 2020 market value up to 18%. +Figure 14 .14Figure 14.Communications tower inspection population density overflight constraint sensitivity study.Flight over more populated areas is expected to significantly increase potential 2020 market value of communications tower inspection. + +Table I . Projected 2020 leading commercial sUAS market sectors. ThroughIa review of 135 proposed sUAS applications, sixteen market sectors were identified where market entry depends upon technologies and ConOps testing that is anticipated to be completed through NASA UTM by 2020.Market SectorDescriptionFarmingCropland Precision Agriculture Pasture/Range ManagementInspection of crops, pasture and property through the use of a variety of imaging sensors Activities such as herd counting, pasture inspection, sick/injured animal identification, birth detection and containment inspection, among othersAerial ApplicationThe spreading of chemicals, water, or other substances on cropsCentralizedInfrastructureInspection ComplexWind Turbine Inspection Turbine blade and superstructure visual inspections Solar Array Inspection Visual and infrared inspection of photovoltaic (PV) solar arrays Dam Inspection Visual inspection of potential hazardous areas such as dam face, spillway, abutments, reservoir, and other structures Infrastructure inspection in power plants, oil rigs, petroleum refineries, Heavy Industry Inspection petrochemical plants, steel mills, pulp and paper mills, cement facilities and coal processing plants Communications Tower Inspection At height inspection of cell, microwave, radio and television towersDecentralizedInfrastructureInspectionPipeline Inspection Power Line Inspection Railroad Inspection Bridge Inspection Railroad Trestle InspectionAirborne inspection of difficult or dangerous pipeline segments Pylon and at-height wire inspection of transmission lines as well as right of way intrusion monitoring Visual inspection of railroad tracks At-height inspection of bridge superstructure and substructure At-height inspection of trestle superstructure and substructureSearch and RescueAirborne communications, detection and equipment transport assetsOtherMaritime Security and Surveillance ConstructionAssets for anti-piracy, weather monitoring, hull inspection, channel navigation, fish spotting and route mapping, among others Airborne surveying, safety and quality inspection, and documenting of build progress +Figure 2. UTM economic impact market analysis approach. +Table II . sUAS general categories and representational vehicle information.IIThis research sought to define generic categories that represent average acquisition and operating costs for sUAS of varying payload and operating capabilities.sUAS CategoryAcquisitionOperatingExample VehicleCostCost<10 lb multipurpose multirotor$4000$753DR X-8 Octocopter25 lb to 55 lb multipurpose multirotor$20,000$150Airbornedrones A-2<55 lb agriculture sprayer multirotor$15,000$100HSE AG-V8A Octocopter>55 lb agriculture sprayer helicopter$150,000$300Yamaha R-MAX<10 lb fixed wing$25,000$150Precision Hawk Lancaster25 lb to 55 lb fixed wing$50,000$500Insitu ScanEagle + Downloaded by NASA AMES RESEARCH CENTER on August 18, 2016 | http://arc.aiaa.org| DOI: 10.2514/6.2016-4373 + + + + +AcknowledgmentsThis material is based upon work supported by the National Aeronautics and Space Administration under Grant Number NNX14AT14H issued through the NASA Education Aeronautics Scholarship Program. + + + +American Institute of Aeronautics and Astronautics + + + + + + + U.S. Unmanned Aerial Systems + + JGertler + + + 2012 + Congressional Research Service + Washington, DC + + + Gertler, J. (2012). U.S. Unmanned Aerial Systems. Washington, DC: Congressional Research Service. + + + + + International Finance Corporation + 10.1596/978-1-4648-0484-7_international_finance_corporation + + + The World Bank Group A to Z 2016 + Fairfax, VA + + The World Bank + 2015 + + + + Teal Group Corporation + + + th ed. + Teal Group Corporation. (2015). World Unmanned Aerial Vehicle Systems: 2015 Market Profile and Forecast (12 th ed.). Fairfax, VA: Teal Group Corporation. + + + + + The Great Reform Act of 1832 and the Political Modernization of England + 10.1086/ahr/100.2.411 + + + The American Historical Review + 1937-5239 + + 112 + 95 + 126 + + Oxford University Press (OUP) + + + § + FAA Modernization and Reform Act of 2012. Pub. L. No. 112-95, § 331(6), 126 + + + + + Unmanned Aerial Vehicles unique cost estimating requirements + + PMalone + + + HApgar + + + SStukes + + + SSterk + + 10.1109/aero.2013.6496852 + + + 2013 IEEE Aerospace Conference + Big Sky, MT + + IEEE + 2013 + + + Malone, P., Apgar, H., Stukes, S., & Sterk, S. (2013). Unmanned Aerial Vehicles Unique Cost Estimating Requirements. IEEE Aerospace Conference. Big Sky, MT: IEEE. + + + + + Supplement analysis, Southpoint power project + + Volpe + + 10.2172/353400 + + 2013 + Office of Scientific and Technical Information (OSTI) + Cambridge, MA + + + John A. Volpe National Transportation Systems Center + + + Volpe. (2013). Unmanned Aircraft System (UAS) Service Demand 2015-2035: Literature Review and Projections of Future Usage (Version 0.1). Cambridge, MA: John A. Volpe National Transportation Systems Center. + + + + + System Technologies for Civilian Missions + + TSkrzypietz + + 10.2514/5.9781624103544.0043.0064 + + + Civil and Commercial Unmanned Aircraft Systems + Potsdam, Germany + + American Institute of Aeronautics and Astronautics, Inc. + 2012 + + + + Skrzypietz, T. (2012). Unmanned Aircraft Systems for Civilian Missions. Potsdam, Germany: Brandenburg Institute for Society and Security. + + + + + FAA Aviation Forecasts: Fiscal Years 1981-1992. Federal Aviation Administration, U.S. Department of Transportation, Washington, D.C. 20591. 1980. 69p + 10.1177/004728758102000159 + + + Journal of Travel Research + Journal of Travel Research + 0047-2875 + 1552-6763 + + 20 + 1 + + 2015 + SAGE Publications + Washington DC + + + Federal Aviation Administration. (2015). FAA Aerospace Forecast: Fiscal Years 2015-2035. Washington DC: Department of Transportation. + + + + + Light Unmanned Aerial Vehicles (UAVs) for Cooperative Inspection of PV Plants + + PaoloBellezzaQuater + + + FrancescoGrimaccia + + + SoniaLeva + + + MarcoMussetta + + + MohammadrezaAghaei + + 10.1109/jphotov.2014.2323714 + + + IEEE Journal of Photovoltaics + IEEE J. Photovoltaics + 2156-3381 + 2156-3403 + + 4 + 4 + + 2014, July + Institute of Electrical and Electronics Engineers (IEEE) + + + Quater, P. B., Grimaccia, F., Leva, S., Mussetta, M., & Aghaei, M. (2014, July). Light Unmanned Aerial Vehicles (UAVs) for Cooperative Inspection of PV Plants. IEEE Journal of Photovoltaics, 4(4), 1107-1113. + + + + + Deployment and Performance of a UAV for Crop Spraying + + DKGiles + + + RCBilling + + + + Chemical Engineering Transactions + + 44 + + 2015 + + + 9 Giles, D. K., & Billing, R. C. (2015). Deployment and Performance of a UAV for Crop Spraying. Chemical Engineering Transactions, 44, 307-312. + + + + + FHWA-GA-1H-12-38. Feasibility Study to Determine the Economic and Operational Benefits of Utilizing Unmanned Aerial Vehicles (UAVs) + + JIrizarry + + + ENJohnson + + + 2014 + Georgia Institute of Technology + Atlanta, GA + + + Irizarry, J., & Johnson, E. N. (2014). FHWA-GA-1H-12-38. Feasibility Study to Determine the Economic and Operational Benefits of Utilizing Unmanned Aerial Vehicles (UAVs). Atlanta, GA: Georgia Institute of Technology. + + + + + Innovative technology summary report: Houdini{trademark} I and II remotely operated vehicle + + GKimchi + + 10.2172/665927 + + 2015 + Office of Scientific and Technical Information (OSTI) + Washington, DC + + + Kimchi, G. et al. (2015). Unmanned Aerial Vehicle Delivery System, US20150120094 A1. Washington, DC: United States Patent and Trademark Office. + + + + + Flight Demonstration of Unmanned Aircraft System (UAS) Traffic Management (UTM) at Technical Capa... + 10.2514/6.2020-2851.vid + + + + 12 National Aeronautics and Space Administration + + American Institute of Aeronautics and Astronautics (AIAA) + 2015. April 12, 2016 + + + 12 National Aeronautics and Space Administration. (2015). Unmanned Aircraft Systems (UAS) Traffic Management (UTM) Fact Sheet. Retrieved April 12, 2016, from http://utm.arc.nasa.gov/docs/utm-factsheet-02-23-16.pdf. + + + + + Unmanned Aircraft Systems (UAS) research and future analysis + + ChrisAWargo + + + GaryCChurch + + + JasonGlaneueski + + + MarkStrout + + 10.1109/aero.2014.6836448 + + + 2014 IEEE Aerospace Conference + Big Sky, MT + + IEEE + 2014 + + + Wargo, C. A., Church, G. C., Glaneueski, J., & Strout, M. (2014). Unmanned Aircraft Systems (UAS) Research and Future Analysis. IEEE Aerospace Conference. Big Sky, MT: IEEE. + + + + + Propulsion Systems Integration for Transonic Unmanned Aircraft System + + DJenkins + + + BVasigh + + 10.2514/6.2021-3550.vid + + 2013 + American Institute of Aeronautics and Astronautics (AIAA) + + + Jenkins, D., & Vasigh, B. (2013). The Economic Impact of Unmanned Aircraft Systems Integration in the United States. Association for Unmanned Vehicle Systems International. + + + + + Endocrine Critical Care + + DLucey + + + KAho + + + LEwing + + + GLHoneycutt + + + ZRosenberg + + 10.1891/9780826177216.0011 + + + + Fast Facts for The Critical Care Nurse + + Springer Publishing Company + 2014. May + 4 + + + + Lucey, D., Aho, K., Ewing, L., Honeycutt, G. L., & Rosenberg, Z. (2014, May). Fast Facts on Unmanned Agriculture. Mission Critical, 4(2), 8-9. Retrieved from http://www.auvsi.org/. + + + + + + MERogers + + Diffusion of Innovations + New York, NY + + Simon & Schuster + 2003 + + + th ed. + Rogers, M. E. (2003). Diffusion of Innovations (5 th ed.). New York, NY: Simon & Schuster. + + + + + Practice for Application of Federal Aviation Administration (FAA) Federal Aviation Regulations Part 21 Requirements to Unmanned Aircraft Systems (UAS) + 10.1520/f2505 + + + Federal Aviation Administration + + ASTM International + 2015 + + + Federal Aviation Administration. (2015). FAA-2015-0150. Operation and Certification of Small Unmanned Aircraft Systems [NPRM]. + + + + + + + DCWashington + + + Department of Transportation + + + Washington, DC: Department of Transportation. + + + + + + diff --git a/file754.txt b/file754.txt new file mode 100644 index 0000000000000000000000000000000000000000..b408dcf4da24a30f1ff3e00436d3889e9b16e23e --- /dev/null +++ b/file754.txt @@ -0,0 +1,216 @@ + + + + +INTRODUCTIONThe Next Generation Air Transportation System (NextGen) is being designed with the expectation that the volume of the traffic will double by 2025 (Joint Development and Planning Office, 2004).In order to handle the expected traffic demand, airport capacity needs to expand dramatically.To gain such capacity at the airport, runways with centerline distances closer than 2500 ft need to be The dual and triple studies analyzed in this paper use a concept developed by NASA in collaboration with the Raytheon Corporation called the Terminal Area Capacity Enhancing Concept (TACEC), which allows paired approaches on runways that are 750 apart in instrument meteorological conditions (Miller et al., 2005).The TACEC concept includes a ground-based processor, which identifies aircraft that could be paired approximately 30 minutes from the terminal airspace boundary.The aircraft are selected for pairing based on several parameters such as relative aircraft performance, arrival direction, and the size of the aircraft's wake.The ground based processor then assigns 4-dimensional (4-D) trajectories to the aircraft in the pair.It is assumed that all aircraft will use differential GPS-enabled, high-precision 4-D flight management system capabilities for the execution of these trajectories.Enhanced cockpit displays provide the trailing pilot with detailed position and some intent information about the lead aircraft, and show a predicted wake for the lead aircraft.The concept uses breakout trajectories that require a less extreme turn than the maneuvers used in other concepts.This concept that was originally developed for dual runways and then extended to triple runways considers wake prediction data to determine when a breakout is required, and provides for a dynamically generated breakout trajectory that changes as the aircraft flies the approach.Most of the previous concepts did not consider wake data in their concepts or displays.This paper provides a comparative analysis of the 2-runway and 3-runway experiments by comparing results of the pairs of aircraft in the triple formation (left and center aircraft or center and right aircraft) with the 2-runway pair (left and right aircraft).Procedures for dual runways involving a leading and trailing aircraft pair are compared to procedures from the 3-runway study, when the piloted aircraft was either the center or the right aircraft in the echelon formation (Figure 1).The comparison between the dual and triple runways is meant to evaluate the dualrunway procedures that were implemented to create new procedures for triple runways.The dual and triple runway procedures are compared on level of accuracy for flying breakouts and differences in workload experienced by the pilots.Results on these factors will provide insight into the human factors issues associated with the different positions of the aircraft in the dual and triple formation. +METHODS +Airport and Airspace DesignBoth the 2-runway and 3-runway studies used a common fictitious airport (KSRT) based on the current Dallas/Fort Worth International Airport (DFW) layout and operations, with the exception of the runways which were set 750 feet apart.The west side of the airport was simulated in its south configuration only (18L and 18R for 2-runways, and 18L, 18C and 18R for 3-runways).Equipage to a CAT-IIIB level was assumed. +Operational ProceduresFlights in the simulation were initiated at about 25 nmi from the airport, with the assumption that they were already placed into aircraft pairs or triples.Approach and departure routes and procedures were similar to those used at DFW.The aircraft flew 4D arrival trajectories and were paired or 'tripled' with aircraft arriving from any of the four meter fixes (NE, NW, SE, SW) located near the edge of the terminal airspace, about 40-60 nmi from the airport.The concept allows for pairing based on aircraft type, performance characteristics and estimated time of arrival.In the study, the pairing was scripted in the traffic scenarios.The aircraft fly 4D trajectories up to a point in the airspace, referred to as the coupling point, designated at 12 nmi from the runway threshold.From the coupling point onwards, the aircraft fly in a formation such that they were coupled for speed.In the 2-runway study, the trailing aircraft precisely maintained temporal spacing of 15 sec with +/-10 sec tolerance for error (a window of 5-25sec), behind the lead aircraft to avoid wake of the lead aircraft (Rossow et al., 2005).The path flown by the trailing aircraft in the 2-runway study involved a slew angle of six degrees to the landing runway.The aircraft became parallel at about two nmi from the runway, as shown in Figure 2. In the 3-runway study, the center aircraft precisely maintained 12 s spacing behind the lead aircraft, and the right aircraft maintained 24 s behind the lead aircraft beyond the coupling point.As shown in Figure 2, the approach paths of the two trailing aircraft were at designated slew angles from the center of the runway -6-deg for the center runway aircraft and 12-deg for the right runway aircraft.All three aircraft turned straight-in for the final approach during the last two nmi from the runway.For both the 2-and 3-runway procedures, onboard automation monitored the paired runway for potential conflicts.Automation also displayed the predicted safe zone from the wake generated by the lead aircraft (and center aircraft for 3runway procedures).Visual and aural alerts were used to alert pilots to the lead (or center) aircraft blunders or the wake of the lead (or center) aircraft drifting towards the aircraft behind it.The navigation displays (Figure 3) depicted the breakout trajectory as a white line, after the aircraft crossed the coupling point.For the 3runways study, the breakout trajectory was shown for both the center and the right aircraft.In both studies, the breakout trajectory was dynamically generated considering wake, traffic, structures, and terrain of airport surroundings.Breakouts were caused by an intentional lead-aircraft blundering towards the following aircraft, or the wake of the lead aircraft drifting towards the following aircraft.Different locations of the breakout on the arrival path required different breakout maneuvers, which change the angle of the escape trajectory on the navigation displays.When the breakout was required at different altitudes on the arrival path, different bank angles for the breakout maneuvers were used and the curvature of the breakout trajectory changed on the navigation displays.The pilots were required to fly the breakout trajectory manually using the flight director when they received an aural and red visual alert.For both the 2-runway and 3-runway studies, the breakout performed above 500 ft altitude required an initial bank angle of 30 deg, and the breakout at an altitude between 200-500 ft required an initial bank angle of 10 deg (Tables 1 and2).The pilots at this stage were instructed to follow the "S" shaped breakout trajectory displayed on the navigation display as accurately as possible (Figure 3).The trajectory was "S" shaped so the final leg of the trajectory became parallel to the runways.The 3-runway study used similar bank angles to those used in the 2-runway study.In addition, the pilot participants flew different headings based on the position in the echelon.The center aircraft (18C) changed its heading to 20-deg and the right aircraft (18R) changed its heading to 40-deg, giving more space to the center aircraft.The aircraft performing the breakout maneuver also climbed to 3000 ft as part of the breakout trajectory.The final leg of the breakout trajectory parallel to the runways was 1.5 nmi abeam for aircraft flying to 18C and 3.0 nmi for aircraft flying to 18R. +Simulation PlatformFor both studies, the human-in-the-loop experiments of breakout maneuvers for paired and triple runways were performed approaches in the Advanced Concepts Flight Simulator (ACFS) located at the NASA Ames Research Center.The ACFS is a motion-based simulator that can be configured to represent current and future cockpits.At the time of this experiment, the simulator had performance characteristics similar to a Boeing 757, but its displays were modified to study advanced flight operational concepts. +ParticipantsThe study participants were recently retired pilots from commercial airlines.All of them were male and all had experience with glass cockpits.Their average pilot experience was about 38 years, and their average number of years since retirement was less than two. +Traffic ScenarioFor the 2-runway study, the traffic scenario involved two aircraft: (1) The ACFS flight simulator as the trailing aircraft (i.e., the ownship) and (2) A scripted Boeing 747-400 as the leading aircraft.For the 3-runway study, the traffic scenario involved three aircraft, where the flight simulator (i.e., the ownship) was either the center or right aircraft.The other two aircraft were scripted.When the ownship was in the center position, the aircraft causing the off-nominal situation was the left-most aircraft.When the ownship was in the right-most position, the aircraft causing the off-nominal maneuver was the center aircraft.The off-nominal event was introduced in the scenarios through lead aircraft intentionally deviating off its trajectory or adverse winds causing its wake to drift towards the following aircraft. +RESULTS AND DISCUSSIONStatistical results on two dependent variables are reported in the analysis of data generated from the experimental runs: (1) Ownship cross track error, collected digitally during the breakout phase of the simulation flight and (2) The pilots' subjective assessments of workload.Data were analyzed using 3-way Factorial Analysis of Variance with three independent variables, with each independent variable having 2 levels: (1) Number of runways (2 vs. 3), (2) Cause of breakout (aircraft deviation and wake) and (3) Location of breakout (high and low altitude). +Cross Track ErrorCross track error, collected by the simulator's digital data collection system, is one measure of trajectory accuracy particularly sensitive to breakout maneuvers.Cross track error was measured by the distance between the actual ownship position and the system-generated breakout trajectory position (i.e., the off-course distance), with both positions shown on the Navigation Display.Hence, less cross track error correlates to higher breakout trajectory conformance.For each simulation run, cross track error was averaged across time from the breakout point to the end of the flight.A statistically significant ANOVA main effect of the number of runways on the ownship's breakout cross track error was found, in comparing the 2-runway and the 3-runway (right ownship) conditions (F=21.92,df=1,15, p<0.001) (Table 3).The directionality of means for this main effect indicates more cross track error under the 3-runway (right ownship) condition as compared to the 2-runway condition.This could be attributed to having 2 aircraft to the left of the ownship during breakout (3-runway, right ownship), creating an increased sense of urgency on the part of the pilot to escape the cause of the off-nominal situation, i.e., the possible additive effect of wake and/or blunder of both aircraft to the left of the ownship might prompt the pilot to overshoot the breakout trajectory further to the right as a safety measure.Some increased cross track error was also observed under the 3-runway (center ownship) condition as compared to the 2-runway condition, but this difference did not reach statistical significance.This lack of statistical significance might reflect the center position of the ownship, which requires that the pilot maintain safe separation with 2 other aircraft -one to the right, and one to the left of the ownship, thereby posing constraints on aircraft movement to either the right or the left, to maintain adequate separation.The pilot-participants pointed out that this prompted them to exercise a larger degree of vigilance in flying the breakout trajectory, which would explain less cross track error, as compared to the 3-runway (right ownship) condition, even though the right ownship is not much safer than the center aircraft.Mean cross track error values generally indicate reasonable levels of accuracy in flying the breakout trajectory.However, maximum values at the end of the distribution for the triple-runway operations (Table 3) might indicate a need for improved training to prevent the occasional overshoot of the breakout trajectory.A statistically significant Number of Runways by Breakout Location (altitude) interaction effect on cross track error was also observed.A larger mean cross track error difference between high and low altitude locations was observed under the 3-runway (right ownship) condition, as compared to the 2-runway condition (F=16.12,df=1,15, p<0.005) (Figure 4).This interaction effect is best understood when one considers that breakout procedures for the higher altitudes are more difficult and that procedures are more complex for the 3-runway operations.As postulated above, the 3-runway centerownship pilot may have exercised special vigilance in flying the breakout trajectory more accurately, due to the aircraft's central location in the triplet echelon, which would account for less cross track error.Since the pilot in the 3-runway (right ownship) condition is mostly concerned about loss of separation with the center aircraft and the possible additive effect of having two aircraft to the left in the breakout formation (wake turbulence and/or track deviation of both aircraft), the pilot may be less concerned about exercising special vigilance in flying the breakout trajectory accurately, but rather, escaping the track deviation or wake turbulence of both aircraft by moving as quickly as possible towards the right, and possibly overshooting the breakout trajectory.Also, since the higher altitude breakout procedures require a more aggressive maneuver (as compared to the lower altitude procedures), the possible tendency for the pilot to overshoot the breakout trajectory further to the right at the higher altitude might reflect a continuation of the already aggressive nature of the required maneuver. +WorkloadParticipants completed the NASA TLX workload questionnaire (Hart and Staveland, 1988) after every run.Data were collected on each of the six TLX workload measures, which were combined to derive a composite workload measure, which ranged from 1 (very low workload) through 7 (very high workload).Table 4 presents statistics on average composite workload, broken down by the number of runways and position of the ownship.Overall, workload can be characterized as moderate.While trends should be viewed with some caution, due to lack of statistical significance, the directionality of means shows increased workload under the 3-runway conditions as compared to the 2-runway condition.This would make sense, due to the increased geometric complexity of the 3-runway procedures and pilots needing to maintain safe separation with 2 other aircraft (as compared to only 1 other aircraft under the 2-runway condition), thereby increasing pilot workload.A statistically significant Number of Runways by Breakout Cause interaction effect was observed, in comparing workload for 2-runways and 3runways (Right Ownship) by Aircraft Deviation and Wake (Figure 5)..43, df=1,15, p<0.005).This effect can be explained by the relative complexity of the 3-runway operations and the unstable nature of wake turbulence from possibly two other aircraft to the left of the ownship.Since having two aircraft to the left of the ownship during breakout (3-runway, right ownship) could create an increased sense of urgency on the part of the pilot to escape the cause of the offnominal situation, i.e., the possible additive effect of wake of both aircraft to the leftFigure 1 :1Figure 1: Echelon formation for triples (shaded area below aircraft shows predicted wake turbulence zone) +Figure 2 :2Figure 2: Final approach geometry for operational procedures for dual and triple runways +Figure 3 :3Figure 3: Navigation Displays for final approach for 2 and 3 parallel runways +Figure 4 :4Figure 4: Number of Runways X Breakout Location Interaction Effect: Cross Track Error (* p<0.005) +Figure 5 :5Figure 5: Comparison of Composite Workload for Wake versus Aircraft Deviation Under 2-runway and 3-runway Conditions (Right Ownship) *p<0.005 Figure 5 shows a larger workload difference between Aircraft Deviation and Wake causes under the 3-runway (Right Ownship) condition, as compared to the 2-runway condition (F=12.43,df=1,15, p<0.005).This effect can be explained by the relative complexity of the 3-runway operations and the unstable nature of wake turbulence from possibly two other aircraft to the left of the ownship.Since having two aircraft to the left of the ownship during breakout (3-runway, right ownship) could create an increased sense of urgency on the part of the pilot to escape the cause of the offnominal situation, i.e., the possible additive effect of wake of both aircraft to the left +Table 1 : Breakout trajectory for dual runways1RunwayBreakout LocationInitial Bank Angle18 R (2-runway)> 500 feet30 °200-500 feet10 ° +Table 2 : Breakout trajectory for triple runways2RunwayBreakoutInitialBankInitial HeadingLocationAngleChange(altitude)18 C [3-Runway> 500 feet30 °20 °(Center Ownship)]200-500 feet10 °20 °18 R [3-Runway-> 500 feet30 °40 °Right Ownship)]200-500 feet10 °40 ° +Table 3 : Breakout Cross Track Error (2-runway vs. 3-runway)3RunwayMean (ft)SEMINMAX2-runway56.253.670.65106.123-runway (center ownship)73.4411.455.83542.423-runway (right ownship)104.3711.4616.26513.29 +Table 4 : Average Composite Workload Statistics by Number of Runways4RunwayMeanSE2-runway2.780.153-runway (Center Ownship)3.690.123-runway (Right Ownship)3.640.12 + + + + + + + + + Development of NASA-TLX (Task Load Index): Results of Empirical and Theoretical Research + + SandraGHart + + + LowellEStaveland + + 10.1016/s0166-4115(08)62386-9 + + + Advances in Psychology + + PHancock + + + NMeshkati + + Amsterdam, The Netherlands + + Elsevier + 1988 + + + + Hart, S., & Staveland, L. (1988). Development of a Multi-Dimensional Workload Rating Scale: Results of Empirical and Theoretical Research, In P. Hancock & N. Meshkati (Eds.), Human Mental Workload, Amsterdam, The Netherlands, Elsevier, 139-183. + + + + + The Next Generation of Air Supply Units + + JanvierEBeaver + + + KarlHering + + 10.1115/rtd2004-66002 + + + Joint Rail + + ASMEDC + 2004. December 12, 2004 + + + Joint Development and Planning Office (2004). The Next Generation Air Transport System Integration National Plan. December 12, 2004. + + + + + Compilation of DOT and FAA Airport Legal Determinations and Opinion Letters as of December 31, 2020 + 10.17226/26202 + Study. DOT/FAA/CT-TN02/01 + + + Simultaneous Offset Instrument Approaches -Newark International Airport: An Airport Feasibility + + Transportation Research Board + 2001 + + + Magyratis, S., Kopardekar, P., Sacco, N. & Carmen, K. (2001). Simultaneous Offset Instrument Approaches -Newark International Airport: An Airport Feasibility Study. DOT/FAA/CT-TN02/01. + + + + + CNS requirements for precision flight in advanced terminal airspace + + MEMiller + + + SDougherty + + + JStella + + + PReddy + + 10.1109/aero.2005.1559471 + + + 2005 IEEE Aerospace Conference + + IEEE + 2005 + + + + Miller, M.E., Dougherty, S., Stella, J. & Reddy, P. (2005). CNS Requirements for Precision Flight in Advanced Terminal Airspace. Aerospace, 2005 IEEE Conference, 1-10. + + + + + Models of Wake-Vortex Spreading Mechanisms and Their Estimated Uncertainties + + VernonRossow + + + GordonHardy + + + LarryMeyn + + 10.2514/6.2005-7353 + AIAA-2005-7353 + + + AIAA 5th ATIO and16th Lighter-Than-Air Sys Tech. and Balloon Systems Conferences + Arlington, VA + + American Institute of Aeronautics and Astronautics + 2005. September 26-28, 2005 + + + Rossow, V. J., Hardy, G. H., and Meyn, L. A. (2005). Models of Wake-Vortex Spreading Mechanisms and Their Estimated Uncertainties, AIAA-2005-7353, ATIO Forum, Arlington, VA, September 26-28, 2005. + + + + + Procedures for off-nominal cases: Very closely spaced parallel runway operations + + SavitaVeram + + + SandyLozito + + + ThomasKozon + + + DeborahBallinger + + + HerbertResnick + + 10.1109/dasc.2008.4702783 + + + 2008 IEEE/AIAA 27th Digital Avionics Systems Conference + + IEEE + 2009. 2009 + + + Eighth USA/Europe Air Traffic Management Research and Development Seminar + + + Verma, S., Lozito,S., Ballinger,D., Kozon, T., Panda, R., Carpenter, D., Wooten, D., Hardy, G., and Resnick, H. (2009). Evaluation of Triple Closely Spaced Parallel Runway Procedures for Off-nominal Cases, Eighth USA/Europe Air Traffic Management Research and Development Seminar (ATM2009). + + + + + Procedures for off-nominal cases: Very closely spaced parallel runway operations + + SavitaVeram + + + SandyLozito + + + ThomasKozon + + + DeborahBallinger + + + HerbertResnick + + 10.1109/dasc.2008.4702783 + + + 2008 IEEE/AIAA 27th Digital Avionics Systems Conference + St. Paul, MN + + IEEE + 2008 + + + Verma, S., Lozito, S., Kozon, T., Ballinger, D., and Resnick, H. (2008) Procedures for Off- Nominal Cases: Very Closely Spaced Parallel Runways Approaches, Digital Avionics System Conference, St. Paul, MN. + + + + + + diff --git a/file755.txt b/file755.txt new file mode 100644 index 0000000000000000000000000000000000000000..e426d50178663f107a9f575fa9bcac7c59d3f721 --- /dev/null +++ b/file755.txt @@ -0,0 +1,340 @@ + + + + +IntroductionManaging departure operations in busy airport and airspace environments, with limited data sharing and system integration, can significantly limit efficiency and predictability.Stakeholders, including air carriers and air navigation service providers, have their own objectives in managing traffic, and often these objectives compete with one another.Also, decisions in managing traffic are often made in a reactive manner with short planning horizons due to operational uncertainties and a lack of common situation awareness between Flight Operators and service providers.For example, during periods when demand exceeds capacity at the airport, service providers manage traffic using the First Come, First Serve (FCFS) paradigm, where they serve the flights that first call in as ready for pushback.With many airlines having similar ticketed departure times, this leads to surface congestion.There is a need for a tool that provides departure metering to prevent surface congestion.Other tools that perform departure metering include A Human-in-the-loop (HITL) study investigated procedures for the Metering Tool [8] and found that ramp managers may have RTMC running all the time, and they can enable or disable metering anytime, depending on their strategy for demand/capacity balancing.When ramp managers decide to enable the time-based metering, they could choose the level of gate holding from three options -'Nominal hold', 'Less hold', or 'More hold.'The 'Less hold' option allows more flights to be on the airport surface (movement area), whereas the 'More hold' option allows the flights to be held at their gates longer, thus resulting in less taxiing delay or less excess queue time on the surface or airport movement area.The 'Nominal hold' option seeks to utilize the existing runway capacity with the available demand and find acceptable levels of excess queue time taken on the surface.These gate hold levels are associated with a metering value that defines the target excess queue time that will be taken in the Airport Movement Area (AMA) as compared to unimpeded transit time from gate to runway.A study [8] found that users preferred the inputs for level of gate hold to be labeled as "Excess Queue Time in AMA" and they chose 12 min as the target for excess queue time with 14 min to mark "less hold" and 10 min to represent "more hold" as shown in Fig. 3.In this paper, we describe the procedures that were adopted for tactical surface metering and the way these evolved.In section 2, we will briefly overview the airport surface and airspace operations at CLT.In section 3, we will describe the tools and interfaces deployed at CLT.In section 4 we will discuss data analysis of procedures and early anecdotal feedback, which led to changes to procedures and the algorithm that were adopted by the users during the field demonstration.Lastly, section V will provide the closing remarks. +CLT Operations OverviewAccording to the recent airport activity report, CLT accommodates about 1,400 operations per day and is the seventh busiest airport in aircraft movements worldwide in 2016 [9].Because CLT is one of the main hub airports for American Airlines (AAL), AAL and its regional air carriers operate nearly 93% of the flights into and out of the airport.The remaining operations are comprised of other regional carriers, mainline flights operated by Southwest, Delta, United and Jet Blue, military flights, business and general aviation, and air cargo.As the dominant carrier, AAL manages all ramp operations at the airport, whereas air traffic on the AMA is controlled by the ATC Tower.As shown in Fig. 1, CLT has three north/south parallel runways (18L/36R, 18C/36C, and 18R/36L) that can support simultaneous independent instrument approaches, and a fourth diagonal runway (5/23) that intersects Runway 18L/36R.The airport operates in either a "North" or "South" flow configuration.The diagonal runway, Runway 23, is used in a South flow configuration for arrivals.Runway 5 (the opposing end) is not used for arrivals or departures during normal daylight/evening operations, but it is used as a taxiway in a North flow operation.However, during North flow operations, Runway 5 is used for both arrivals and departures when North flow nighttime noise abatement procedures are in effect.Traffic at CLT is characterized by definite peaks and valleys.There are clear distinctions between departure and arrival banks throughout the day.Banks are peak periods of departure or arrivals due to the hub-and-spoke operations.Each departure and arrival bank takes approximately an hour with a slight overlap existing between banks.Ramp Control strives to clear the departures from the gates before an arrival bank builds up, so that ramp congestion and gate conflicts can be minimized.The ramp area is divided into four sectors (e.g., West, South, East, and North sectors).The corresponding ramp controller controls the traffic in each sector.The ramp operations at CLT are constrained due to physical limitations of the ramp, such as limited ramp space with alleys between concourses, single-direction taxiways, and limited holding areas (hardstands). +Metering Tool User Interfaces and ProceduresThis section describes the Metering Tool's interface and procedures it was deployed at CLT in the fall of 2017.The metering tool was used by ramp traffic controllers and managers and to depict the advisories associated with departure metering. +Ramp Traffic Console (RTC)The Ramp Traffic Console (RTC) and Ramp Manager Traffic Console (RMTC) are decision support tools developed for the ramp controllers and ramp managers, respectively.These tools provide a map display that depicts the ramp area with flight strips positioned at each gate for departures.These flight strips provide information such as call sign, aircraft type, and departure fix.The ramp controllers can interact with the tool to provide flight intent information, such as pushback, holding a flight, changing the assigned spot (point where control is handed over to FAA controlled from the ramp/airline), changing its gate, and marking the flight if it is sent to the hardstand.Double clicking on the flight strip allows the user to open the Flight Menu where the user can change assignment of a flight's spot, gate, runway, and mark it as temporarily out of service or mark it as being sent to the hardstand.Gate pushback intent information can also be provided by the ramp controller by right clicking the flight and selecting pushback in the drop-down menu, which is marked "pushback-cleared" with an engine symbol.Similarly, the right click drop down menu provides the user with an option to hold the flight by putting a red border around the flight strip (see Fig. 2).The color of the flight strips and icons show the flight direction they are going to, the blue strips have destinations in the east direction whereas the brown are flying in the west direction.Arrivals are depicted as green color aircraft icons.Flights that are moving and tracked are shown as solid aircraft icons and those that are moving but not detected by surveillance are shown as hollow aircraft icons (see Fig. 2).Tactical Surface Scheduler/ Metering Tool advisories are shown next to the flight strips (see Fig. 3). +Ramp Manager Traffic Console (RMTC)The ramp manager has the ability to enable Time-Based Metering, i.e., the Metering Tool, via the user interface provided by the RMTC (see Fig. 4).The RMTC interface is used to set the Airport Movement Area (AMA) excess queue time using this interface.The target of 12 min excess queue time means that flights will be taking up to 12 min of predicted excess taxi time (or delay) in the airport movement area and any remaining time at the gates.The higher the excess queue time, the lower the amount of gate hold taken by that flight.These values for the target, upper and lower thresholds were investigated and determined in a HITL at NASA Ames [8].These target excess queue values affect the output of the metering tool and its pushback advisories shown on the RTC. +Tactical Scheduler for SurfaceThe surface metering tool calculates Target Off-Block Times (TOBT) and provides gate hold recommendations to the ramp controller.For each departure flight, the tactical scheduler generates the Target Takeoff Time (TTOT) that would meet constraints, including runway separation criteria and Traffic Management Initiatives (TMI) constraints.Once the TTOT is generated, the TOBT is calculated according to the delay propagation formula: TOBT = max[EOBT, TTOT -UTT -MeteringValue] where UTT is the unimpeded transit time from gate to runway.After calculating the TOBT, the TMAT is scheduled as TMAT = TOBT + URT, where URT is the unimpeded ramp transit from gate to spot.This results in all the delay being absorbed in the AMA and no delay modeled or taken in the ramp.Based on the flight's TOBT, a gate hold recommendation is provided on the RTC.The Metering Value or excess queue time is specified in the tactical scheduler's delay propagation logic and is used for calculation of TMATs and TOBTs.The purpose of the metering value is to control the maximum amount of predicted excess queue time that the flights are allowed to experience.The larger delay buffer causes the flights to spend more time in the queue or AMA before takeoff, and therefore, allows the aircraft to push back earlier from the gate.The target excess queue time (as shown in Fig. 3) selected by the ramp manager determines the value of delay buffer.The metering tool considers the demand and capacity imbalance for each departure runway before providing guidance for the flights.It is often observed that sometimes there are no gate hold advisories provided for flights departing in one direction, whereas gate hold advisories are provided for flights departing in the other direction.Flights can be marked as exempt from metering or as a priority flights on the RTC or the RMTC, and the metering tool treats them accordingly.International and General Aviation (GA) flights may also be marked as exempt from metering.The tactical scheduler regards EOBT as a flight's predicted ready time and uses that to generate gate hold advisories.EOBT is being calculated by the airline based on various factors such as percentage of passengers boarded, baggage loaded and more.The tactical surface scheduler allocates runway departure slots on the timeline according to the flight's schedule, with an order of consideration applied, and based on the quality of the flight's EOBT.The order of consideration dictates in the tactical surface scheduler the order in placing flights in different sequential groups based on their predictability in runway time prediction, which improves as the flight progresses in departure process, since the look-ahead time to the runway decreases.The definitions of the groups are shown in Table I.Flights that are further than 10 minutes from their EOBTs or have poor quality EOBTs (i.e., high prediction errors) are marked in the Uncertain group.The flight is considered to be part of the Planning group when it is 10 minutes from its EOBT.Gate hold advisories or push advisories will be shown on the RTC for the flights in the Planning group.When the pilot calls in ready to push, the ramp controller is expected to mark the flight strip for pushback or hold according to the advisory shown on the display, and at this point the scheduler marks the flight in the Ready group.When the flight is cleared for pushback by the ramp controller, it is considered to be in the Out group, and in the Taxi group when it starts taxiing.Similarly, it is considered in the Queue group when it is waiting in a queue at the runway getting ready for take-off.The RTC shows a hashtag for flights in the Uncertain group (Fig. 4) instead of providing any gate hold advisories.This is done to avoid too many fluctuations in the gate hold advisories due to the uncertainty in flight ready time.However, this does not prevent a pilot from calling in for pushback.When this happens, the ramp controller can click the hashtag, and the tactical scheduler instantaneously returns the gate hold advisory and displays it on the RTC.Flights automatically move from the Ready or Planning group to the uncertain group, if the flight called ready to push but did not pushback 5 minutes after its TOBT or it did not call 5 min past its EOBT respectively.The tactical surface metering tool frequently updates (i.e., every 10 seconds) and adjusts the schedule to accommodate uncertainties and changes in the traffic situation. +Data Analysis and System Health (DASH)Data Analysis and System Health is a tool that provides several metrics in real time.This tool is used to look at predicted excess queue times in future, manipulate the target, upper, and lower thresholds for the excess queue times and see how many flights will be impacted when metering is enabled.Fig. 5 shows the ability to manipulate the thresholds in the bottom bar of the graph.The green middle line depicts the target excess queue time, the top red line shows the upper threshold for the excess queue time and the bottom line shows the lower threshold for excess queue time.The flights are depicted as dots on this chart and the number of dots above the upper threshold will likely have a gate hold time displayed as a pushback advisory, and those below the lower threshold do not display a pushback advisory on the RTC.When metering is enabled and advisories are being shown, every aircraft going to a runway that is being metered will show an advisory.For aircraft in this figure that are below the target queue, the advisory will be such that TOBT is the same as EOBT.For aircraft above the target queue, the advisory will be such that TOBT is greater than EOBT.In this verbal discussion session, the ramp mangers and TMCs discussed the demand and capacity imbalance situation by using the predicted excess queue time graphs in the DASH tool opened inside a "Metering What-if" system.They collaboratively determined the desired peak hold level and modified the target, upper and lower thresholds of excess queue time in the DASH tool.At about 0910, the Ramp manager verified the runway configuration and utilization plan intended for the upcoming Bank-2.They ensured that the airport and runway utilization scheme was an input into the ATD-2 tools so that tactical surface scheduler would have the correct predictions for capacity.The Ramp Manager enabled metering and was also responsible for making entries for the target, upper and lower thresholds in the RMTC.Once metering was enabled, all ATD-2 tools in the ramp and ATC Tower received a notification about the same.At this point, Ramp Managers also ensured that heavy jet aircraft and international flights were exempted from Metering. +Initial Data Analysis and Feedback from Users at CLTThis section describes the verbal feedback received from the users and some data analysis that depicts the changes made to the procedures early on.The data analyses were performed on data collected over a two-month period from November 29, 2017 to January 29, 2018.If metering was not used on particular days or if it was not collected correctly, those days were eliminated from these analyses.Although data was analyzed for both South and North Flows, only North Flow data is being reported here since it was the predominant configuration that the airport was in during the two-month period.Initial Gate Hold Recommendations: Initially, the ramp controllers and managers indicated that when Surface Metering was enabled they observed that flights received gate hold recommendations even when there was little or no delay on the surface.The ramp controllers are used to pushing flights early rather than late to make room for arrivals, the early recommendations regarding gate hold provided guidance that was contrary to their usual way of operating.On further investigation, it was found that the system was only considering the predicted queue for all aircraft at the gate whose best Estimated Off Block Time was within 10 min of current time.Since the actual physical queue was not considered in this calculation, metering was triggered early based on the algorithm's predictions.This did not allow the queue to properly build up.Also the advisories correctly computed gate hold times for aircraft against EOBT but this alarmed the controllers since they did not expect to hold those flights during slow traffic.To allow flights to pushback early in the bank and not be held when there was no traffic delay on the airport surface, the ramp managers were guided to make the upper threshold as +3 or +5 minutes above the target excess queue time (Fig. 6 shows the target and threshold values for the inputs that the users made).The high upper threshold values ensured that even though metering was enabled early on, display of pushback advisories was not triggered until the predicted excess queue time built up above the upper threshold, which meant that the onset of pushback advisories was delayed.Fig. 8 shows that metering was mostly triggered within 20 minutes from the time metering was enabled, which was usually earlier or close to the start of Bank-2, i.e., around 0900 local time.This confirmed the problem the ramp controllers reported regarding the gate hold recommendations coming early in the bank.A change in the metering algorithm was made and included in a subsequent software release.In addition to the predicted queue above the upper threshold for all aircraft at the gate whose UOBT was within 10 min of current time, the new metering algorithm also detected an active aircraft that was off the gate with a queue time greater than or equal to the target excess queue time.This change led to flights experiencing gate holds later in the bank as shown in Fig. 8. Figure 7 also shows that the values of the target, upper and lower thresholds input into the system after the algorithm change were lowered, since it was no longer necessary to have a high upper threshold with the change in the display triggering logic.It was also reported by the users in the field that many flights depicted an advisory or a recommended gate hold time that was regarded as very high.This was the case because these flights had their ready times several minutes away, and the system had calculated their TOBT as greater than EOBT.For example, if an aircraft is being scheduled with an 8-minute gate hold (i.e., TOBT is 8 minutes after EOBT) and EOBT is 7 minutes in the future, this will result in an 8+7=15 min advisory.In this case, ramp managers were guided to advise the ramp controllers to take a look at the ready time or EOBT in the flights strips on RTC via the Flight Menu.The Flight Menu is available one level deep inside the flight strip, and can be accessed by double clicking on the flight to open its flight menu.This information helped the ramp controllers to decide when to push the flights if the flights called earlier than their EOBT, given that airline policy does not expect flights to pushback earlier than 5 min to scheduled departure time.Several user interface changes were also explored to address this problem of not showing the ready times or EOBTs to the users in the flight strips but none have been implemented yet since the EOBT accuracy is variable. .The number of flights subjected to metering is much higher than the flights that were actually held at the gates by the ramp controllers.The flights are considered as subject to metering when the pilot calls, metering is enabled and advisories are shown for the runway the aircraft was scheduled for.An advisory as mentioned earlier could be a push advisory or one with a recommended gate hold time.The number of flights actually held for metering is lower than the flights subjected to metering since the ramp controllers were allowed to push the flights up to 2 minutes before the recommended gate hold expired.Also in the early days of implementation the advisories were not followed due to the flights being held early in the bank when there was little or no delay in the surface traffic. +SummaryThe Metering tool was deployed at CLT in the Fall of 2017 and metering procedures were collaboratively planned by the airline ramp control and FAA's ATCT.They collaboratively selected the target excess queue time that was used as input into the system.The initial deployment revealed an issue with the metering tool holding flights earlier in the bank when there was minimal delay associated with active flights on the surface, which was not operationally desired.The users were instructed to manage this situation by selecting higher upper threshold values.The algorithm was changed to ensure that not only predicted excess queue time but also actual queue time was taken into account for flights that were off the gate.This helped solved several problems, such as unnecessary gate hold times assigned to flights early in the bank.The users required refresher training to remind them to try to pay attention to the EOBTs and follow the advisories and airline policy regarding departures calling in early.This research brought a paradigm change for the controllers, who started paying attention to EOBTs and advisories instead of pushing every flight as they called thus reducing surface congestion.The algorithm change is expected to improve the compliance to the advisories, and hence reduce congestion.Future versions of this algorithm, which NASA continues to test in CLT, will focus on trying to meet target movement area entry times or spot times and not just pushback times.Fig. 1 .1Fig. 1.CLT airport plan view +Fig. 2 .Fig. 3 .23Fig. 2. Different states for flight strips and icons on RTC +Fig. 4 .4Fig. 4. Window on RMTC to set time-based metering and level of AMA Excess Queue Time +Fig. 5 .5Fig. 5. Predicted Excess Queue Time Monitor (DASH) used for assessing impact of target values and thresholds. +Fig. 6 .Fig. 7 .67Fig. 6.Target and Thresholds set by users at CLT (north Flow only) +Fig. 8 .8Fig. 8. Number of Flights impacted by metering for the entire airportGate Hold Advisories.The recommended gate hold or pushback advisories are shown in Fig.9.The mean hold advisory which is calculated as time when the pilot calls in ready to pushback subtracted from the target off block time, is below 10 minutes for the most part with a few outliers.The ramp controllers actually held the flights less than 10 minutes.They were provided guidance to pushback flights within two minutes of the advisory.The peak advisories are a function of many different factors such as traffic demand, surface congestion, and whether or not aircraft called ready while in the uncertain group.If an aircraft that is in the uncertain group (Table1) calls the ramp during peak traffic, it is very likely they will receive a large hold advisory.The peak advisories have to be evaluated in light of the number of flights that were subject to metering. +Fig. 9 .9Fig. 9. Recommended Gate Hold Advisories shown during the two-month period +Table 1 .1Definitions of Scheduling GroupsGroupDefinitionInterface (Fig. 4)UncertainFlights with poor quality EOBT or EOBT -Hashtagcurrent time > 10 minPlanningFlights within 10 min of EOBT (i.e., EOBTAdvisory-current time <= 10 min)ReadyFlights that have called in ready forAdvisorypushbackOutFlights that are in pushback stateEngine symbol + + + + + + + + + Assessing the impacts of the JFK Ground Management Program + + StevenStroiney + + + BenjaminLevy + + + HarshadKhadilkar + + + HamsaBalakrishnan + + 10.1109/dasc.2013.6712508 + + + 2013 IEEE/AIAA 32nd Digital Avionics Systems Conference (DASC) + Syracuse, NY + + IEEE + October 2013 + + + S. Stroiney, B. Levy, H. Khadilkar, and H. Balakrishnan, "Assessing the impacts of the JFK ground management program," 32nd Digital Avionics Systems Conference (DASC), Syracuse, NY, October 2013. + + + + + Performance Evaluation of SARDA: An Individual Aircraft-based Advisory Concept for Surface Management + + YoonJung + + + TyHoang + + + MiwaHayashi + + + WaqarMalik + + + LeonardTobias + + + GautamGupta + + 10.2514/atcq.22.3.195 + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 22 + 3 + + 2015 + American Institute of Aeronautics and Astronautics (AIAA) + + + Y. Jung, W. Malik, L. Tobias, G. Gupta, T. Hoang, et al., "Performance evaluation of SARDA: an individual aircraft-based advisory concept for surface management," Air Traffic Control Quarterly, Vol. 22, Number 3, 2015, pp. 195-221. + + + + + Call for Papers + + MHayashi + + + THoang + + + YJung + + + MMalik + + + HLee + + 10.1027/2192-0923/a000067 + + + Aviation Psychology and Applied Human Factors + Aviation Psychology and Applied Human Factors + 2192-0923 + 2192-0931 + + 4 + 2 + + June 23-26, 2015 + Hogrefe Publishing Group + Lisbon, Portugal + + + M. Hayashi, T. Hoang, Y. Jung, M. Malik, H. Lee, et al., "Evaluation of pushback decision- support tool concept for Charlotte Douglas International Airport ramp operations," 11th USA/Europe Air Traffic Management R&D Seminar (ATM2015), Lisbon, Portugal, June 23-26, 2015. + + + + + Airspace Technology Demonstration 2 (ATD-2) Phase 1 Concept of Use (ConUse) + + YJung + + + SEngelland + + + ACapps + + + RCoppenbarger + + + BHooey + + + SSharma + + + LStevens + + + SVerma + + + GLohr + + + EChevalley + + + VDulchinos + + + WMalik + + + LRuszkowski + + + + August 2017 + + + in review + Jung, Y., Engelland, S., Capps, A., Coppenbarger, R., Hooey, B., Sharma, S., Stevens, L., Verma, S., Lohr, G., Chevalley, E., Dulchinos, V., Malik, W., and Ruszkowski, L., "Airspace Technology Demonstration 2 (ATD-2) Phase 1 Concept of Use (ConUse)," NASA/TM-2017-xxxxxx, August 2017 (in review). + + + + + + SEngelland + + + ACapps + + + KDay + + + MKistler + + + FGaither + + NASA/TM-2013-216533 + Precision Departure Release Capability (PDRC) final report + + June 2013 + + + S. Engelland, A. Capps, K. Day, M. Kistler, F. Gaither, et al., "Precision Departure Release Capability (PDRC) final report," NASA/TM-2013-216533, June 2013. + + + + + Evaluation of the controller-managed spacing tools, flight-deck Interval management and terminal area metering capabilities for the ATM Technology Demonstration #1 + + JThipphavong + + + JJung + + + HSwenson + + + KWitzberger + + + LMartin + + + + 10th USA/Europe ATM R&D Seminar (ATM2013) + Chicago, Illinois + + June 2013 + + + + J. Thipphavong, J. Jung, H. Swenson, K. Witzberger, L. Martin, et al., "Evaluation of the controller-managed spacing tools, flight-deck Interval management and terminal area metering capabilities for the ATM Technology Demonstration #1," 10th USA/Europe ATM R&D Seminar (ATM2013), Chicago, Illinois, 10-13 June 2013. + + + + + An Integrated Collaborative Decision Making and Tactical Advisory Concept for Airport Surface Operations Management + + GautamGupta + + + WaqarMalik + + + YoonJung + + 10.2514/6.2012-5651 + + + 12th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference and 14th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference + + American Institute of Aeronautics and Astronautics + July, 2013 + + + FAA Air Traffic Organization Surface Operations Directorate + FAA Air Traffic Organization Surface Operations Directorate, "U.S. Airport Surface Collaborative Decision Making Concept of Operations (ConOps) in the Near-Term: Application of the Surface Concept at United States Airports," July, 2013. + + + + + Evaluation of a tactical surface metering tool for Charlotte Douglas international airport via human-in-the-loop simulation + + SavitaVerma + + + HanbongLee + + + LynneMartin + + + LindsayStevens + + + YoonJung + + + VictoriaDulchinos + + + EricChevalley + + + KimJobe + + + BonnyParke + + 10.1109/dasc.2017.8102046 + + + 2017 IEEE/AIAA 36th Digital Avionics Systems Conference (DASC) + St. Petersburg, FL + + IEEE + Oct 2017 + + + Verma, S.; Lee, Hanbong, Martin, L., Stevens, L. Jung, Y., Dulchinos, V., Chevalley, E., Jobe, K., Parke, B. et al., "Evaluation of a Tactical Surface Metering Tool for Charlotte Douglas International Airport via Human-In-The-Loop Simulation," 36 th Digital Avionics Systems Conference (DASC), St. Petersburg, FL, Oct 2017 + + + + + Impact of General Aviation Operations on Airport Performance Through Fast-Time Simulations at Charlotte-Douglas International Airport + + ZhifanZhu + + + VaishaliAHosagrahara + + + HanbongLee + + + YoonCJung + + + DeborahLBakowski + + 10.2514/6.2020-2916 + + + + AIAA AVIATION 2020 FORUM + + American Institute of Aeronautics and Astronautics + April 2017 + + + Charlotte Douglas International Airport, Fast Facts and Aviation Activity Reports, April 2017. http://www.cltairport.com/News/Pages/FactsandFigures.aspx + + + + + + diff --git a/file756.txt b/file756.txt new file mode 100644 index 0000000000000000000000000000000000000000..32edd3e588b2e23cad4047c02486f3281b4b1ac9 --- /dev/null +++ b/file756.txt @@ -0,0 +1,495 @@ + + + + +IntroductionReduction in air traffic capacity is the biggest challenge that airports must address with closely spaced parallel runways when visual approaches are not possible due to poor visibility [1].The FAA's NextGen program aims to maintain visual capacities under all weather conditions at airports with closely spaced parallel runways.Previous concepts investigated safety issues related to parallel runway operations but did not examine the information and procedures for pairing aircraft.The authors have conducted high fidelity flight simulation studies to investigate the safety issues associated with parallel approaches that may require aircraft to perform breakout maneuvers due to hazardous conditions [2,3] such as the wake of lead aircraft drifting towards the follower or the lead aircraft blundering towards the follower.In addition, the role of air traffic control in aircraft pairing for simultaneous approaches was explored [4], including the examination of controller responsibilities and communication tasks.The next logical step is to investigate the integrated role of the controller and pilot in pairing aircraft for simultaneous approaches, which was the intent of the current study.This high fidelity human-in-the-loop simulation investigates the integrated dynamic role of controllers and pilots for pairing aircraft to parallel runways for simultaneous approaches.This paper will focus on only the controller's role specifically in terms of team performance, communication, and potential automation induced complacency.The results of the measures pertaining to both the controller and pilot have been published elsewhere [5].Hence, the objective of this paper is to describe the automation, procedures, information requirements, and other subjective data for controllers only, when pairing aircraft for simultaneous approaches in an integrated study of controllers and pilots. +BackgroundThe FAA has successfully conducted independent approaches to parallel runways for over forty years using the Instrument Landing System (ILS) navigation and terminal radar monitoring [1].Some airports, like San Francisco International (SFO), can support approximately 60 landings per hour on two parallel runways that are 750 ft apart using visual approaches, and approximately 45 landings under Simultaneous Offset Instrument Approaches (SOIA) under limited cloud ceilingvisual meteorological conditions (VMC).As visibility degrades further, the current navigation and surveillance system, as well as the existing procedures, cannot support SOIA approaches, dramatically reducing the landing rate [6].Previous human-in-the-loop studies have explored Very Closely Spaced Parallel Runways (VCSPR) operations from the flight-deck perspective.One study examined pilot responses to VCSPR operations using the Airborne Information for Lateral Spacing (AILS) concept [7].This concept requires technologies that enable the use of precise navigation and surveillance data, as well as technology for the detection of blunders.Further simulations have been conducted by NASA to examine pilot procedures for paired approaches on runways that are 750 ft apart in instrument meteorological conditions (IMC) [2].Enhanced cockpit displays that depict both traffic and wake information were provided to the flight crew for these operations.The results from these investigations revealed that even in the blunder cases, no loss of separation was observed and the breakout trajectory was flown accurately.Also, pilot workload was manageable, and an adequate level of situation awareness (SA) was maintained.Previous research has also examined the role of the controller in parallel runway operations.Under SOIA, the controller has positive control over the aircraft until the pilot breaks through the clouds and the follower aircraft has visual contact with the leading aircraft [6].Under AILS, the final approach controller has positive control over the aircraft pair until the trailing aircraft is given a clearance for the AILS approach [8].Previous studies by the authors have explored procedures for controllers to pair aircraft under different levels of automation [4].Different levels of pairing automation were examined with respect to workload, situation awareness and various operational factors.The study found the most favorable controller feedback when they were given more flexibility, i.e., to either select pairs offered to them by the pairing scheduler or to create their own pairs.Another VCSPR concept known as Terminal Area Capacity Enhancing Concept (TACEC) [9] was collaboratively developed by Raytheon and NASA Ames Research Center.TACEC defines a set of automation-assisted procedures that can be used for conducting simultaneous instrument approaches to two or even three closely spaced parallel runways that are 750 ft apart.The concept defines a safe zone behind the leader aircraft where the trailing aircraft is protected from the wake of the leader.The suggested safe trailing distance for the follower aircraft is in a window of 5s to 25s behind the lead aircraft, with 15s representing the optimal temporal distance and +/-10s representing the tolerance [10].The goal for both controllers and pilots is to bring the aircraft in this wake-safe window at the "coupling point."The coupling point is defined as a point in airspace 12nmi from the runway threshold, where the aircraft achieve the desired wake-safe temporal spacing, as described above.Pilot procedures and information requirements for TACEC were explored in several studies, and controller procedures were examined in a separate investigation [4].The previous work provided the framework for the current study, which investigates VCSPR operations with both pilot and controller procedures integrated into the same human-in-the-loop simulation experiment.The remainder of the paper will discuss the methods and results for the controller participants in this simulation. +Method +ParticipantsThe participants were three teams of retired controllers from the Northern California Terminal Radar Approach Control (TRACON) facility.Each controller team participated with two glass-cockpit qualified flight crews.The total of number controllers and flight crews were nine and six respectively.Each controller team consisted of three controllers and each flight crew consisted of a Captain and a First Officer.All participants had at least 10 years of experience in their respective fields.The study was run for two days per flight crew with each ATC team participating for four days.The pairing procedures were developed with the assistance of pilot and air traffic control subject matter experts.The closely spaced parallel runway operations on 28L and 28R in use at SFO were simulated.Participants were briefed and trained on the pairing concept, the new displays, and their automation tools.Pseudo pilots controlled other aircraft targets in the scenarios to add realism.The data of the pseudo pilots' aircraft were not included in the analyses.The piloted crews always flew as the following aircraft, so there was always an opportunity to pair with another aircraft.All participants completed questionnaires and took part in a debrief session at the end of the study.During the trials, observers were positioned to watch the controllers' operations, and assessed the teamwork behavior ratings (described in the results section), took notes of controllers' verbal communication with each other, and recorded any other observations. +AirspaceOnly arrival traffic was simulated in this study.Two arrivals flows from the east, Yosemite and Modesto, one flow from the north, Point Reyes, one from the south, Big Sur, and one from the west, Oceanic, were simulated (Figure 1).The arrival flows were similar to those used in the current airspace, except every arrival flow split allowing the aircraft to land its pre-specified runway, either 28R or 28L.The simulation used an arrival rate of 60 aircraft per hour.The parallel runways 28L and 28R were separated by 750ft, and were used for arrivals.The visibility was assumed to be low, runway visual range of 2nmi at 400ft in all the traffic scenarios.Two comparable traffic scenarios were exercised during eight data collection runs with a Visual Flight Rules (VFR) level of traffic.The scenarios were scripted to simulate an upstream scheduler that metered traffic into the terminal area. +Controller ProceduresThree controller positions were simulated for the study, namely, Area Coordinator, Boulder Sector and Niles Sector.The Boulder controller managed the north (Point Reyes), south (Big Sur) and west (Oceanic) traffic flows.The Niles Controller managed the arrival flows from the east, Yosemite and Modesto.The coordinator position was responsible for the creation of pairs in the two sectors, Niles and Boulder.The goal of the pairing procedure was to have the trailing aircraft reach the coupling point at 5 to 25s behind the lead aircraft.The Area Coordinator could pair aircraft from any of the five arrival streams but not the same stream to avoid an overtake situation.The sector controllers were responsible for maintaining the pairs to the 'coupling point' (12 nmi from the runway threshold) with the desired intra-pair spacing of 5-25s.They were allowed to use speed adjustments only to achieve pairing and spacing.The flight deck of the following aircraft had speed control algorithms that allowed the flight deck to adjust speeds automatically in order to come behind the lead aircraft in the wake-safe zone of 5-25s.The controller was not allowed to manipulate the speeds of the following aircraft, unless pairing was cancelled.However, the controllers had more direct control over the lead aircraft.The procedure to manipulate speeds on the follower required controllers to cancel the pair and then provide speed commands to the following aircraft.If they did not wish to cancel the pair, they manipulated the speeds on the lead aircraft, and eventually the speed algorithm on the follower reacted and accordingly adjusted the speeds on the follower.Based on the findings of previous research [4], a level of automation was selected for the pairing tool, in which the automation suggested pairs of aircraft in the Pairing Table (Figure 2).The Area Coordinator then had the option of either selecting one of the suggested pairs, or manually overriding the pairs suggested by automation and selecting an alternate pair.The main goal for the coordinator was to evaluate pairs offered by the automation to ensure the two aircraft were capable of landing between 5 and 25s of each other.The coordinator used the timeline (Figure 3) to evaluate and select aircraft that appeared to be natural pairs, such that their times to the runway-thresholds were within 30-60s from each other.To finalize a pair, the coordinator evaluated the pair suggested by the automation against the timeline.If the pair was evaluated as acceptable, the coordinator sent a data link message to the two aircraft.When the pilots of both aircraft acknowledged the pairing, the aircraft call signs turned green in the pairing table.The pairing table in the sector controller's display contained only finalized pairs.Both aircraft in the pair were then given an approach clearance electronically by the sector controller who owned the trailing aircraft in the pair.The approach clearance was given at about 14 nmi from the threshold.It was found necessary that the two aircraft in the pair receive the approach clearance at the same time to ensure that they make the 15s temporal separation at the 'coupling point.'The approach clearance also implicitly delegated separation authority to the flight-deck.Aircraft pairs that were out of conformance could only be given approach clearances via voice.Once an approach clearance was provided, the aircraft changed color to blue in the pairing table.This color coding helped the controllers manage information about the pair.If a pair lost conformance, controllers had to perform any of the following three options -1) re-establish the pair after making speed adjustments (if possible), 2) land the planes as singles, or 3) vector them away and return them back to the flow upstream.Any of the controller positions could cancel a pair, by highlighting the pair in the pairing table and pressing the delete button.All the three controller positions had a pairing table, which listed all pairs in the order of their Estimated Times of Arrival (ETAs), with a continually-updated timeline (configured to show the ETAs of the aircraft to the two parallel runways), and a conformance monitoring tool (Figure 4), which displayed two bars on the following aircraft to show the leading and trailing edge of the 5-25s conformance envelope. +Results and DiscussionThe study goal was to explore the dynamic and integrated role of controllers and pilots for aircraft pairing on simultaneous arrivals.Results on the metrics of throughput, controller workload and controller situation awareness have been reported elsewhere [5].Hence, these results will only be described briefly.The remainder of this paper will focus, in greater detail, on the other metrics such as team behaviors as reported by the observers, trust in automation and other subjective feedback received by the controllers, which help to define the air traffic control information requirements and procedures.The study aims to provide results on team behavior to investigate the changes induced due to the introduction of the Area Coordinator position and the task of pairing aircraft assigned to them.Similarly, inter-controller communication is reported to explore any changes in communication brought about by the introduction of new pairing automation and procedures that involved pairing aircraft from different sectors.Potential for complacency towards the new pairing automation and procedures have also been described in this paper to assess controller's level of trust in the new automation.The metrics presented here have been averaged across the three controller positions. +Prior +Results: Throughput, Workload, Situation AwarenessThe controllers helped achieve the desired VFR throughput by pairing 30 aircraft and canceling only one pair (on average) in any 30 min run.Although the objective of the controller was to land as many pairs as possible, having a small number of singles or unpaired aircraft (e.g., canceled pairs) helps with efficiency, particularly in cases when an aircraft was vectored or had a go-around and had to be reintegrated back into the flow.The ATWIT (Air Traffic Workload Input Technique) [11] was used to collect subjective controller workload assessments during the course of the simulation runs.While there were no statistically significant differences between the positions on subjective workload, overall ratings indicated a low and manageable level of controller workload, on average.However, the variability in the workload rating distribution suggested that workload was occasionally high enough to prevent vigilance decrement.Controller situation awareness data were collected using the Situation Awareness Rating Technique (SART) [12].ANOVA results indicated high controller situation awareness across all the positions.No statistically significant differences were found between the positions. +Team Behavior DataThree experiment observers used an adapted version of the Anti-air Teamwork Observation Measure (ATOM) [13] to provide an assessment of controller team behaviors.ATOM consists of 15 items that measure six dimensions of teamwork, namely, communication, monitoring, feedback, backup, coordination, and team orientation.The authors felt the need to assess the impact of the new position-Area Coordinator required for pairing on team behavior.The adapted version of ATOM used in this study has a reduced number of questions/items in the overall scale, but collectively, all items map to the same 6 dimensions of teamwork.It is designed to be used by observers who have operational knowledge of participants' tasks.Three observers used the scale to observe the team behavior exhibited by the three controller positions.The scale is a behaviorally anchored 7-point Likert scale that is used by the observers to capture poor-team behavior on one end of the scale (1 on the scale) and good team behavior on the other end of the scale (7 on the scale).Table 1 shows the average team behavior across all positions.Overall the observers rated the team at mid-point to above mid-point level on team behavior.In the absence of other data on team behavior in similar air traffic management setting, it is difficult to interpret the mid-point range as being average or not.The items 'Providing Guidance', 'Error Correction' and 'Providing & Requesting Backup' were found to be relatively low.The item 'Stating Priorities' was found to be the lowest rated item amongst all the ratings.These items particularly depict the roles and responsibilities that the controllers assumed while performing the pairing task.This could mean that the controller-participants did not regard stating priorities or providing guidance for another controller as their job unless they are in a supervisor position.We found that the Area Coordinator position sometimes took that role, or sometimes, the expert in the group assumed that role.More insight into inter-controller communication data is provided in the following section. +Inter-Controller CommunicationsThe Adaptive Architectures for Command and Control (A2C2) technique [14] was used to assess semantic and quantitative aspects of inter-controller verbal communications.All inter-controller communication was recorded by the observers stationed at every controller position.These communications were then categorized as 'Requests' or 'Transfers' using the form shown in Figure 5.The number of transfers and requests assesses the push and pull of information within the team.Push refers to information being proactively offered and pull refers to information requested or actively sought.Within the 'Request' and 'Transfer' category the items were further categorized for information, action, and coordination.Since the categories 'action' and 'coordination' were hard to separate in the terminal environment, they were merged as 'coordination' only.Thus the current investigation used only four of the communication categories provided by the A2C2, namely, information requests, information transfers, coordination requests, and coordination transfers. +Quantitative communication analysis through the A2C2 technique involves an anticipation ratio.The anticipation ratio measures efficiency of communication for effective team performance [14].The ratio is the number of communication transfers to number of communication requests.A ratio greater than one indicates that more information is being sent than asked for and information needs are anticipated before requests occur.A higher ratio indicates more information is being provided than has been requested.The ratios help understand the nature of the communication that is being proactively used to achieve the goals of the tasks.2), which led to an anticipation ratio for coordination to be 4.14, which is almost double the anticipation ratio for information (2.18).It is difficult to interpret what an anticipation ratio of 4.14 means.However, it's safe to say that the level of coordination in this study was double that of level of information proactively provided by the controller.The conditions under which the overall number of communication increased involved an aircraft pair going out of conformance or the necessity to vector aircraft out of approach routes and merge it back into the arrival flow.Also, higher levels of coordination were required to handle inter-pair spacing when the lead and following aircraft were in different sectors. +Team Id____ Run_____ Date____ Observer______Chi-square analyses yielded no significant differences between the positions or the scenarios on the various categories of communication. +Controller Feedback on CommunicationsThe controllers provided feedback on levels of communication and coordination during the debrief sessions.They indicated that coordination was used quite frequently when the lead and following aircraft were in different sectors.Controllers reported that radio communication between controllers and pilots increased when an aircraft pair had to be canceled and vectored.Those situations were often marked as high workload for the controllers.Approach clearances were sent via data link, which had the effect of reducing radio communication.However, the procedures required one controller to give approach clearances for two aircraft and sometimes, they were in different sectors.In this situation, there was increased verbal coordination with the other sector controller because the aircraft being cleared for approach was not owned by the same controller.Some controllers mentioned that they still preferred to give the approach clearance verbally instead of using datalink, despite the increase in workload.The verbal clearance issuance would provide the controllers the assurance that the pilots are aware of their responsibility for selfspacing from the aircraft in front of them.The controller participants agreed that although the level of overall radio communication was reduced between the controllers and the pilots, some of the verbal communications between the controllers increased. +Complacency Potential Factor in Pairing AutomationA Complacency Potential Rating Scale was used to collect data on automation-induced complacency [15].Wiener [16] defined complacency as "a psychological state characterized by a low index of suspicion."Automation is often identified as a significant factor that induces complacency.Procedures, roles and responsibilities are also potential factors that induce complacency.According to Wickens [17], reliability in automation engenders excessive trust and over-reliance in pilots.Singh et.al. [15] identified four factors that may be related to over-trust or complacency in automation.These are confidence, reliance, trust, and safety in automation.Some examples of scale items that measure different constructs are shown in Table 3 below.The Complacency Potential Rating Scale was adapted and used to collect data for all the three controller positions.The adapted scale for the four dimensions is provided in the Appendix.The scale uses a 5 point Likert scale that ranges from 'strongly disagree' to 'strongly agree'.For some of the questions in the rating scale are reversed to ensure reliability in the responses.The scale was adapted to ask questions about the pairing automation and procedures that the controllers used.It was found that the controllers reported trust and confidence about the pairing automation (Figure 7).The controllers also rated the pairing procedures and conformance bars as highly safe.However, the controllers did not rate the pairing automation as highly on the Reliance scale.This provides some insight into the way the automation was used by the controllers.In general they believed that pairing automation was better than manually pairing aircraft or monitoring aircraft for conformance.The pairing automation suggested pairs to the controller, which they could manually override at any time.Since they had little experience with the pairing automation, they did not assume that the automation always selected the best aircraft pairs.Rather, they evaluated every pair against the timeline before finalizing the pairs for simultaneous arrivals. +Reliance"How many safeguards does conformance monitoring provide against error?e.g., miscalculations for the leader or follower." +Trust"Which method do you think is more likely to be correct-manually monitoring or using automation for conformance monitoring?" +Safety"Given the choice between using automated conformance monitoring or manual monitoring to monitoring to ensure 15 s between aircraft, which would you use?" +Controller InterfacesThe controllers were also asked questions on the ease with which they derived information from the displays on certain functionalities (Table 3).They rated the questions on a 5 point scale where 1 represented 'very difficult' and 5 represented 'very easy.'The questions on responsibility and display confusion used a reversed scale.Overall the controllers felt that they could easily create a pair.Sometimes they found it difficult to locate the leader, especially when the aircraft was not in their sector.A 'locate' button on the pairs table was provided, but it involved multiple steps, where the controller first selected the pair on the pairing table, and then pressed the locate button.The controllers reported that this function had too many steps.The controllers also reported canceling the pair, which involved similar steps as the locate function, also a cumbersome multi-step procedure.Finally, the controllers experienced little confusion over display features or roles and responsibilities amongst themselves or between air and ground. +Controller FeedbackThis section describes the feedback provided by the controllers during debriefs and observations made by the observers.The controllers reported that they were able to utilize all the tools provided, e.g., the pairing table, conformance monitoring, timeline for the pairing task, etc.They did report that they used the tools in different ways at the various positions.For example, the Area Coordinator used the timeline and pairing table extensively, whereas the sector controllers used the conformance monitoring and timeline more often.The pairing table was used to issue clearances by the Niles and Boulder controllers, while the conformance bars provided support for appropriate spacing between the aircraft and meeting the goal of achieving the 15s temporal separation between the two aircraft.As mentioned earlier, the controllers found the 'locate' and the 'cancel' features on the display very cumbersome.A different procedure for the locate function was recommended by the controllers.They suggested using a mouse-over or dwell-over on any aircraft that should highlight both aircraft in the pair.In the experiment the dwell-over was used to bring up the conformance bars on the following aircraft.Their suggestion was to use the same dwellover to locate the aircraft as well.The ability to re-sequence or re-establish an aircraft pair after the pair had been canceled and the aircraft had been vectored was also found to be very difficult for the controllers.A contributing factor to this difficulty was that the timelines did not always update the vectoring aircraft accurately to allow the controller to pair the aircraft.As for sending approach clearances via data link, it had the positive effect of reducing controllerpilot communications.This also increased the intercontroller coordination because the controller sending the clearance did not always have ownership or control over both the aircraft.The controllers also suggested that the controller owning the lead aircraft should send the clearance to both the aircraft in the pair instead of the owner of the following aircraft.This was suggested because as discussed earlier, the controller who owns the lead aircraft indirectly also has control over the speed of the following aircraft.The controllers reported that they liked the manual override feature because it gave them flexibility and control especially when the automation offered pairs in a sequence they did not prefer.Sometimes, they reversed the sequence of the leader and follower to achieve better inter-pair spacing, for which they were still responsible.This is because making the natural leader, a follower forced the speed control on the natural follower's flight deck to go behind the chosen leader, creating extra spacing in front of the leader.The controllers mentioned that having aircraft pairs distributed between the two sectors made the conformance tasks extremely difficult.Often, to achieve inter-pair spacing, they had to request the controller of the lead aircraft to adjust the speed to keep the pair intact.They suggested the ability to make small speed changes on the following aircraft, without canceling the pair, would be a useful feature.The presence of automation on the flight deck to manage inter-pair spacing may also have created confusion between pilots and controllers over who had the final authority and responsibility over a pair.The controllers assumed that they would be notified by the flight deck if pilots could not achieve the 15 s temporal spacing behind the lead aircraft.However, the pilots closely monitored their lead aircraft's speeds in order to maintain their own conformance, but also expected that the controllers might cancel their pair at any time. +Summary and ConclusionsA high fidelity human-in-the-loop simulation experiment was conducted to investigate the integrated dynamic role of controllers and pilots for pairing aircraft to closely spaced parallel runways for simultaneous approaches.Since the results of the integrated measures pertaining to both the controller and pilot have been published elsewhere [5], this paper reports the results pertaining to the controller's role in this investigation, focusing on controller team performance, controller communications, potential complacency factors and controller feedback.Results show that the controllers were able to achieve the desired VFR-level throughput during low-visibility, low-ceiling using the pairing automation.Results also suggest that the controllers experienced manageable workload and a high level of situation awareness.Results provide evidence as per the observers, the Area Coordinator and two sector controllers performed capably as a team, with most of the 11 team behavior analysis scales being in the mid-point and above mid-point range.Analysis of intercontroller communications indicate that the controllers may be anticipating information needs even before requests occur, and that controllers tend to spend more time communicating among controllers when pairing aircraft from different sectors or under aircraft-breakout conditions.The conformance monitoring tools such as the conformance bars aid the controllers in this proactive form of communication.Complacency-Potential analyses show that the controllers generally report trust in the automation, although they did not assume that the automation would always select the best aircraft pair, and they would carefully evaluate each pair against the timeline before making a final selection.Results were somewhat mixed on the ease of deriving information from the automation and the displays.The controllers made specific suggestions to increase the usability of the system (e.g., to improve the 'locate' function by simplifying the procedures).Overall, results of this investigation show the potential promise of the air traffic control pairing automation tested, pending future research and system enhancements.While controllers were able to use the system quite capably and safely, and liked many features of the automation (e.g., the manual override in selecting aircraft pairs), they did provide suggestions for improvement.Controller feedback suggested that locate, cancel, re-sequence, pairreassignment were found to be somewhat cumbersome to use, so it seems that careful attention is needed to address these and other issues, prior to the full implementation of the automation.Finally, there appeared to be some confusion over who had the final authority and responsibility over the aircraft pairs, when separation conformance tasks were shared between the flight-deck and air traffic control.This issue must be fully addressed, prior to the introduction of any new automation system into the real-life milieu of air traffic control operations.Figure 1 .1Figure 1.SFO Airspace. +Figure 2 :2Figure 2: Partial view of the finalized pairs in the pairs table. +Figure 3 .3Figure 3. Timeline showing aircraft scheduled for the two runways 28L and 28R.Example of natural pair QFA83 and SWA246. +Figure 4 .4Figure 4. Conformance monitoring bars. +Figure 5 .5Figure 5. Example of Matrix used to capture Verbal CommunicationMost of the communication instances were directed towards coordination (Table2), which led to an anticipation ratio for coordination to be 4.14, which is almost double the anticipation ratio for information (2.18).It is difficult to interpret what an anticipation ratio of 4.14 means.However, it's safe to say that the level of coordination in this study was double that of level of information proactively provided by the controller.The conditions under which the overall number of communication increased involved an aircraft pair going out of conformance or the necessity to vector aircraft out of approach routes and merge it back into the arrival flow.Also, higher levels of coordination were +Figure 7 .7Figure 7. Complacency Potential Rating for the pairing tool and procedures +Table 1 : Means and Standard Deviations of Team Behavior Ratings Mean SD1Seeking Sources4.000.89Passing Information4.100.87Situation Update4.200.87Using Proper Phraseology4.800.51Providing Guidance3.800.97Stating Priorities3.101.13Completeness of Reports4.600.64Brevity4.400.94Clarity4.800.50Error Correction3.700.93Providing & Requesting Backup3.700.87 +Table 2 . Frequencies and Anticipation Ratios for Inter Controller Communications (Mean values and Standard Deviations)2CATEGORYMEANSDNumberof73.0035.50information transfersNumberof40.5024.90information requestsInformation2.181.02anticipation ratioNumberof95.0053.70coordination transfersNumberof26.5018.40coordination requestsCoordination4.141.17Anticipation ratioTotal number of168.0087.90transfersTotal number of66.8042.60requestsAnticipation ratio for2.901.04all communications +Table 3 . Examples for Complacency Potential rating Scale3Confidence"Theconformancemonitoringfunction is reliable""Theconformancemonitoringautomation is safe compared tomonitoring aircraft manually." +Table 4 . Ease of deriving information on different pairing function procedures (Means and Standard Deviation)4FunctionalityMeanSDratingCreating a pair4.700.50Locating leaders4.001.00Locating trailers4.300.70Locating pairs4.600.50Conformance4.201.00monitoring of your pairCanceling a pair3.401.10Sendingapproach4.201.00clearanceSending handoffs3.701.00Responsibility1.700.50confusion (reversed)Displayconfusion1.900.60(reversed) + + + + + + + + + + + Frequency and severity of deviations during simultaneous independent approaches to parallel runways + + SebastianVMassimini + + + GerryCMcneil + + + NoamTene + + 10.1109/icnsurv.2010.5503313 + + + 2010 Integrated Communications, Navigation, and Surveillance Conference Proceedings + Mclean, VA + + IEEE + 2006 + + + + Massimini, S. (2006) Simultaneous independent and dependent parallel instrument approaches: Assumptions, analysis, and rationale. Mitre Product: MP 06W0000127, Case number 06-0984. Mclean, VA. + + + + + Comparison of manual and autopilot breakout maneuvers with three closely spaced parallel runway approaches + + SavitaVerma + + + SandraLozito + + + ThomasKozon + + + GordonHardy + + + HerbertResnick + + 10.1109/dasc.2009.5347457 + + + 2009 IEEE/AIAA 28th Digital Avionics Systems Conference + Orlando, FL + + IEEE + 2009 + + + Verma, S.A., Lozito, S., Ballinger, D., Kozon, T., Hardy, G. & Resnick, H. (2009) "Comparison of Manual and Autopilot Breakout Maneuvers for Three Closely Spaced Parallel Runways approaches." Digital Avionics Systems Conference, Orlando, FL. + + + + + Case Study of Paired Approach Procedure to Closely Spaced Parallel Runways + + JonathanHammer + + 10.2514/atcq.8.3.223 + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 8 + 3 + + 2000 + American Institute of Aeronautics and Astronautics (AIAA) + + + Hammer, J. (2000). Case study of paired Approach procedure to closely spaced parallel runways. Air Traffic Control Quarterly, 8 (3), 223- 252. + + + + + Preliminary Guidelines on Controller's Procedures for Pairing Aircraft for Simultaneous Approaches under Different Levels of Automation + + SavitaVerma + + + ThomasKozon + + + DeborahBallinger + + 10.1201/ebk1439834916-c52 + + + Advances in Human Factors and Ergonomics Series + Orlando + + CRC Press + 2010. July 2010 + + + + Verma, S., Kozon, T., & Ballinger, D (2010). Preliminary Guidelines on Controller's Procedures for Pairing Aircraft for Simultaneous Approaches under Different Levels of Automation. Applied Human factors and Ergonomics, Orlando, July 2010. + + + + + Role of the controller in an integrated pilot-controller study for parallel approaches + + SavvyVerma + + + ThomasKozon + + + DebbiBallinger + + + SandyLozito + + + ShobanaSubramanian + + 10.1109/dasc.2011.6096205 + + + 2011 IEEE/AIAA 30th Digital Avionics Systems Conference + Berlin, Germany + + IEEE + 2011. June, 2011 + + + Verma, S., Lozito, S., Martin, L., Kozon, T., Kaneshige, J., Sharma, S., Ballinger, D. & Cheng, L. (2011). Integrated Pilot and Controller Procedures: Aircraft Pairing for Simultaneous Approaches to Closely Spaced Parallel Runways. Ninth USA/Europe Air Traffic Management Research and Development Seminar (ATM2011), Berlin, Germany, June, 2011. + + + + + Compilation of DOT and FAA Airport Legal Determinations and Opinion Letters as of December 31, 2020 + 10.17226/26202 + + + + Simultaneous Offset Instrument Approaches -Newark International Airport: An Airport Feasibility Study + + Transportation Research Board + 2001 + + + Magyratis, S., Kopardekar, P., Sacco, N., & Carmen, K. (2001). Simultaneous Offset Instrument Approaches -Newark International Airport: An Airport Feasibility Study, DOT/FAA/CT-TN02/01. + + + + + NASA administrator announces resignation + + TAbbott + + + DElliot + + 10.1016/s0045-8732(01)80042-5 + Concept. NASA/TP-2001-210665 + + + COSPAR Information Bulletin + COSPAR Information Bulletin + 0045-8732 + + 2001 + 152 + + 2001 + Elsevier BV + + + Abbott, T. & Elliot, D. (2001). Simulator Evaluation of Airborne Information for lateral Spacing (AILS) Concept. NASA/TP-2001-210665. + + + + + + MWaller + + + TDoyle + + + FMcgee + + Analysis of the Role of ATC in the AILS Concept, NASA/TM-2000-210091.Symposium on Aviation Psychology + Columbus, OH + + 2000 + + + Waller, M., Doyle, T. & McGee,F. (2000). Analysis of the Role of ATC in the AILS Concept, NASA/TM-2000-210091.Symposium on Aviation Psychology, Columbus, OH. + + + + + CNS requirements for precision flight in advanced terminal airspace + + MEMiller + + + SDougherty + + + JStella + + + PReddy + + 10.1109/aero.2005.1559471 + + + 2005 IEEE Aerospace Conference + + IEEE + + + + Miller, M., Dougherty, S., Stella, J., & + + + + + CNS Requirements for Precision Flight in Advanced Terminal Airspace + + PReddy + + + + Aerospace, 2005 IEEE Conference + + 2005 + + + + Reddy, P. (2005) CNS Requirements for Precision Flight in Advanced Terminal Airspace," Aerospace, 2005 IEEE Conference, pp. 1-10. + + + + + Models of Wake-Vortex Spreading Mechanisms and Their Estimated Uncertainties + + VernonRossow + + + GordonHardy + + + LarryMeyn + + 10.2514/6.2005-7353 + + + AIAA 5th ATIO and16th Lighter-Than-Air Sys Tech. and Balloon Systems Conferences + + American Institute of Aeronautics and Astronautics + + + + Rossow, V. J., Hardy, G. H., & Meyn, L. + + + + + Models of Wake-Vortex Spreading Mechanisms and Their Estimated Uncertainties + + A + + AIAA-2005-7353 + + + ATIO Forum + + 2005 + Arlington, VA + + + A. (2005) Models of Wake-Vortex Spreading Mechanisms and Their Estimated Uncertainties, AIAA-2005-7353, ATIO Forum, Arlington, VA. + + + + + Air Traffic Controller Workload: An examination of workload probe (FAA/CT-TN90/60) + + ESStein + + + 1985 + FAA + Atlantic City, New Jersey + + + Stein, E.S. (1985). Air Traffic Controller Workload: An examination of workload probe (FAA/CT-TN90/60). Atlantic City, New Jersey: FAA. + + + + + Situation Awareness as a Predictor of Performance for En Route Air Traffic Controllers + + FrancisTDurso + + + CarlaAHackworth + + + ToddRTruitt + + + JerryCrutchfield + + + DankoNikolic + + + CarolAManning + + 10.2514/atcq.6.1.1 + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 6 + 1 + + 1998 + American Institute of Aeronautics and Astronautics (AIAA) + + + Durso, F., Hackworth, C., Truitt, R., Cruthfield, J. Nikolic, D., Manning, C. (1998). Situation Awareness as a Predictor of Performance for En Route Air Traffic Controllers, Air Traffic Control Quarterly, 6(1), pp.1-20. + + + + + Online anti-air warfare team observation measure + + KSmith + + + + Orlando FL: Naval Air warfare Center Training System Division + + 1994 + + + unpublished paper + Smith, K. (1994). Online anti-air warfare team observation measure: (unpublished paper). Orlando FL: Naval Air warfare Center Training System Division. + + + + + Measures for Evaluation of Team Processes and Performance in Experiments and Exercises. Paper presented at the 6th International CCRTS + + EEEntin + + + EBEntin + + + 2001. June 19-21 + Annapolis, Maryland + + + Entin, E. E., & Entin, E. B. (2001, June 19-21). Measures for Evaluation of Team Processes and Performance in Experiments and Exercises. Paper presented at the 6th International CCRTS, Annapolis, Maryland. + + + + + Automation- Induced "Complacency": Development of the Complacency-Potential Rating Scale + + IndramaniLSingh + + + RobertMolloy + + + RajaParasuraman + + 10.1207/s15327108ijap0302_2 + + + The International Journal of Aviation Psychology + The International Journal of Aviation Psychology + 1050-8414 + 1532-7108 + + 3 + 2 + + 1993 + Informa UK Limited + + + Singh, I., Molloy, R. and Parasuraman, R. (1993). Automation-Induced "Complacency": Development of the Complacency-Potential Rating Scale. The International Journal of Aviation Psychology, 3(2), 111-122. + + + + + Complacency: Is the term useful for air safety? + + ELWiener + + + + Proceedings of the 26 th Corporate Aviation Safety Seminar + the 26 th Corporate Aviation Safety SeminarDenver, CO + + Flight Safety Foundation, Inc + 1981 + + + + Wiener, E.L. (1981). Complacency: Is the term useful for air safety? In Proceedings of the 26 th Corporate Aviation Safety Seminar (pp.116- 125). Denver, CO: Flight Safety Foundation, Inc. + + + + + Engineering Psychology and Human Performance + + ChristopherDWickens + + + JustinGHollands + + + SimonBanbury + + + RajaParasuraman + + 10.4324/9781315665177 + + 1992 + Psychology Press + New York + + + Wickens, C.D. (1992). Engineering psychology and human performance. New York: Harper Collins. + + + + + + diff --git a/file757.txt b/file757.txt new file mode 100644 index 0000000000000000000000000000000000000000..c7235606cadbd3b9378c0fb4330081ae3037c17f --- /dev/null +++ b/file757.txt @@ -0,0 +1,379 @@ + + + + +I. INTRODUCTIONManaging departure operations in busy airport and airspace environments, with limited data sharing and system integration, can significantly reduce efficiency and predictability.Stakeholders, including air carriers and air navigation service providers, have their own objectives in managing traffic, and often these objectives compete with one another.Also, decisions in managing traffic are often made in a reactive manner with short planning horizons due to operational uncertainties and a lack of common situation awareness between Flight Operators and service providers.For example, during periods when demand exceeds capacity at the airport, service providers manage traffic using the First Come, First Serve (FCFS) paradigm, where they serve the flights that first call in as ready for pushback.With many airlines having similar ticketed departure times, this leads to surface congestion.A departure metering tool that could meter the traffic while considering arrivals, runway crossings, etc. in a tactical manner could potentially alleviate the problem.Research on one such tool, the Spot and Runway Departure Advisor (SARDA) [1,2] was conducted at NASA Ames Research Center.Other tools that perform departure metering include those deployed at sites such as John F. Kennedy (JFK) airport and are focused on providing Target Movement Area entry Times (TMATs) to the airlines several hours in advance, making the tool primarily strategic in nature [3].The tool deployed at JFK provides departure metering capability with a longer planning horizon, i.e., several hours into the future, and it also allows the users to update flight ready times and request swap flights as they know the situation better.There exists a need for a departure metering tool that is more tactical in nature and can handle the changing demand and capacity over a relatively short time horizon.NASA is collaborating with the Federal Aviation Administration (FAA) and aviation industry partners to develop and demonstrate new concepts and technologies to solve some of these complex problems in the Integrated Arrival, Departure, and Surface (IADS) traffic management capabilities under the Airspace Technology Demonstration 2 (ATD-2) subproject.The primary goal of ATD-2 is to improve the predictability and the operational efficiency of the air traffic system in metroplex environments while maintaining or improving throughput by enhancing and integrating arrival, departure, and surface prediction, scheduling, and management systems.The IADS capabilities defined in the ATD-2 project are built upon the previous NASA research, including the Spot and Runway Departure Advisor (SARDA) [1,2], the Precision Departure Release Capability (PDRC) [4], and the Terminal Sequencing and Spacing (TSAS) capability [5].Benefit analysis results indicated substantial opportunities to reduce taxi delays for both departures and arrivals and increase throughput and predictability by integrating these capabilities [6].The ATD-2 subproject is a five-year research activity that will run through 2020.In Phase 1 of the project, the Baseline IADS capability will be demonstrated at Charlotte Douglas International Airport (CLT) in 2017.In this first phase, the tactical surface scheduling capability and the user interfaces for ramp controllers and ramp traffic managers will be implemented for ramp operations.The tactical surface scheduler or metering tool was developed based on previous research conducted on NASA's SARDA tool [1,2].Its key capability is the initial integrated system of tactical surface scheduling that incorporates Surface Collaborative Decision Making (S-CDM) [7,8] principles and tactical departure scheduling to an enroute meter point that will help insert departures into the overhead departure stream.This paper focuses on the Tactical Surface Scheduler, also referred to as the Metering Tool.The tool described in this paper is tactical in nature and works over a relatively short time horizon.It is meant to provide the airline ramp controller with aircraft pushback advisories that reduce surface congestion and to respond to surface and airspace constraints that become known with greater certainty in the next 10 minutes.For this purpose, the tactical surface metering tool first estimates the capacity of current and near-future runway resources from flight schedule and surveillance data.With demand forecasts and predicted taxi trajectories, this tool computes an efficient runway schedule of aircraft within the planning horizon based on their flight readiness, Earliest Off-Block Times (EOBTs), according to a ration by schedule (RBS) rule.The tool then generates gate pushback and recommended hold time advisories to meet the runway schedule.These advisories are shown on the user interfaces for the ramp controller and the ramp traffic manager, called Ramp Traffic Console (RTC) and Ramp Manger Traffic Console (RMTC), respectively.RTC and RMTC were developed as part of the SARDA project and the research found they could successfully replace the paper strips for the ramp controllers [2].The tactical scheduler is expected to run all the time, but the ramp manager can turn the metering on and off, according to their strategy for demand/capacity balancing.When the ramp manager decides to turn on time-based metering, he or she can choose the level of gate holding from three options -'Nominal hold,' 'Less hold,' or 'More hold' -depending on the traffic situation.The 'Nominal hold' option seeks to utilize the existing runway capacity with the available demand.It aims to provide a gate hold level that is associated with 'nominal' or acceptable queues in the Airport Movement Area (AMA).Discussions with Subject Matter Experts (SMEs) at CLT provided eight aircraft in a queue as a good starting number for quantifying the 'nominal' level of hold.This number became the basis for the experiment matrix discussed later.The 'Less hold' option allows flights to spend more time on the airport surface (movement area) when compared to nominal level, whereas the 'More hold' option allows the flights to be held at their gates longer, thus resulting in less delay or excess queue time on the surface or movement area.These gate hold levels are associated with a metering value that defines the level of delay or excess queue time that will be incurred in the AMA.In March 2017, NASA conducted a human-in-the-loop (HITL) simulation that integrated airspace and surface operations for CLT airport.The objectives of the HITL simulation were to evaluate the operational procedures and information requirements for the tactical surface metering tool, APproval REQuest (APREQ)/Call For Release (CFR) procedures between the Air Traffic Control Tower (ATC-T) and the Air Route Traffic Control Center (or Center), and data exchange elements between the Ramp and the ATC Tower.The results of the APREQ/CFR procedures are discussed elsewhere [9].One of main goals in this simulation was determining the parameters to set the level of gate holding for the tactical surface scheduler's delay propagation logic.As described above, three different levels of gate holding (and the resultant excess queue time) were manipulated for each scenario in the HITL simulation.This paper evaluates and describes the effectiveness of the tactical surface metering tool and the results of the calibration of the gate holds by analyzing the HITL simulation results for CLT.Section II briefly provides an overview of the airport surface and airspace operations at CLT, the target test site for the simulation.Section III describes the HITL simulation method, including traffic scenarios, participants, simulation facilities and equipment, and the surface metering tool.Sections IV and V discuss the objective and subjective performance metrics from the simulation.Lastly, Section VI provides the closing remarks. +II. CLT OPERATIONS OVERVIEWAccording to the recent airport activity report, CLT accommodates about 1,400 operations per day and was the seventh busiest airport in aircraft movements worldwide in 2016 [10].Because CLT is one of the main hub airports for American Airlines (AAL), AAL and its regional air carriers operate nearly 93% of the flights into and out of the airport.The remaining operations are comprised of other regional carriers, mainline flights operated by Southwest, Delta, United and Jet Blue, military flights, business and general aviation, and air cargo.As the dominant carrier, AAL manages all ramp operations at the airport, whereas air traffic on the airport movement area (AMA) is controlled by the ATC Tower (ATC-T).As shown in Fig. 1, CLT has three north/south parallel runways (18L/36R, 18C/36C, and 18R/36L) that can support simultaneous independent instrument approaches, and a fourth diagonal runway (5/23) that intersects Runway 18L/36R.The airport operates in either a "North" or "South" flow configuration.The diagonal runway, Runway 23, is used in a South flow configuration for arrivals.Runway 5 (the opposing end) is not used for arrivals or departures during normal daylight/evening operations, but it is used as a taxiway in a Traffic at CLT is characterized by definite peaks and valleys.There are clear distinctions between departure and arrival banks throughout the day.Each departure and arrival bank takes approximately an hour with a slight overlap existing between banks.Ramp Control strives to clear the departures from the gates before an arrival bank builds up, so that ramp congestion and gate conflicts can be minimized.The ramp area is divided into four sectors (e.g., West, South, East, and North sectors).The corresponding ramp controller controls the traffic in each sector.The ramp operations at CLT are constrained due to physical limitations of the ramp, such as limited ramp space with alleys between concourses, single-direction taxiways, and limited holding areas (hardstands). +III. HUMAN-IN-THE-LOOP SIMULATION METHODThis study evaluates the Tactical Surface Scheduler or Metering Tool and determine the level of gate holds that are acceptable to both the Ramp and ATC Tower.This research effort involved both retired and active ramp controllers, one ramp manager from CLT and several pseudo pilots to effectively manage traffic.This high-fidelity simulation was conducted in the NASA Ames' Future Flight Central (FFC) that can generate a 360-degree out-the-window view of the airport.This study also simulated FAA's Air Traffic Control Tower (ATC-T), where the participants set runway utilization intent into the system that enables the system to automatically estimate runway capacity.From the estimated runway capacity, the surface metering tool generates Target Off Block Times (TOBTs) for individual flights and provides the controllers with its pushback advisories to throttle demand that results in mitigating surface congestion.These TOBTs are depicted on the decision support tool-Ramp Traffic Console (RTC) described in the next section. +Tools and Equipment 1. Ramp Traffic Console (RTC) and Ramp Manager Traffic Console (RMTC)The Ramp Traffic Console (RTC) and Ramp Manager Traffic Console (RMTC) are decision support tools developed for the ramp controllers and ramp managers, respectively.These tools provide a display that depicts the map of the ramp area with flight strips positioned at each gate for departures.The ramp controllers can provide flight intent information, such as pushback, holding a flight, changing the spot the flight is going to, changing its gate, and marking the flight if it is sent to the hard stand, by interacting with the tool.Double clicking on the flight strip allows the user to open the Flight Menu where the user can change assignment of a flight's spot, gate, or runway, and mark it as temporarily out of service or mark it as being sent to the hardstand.Gate pushback intent information can also be provided by the ramp controller: swiping the flight strip away from the gate marks the flight as pushback cleared, depicted with an engine symbol, whereas swiping the flight strip towards the gate marks the flight on hold by putting a red border around the flight strip (see Fig. 2).The color of the flight strips and icons shows the direction they are going to, the blue strips have destinations in the east direction whereas the brown are flying in the west direction.Arrivals are depicted as green color aircraft icons.Flights that are moving and tracked are shown as solid aircraft icons and those that are moving but not detected by surveillance are shown as hollow aircraft icons (see Fig. 2).Tactical Surface Scheduler/ Metering Tool recommended advisories are shown next to the flight strips (see Fig. 5).The ramp manager has the ability to turn on Time-Based Metering, i.e., the Metering Tool, via the user interface provided by the RMTC (see Fig. 3).If the Time-Based Metering option is selected, the ramp manager is required to select the preferred hold level as shown in the same user interface.The input made by the ramp manager regarding gate holds provides the variable for this research.Details on the Metering Tool and how the advisories are depicted on the flight strips on RTC is described in the following section on the Metering Tool.The details on the definition of these hold levels is described in the Experiment Matrix subsection. +Metering Tool/ Tactical Scheduler for SurfaceThe surface metering tool calculates Target Off-Block Times (TOBT) and provides gate hold recommendations to the ramp controller.For each departure flight, the tactical scheduler generates the Target Takeoff Time (TTOT) that would meet constraints, including runway separation criteria and TMI constraints.Next, the time a flight is required to be at a spot, called TMAT (Target Movement Area entry Time), is computed by subtracting the nominal or undelayed taxi time in the AMA with a delay buffer from the TTOT.Similarly, the nominal taxi time in the Ramp with a delay buffer is subtracted from TMAT to get TOBT for flights that are being metered.Based on the flight's TOBT, a gate hold recommendation is provided on RTC.The delay buffer, also called 'metering value,' is specified in the tactical scheduler's delay propagation logic and is used for calculation of TMATs and TOBTs.The purpose of the metering value is to control the amount of excess queue time that the flights are predicted to experience in the AMA.The larger delay buffer causes the flights to spend more time in the queue or AMA before takeoff, and therefore, allows the aircraft to push back earlier from the gate.The gate hold level (as shown in Fig. 3) selected by the ramp manager determines the value of this delay buffer.Flights can be marked as exempted from metering or as a priority flight on RTC or RMTC, and the metering tool treats them accordingly.International and General Aviation (GA) flights may also be marked as exempt from metering.The tactical scheduler regards EOBT as a flight's ready time and uses that to generate gate hold advisories.EOBT is calculated by the airline based on various factors such as percentage of passengers boarded, baggage loaded and more.The tactical surface scheduler allocates runway departure slots on the timeline according to the flight's schedule, with the order of consideration applied based on the quality of the flight's EOBT.The tactical surface scheduler places flights in different groups based on their predictability in runway time prediction, i.e., Uncertain/Planning/Ready/Out/Taxi/Queue in ascending order.The definitions of the groups are shown in Table I.Flights that are further than 10 minutes from their EOBTs or have poor quality EOBTs (i.e., high prediction errors) are marked in the Uncertain group (Fig. 5).The flight is considered to be part of the Planning group when it is 10 minutes from its EOBT.Gate hold advisories will be shown on RTC for the flights in the Planning group.When the pilot calls in ready to push, the ramp controller is expected to swipe the flight strip for pushback or hold according to the advisory shown on the display, and at this point the scheduler marks the flight in the Ready group (see Fig. 5).When the flight is cleared for pushback by the ramp controller, it is considered to be in the Out group, and in the Taxi group when it starts taxiing.Similarly, it is considered in the Queue group when it is waiting in a queue at the runway getting ready for take-off.The RTC shows a hashtag for flights in Uncertain group (Fig. 5) instead of providing a gate hold advisory.This is done to avoid fluctuations in the gate hold advisories due to the uncertainty in flight ready time.However, this does not prevent a pilot from calling in for pushback.When this happens the ramp controller can click the hashtag, and the tactical scheduler instantaneously returns the gate hold advisory and display on RTC.Flights can be moved from the Ready group to the Uncertain group, if the flight called ready to push but did not pushback within five-minutes after its TOBT.The tactical surface metering tool updates every 10 seconds and adjusts the schedule to accommodate uncertainties and changes in the traffic situation. +Airspace and ScenarioThe tactical surface scheduler and level of holds were tested using a simulation of CLT surface operations.The CLT Ramp area is a south facing, large, semicircular area situated between runways 18L and 18C, see Fig. 1.The ramp surrounds five concourses that make up the passenger terminals at CLT, and comprises seven alleys, with 99 gates at which the airplanes park, and two hardstand areas, one on the southwest corner and the other on the northwest corner.CLT ramp control has four sectors marked as West, South, East and North as shown in Fig. 6.In the simulation, the ramp manager was seated between the West and South sector positions in the ramp.The ramp tower was simulated in the high-fidelity Future Flight Central (FFC) tower simulation facility at NASA Ames Research Center.The tower simulator offers a 360-degree field of view provided by twelve projectors giving a realistic moving image to the viewer.Simulation scenarios were designed to have a concentrated mix of traffic with, on average during the hour-long run under clear weather, Instrument Meteorological Conditions (IMC).The North-flow scenario had traffic fed to controllers at a rate of 75 arrivals and 65 departures per hour.In the South-flow scenario, traffic was fed at a rate of 92 arrivals and 80 departures per hour.These rates match the operational rates at CLT airport.The actual number of flights that departed and landed is shown in Table III.Several additional events that CLT Ramp controllers regularly have to work with and that impact the metering tool's schedule, were also built into the scenarios reflecting a number of gate conflicts, two requested changes in taxi route, and an unanticipated delay pushback (e.g., due to maintenance).Several flights were subject to tactical TMIs such as APREQ/CFRs and strategic TMIs such as EDCTs.In this simulation, operations in the airspace surrounding the CLT ATC-T and TRACON were simulated via a mini tower created using eight monitors that provided a 220-deg field of view of the airport where the eye-point corresponded to that of the ATC-T. +ParticipantsFive ramp controllers took part in the simulation, two were retired and three were current American Airlines CLT ramp personnel.Participants' years of experience as active ramp controllers (excluding training) ranged from 1-11 years (M=4.2,SD=4.0).All were generally experienced in aviation, having either worked in an ATC Tower or worked for the airline in other capacities before working as a ramp controller.In the simulation, four of the participants served as ramp controllers and one of them worked as a ramp manager for the duration of the experiment, while the other four participants rotated through the four ramp controller positions.The controller participants were paired with four pseudo pilots who conducted standard pilot pushback and taxi tasks, controlling the aircraft in accordance with controller instructions via simulated radio communication.The Traffic Management Coordinator (TMC) position in the ATC-T was also staffed by active TMCs from CLT.The ATC-T TMC's primarily role was to exercise and evaluate the APREQ process using the tools.But they also evaluated the level of holds that the flights exercised at the gates and how it impacted the queue in the AMA.There were also four ATC Tower controllers that managed the flights in the AMA. +Experiment MatrixThe independent experiment variables used to evaluate and calibrate the tactical surface scheduler/metering tool were runway configuration and level of hold/metering value.The runway configuration used in the simulation scenarios were North and South flow.The second independent variable, level of hold/metering value used in the metering tool, was set as 8, 10 and 12 minutes.As explained earlier, the nominal value suggested by the subject matter experts was eight, so that was used as the starting point for the metering values that were tested.The metering value specified the delay buffer or excess queue time or taxi-out time taken on the airport surface, including both ramp and AMA.The smaller metering value was associated with larger gate holds.For example, the metering value of 8 minutes was associated with 'more gate holds' and 12 minutes was associated with 'less gate holds.'The metering value of 10 minutes was associated with 'nominal hold.'These values were entered to the tactical scheduler via the 'Set Metering Mode' interface of the tool as shown in Fig. 3.One of the purposes of this study was to evaluate the metering value for the tool by getting feedback from both the ramp personnel and the ATC Tower TMCs.Table II shows the experiment matrix and the associated six simulation runs that were exercised in a random fashion.IV.SIMULATION RESULTSBoth objective and subjective data from the HITL simulation were analyzed to evaluate the effect of the metering value in the Metering Tool's delay propagation logic.These were exercised under the two runway configurations -North and South flow. +Objective MeasurementsTable III shows simulation run information for each of six runs, runway configuration, metering value, run duration, number of departures and arrivals.Each run has different simulation run time ranging from 50 to 67 minutes due to limited simulation schedule.As shown in the table, these various run durations resulted in different numbers of departures and arrivals that have completed takeoffs and reached the gates, respectively.During the simulation, the metering tool provided ramp controllers with pushback advisories, which can be either immediate push or n minutes of gate holding.The gate holding times of departures and TOBT compliance of ramp controllers were evaluated by looking at the relationship of the EOBTs, TOBTs and Actual Off-Block Times (AOBTs) of individual departures.In this study, the gate holding time of a departing aircraft is calculated by subtracting the EOBT from the AOBT.The TOBT compliance is measured by the difference between the AOBT and TOBT values.Fig. 7 shows the mean gate holding time and the TOBT compliance for each run, with whiskers representing standard deviations.Fig. 7 shows that there is a noticeable decrease in the gate holding times for South flow runs as the metering value increases from 8 to 12 minutes.This is as expected from the tactical scheduler (i.e., the more gate holding with the lower metering value).For North flow runs, however, it seems that the gate holding time is not Fig. 7. Mean time difference between AOBT and EOBT/TOBT for each run associated with the metering value, nor is it affected by other factors varied in the study.The time differences between AOBTs and TOBTs are within one minute for all runs in Fig. 7.This indicates that the ramp controllers tried to follow the pushback advisories from the tactical scheduler, unless there was either a safety issue or a TMI constraint involved.In fact, the negative mean values result from the EDCT and/or APREQ/CFR flights, which ramp controllers tend to push back earlier than the recommended pushback times so as to meet the given scheduled release times.When these TMI flights are excluded from the analysis, the mean difference between AOBT and TOBT are closer to zero as seen for South 12 (S_12) condition.Taxi-out times in the ramp area and AMA are illustrated in Fig. 8 with whiskers showing the standard deviations of the total taxi-out times.As the metering value increases it was expected that the taxi-out time in the movement area would increase because more flights would wait in the departure queue instead of being held at the gates.In South flow runs, there is slight increase in the mean AMA taxi times (4.4 min for S_8, 5.4 min for S_10, and 5.8 min for S_12) associated with the larger metering value.In North flow runs, on the other hand, the ramp taxi times appear to be similar, but there is no clear trend on the AMA taxi times with the metering value.It seems that the taxi-out times are affected by other factors, such as simulation run time, changes of runway assignment, and TMI constraints.For instance, the total taxi time for the N_10 run is relatively short, and the shorter taxi-out time may be related to the fewer departures that completed takeoffs within the shorter run duration.Since the traffic scenarios represent one bank having a peak, the taxi delay can be further propagated to the flights scheduled in the later time window after the peak.This result depicts that the change in metering value has the potential for distributing the delay differently between the AMA and ramp area, because the metering value impacts the delay taken in the AMA.The stacked bars in Fig. 9 show the ramp and AMA taxi-in times per arriving aircraft for each run.For South flow runs, the ramp taxi-in time decreases as the metering value increases.This can be explained by the interaction between departures and arrivals for gate utilization.More holding at the gate for departures can cause arrivals assigned to the same or adjacent gates to be delayed to avoid gate conflicts, leading to the increased taxi-in time (S_8 in Fig. 9).For North flow runs, however, it seems that the taxi times for arrivals are affected by other factors, such as run duration, runway changes, and interaction with departures.Since the taxi distance from Runway 18R/36L to the main terminal is relatively long, runway changes by a tower controller can impact the average taxi-in times significantly.To assess departure runway throughput performance, the number of flights that take off in a given time period was compared (Fig. 10 and11).There was no significant difference in the runway throughput for the different metering values.This implies that the tactical surface metering can maintain the current runway throughput without any loss of runway usage.Data was collected on the number of departures taxiing in the movement area in order to measure the congestion level on the surface, as shown in Fig. 10 and 11.This metric represents the departure queue size for takeoffs.In the beginning of the bank in the scenarios, the surface counts look very similar, regardless of the metering value.However, as the traffic demand gets close to the peak, the aircraft counts in the AMA vary by the metering value.In both North and South flow operations, more departures are observed in AMA when the metering value is higher.This is a direct result from surface metering with different levels of gate holding.The analysis results on the gate holding and taxi times shown above imply that those metrics are affected by the simulation run time.Fig. 12 plots the run duration with the corresponding mean gate holding and taxi-out times.Although the sample size is very limited, the mean values of taxi out times seem proportional to the run duration for North and South flow runs.Gate Hold times increase with the length of the simulation for North Flow but this trend is not clear in the South Flow.As seen in Figs. 10 and11, the departure demand increases when the simulation time progresses.In the later part of the bank, therefore, the mean taxiout time could increase, and the tactical scheduler would put more holding at the gates to mitigate the surface congestion.Whereas the previous HITL simulation for SARDA [2] had assumed a fixed runway assignment, our simulation allowed tower controllers to change the assigned runway, if needed.This change can also affect the airport performance, such as runway throughput, queue lengths, and taxi times.Table IV shows the number of runway changes between Eastbound (18L/36R) and Westbound (18C/36C) departure runways for each run.For runway balancing and airport efficiency, frequent runway changes were made by the Tower or TRACON TMCs during the simulation runs.In North flow cases, for example, Runway 36R is used for the mixed operations of departures and arrivals.When consecutive arrivals were expected, several departures originally bound for 36R were sent to the Westbound runway (Runway 36C), which is mostly used for departures, but has a longer taxi time.This can explain one reason why the N_10 case shows the shorter taxi-out time on average, compared to other runs.Similarly, the runway changes for arrivals can make some impacts on the taxi-in times as well. +Subjective measurementThe study also collected subjective data such as workload, situational awareness, acceptability of the tools and the advisories.In general, no statistical differences were seen between the North and South flow configurations.The results have been aggregated to focus on the metering value only.Participants provided workload ratings at the end of every run using the NASA Task load index (TLX) on a scale of 1 (low) Fig. 13.Aggregate workload ratings by metering value to 5 (high).Fig. 13 shows the mean ratings for the subscales of workload provided by the ramp controllers and manager based on their perception of their workload at the busiest time in the run.The graph shows similar mean ratings between the different metering values (gate hold time) conditions.The trend shows that the participants perceived slightly higher mental demand, physical demand, and time pressure in the lower metering value (8 min) because it translates to higher level of holds at the gates.The trend also shows that the participants perceived lower demand, better success (reverse scale) and lower frustration with the higher metering value of 12 min associated with the lower gate holds.This data was further substantiated with verbal feedback from the participants who referred to the runs with the higher metering value (12 min) as "normal operations in the field."Participants also provided subjective situational awareness ratings using the Situational Awareness Rating Technique (SART) that uses three subscales: understanding of the situation, demand on attention and attentional resources provided by the displays where 1 is at the lowest end and 5 is at the the highest end of the scale.Fig. 14 shows that similar mean ratings between the different metering values.However, the trend shows a slight improvement in the situational awareness on the two subscales -understanding and attentional resources for the metering value of 12 min associated with the lower gate holds.Attentional demands were similar under all metering value conditions.Participants were asked to rate how often they found the gate hold recommendations "making sense."The results of the data on that question are provided in Fig. 15.The smaller metering value provided longer gate holds and was also seen as "making sense" when compared to the other conditions, even though none of these values were statistically significant.In their verbal feedback, the users mentioned that the gate hold times were "just" right for all the three conditions, most of times and they generally complied with them.They mentioned that they would have changed the gate hold times on one or two flights only in the entire scenario runs.Participants from the ramp and ATC Tower were asked to assess the acceptability of the metering tool by asking the following questions at the end of each run.Ramp controllers were asked -"During your busiest time, how acceptable was the departure demand at the spots?"The Tower participants were asked -"During your busiest time, how acceptable was the flow of departing aircraft onto the AMA (out from the spots)?" Their responses are shown in Fig. 16.There were no statistically significant differences between the diferrent metering values.The graph shows that the Ramp personnel found the metering value of 10 min as slightly more acceptable whereas the ATC Tower participants found the metering value of 12 min (less gate holds) as more acceptable.The participants were asked to assess the acceptability of the departure queue at their busiest time, at end of each run.The metering value directly impacted the departure queue size in the AMA.The data is depicted in Fig. 17, it shows that overall both ramp and tower participants found the queue close to 'just right'.In their verbal feedback, both sets of partcipants mentioned that they preffered the queue size with the metering value of 12 min.In most cases the metering tool's gate hold recommendations did not exceed 10 min.The objective of the study was to evaluate the tactical surface metering tool and evaluate the metering value associated with delay propagation.For the purpose of the evaluation, the metering tool was tested under three different metering values (8, 10 and 12 min) that were associated with the gate hold levels.The higher the metering value, the lower the corresponding gate hold levels recommended by the metering tool on the RTC decision support tool.To establish the best metering value that could be used as a nominal metering value, six runs were conducted in the simulated CLT airport and airspace in both the North and South flow airport configurations.Objective results verified that higher metering values resulted in lower gate hold times.Ramp controllers adhered to the gate hold recommendations as much as possible unless there was a safety issue or TMI constraint to meet.This compliance was shown by both objective and subjective results.They found the gate hold times as 'just right' in all conditions.It was found that there is a potential for distributing the delay differently between the AMA and ramp area, because the metering value impacts the delay taken in the AMA.Clear trends could not be established due to lack of sufficient data points and simulation run duration.The higher metering value had the potential for increasing the taxi time in the AMA, and decreasing the taxi time in the ramp or non-movement area.It was also found that there was no significant difference in the runway throughput when analyzed for metering value.This implies that the tactical surface metering can maintain the current runway throughput without any loss in runway usage.This is an important finding because it means that the longer gate holds only change the distribution of where the excess taxiout time is taken -gate or runway queue, but does not impact runway throughput.Subjective results show that both ramp controllers and ATC-T TMCs favored the higher metering values.In this collaborative effort both the ramp and ATC Tower personnel decided that using a metering value of 12 min as the nominal level of hold would be a good starting point for the ATD-2 IADS systems at the time of deployment in the operational environment.Workload and situational awareness had similar mean ratings across the metering values, but trends did show slight improvement in both metrics as metering value increased.During the verbal debrief sessions, controllers and TMCs provided suggestions for improving RTC and the metering tool.They mentioned losing awareness of the flight's pushback status when the flights transferred to the Uncertain group, where flights were categorized when they did not pushback within five minutes after having been cleared.The authors also found a scheduler design issue that the metering tool always metered flights to all runways even if only one runway experiences demand capacity imbalance.This issue was found when the participants pointed that the tool was metering to a runway that had no demand.The ramp personnel also expressed the need for a predictive tool that could help them decide when to turn the metering on and off, and even play with the different metering values to see the effect of the tool on gate holds.Subject feedback was valuable and helped with further improving the system. +VI. CONCLUSIONSThis study focused on studying the effectiveness of the metering tool during a simulation evaluation, and to establish a metering value that was acceptable to both the FAA ATC Tower and American Airlines Ramp based at CLT.Both groups preferred the higher metering value as the 'nominal hold' in the metering tool.In general, the participants found the metering tool and its advisories acceptable.They provided feedback on improving the metering tool and the decision support tool.It is expected that this tool will undergo another level of calibration when used in the operational environment.Fig. 1 .1Fig. 1.CLT airport plan view North flow operation.However, during North flow operations, Runway 5 is used for both arrivals and departures when North flow night-time noise abatement procedures are in effect. +Situated between the Washington DC metroplex and the Hartsfield-Jackson Atlanta International Airport (ATL), CLT underlies one of the busiest air corridors in the U.S.This location significantly influences operations at CLT because many flights from CLT are destined to constrained airspace and airports on the East Coast.That makes CLT the subject of frequent traffic flow management constraints for managing overhead stream insertion for flights heading to both the Washington metroplex and New York metroplex areas.Various Traffic Management Initiatives (TMIs) are used to regulate air traffic flows for managing imbalances between demand and capacity in the National Airspace System (NAS).TMIs can be divided into strategic and tactical categories, based on the impact level of the constraint and who initiates the restriction.An example of a strategic TMI is where flights are assigned departure times, known as Expect Departure Clearance Times (EDCTs), which in turn regulate their arrival time at the impacted airport.Tactical TMIs are issued by local facility traffic management personnel such as Center, Terminal Radar Approach Control (TRACON), and ATC Tower traffic management coordinators.Tactical TMIs resolve local demand/capacity imbalances in the NAS.Two widely used tactical TMIs are Miles-in-Trail (MIT) and APREQ/CFR restrictions. +Fig. 2 .2Fig. 2. Different states for flight strips and icons on RTC +Fig. 3 .3Fig. 3. Window on RMTC to set time based metering and level of holds +Fig. 4 .4Fig. 4. Metering Tool advisories on RTC +Fig. 5 .5Fig. 5. Different Metering tool groups and associated advisories on RT +Fig. 6 .6Fig. 6.Sectors in CLT Ramp Area +Fig. 8 .8Fig. 8. Average taxi-out time in ramp area and AMA for each run +Fig. 9 .9Fig. 9. Average taxi-in time in ramp area and AMA for each run +Fig. 10 .10Fig. 10.Number of departures taxiing in AMA for North flow runs +Fig. 12 .12Fig. 12. Relations between run duration and gate holding/taxi-out times +Fig. 14 .14Fig. 14.Aggregate Situational Awareness ratings by metering value +Fig. 15 .15Fig. 15.Ratings for did the 'recommended' gate hold times make sense? +Fig. 16 .16Fig. 16.Ratings for Accepatibility of departure demand by ramp and tower participants +Fig. 17 .17Fig. 17.Ratings for Accepatibility of departure queue by ramp and tower participants V. DISCUSSION +TABLE I .IDEFINITIONS OF SCHEDLING GROUPSGroupDefinitionUncertainFlights with poor quality EOBT or EOBT -currenttime > 10 minPlanningFlights within 10 min of EOBT (i.e., EOBT -currenttime <= 10 min)ReadyFlights that have called in ready for pushbackOutFlights that are in pushback stateTaxiFlights that are cleared for taxiQueueFlights waiting in the runway queue +TABLE II . EXPERIMENT MATRIX Metering Value Level of Gate Hold Runway ConfigurationIINorthSouth8 minMoreN_8S_810 minNormalN_10S_1012 minLessN_12S_12 +TABLE IIIIII.SIMULATION RUN INFORMATIONRunRunwayMeteringRunDepartureArrivalnameConfigurationvaluedurationnumbernumber(min)(min)(OFF)(IN)N_8866.34438N_10North flow1050.22726N_121267.25450S_8853.44228S_10South flow1052.44134S_121256.34943 +TABLE IV .IVDEPARTURE RUNWAY CHANGES DURING SIMULATIONRunway changesN_8N_10N_12S_8S_10S_12Eastbound to Westbound949569Westbound to Eastbound003312Total94128711 + + + + +ACKNOWLEDGMENTThe authors thank all the participants in the human-in-theloop simulation, including active ramp controllers and a ramp manager from American Airlines, CLT FAA ATC Tower TMCs, & FLMs, retired tower controllers, and pseudo pilots. + + + + + + + + + Performance Evaluation of SARDA: An Individual Aircraft-based Advisory Concept for Surface Management + + YoonJung + + + TyHoang + + + MiwaHayashi + + + WaqarMalik + + + LeonardTobias + + + GautamGupta + + 10.2514/atcq.22.3.195 + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 22 + 3 + + 2015 + American Institute of Aeronautics and Astronautics (AIAA) + + + Y. Jung, W. Malik, L. Tobias, G. Gupta, T. Hoang, et al., "Performance evaluation of SARDA: an individual aircraft-based advisory concept for surface management," Air Traffic Control Quarterly, Vol. 22, Number 3, 2015, pp. 195-221. + + + + + Call for Papers + + MHayashi + + + THoang + + + YJung + + + MMalik + + + HLee + + 10.1027/2192-0923/a000067 + + + Aviation Psychology and Applied Human Factors + Aviation Psychology and Applied Human Factors + 2192-0923 + 2192-0931 + + 4 + 2 + + June 23-26, 2015 + Hogrefe Publishing Group + Lisbon, Portugal + + + M. Hayashi, T. Hoang, Y. Jung, M. Malik, H. Lee, et al., "Evaluation of pushback decision-support tool concept for Charlotte Douglas International Airport ramp operations," 11th USA/Europe Air Traffic Management R&D Seminar (ATM2015), Lisbon, Portugal, June 23-26, 2015. + + + + + Assessing the impacts of the JFK Ground Management Program + + StevenStroiney + + + BenjaminLevy + + + HarshadKhadilkar + + + HamsaBalakrishnan + + 10.1109/dasc.2013.6712508 + + + 2013 IEEE/AIAA 32nd Digital Avionics Systems Conference (DASC) + Syracuse, NY + + IEEE + October 2013 + + + S. Stroiney, B. Levy, H. Khadilkar, and H. Balakrishnan, "Assessing the impacts of the JFK ground management program," 32nd Digital Avionics Systems Conference (DASC), Syracuse, NY, October 2013. + + + + + + SEngelland + + + ACapps + + + KDay + + + MKistler + + + FGaither + + NASA/TM-2013-216533 + Precision Departure Release Capability (PDRC) final report + + June 2013 + + + S. Engelland, A. Capps, K. Day, M. Kistler, F. Gaither, et al., "Precision Departure Release Capability (PDRC) final report," NASA/TM-2013-216533, June 2013. + + + + + Evaluation of the controller-managed spacing tools, flight-deck Interval management and terminal area metering capabilities for the ATM Technology Demonstration #1 + + JThipphavong + + + JJung + + + HSwenson + + + KWitzberger + + + LMartin + + + + 10th USA/Europe ATM R&D Seminar (ATM2013) + Chicago, Illinois + + June 2013 + + + + J. Thipphavong, J. Jung, H. Swenson, K. Witzberger, L. Martin, et al., "Evaluation of the controller-managed spacing tools, flight-deck Interval management and terminal area metering capabilities for the ATM Technology Demonstration #1," 10th USA/Europe ATM R&D Seminar (ATM2013), Chicago, Illinois, 10-13 June 2013. + + + + + Benefit opportunities for integrated surface and airspace departure scheduling: A study of operations at Charlotte-Douglas International Airport + + RichCoppenbarger + + + YoonJung + + + TomKozon + + + AmirFarrahi + + + WaqarMalik + + + HanbongLee + + + EricChevalley + + + MattKistler + + 10.1109/dasc.2016.7778084 + + + 2016 IEEE/AIAA 35th Digital Avionics Systems Conference (DASC) + Sacramento, California + + IEEE + September 25-29, 2016 + + + R. Coppenbarger, Y. Jung, E. Chevalley, T. Kozon, A. Farrahi, et al., "Benefit opportunities for integrated surface and airspace departure scheduling: a study of operations at Charlotte-Douglas International Airport," 35th Digital Avionics Systems Conference (DASC), Sacramento, California, September 25-29, 2016. + + + + + Procedures for preparation, printing, and distribution of UMTRA Project National Environmental Policy Act documents + 10.2172/6345062 + + + Processes, Procedures, and Policy + esses, edures, and Policy + + Office of Scientific and Technical Information (OSTI) + + + + FAA Surface Operations Office, "Processes, Procedures, and Policy + + + + + Frontmatter + + Initiative + + 10.1515/9783112480144-fm + + + Seattle, Washington, USA, 21 June – 2 July, 1971 + + De Gruyter + July 2014 + + + + Initiative," July Plenary Meeting, July 2014. + + + + + An Integrated Collaborative Decision Making and Tactical Advisory Concept for Airport Surface Operations Management + + GautamGupta + + + WaqarMalik + + + YoonJung + + 10.2514/6.2012-5651 + + + 12th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference and 14th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference + + American Institute of Aeronautics and Astronautics + July, 2013 + + + FAA Air Traffic Organization Surface Operations Directorate + FAA Air Traffic Organization Surface Operations Directorate, "U.S. Airport Surface Collaborative Decision Making Concept of Operations (ConOps) in the Near-Term: Application of the Surface Concept at United States Airports," July, 2013. + + + + + Evaluation of approval request/call for release coordination procedures for Charlotte Douglas international airport + + LindsayStevens + + + LynneMartin + + + EricChevalley + + + HanbongLee + + + KimberlyJobe + + + SavitaVerma + + + BonnyParke + + + VictoriaDulchinos + + 10.1109/dasc.2017.8101983 + + + 2017 IEEE/AIAA 36th Digital Avionics Systems Conference (DASC) + + IEEE + + + + in press + L. Stevens, L. Martin, K. Jobe, B. Parke, E. Chevalley, et al., "Evaluation of Approval Request /Call for Release coordination procedures for Charlotte Douglas International Airport," 36 th Digital Avionics Systems Conference (DASC), in press. + + + + + Impact of General Aviation Operations on Airport Performance Through Fast-Time Simulations at Charlotte-Douglas International Airport + + ZhifanZhu + + + VaishaliAHosagrahara + + + HanbongLee + + + YoonCJung + + + DeborahLBakowski + + 10.2514/6.2020-2916 + + + + AIAA AVIATION 2020 FORUM + + American Institute of Aeronautics and Astronautics + April 2017 + + + Charlotte Douglas International Airport, Fast Facts and Aviation Activity Reports, April 2017. http://www.cltairport.com/News/Pages/FactsandFigures.aspx + + + + + + diff --git a/file758.txt b/file758.txt new file mode 100644 index 0000000000000000000000000000000000000000..cfe631516982cbe3e9c7c674d81939733ce06910 --- /dev/null +++ b/file758.txt @@ -0,0 +1,385 @@ + + + + +I. Nomenclature +II. IntroductionThe National Airspace System continues to grow in complexity and demand resulting in more delays and increased pressure on current resources.Recently, there has been a deluge of new entrants in the NAS that are putting more demands on existing facilities and resources.Most of the research in the United States and European Skies [1] related to the urban areas has focused on the new entrant small Unmanned Aerial Systems (sUAS) flights, their integration with the airspace, and building safe operations in densely populated areas.In the United States, NASA, the FAA, industry, and academia have been actively working toward the development and demonstration of a UAS Traffic Management (UTM) system for sUAS operations in low-altitude airspace [2].Increasing road congestion has led to a growing interest in UAM.The goal of Urban Air Mobility (UAM) is to move people and cargo through metropolitan areas via safe and efficient air traffic operations [3,4].UAM is expected to improve mobility for the general public, decongest road traffic, reduce transport time and reduce the strain on existing public transport networks [5,6].There is a great deal of investment made by industry, government and academia who are researching Urban Air Mobility.Uber Elevate, one of the Industry partners, has published a white paper describing their vision for an air taxi service [4].It includes an analysis of the feasibility of using electric Vertical Takeoff and Landing (eVTOL) vehicles for air taxi services, the economics of the air taxi market, and the ground infrastructure (e.g., vertiports, charging systems) required for air taxi operations.Additionally, it discusses several crucial airspace integration challenges for UAM, such as efficient sequencing and scheduling of eVTOLs into and out of vertiports (similar to heliport with several helipads) and interoperability between vehicles.There exist several other challenges to UAM, such as integration of procedures with existing airspace and the airport, noise levels that are acceptable to the general public, public safety, public acceptance, vehicle certification, and many more..Previous studies on UAM have focused on fast-time simulations of the routes that are separated via a separation service and network of routes [7].Similarly, research done collaboratively by US and Europe has also primarily been conducted using fast-time simulations and has focused on the approach profile for these innovative eVTOL aircraft, vertiports and battery life [8] and more.Other studies, including Market studies, have focused on the demand for UAM and population demographics who can afford air taxi, using surveys to estimate the commuter's willingness to pay for such flights [9].In collaboration with partners and stakeholders over the next few years, NASA will develop detailed concepts of operations for UAM airspace integration at different stages of operational maturity.The evolution of the UAM operational concept is expected to start with low-tempo, low-density flights along with a set of fixed routes referred to as "Emergent operations."These operations will then morph into "Early Expanded UAM operations," characterized by higher-tempo, higher-density flights in a small network of vertiports feeding a common hub location and managed by UAM operators and third-party services.These will be followed by "Mature UAM operations," characterized by high-tempo, high-density flights in a network with multiple hub locations, potentially with orders-of-magnitude more vehicles and operations in an area that are currently supported in the NAS [5].NASA is currently exploring how the UTM paradigm can be applied to UAM as concepts are being developed and evaluated.It is also expected that the following guidelines for UAM operations as part of the On-Demand Mobility [3] effort will need to be considered: 1.Does not require additional ATC infrastructure 2. Does not impose additional workload on ATC 3. Does not restrict operations of traditional airspace users 4. Will meet appropriate safety thresholds and requirements 5. Will prioritize operational scalability 6.Will allow flexibility where possible and structure where necessary Since UAM operations are likely to occur in metropolitan areas that will require UAM flights to access controlled airspace (Class B, C and D) as also described in previous papers [5].It is also likely that near-term UAM operations will be required to comply with current airspace rules and regulations and that they will be very similar to, if not the same as, current Visual Flight Rules (VFR) operations.Furthermore, it is expected that these initial UAM flights will be conducted by human pilots on board the UAM aircraft at all times and will have ATC providing services as in current operations [5].This paper explores potential routes and procedures in a Human-in-the-Loop (HITL) experiment that could be applied in the near-term to allow integration of UAM flights into the airspace, as well as into a large airport.The purpose of this exploration is to help identify operational constraints and capabilities in currentday operations.The rest of the paper is organized as follows: Section III will describe the methodology used to explore these operations with a focus on airspace and experimental details and metrics planned for the study.Section IV will focus on data analysis, Section V will provide a discussion of the results and Section VI will summarize, provide conclusions and next steps. +III. MethodologyThe research focuses on designing and testing routes and procedures for UAM operations that access Class B, C and D airspace.As mentioned earlier, UAM operations will be expected to comply with current airspace rules and regulations.Since the test airspace of the Dallas Fort Worth area does not include a Class C airport, only Class B and D airspace procedures were investigated. +A. Experiment MatrixThe two variables explored in this research study were levels of traffic and routes with different communications.Three different communication procedures including Letter Of Agreements (LOA) were evaluated with three different levels of UAM traffic shown in Table 1 Table 1.Experimental Matrix +Level of UAM trafficThe level of UAM traffic was varied across conditions, where low traffic was characterized by UAM flights temporally separated by 90 s, medium traffic at 60 s, and high traffic was temporally spaced at 45 s.Table 2 shows the equivalent enroute distance spacing of each traffic level assuming that all flights were flying at 130 knots, and the total number of flights simulated within a 40-minutes run to achieve each traffic level.Table 3 shows the number of traditional commercial flights utilizing Dallas Fort Worth (DFW), Dallas Love (DAL) and Addison (ADS) airports that were also included as background traffic in all the runs.This background traffic was kept constant across the different conditions. +Low Traffic +Medium TrafficHigh +Routes and ProceduresThe scenarios C1 to C3 in Table 1 refer to Baseline conditions, which use current-day helicopter routes as shown in Figure 1.The current-day helicopter routes are generally set along the highways for visual reference and have no speed or altitude restrictions.The communications involved in these conditions were the same as current-day, where the helicopter or UAM pilot needs to ask permission to enter Class B airspace from the air traffic controller prior to entrance and provide the full planned route or general intent.The controller is required to assign beacon codes, read back the entire route clearance, and ensure that this UAM traffic stays separated from the traditional traffic.The controller may make traffic calls for wake turbulence or for notice of other traffic to both helicopter or UAM pilot and traditional traffic when required.The scenarios CL1 to CL3 in Table 1 refer to the Current Routes with LOA conditions, which used current-day routes with a newly introduced LOA that was meant to help reduce communication with the UAM flights.LOAs can help reduce verbiage, define the available routes precisely, and separate the routes from each other.The LOA made provisions for the UAM flight to have pre-assigned beacon codes to signatory operators who sign and agree to the terms of the LOA.It also created route codes or names for flights flying the same origin and destination pairs, which the signatories would use when entering Class B airspace.This would help reduce the verbiage involved in providing the full route description or flight intent on entry into Class B and D airspaces.In the CL1 to CL 3 scenarios, the current-day helicopter routes were assigned altitudes and cruise speeds so that they could be strategically separated.The LOA also required pilots/signatories to automatically change frequency when exiting Class B Airspace (CBA).However, frequency changes to different sectors were not automatic and still required the controller to make sector handoffs.Also, where possible, point-outs were included in the LOA.For example, any UAM flights flying Spine road (the route between East and West complexes at DFW) had point-outs between the DFW East and West Towers.The final set of scenarios, M1 to M3 in Table 1, referred to as Modified Routes with LOA, used a modified version of the current helicopter routes.These modified routes (see Figure 2) were designed to avoid approach and departure paths for traditional flights, any common Temporary Flight Restrictions (TFRs), heavily populated areas, and more.These modified routes were also assigned altitudes and cruise speeds as per the LOA.For example, in Figure 2 the red, yellow, green, and blue routes set UAM flights at 400, 500, 600, and 1000 ft AGL, respectively.All other elements of the LOA were also in effect for this condition.The modified routes also included new routes that could be used bidirectionally such as the Central Expressway and I-30.During the simulation, any participant position could deny flights into Class B airspace due to workload concerns.In the operational world, these flights would be worked into the sectors, but in the simulation environment they were removed by the pilot on controllers' command based on their demands and an input regarding the reason for the removal was also made into the system. +B. Test AirspaceThe test airspace that was simulated for this research was managed by the DFW East Tower in South flow only.UAM traffic using helicopter routes flew 500 AGL, or 1100 MSL in one direction and 1000 AGL, or 1600 MSL in the other direction.In addition, DAL and ADS airspace were also simulated.DAL is part of Class B airspace, where ADS represents Class D airspace.The initial set of routes investigated in this study were published helicopter routes in the DFW area.Figure 1 shows Class B airspace in the DFW area and the origin/destination city pairs (depicted as red arrows in Figure 1) where UAM flights flew along with helicopter routes shown in blue.For the purpose of the test, helicopter routes spanning the east complex of DFW airport, DAL and ADS were evaluated.The simulation was conducted at NASA Ames Research Center's Air Traffic Control (ATC) laboratory.This laboratory uses Multi Aircraft Control System (MACS) as its primary software that emulates Standard Terminal Automation Replacement System (STARS) and is used as a rapid prototyping tool for various studies.For this study, the laboratory was configured to represent DFW Tower, DAL Tower and ADS Tower via MACS displays.All the traffic outside the towers were handled by "ghost" confederate positions.The pseudo-pilots and confederates who flew these flights also used MACS in a separate area of the laboratory and communicated over the radio. +C. Equipment and Facilities +D. ParticipantsParticipant positions for this research effort included DFW Tower Local East-3, or DFW LE-3, DAL Helicopter, or DAL Helo position, and ADS Tower controller.Sectors surrounding the positions included DFW Local East-1, DFW Local West, DAL Tower position, and D10 TRACON.These positions were simulated as "ghost" confederate positions.There were a total number of six participants, four were recently retired from the DFW area and two others who were retired from Northern California Terminal Radar Control (TRACON) who staffed the ADS position.All participants rotated between the confederate and participant positions at the start of each new trial.The D10 confederate position was staffed by an in-house controller.All participants completed all experimental conditions in both the confederate and participant positions. +E. Data CollectionData were collected over five days.The simulations were conducted in blocks starting with Baseline (Current Routes) condition, followed by Current Routes with LOA and Modified Routes with LOA.Each condition had three levels of traffic that were repeated, so every condition ended with six runs and was conducted as a block of runs.Every run lasted about 40 min.Prior to each block condition, the participants spent two to three hours training on the next condition.During the study, MACS' built-in data collection system captured all relevant data, including all aircraft states, trajectories, automation and operator events.All displays were recorded by a commercial screen capture product.At the end of each run, participants responded to a questionnaire that included questions on workload, situational awareness, and acceptability of routes and procedures.At the end of each block of runs and the whole simulation, the participants responded to surveys that focused on the specific condition used for the six-run block and all the conditions at the end of the simulation; respectively.All questionnaires (e.g., post-run, post-block and postsimulation) were administered electronically at the participant position. +IV. ResultsThe results presented in this paper include desirable trends of UAM aircraft count metrics by each controller position (DFW LE-3, DAL Helo, and ADS), mean lateral separation between UAM flights by position, controller workload, and qualitative feedback on routes, procedures and communications.Descriptive statistics are described in this section, no inferential statistics were conducted due to small number of participants.For detailed descriptions of Workload and Situational awareness and other subjective feedback, the reader is referred to another paper [12]. +A. UAM Aircraft CountUAM Aircraft count is described using three metrics, the average UAM aircraft count, average number of UAM flights that were removed by the controller when they denied access to Class B airspace, and the percentage of flights that were managed by the controller as compared to what was planned for the sector for the simulation run, this number was reduced due to the removal of flights.The amount of traditional traffic managed by all the positions was the same across all the runs as shown in Table 3.The metric related to average aircraft count refers to average number of UAM flights managed at any given time by a position.The average numbers of UAM flights every second were calculated and averaged over 5 min intervals as well as one average across the entire 40 min run. +DFW Local East 3 ControllerFigure 3 shows the average total UAM aircraft that were managed by the DFW Local East-3 position at any given time for the three conditions.The DFW LE-3 position is primarily responsible for arrivals into Runway 17L.In addition, they managed UAM flights arriving from Frisco to DFW, and also managed departures from DFW into DAL as well as departures to ADS.The total number of UAM flights managed by this position, in the low traffic condition was approximately seven aircraft across the three conditions shown on x-axis.In the medium level traffic, the total average increased from Baseline condition (M=9.6,SD=3.1) to Current Routes with LOA (M=10.8,SD =3.5) and stayed about the same for the Modified Routes (M=10.4,SD= 3.1) relative to Current routes with LOA condition.In the high traffic levels, average number of flights increased from Baseline condition (M = 10.6,SD = 2.8) to Current Routes with LOA condition (M = 15.0,SD = 4.8) and stayed about the same for the Modified Route condition (M = 15.4,SD = 4.4) relative to Current routes with LOA condition.The following trends were observed: the number of UAM aircraft managed by the DFW LE3 position increased as traffic levels increased from low to high.There was an increase in number of UAM flights that were managed due to modification of routes and procedures.It was also observed that LOA seems to have a bigger impact on the number of UAM flights that were managed, than modification of routes.For example, in the high traffic condition, number of UAM flights managed changed from 10 in Baseline condition to 15 in Current Routes with LOA, but stayed about the same for Modified Routes with LOA condition relative to Current Routes with LOA condition.with LOA under High traffic condition.This does indicate that the modification of routes and procedures led to the removal of fewer flights and this had an impact on the percentage of UAM flights that the controller managed during the 40 minute long run as shown in Figure 4.This percentage was lowest for the Baseline condition.For example in the Baseline and High traffic condition, approximately 67.5% of flights were managed.However, in the Current with LOA and Modified conditions with high traffic, the controllers managed almost 100% of the UAM flights planned for these runs.As mentioned earlier, the LOA was intended to reduce communication load on the controller and allow handling of more UAM flights than the Baseline condition without LOA, which uses current day communications due to absence of LOA. +DAL Helicopter (Helo) PositionA similar analysis of the average number of UAM aircraft managed by the DAL Helo position controller, the average number of UAM flights removed, and percentage of flights managed are described in this section.DAL Helo position was responsible for helicopter traffic in and around DAL and Dallas Downtown, all of which were positively controlled since they fall in Class B airspace.In this HITL study, UAM flights into and out of the DAL airport, as well as into and out of Downtown Dallas were managed by DAL Helo position.There are several helicopter routes that merge at Downtown Dallas (see Figure 2).In the conditions where LOA was assumed, the area around Downtown Dallas was created as a non-movement area, and the UAM flights contacted DAL Helo position when they reached one of the transition points defined in the non-movement area.The rationale for creating the non-movement area was to allow UAM flights to move between vertiports inside the downtown area without getting approval from the tower controller, thus reducing communication workload on the controller.Figure 5 shows box and whisker plots for the average number of UAM flights that were handled by this position at any given time.We see a similar trend for DFW LE-3 in Figure 6, where in general, there is an increase in UAM traffic managed by the DAL Helo controllers as traffic level increases, except in the Baseline condition.We also see a general increase in the UAM traffic managed across conditions.For example, under the high traffic condition, there is a trend for increase in the average UAM traffic managed across conditions -Baseline condition (M = 6.0,SD = 1.9),Current routes with LOA (M = 9.1, SD = 3.5), Modified Routes with LOA (M = 11.5, SD = 3.1).The DAL Helo position was able manage the highest level of traffic, approximately 15 UAM flights, under Modified Routes with LOA.The highest number of UAM flights that were removed from DAL Helo was in the Baseline condition.As many as 44 flights were removed from the DAL Helo's sector in the Baseline under High traffic condition, which decreased to about 34 flights in the Current Routes with LOA under High traffic condition and further decreased to 12 in the Modified Routes High traffic condition.Figure 6 illustrates the percentage of flights that were managed by the DAL Helo position.This position managed fewer flights in the Baseline condition, especially under high traffic (42%), which improved in the Current Routes with LOA condition under high traffic (65%).Almost 90% of the UAM traffic was managed under the Modified Routes with LOA conditions as planned, irrespective of the UAM traffic levels.The big improvement in percentage of traffic managed under the Modified Routes conditions was possible due to the elimination of some unusable, original helicopter routes such as Tollway and I-35 that created conflicts with approaches into DAL and departures from both DAL and ADS. +Addison Tower PositionA similar analysis of average number of UAM aircraft managed by the controller, the average number of UAM flights removed, and percentage of flights managed by the ADS Tower position are described in this section.ADS Tower is responsible for traffic into ADS airport, which is Class D airspace that mostly manages flights flying under Visual Flight Rules (VFR).In this airspace, it is necessary for the controller to establish a two way communication with the UAM flight or any other flight, but radar identification is not required.Figure 7 shows that on an average this position managed about three UAM flights in addition to their background traffic.The number of UAM flights managed between the Baseline (M = 2.6, SD = 1.8) and Current routes with LOA high traffic level conditions stayed about the same, whereas the Modified Routes with LOA featuring high traffic level saw the average traffic managed by the controller relatively decrease (M = 1.5, SD = 0.6).We see this decrease in traffic because the modification of the routes removed some routes going through ADS.For example, the Tollway was removed as the controllers referred to it as unusable due its close proximity to DAL arrivals and departures as well as ADS departures.It was observed that flights were mostly removed due to unusable routes in Baseline condition, or roughly four flights in the high traffic level, and in the Current Routes with LOA about six flights were unable to use the routes in the high traffic level, whereas no flights were removed in the Modified Routes with LOA conditions irrespective of traffic level.This had an impact on the percentage of flights managed by the controller as compared to flights planned through their sector as shown in Figure 8.The least amount of traffic managed by the controller was 75% for Current Routes with LOA high traffic level condition.The highest amount of traffic managed by ADS tower was 100% traffic planned in all the Modified Routes with LOA conditions, which was less than the other conditions due to the modifications of the routes. +B. Actual Separation between UAM flightsThe actual mean lateral separation between UAM flights that were on the same route and in the same direction was also measured.There is no standard separation requirement for helicopter or UAM flights.For simplicity, only one route was chosen for each position and was investigated across the different conditions.Also, there was not enough traffic in the ADS Tower's sector on any given route that could be compared across conditions.Thus, the two positions analyzed in this section are DFW LE-3 and DAL Helo positions.The route from Frisco to DFW airport is analyzed for the DFW LE-3 controller and the route from Downtown Dallas to McKinney is analyzed for the DAL Helo position.Box and whisker plots as shown in Figure 9 and 10 show the mean actual lateral separation between the flights, the maximum and minimum lateral separation scores, as well as their variability, shown by the whiskers.The dots/ circles represent individual scores. +DFW LE-3 controllerFigure 9 shows the average lateral separation between flights on the route from Frisco to DFW, which was managed by DFW LE-3 controllers.Average separation between flights on this route was high between flights in the Baseline with low traffic condition (M = 5.1 nm, SD = 2.3 nm), and was about the same in the low traffic scenarios for Current Routes with LOA condition (M = 4.6 nm, SD = 0.7 nm) and Modified Routes with LOA condition (M = 4.5 nm, SD = 0.6 nm).With medium traffic levels, the separation was approximately the same across the three conditions -Baseline condition (M = 3.3 nm, SD = 0.6 nm), Current routes with LOA (M = 3.0 nm, SD = 0.7 nm) and Modified Routes with LOA (M = 3.3 nm, SD = 0.6 nm).However, the range of the scores depicted by the size of the box shows that the range of lateral separation on the routes decreases for the Current Routes with LOA condition and stays the same between Current Routes with LOA and Modified Routes with LOA runs.In the heavy traffic condition, the separation maintained by the flights on this route was about two nautical miles across conditions.It was observed that the DFW LE-3 controller tried to handle more UAM traffic in the high traffic conditions than they could provide service to especially in the Baseline condition by not following the full communication protocols.This in turn would explain their smaller spacing between flights in the High Traffic condition. +DAL-Helo positionFigure 10 shows the actual mean lateral separation between flights on the Downtown to McKinney route in the DAL Helo's sector.In the Low traffic condition, we see that the flights are able to fly closer and their spacing is reduced from the Baseline (M = 7.4 nm, SD = 4.3 nm) to Current Routes with LOA (M = 5.9 nm, SD = 1.4 nm) and again increases in the Modified Routes with LOA condition (M = 7.4 nm, SD = 4.6 nm) relative to Current Routes with LOA.Under medium traffic, the three conditions have about the same average separation between UAM flights but their variability decreases from the Baseline, (M = 4.4 nm, SD = 1.4 nm), to the Current Routes with LOA (M = 4.6 nm, SD = 0.9 nm) and again increases for the Modified routes with LOA conditions (M = 4.9 nm, SD = 1.8 nm) relative to Baseline condition.Under the High Traffic condition, the mean and variation of separation between the UAM flights are: Baseline (M = 4.0 nm, SD = 1.4 nm), Current routes with LOA (M = 4.17 nm, SD = 1.1 nm), and Modified Routes with LOA condition (M = 3.7 nm, SD = 0.6 nm).It is seen that there is a trend for similar separation across the three conditions in the High Traffic runs.The trend also shows a slight decrease in variability of the separation, which means that there were no big gaps between most flights on that route especially under the Modified Routes with LOA conditions. +Self-Reported WorkloadWorkload was measured using several different tools and scales such as the Workload Assessment Keyboard (WAK) and NASA Task Load Index (TLX).These workload metrics have been discussed in a different paper [12].This section discusses the responses that different controller positions provided at the end of the block of runs pertaining to a condition.They responded to the statement: "My workload level negatively impacted my performance."A score of 1 indicated "strongly disagree" and a score of 7 referred to "strongly agree" on a scale of 1 to 7. The pattern of results is depicted in Figure 11 for different conditions as experienced by all controller positions.There is a trend seen in the responses here, such that the workload perceived under the Baseline condition (M = 4.7, SD = 1.5) was reported as negatively affecting performance (higher workload) when compared to the other conditions Current Routes with LOA (M = 3.8, SD = 1.7) and Modified Routes with LOA (M = 3, SD = 1.7).The Modified Routes with LOA helped the controllers keep their workload under control without severely affecting their performance.Similarly, when asked about whether their workload was operationally acceptable (see Figure 12), participants rated it at about the same level for Baseline condition (M = 4.3, SD = 2.2) and Current Routes with LOA condition (M = 4.3, SD = 1.9), whereas modified routes with LOA (M = 5, SD = 1.7) were considered as more operationally acceptable than other conditions, mostly since there was improved task efficiency and reduced communications.However, Figure 13 shows that for the same question the responses varied by the controller position.The DFW LE-3 position controllers that managed arrivals on runway 17L did not perceive their workload to be operationally acceptable, whereas the controllers positions at smaller airports (DAL and the Class D airport ADS) perceived the workload to be relatively more acceptable possibly due to less volume of operations they handle, compared to what DFW handles for traditional traffic. +CommunicationsFigure 14 illustrates the controller response to the question whether communications were manageable and was asked after a block of runs pertaining to a condition.The participants provided a rating on the Likert scale where a score of 1 was equivalent to "not manageable" and 5 referred to "manageable."The scores were averaged across the three positions.As shown in Figure 14, the controllers reported that communications were at slightly above medium levels of manageability across the conditions.This is an interesting subjective result, when analyzed along with the average number of flights managed by the controllers (previously seen in Figures 3, 5, and7).We see there was a trend where average aircraft increased as we moved from Baseline to the Modified Routes with LOA conditions when compared to Baseline condition, except for ADS Tower controller.In general, even though the controllers managed higher numbers of UAM flights, they reported the same level of communications manageability across conditions.This can be interpreted such that LOAs helped to keep communications at a relatively similar level while increasing the number of flights that were managed under the Modified Routes with LOA condition.The subjective feedback provided about communications manageability was supported by an additional metric computed from actual audio transmissions, percentage of time controllers spent with listening to, responding to, and initiating communications with pilots over the radio, as shown in Figure 15.The controllers on average spent almost 55% of their time communicating in Baseline condition.We also see a trend where the introduction of the LOA reduces communications in both the Current with LOA conditions (43%) and Modified with LOA conditions at 46% relative to Baseline condition the time spent communicating, possibly due to increased UAM traffic managed under these conditions. +Routes and ProceduresParticipants were asked to respond to several subjective questions regarding routes and procedures.Some questions were asked during post-run surveys and others were asked during post-block surveys.Figure 16 shows responses to a question on how operationally acceptable the routes were, which was asked after every run.The figure shows that there is a trend for improved acceptability towards Modified Routes with LOA and least acceptability for the Baseline condition, where current day routes and communication procedures were used by the controllers for managing UAM flights.The reported that the current helicopter routes were not ideal for high volume UAM operations.Figure 17 shows average responses towards the question regarding whether the routes used during the run require improvement or not.A score of 7 indicated necessary improvement, whereas a score of 1 meant that no improvement was needed.As expected, the controllers reported that lesser improvement was needed for Modified routes with LOA, whereas more improvement was required for the current routes both in the Baseline and the Current Routes with LOA conditions.The current helicopter routes that were used for UAM operations were analyzed by Subject Matter Experts (SMEs) and had several issues, such as proximity to approach and departure paths for traditional traffic that required controller attention and affected their workload.Figure 18 shows whether controllers perceived the UAM routes as increasing task complexity, where a score of 1 indicated that they did not increase task complexity and a score of 7 indicated significant task complexity.There is a trend observed in the data shown in Figure 18, where there is a decrease in task complexity offered for the Modified Routes with LOA condition.Roughly similar levels of task complexity occurred for the conditions that utilized current helicopter routes (the Baseline and the Current Routes with LOA conditions).Most controllers reported that Modified Routes with LOA improved efficiency because the tasks were somewhat simplified with the help of modifications to the original helicopter routes and elimination of some communications via the LOA.Even though the LOA helped the condition, the Current Routes with LOA condition was not able to remove the complexity of routes due to their proximity to traditional traffic and intersecting helicopter routes that required separation, thus increasing communication load on the controllers.The controllers reported their scores on this question during post-block surveys, where a score of 1 referred to low efficiency and 7 referred to high efficiency.The scores were averaged across all the controllers and level of traffic.In Figure 19, we see a trend across the conditions where the participants rated that an improvement of efficiency provided by routes and procedures from the Baseline conditions, to Current Routes with LOA conditions to the Modified Routes with LOA conditions.The controllers reported that the LOA played a big role in separating the traffic on current routes and reducing communication load.The Modified routes with LOA conditions further reduced communication by eliminating the need for traffic calls, since the UAM routes were separated from approach and departures paths.The results on the efficiency provided by the routes is further explored for the different controllers as shown in Figure 20.The DFW LE-3 controller, whose primary responsibility is to manage 17L arrivals, reported that the UAM routes distract from the primary task.Since the UAM routes were in the South of the airport, this addition increased the field of view or scanning area, thus making it hard to pay attention to all the routes.The DAL Helo position participants did comment that the UAM routes were more usable after the modifications and improved efficiency.For example, using Central Highway as a route allowed for greater efficiency than using Tollway, which always required re-routing of UAM traffic to avoid traditional traffic.ADS Local found the routes most usable and efficient.This is most likely because most of their traffic in Class D airspace is VFR traffic and it did not impact the rest of their traffic except for the Tollway route. +V. DiscussionThe experiment reported in this paper investigated routes and procedures for UAM near-term operations, with a focus on investigating operational capabilities and limitations.For the purpose of this research, we started with current airspace constructs such as helicopter routes and current communication procedures for entering Class B and D airspaces.There were two other conditions investigated.The second condition included the current helicopter routes, but reduced communication via a Letter of Agreement (LOA).The third condition was to use modified helicopter routes with LOA that aimed to separate UAM traffic from traditional commercial traffic.It was hypothesized that the modification of the routes and the introduction of the LOA will reduce workload and communications' load.Controllers reported that the LOAs, which assigned route names, pre-assigned beacon codes, and called out pointouts, simplified entry and exit communications procedures for Class B and Class D airspaces, and allowed them to manage more UAM traffic in addition to their primary responsibility of managing traditional traffic.Controllers also mentioned that the introduction of the LOA affected the number of UAM flights that they could manage, which were more flights managed than what the modification of the routes helped achieve.The number of UAM flights that the controller could manage improved from the conditions that used the current routes (e.g., the Baseline and Current Routes with LOA conditions) to the Modified Routes with LOA.A similar result was seen for the percentage of flights managed of those planned for that sector.The Modified Routes with LOA allowed DAL Helo to manage almost 90% of the UAM flights that were planned, versus about 70% managed using Current Routes with LOA condition.This notable improvement was seen due to the fact that there were many unusable routes that were removed in the Modified Route conditions.The modifications to the routes and procedures allowed DFW LE-3 and ADS to manage all UAM flights to full capacity as planned.The reason DAL Helo position could not handle as many flights using the current helicopter routes was due the fact that there were a couple of routes that were unusable due to their proximity to approach and departure paths of traditional traffic.The separation between flights also decreased between the Baseline conditions and the conditions with LOA.Some effects of social desirability were also seen where the DFW LE-3 controllers attempted to manage more flights in the high traffic scenario than they could especially under the Baseline condition, where they did not follow full communication protocols to the UAM flights as expected.Controllers reported that workload did not negatively impact their performance under the Modified Routes with LOA condition and found the workload to be operationally acceptable.The controllers at the large airport, or DFW LE-3, did not report that their workload was operationally manageable.During debriefs, they mentioned that the UAM flights and their routes were so widespread, that it took their attention away from their primary task of managing arrivals to the runways.They also mentioned that this task of managing additional UAM flights would be difficult given their current workload.Feedback regarding communications showed that controllers spent almost 50% of their time communicating with UAM flights and traditional commercial traffic in the Baseline condition.The amount of time spent on communication was reduced due to the introduction of the LOA, which was possibly the reason why the controllers mentioned that the communications were manageable across all the conditions.The LOA is a great tool available for reducing communications, but cannot completely eliminate communication load for controllers.Scalability of UAM operations will be hard to achieve if operations are limited to voice communications between the controller and the flight deck.The participants of the study found the modifications to the current routes and introduction of LOA reduce task complexity and improved efficiency.The modified routes were better separated from traditional traffic and help reduce workload.Thus, the modified routes did outperform the current routes on the various metrics and were considered by different stakeholders to be the direction that the UAM routes could take in the future. +VI. ConclusionIn this research, we considered the guidelines provided for UAM operations [3] to the extent possible.In order to avoid the requirement for additional ATC infrastructure, this research started with current helicopter routes and LOAs that have precedence in the National Airspace (NAS).Also, the authors did not pose any restrictions on traditional airspace users, the modifications were suggested for helicopter routes and not other commercial traffic's approach or departure paths.The research was planned to explore scalability and workload impacts when designing the routes and LOA.However, more considerations may need to be given to safety thresholds and allowing flexibility in the operations.It was found that current day helicopter routes and communication procedures can support near-term UAM operations, but may not be scalable.Controller workload managing traffic at large airports like DFW is already high enough that we need to re-think roles and responsibilities of air traffic controllers, operators of UAM flights, and the flight deck.In the study utilizing the letter of agreement did accomplish its purpose of reducing communications for UAM flights, but use of voice communications still remains a big factor in controller workload as well as scalability of UAM operations.Future work will explore how UTM paradigm used for managing sUAS will work for UAM operations.The service oriented architecture offered by the UTM paradigm has the potential to be very lucrative.However, more investigation is required for using this paradigm for UAM operations that are likely to occur above 500 ft AGL and pose a greater risk than smaller drones that do not carry passengers.Control DAL = Dallas Love Field Airport DFW = Dallas Fort Worth International Airport eVTOL = electric Vertical Takeoff & Landing HITL = Human-In-The-Loop IFR = +Figure 2 .2Figure 2. Modified Routes in DFW Area (red routes are at 400ft, yellow routes at 500ft, green at 600ft and Blue routes are at 1000 ft AGL) +Figure 1 .1Figure 1.Class B Airpsace in DFW as shown by black boundary, the red arrows show origin destination points and blue routes are helicopter routes +Figure 3 .Figure 4 .34Figure 3. Average number of UAM managed DFW LE-3 controllers at any given time, the circles show the data points +Figure 5 .Figure 6 .56Figure 5. Average number of UAM flights managed by DAL Helicopter controllers at any given time, the circles show the data points +Figure 7 .Figure 8 .78Figure 7. Average number of UAM flights managed at any time by ADS Tower position +Figure 9 .Figure 10 .910Figure 9. Actual average lateral UAM separation on the route Frisco to DFW managed by DFW LE-3 position, where the dots show data points +Figure 11 .Figure 12 .1112Figure 11.Average responses to "My workload level negatively impacted my performance" (1 = Strongly Disagree and 7 = Strongly Agree) by condition +Figure 14 .14Figure 14.Average responses to "My communications were manageable" (1 = Not Manageable, 5 = Manageable) by condition +Figure 16 .Figure 17 .1617Figure 16.Average responses to "The routes used operationally acceptable" (1 = Not Acceptable, 7 = Acceptable) by condition +Figure 18 .18Figure 18.Average responses to "Did UAM routes increase task complexity?" (1= Not Complex, 7 = Complex) by condition +Figure 19 .19Figure 19.Average responses to "Did UAM routes support efficiency?" (1 = Low, 7 = High) by condition + +Table 2 Definition of UAM Traffic Level Helicopter Routes UAM Traffic Level2Traffic +Table 3 . Traditional Background traffic at the different airports simulated over the 40 min run3 + + + + +AcknowledgmentsThe authors would like to offer special acknowledgment to our Subject Matter Experts, Dan Woods, Wayne Bridges, Fred Peet and Dean Krause who provided us insights into the operations at DFW and helicopter operations.We are also grateful to Dan Woods for helping us design modified routes for this study and create all the UAM traffic scenarios. + + + + + + + + + Density based Management Concept for Urban Air Traffic + + DagiGeister + + + BerndKorn + + 10.1109/dasc.2018.8569491 + + + 2018 IEEE/AIAA 37th Digital Avionics Systems Conference (DASC) + London + + IEEE + 2018. Sept 2018 + + + Geister, G. & Korn, B. (2018). Density based management Concept for Urban Air Traffic. Digital Avionics System Conference, London, Sept 2018 + + + + + UAS Traffic Management (UTM) Concept of Operations to Safely Enable Low Altitude Flight Operations + + PKopardekar + + + JRios + + + TPrevot + + + MJohnson + + + JJung + + + JERobinson + + 10.2514/6.2016-3292 + + + 16th AIAA Aviation Technology, Integration, and Operations Conference, AIAA AVIATION Forum + + AIAA + 2016-3292, Jun. 2016 + + + Kopardekar, P., Rios, J., Prevot, T., Johnson, M., Jung, J., and Robinson, J. E., "UAS Traffic Management (UTM) Concept of Operations to Safely Enable Low Altitude Flight Operations," 16th AIAA Aviation Technology, Integration, and Operations Conference, AIAA AVIATION Forum, AIAA 2016-3292, Jun. 2016. doi: 10.2514/6.2016-3292 + + + + + Enabling Airspace Integration for High-Density On-Demand Mobility Operations + + EricRMueller + + + ParmialHKopardekar + + + KennethHGoodrich + + 10.2514/6.2017-3086 + AIAA 2017-3086 + + + 17th AIAA Aviation Technology, Integration, and Operations Conference + Denver, Colorado + + American Institute of Aeronautics and Astronautics + 2017. June 2017 + 5 + + + Mueller, E.; Kopardekar, P. & Goodrich, K. (2017). Enabling Airspace Integration for High-Density On-Demand Mobility Operations. 17th AIAA Aviation Technology, Integration, and Operations Conference. 5-9 June 2017, Denver, Colorado. AIAA 2017-3086 + + + + + Fast-Forwarding to a Future of On-Demand Urban Air Transportation + + UberElevate + + 10.3030/774086 + + + 2017 + Publications Office of the European Union + + + Uber Elevate, "Fast-Forwarding to a Future of On-Demand Urban Air Transportation," 2017, https://www.uber.com/elevate.pdf. + + + + + Urban Air Mobility Airspace Integration Concepts and Considerations + + DavidPThipphavong + + + RafaelApaza + + + BryanBarmore + + + VernolBattiste + + + BarbaraBurian + + + QuangDao + + + MichaelFeary + + + SusieGo + + + KennethHGoodrich + + + JeffreyHomola + + + HusniRIdris + + + ParimalHKopardekar + + + JoelBLachter + + + NatashaANeogi + + + HokKwanNg + + + RosaMOseguera-Lohr + + + MichaelDPatterson + + + SavitaAVerma + + 10.2514/6.2018-3676 + + + 2018 Aviation Technology, Integration, and Operations Conference + Atlanta, Georgia + + American Institute of Aeronautics and Astronautics + 2018. 25 th -29 th June 2018 + + + Thipphavong, D.P. et.al. (2018) Urban Air Mobility Airspace Integration Concepts and Considerations. 18th AIAA Aviation Technology, Integration, and Operations Conference. 25 th -29 th June 2018, Atlanta, Georgia. + + + + + Simulation Evaluations of an Autonomous Urban Air Mobility Network Management and Separation Service + + ChristabelleBosson + + + ToddALauderdale + + 10.2514/6.2018-3365 + + + 2018 Aviation Technology, Integration, and Operations Conference + Atlanta, Georgia + + American Institute of Aeronautics and Astronautics + 2018. 25 th -29 th June 2018 + + + Bosson, C. & Lauderdale, T. (2018). Simulation Evaluations of an Autonomous Urban Air Mobility Network Management and Separation Service. 18th AIAA Aviation Technology, Integration, and Operations Conference. 25 th -29 th June 2018, Atlanta, Georgia. + + + + + eVTOL Arrival Sequencing and Scheduling for On-Demand Urban Air Mobility + + ImkeCKleinbekman + + + MihaelaAMitici + + + PengWei + + 10.1109/dasc.2018.8569645 + + + 2018 IEEE/AIAA 37th Digital Avionics Systems Conference (DASC) + London + + IEEE + 2018. Sept 2018 + + + Kleinbekman, I.C.; Mitici, M. A. & Wei, P. (2018). eVTOL Arrival Sequencing and Scheduling for On-Demand Urban Air Mobility. Digital Avionics System Conference, London, Sept 2018. + + + + + If you Fly it + + LAGarrow + + + BGerman + + + PMokhtarian + + + MDaskilewicz + + + THDouthat + + + RBinder + + AIAA Paper 2018-2882 + + + Will Commuters Come? A Survey to Model Demand for eVTOL URBAN Air Trips + + 2018. June 2018 + + + Garrow, L. A.; German, B.; Mokhtarian, P.; Daskilewicz, M.; Douthat, T.H. & Binder, R. (2018) If you Fly it, Will Commuters Come? A Survey to Model Demand for eVTOL URBAN Air Trips. AIAA Paper 2018-2882, June 2018 + + + + + NASA UAS Traffic Management National Campaign-Operations Across Six UAS Test Sites + + JRios + + + DMulfinger + + + JHomola + + + PVenkatesan + + 10.1109/DASC.2016.7778080[22 + + 2016, Sep. 2016 + + + Rios, J., Mulfinger, D., Homola, J., Venkatesan, P., "NASA UAS Traffic Management National Campaign-Operations Across Six UAS Test Sites," IEEE-DASC 2016, Sep. 2016. doi: 10.1109/DASC.2016.7778080 [22] + + + + + Unmanned Aircraft Systems (UAS) Traffic Management (UTM) National Campaign II + + ArwaSAweiss + + + BrandonDOwens + + + JosephRios + + + JeffreyRHomola + + + ChristophPMohlenbrink + + 10.2514/6.2018-1727 + + + 2018 AIAA Information Systems-AIAA Infotech @ Aerospace + + American Institute of Aeronautics and Astronautics + 2018. 2018-1727, Jan. 2018 + 23 + + + Aweiss, A. S., Owens, B. D., Rios, J. L., Homola, J., Mohlenbrink, C. P., "Unmanned Aircraft Systems (UAS) Traffic Management (UTM) National Campaign II", 2018 AIAA Information Systems-AIAA Infotech @ Aerospace, AIAA SciTech Forum, AIAA 2018-1727, Jan. 2018. doi: 10.2514/6.2018-1727 [23] Federal Highway Administration, + + + + + Exploring human factors issues for urban air mobility operations + + TamsynEEdwards + + + SavitaVerma + + + JillianKeeler + + 10.2514/6.2019-3629 + + + AIAA Aviation 2019 Forum + Texas USA + + American Institute of Aeronautics and Astronautics + June 17-21 + + + AIAA: Aviation + Edwards, T., Verma, S., Keeler, J. Exploring human factors issues for urban air mobility operations (In Press). AIAA: Aviation, June 17-21, Texas USA. + + + + + + diff --git a/file759.txt b/file759.txt new file mode 100644 index 0000000000000000000000000000000000000000..ce83f3fc0bea318b63a0ee7b8efe1422248d9251 --- /dev/null +++ b/file759.txt @@ -0,0 +1,324 @@ + + + + +INTRODUCTIONOperations on closely spaced parallel runways have been prevalent in the National Airspace (NAS) for about 40 years.There are several concepts in development and in operational use that define procedures for operations on parallel runways.One concept under development has been Airborne Information for Lateral Spacing (AILS) [Abbott et al., 2001].Simultaneous Offset Instrument Approach (SOIA) [Magyratis et al., 2001] is currently used at airports like San Francisco International airport.Both concepts support arrivals on runways that are only 750 ft apart and assume that air traffic control will pair the appropriate aircraft for simultaneous landings.However, no tool or formal process exists to facilitate the pairing of aircraft.This paper will evaluate the role of the controller for pairing aircraft under different levels of automation using another pairing concept.The levels of automation define how much functionality the tool and interface provide to facilitate pairing aircraft for simultaneous approaches on parallel runways 750 ft apart. +BACKGROUNDThe FAA recognizes that significant capacity is lost when simultaneous operations performed under visual conditions are not operational under poor weather conditions.The FAA, as a part of its NextGen plan [FAA, 2008], aims to reduce the allowable spacing between runways used for simultaneous operations in poor visibility, currently 4300 ft., by revising standards and improving technologies.Several concepts that address the revision of separation standards and new technologies include SOIA, AILS and Terminal Area Capacity Enhancing Concept.The role of the air traffic controller during simultaneous approaches is different for each of the above mentioned concepts.Under SOIA, the controller has positive control over the aircraft until they break through the clouds and the follower aircraft has visual contact with the leading aircraft, at which time separation authority is delegated to the flight deck.Under the AILS concept, the final approach controller has positive control over the aircraft pair until the flight deck of the trailing aircraft is given a clearance for AILS approach.This clearance is given by the final approach controller just prior to transfer of communications to the tower.Once the AILS clearance is given the trailing aircraft crew is responsible for maintaining separation from traffic on the adjacent parallel approach, while ATC remains responsible for longitudinal separation of in-trail traffic operating in the same stream [Waller, et al., 2000].The concept investigated in the current study, called Terminal Area Capacity Enhancing Concept (TACEC) [Miller, et. al., 2005], was collaboratively developed by Raytheon and NASA Ames Research Center.TACEC is a technique that can be used for conducting simultaneous instrument approaches to two or even three closely-spaced parallel runways that are 750 ft apart.TACEC operations could double the landing capacity of airports with closely-spaced parallel runways (closer than 2500 ft) during low visibility conditions, approaching arrival rates comparable to visual approach operations.The concept defines a safe zone behind the leader (range 5s to 25s) where the trailing aircraft is protected from the wake of the leader.The trailing aircraft flies an approach with a 6 degree slew, and at a coupling point, which is about 12 nmi from the runway threshold, the two aircraft are linked, with the trailing plane using flight deck automation to control speed and maintain precise spacing of 15 sec in trail behind the leader (Figure X.1).The concept assumes Differential Global Positioning System (DGPS), augmented ADS-B, 4 dimensional flight management system, wind detection sensors onboard the aircraft, and cockpit automation that are not extant in today's NAS.All the concepts discussed assume that the air traffic controller will assign aircraft to pairs with the knowledge that they are properly equipped.Given this, the TACEC research explores the role of the air traffic controller in assigning aircraft to pairs so they can perform simultaneous approaches.The pairing of aircraft was done under different levels of automation in order to investigate the appropriate human/automation mix for the given task.Previous research that explored the role of the controller under different levels of automation included a simulation study by Slattery et al., [1995] who examined the effects of the Final Approach Spacing Tool (FAST) with aircraft landing simultaneously on parallel runways.The simulation contained various combinations of aircraft, equipped and unequipped with advanced navigation systems.Similarly, another study [McAnnulty and Zingale, 2005] investigated the effect of using a Cockpit Display Traffic Information (CDTI), for enhanced visual operations, on controller workload and situation awareness.They found that advanced concepts involving the use of more sophisticated CDTI functions require modifications to current procedures and additional controller workstation tools.Verma et.al. (2009) also investigated the pilot procedures for breakout maneuvers for simultaneous arrivals that were flown under the manual and auto pilot flight control, but did not explore procedures for controllers.The work described in this paper involves a simulation experiment to examine the role of controllers using a pairing tool, under four different levels of automation, to assign pairs for simultaneous approaches to runways 750 ft apart.A new ATC position, the area coordinator, was added and given responsibility for pairing the aircraft.The simulation also included flight deck automation on the following aircraft that enabled pilots to achieve precise 15 s in trail spacing between the leader and the follower at the coupling point (Figure 1).Results from the different levels of the human/automation mix are presented with the dependent variables of (1) the number of aircraft pairs made by the controllers, (2) the number of aircraft pairs that had to be broken and brought in as single aircraft to prevent potential loss of separation, and (3) controller workload.The experimental approach section defines the airspace used, the scenarios, and the experimental setup.The results and discussion section focuses on the description of the metrics collected and analyzed. +EXPERIMENTAL APPROACH AIRSPACE ORGANIZATIONSan Francisco International (SFO) airport was used as the test bed.SFO has parallel runways, 28L and 28R that are used for all arrival streams.The traffic scenarios consisted of four arrivals streams -Yosem and Modesto from the east, Point Reyes from the north, and Big Sur from the south.The airspace was modified to split the route to the two coupling points (CP28L and CP28R) on each of the four streams.This would allow for runway changes and for aircraft from the same stream to be paired.The routes were modified so they were de-conflicted and were set up with a Required Navigation Performance (RNP) of 1.14 nmi meaning that standard separation was not applied.Instead, the closest distance between the routes before the coupling point was 1.14.The RNP level after the coupling point was 0.01.For this study, the two approach sectors, Niles and Boulder, were configured such that the controller was responsible for the airspace from the TRACON boundary up to the coupling point which is at 4000 ft.AGL.The Niles Sector managed traffic from the two east-side routes-Yosem and Modesto.The Boulder sector managed the routes from the north and south -Point Reyes, and Big Sur respectively. +Traffic ScenariosTwo different traffic scenarios were used for the simulation.Both the scenarios were equivalent in the rate of arrivals, (approximately 60 arrivals per hour), to the rate of arrivals under visual flight rules (VFR) conditions at SFO.The scenarios also approximated the current distribution of traffic across the four arrival routes simulated for the study. +TEST CONDITIONSThe study included a pairing interface on the Standard Terminal Automation Replacement System (STARS) display, and an algorithm that created pairs.To manipulate the level of automation used for the study, the capabilities of the pairing algorithm and the pairing interface were varied.The role of creating aircraft pairs for simultaneous approaches was assigned to the area coordinator who looked beyond the TRACON boundary, with the sector controllers managing the pair inside the TRACON boundary such that the follower arrived 15 sec behind the lead aircraft at the coupling point.The controllers were also responsible for standard separation between the pairs.They were allowed to use speed only to manage the flow and create adequate separation; vectoring of aircraft was not allowed in any of the conditions.In the no-automation condition, the area coordinator used current day technologies and flight strips to make pairs and communicate them to the two sector controllers.There was no pairing algorithm or controller interface to assist the area coordinator with creating pairs for simultaneous approaches.The goal for the sector controllers (Niles and Boulder) was to bring the trailing aircraft slightly behind the lead aircraft at the coupling point sans automation.In the next level of automation (Mixed-1) the area coordinator was responsible for creating pairs, using an interface provided on the STARS display.The area coordinator was able to mouse over the data tag and click on a lead aircraft and a following aircraft to create pairs in the "pairs table" -a new feature added to the STARS display.The area coordinator sent a data link message with pairing information to the two flight decks and waited for an acknowledgement from the pilots.After both acknowledgements were received, a finalized pair was displayed on the area coordinator's and both sector controllers' displays.Under all automated conditions, merging and spacing flight deck automation was used on the simulated flights to achieve 15s temporal distance between leader and follower at the coupling point without the intervention of the controller.The Mixed-2 condition increased automation.In this condition, the area coordinator selected a leader and a pairing algorithm provided up to three options for trailing aircraft in the "pairs table."The area coordinator evaluated each option offered by the automation against the timeline and finalized the best option by sending datalink messages to the aircraft as in Mixed 1.The Full Automation condition further increased the role of automation and reduced the role of the human for the pairing task.The pairing algorithm offered one best option for aircraft pairs to the area coordinator, who finalized the pairs by sending the datalink message after evaluating the pair against the timeline. +MethodologyThe experiment was a 3x2 within subjects design, with three controller positions and two scenarios.The three controller positions consisted of one area coordinator, and Niles and Boulder sector controllers.The three participant controllers on each team rotated between the three positions.A total of 24 runs (4 conditions x 6 runs each) were conducted per week for two weeks, with a different team of recently retired controllers each week.The four experimental conditions were not randomly distributed.All six runs for every condition were conducted before the participants were trained on the procedures for the next condition and training always preceded actual data collection runs.This was done to avoid confusion between the different procedures and displays used in the four conditions. +EQUIPMENT/ DISPLAYSThe simulation used the Multi Aircraft Control Systems (MACS) simulation environment including a STARS display that could be used in the Terminal Radar Approach Control (TRACON).MACS is an aircraft target generator system [Prevot et. al, 2004] that provides current controller displays and can be used for rapid prototyping of new displays.The Airborne Spacing for Terminal Area Routes (ASTAR) modeled flight deck merging and spacing to achieve the 15 sec in trail interval between the lead and following aircraft at the coupling point.ASTAR builds 4D trajectories for both the ownship and the lead aircraft approaching the adjacent runway [Barmore et al., 2008], then provides target speed inputs to the follower's FMS, to achieve the assigned temporal spacing between the leader and follower.A pairing algorithm was integrated with MACS to identify overlapping Estimated Time of Arrivals (ETAs) between aircraft and chose possible pairs (in Mixed 2) or best pairs (in Full Automation).The window of opportunity for pairing was reduced as the aircraft moved closer to the airport, and the distance that could be made up or lost by speed adjustment shortened.The pairing algorithm assessed and offered pairs that could be made by changing the arrival runway for one aircraft [Kupfer, 2009]. +TOOLS FOR DATA COLLECTIONAll participants completed a demographic survey that included information such as age, experience at different facilities etc. before the simulation started.Controller workload data was also collected using the Workload Assessment Keypad (WAK).Metrics such as situation awareness, intra pair spacing and others were analyzed but not presented in this paper. +RESULTS AND DISCUSSIONThe data analysis paradigm used two independent variables, consistent with experiment procedures described earlier.The independent variable of automation condition had four levels: no automation, mixed automation1 (mixed1), mixed automation2 (mixed2), and full automation.The independent variable of controller position had three levels: Boulder, Niles, and area coordinator.The effects of these two independent variables on the three dependent variables are described in this section.The three dependent variables include controller workload, number of aircraft pairs, and number of deleted aircraft pairs. +CONTROLLER WORKLOADParticipants recorded their current level of workload by pressing a key on the electronic Workload Assessment Keypad (WAK) [Stein, 1985] at 5 minute intervals throughout the simulation runs.Workload assessments are subjective and could range from 1 (very low workload) to 7 (very high workload). +Workload By Controller Position and Automation LevelAnalysis of variance results indicated a significant main effect of position on controller workload, F(2,4)=11.56,p<0.05 (Figure X.2). +FIGURE X.2 Controller Workload by PositionWhile overall workload across all positions and conditions was low (mean=2.5,SD=1.0), Figure 2 shows that the area coordinator had a higher level of workload relative to the other two positions.Post-hoc analysis yielded a statistically significant difference between the area coordinator and Boulder, F(1,5)=25.27,p<0.01, and the area coordinator and Niles, F(1,5)=25.55,p<0.01.This finding is not surprising since the area coordinator is responsible for the area covered by multiple sectors, pairing the aircraft under different positions, and monitoring the aircraft pairs and the flow.In this sense, the area coordinator is required to perform a higher level of multi-tasking relative to the other two positions.Also, the experiment procedures did not allow the sector controllers to form pairs if the area coordinator was too busy.Similarly, the procedures did not allow the area coordinator position to break pairs or directly swap runways for any aircraft -this had to be done through the sector controllers who had ownership of the aircraft.Again, this increased workload suggests the need for additional fine-tuning of the area coordinator's responsibilities.While statistically significant, the mean difference of less than 1 scale point between positions might also serve to reinforce the main finding, which is that workload was found to be consistently low across all positions.To further illuminate position differences, analysis of the current study factors is currently underway which explores the sub-components that contribute to overall workload.Workload was also found to be low across the four automation conditions, with the Mixed-2 condition showing the highest workload (mean=2.8,stdev=1.2) and the Full automation condition showing the lowest workload (mean=2.3,stdev=0.7).The higher workload level in Mixed-2 may reflect the excessive task load, which was substantiated with participants' open-ended feedback.However, this result should be viewed as preliminary, since the range was less than 1 scale point.Again, further analysis on workload sub-components might help to illuminate this finding, which may provide an excellent input for the heuristics used by the pairing algorithm. +CREATING AIRCRAFT PAIRSAnalysis of variance results yielded a significant main effect of automation on the number of aircraft pairs made by the study participants, F(3,44)=4.69,p<0.01.A Tukey HSD post-hoc analysis yielded a statistically significant difference between the Mixed-1 and Mixed-2 conditions (mean difference=4.7,std error=1.266,p<0.01).Clearly, the controller-participants were more productive in making aircraft pairs under the Mixed-1 condition (mean=18.1,sd=3.3), as compared to the Mixed-2 (mean=13.4,sd=3.9) and the full automation (mean=16.0,sd=2.7)conditions.The controllers used their own judgment in creating pairs under the Mixed-1 and No-automation conditions.However, Mixed-1 provided the option of an alternative interface that eliminated the process of writing pairs on flight strips.The Mixed-2 condition required the participants to evaluate several pairs before choosing one -a requirement absent from the Mixed-1 condition.Also, the Mixed-1 condition allowed the controller-participants the greatest level of flexibility in aircraft pairing procedures.Controller-participant feedback on the Mixed-1 condition indicated that the display and flight deck automation were very helpful in making aircraft pairs, while being allowed to use their own judgment to create pairs meant they were not constrained by the automation.The Mixed-2 and full automation conditions sometimes frustrated the controllers if they did not agree with the pairs suggested by the automation, which may have contributed to the relatively low number of aircraft pairs made under those conditions.During discussions, controllers indicated preference for the Mixed-1 condition and expressed a desire for another condition where automation would suggest one good pair while a manual override was allowed.Another heuristic for the algorithm would be to not show pairs that would likely be unacceptable to the controller. +BREAKING AIRCRAFT PAIRS BY AUTOMATION LEVELSome aircraft pairs were broken by the controller-participants because flight deck automation could not achieve 15 s temporal separation at the coupling point and standard separation between the aircraft was not possible (Table X.1).It is interesting to note the relatively small percentage aircraft pairs broken under the No automation condition, possibly because the controllers had the goal to bring the aircraft slightly behind each other and they achieved it through speed intervention.In the Mixed-1 condition, the area coordinator created the pairs using the display tools.The higher number of pairs broken under Mixed-1 may have been caused by the flight deck automation's speed manipulation constraints, making it impossible to drive the following aircraft to meet the temporal separation of 15s at the coupling point for some pairs created by the area coordinator. +CONCLUSIONSThis simulation study examined the human-automation mix for pairing aircraft for simultaneous approaches to closely spaced parallel runways under different levels of automation.Four levels of automation and three controller positions were examined, and results include analyses of controller-participant workload, the number of pairs made by the controller-participants, and the number of pairs that were broken before the aircraft landed.Results show that the controller-participants were most productive in forming pairs under the Mixed-1 condition where they used their own judgment to create pairs and used the automation as an interface and for communicating the pairs information to the controllers.Under the Mixed-2 and Full conditions, the study participants did not perform as well on the number of pairs created because the pairing algorithm suggested pairs that were not acceptable to the controller.The heuristics for the pairing algorithm need further refinement.Allowing the controller to have the final say and override any pairing suggestion made by algorithm will be the key for maintaining flexibility for the controller.Finally, while controller workload remained at a manageable level across all automation levels and controller positions, there was higher workload under the Mixed-2 condition and for the area coordinator position, which may suggest the need for additional fine-tuning of the pairing procedures and the area coordinator's responsibilities.FIGURE X. 11FIGURE X.1 Example of aircraft geometries for the concept under investigation. +Table X .1XPercentage of Aircraft Pairs Deleted by Automation Levelconditionpercentage of aircraft pairs brokenNo automation7.6Mixed-1 automation15.7Mixed-2 automation15.0Full automation11.5 + + + + + + + + + NASA administrator announces resignation + + TAbbott + + + DElliot + + 10.1016/s0045-8732(01)80042-5 + Concept. NASA/TP-2001-210665 + + + COSPAR Information Bulletin + COSPAR Information Bulletin + 0045-8732 + + 2001 + 152 + + 2001 + Elsevier BV + + + Abbott, T. & Elliot, D. (2001). Simulator Evaluation of Airborne Information for lateral Spacing (AILS) Concept. NASA/TP-2001-210665. + + + + + Simulation Results for Airborne Precision Spacing along Continuous Descent Arrivals + + BryanBarmore + + + WilliamCapron + + + TerenceAbbott + + + BrianBaxley + + 10.2514/6.2008-8931 + AIAA 2008-8931 + + + The 26th Congress of ICAS and 8th AIAA ATIO + + American Institute of Aeronautics and Astronautics + 2008. 2008 + + + Barmore, B.E., Abbott, T.S.,Capron, W.R., Baxley, B.T. (2008) "Simulation Results for Airborne precision Spacing along with Continuous Descent arrivals." ATIO 2008, AIAA 2008-8931. + + + + + Distributing net-enabled federal aviation administration (FAA) weather data + + MarkSimons + + 10.1109/icnsurv.2008.4559189 + + + + 2008 Integrated Communications, Navigation and Surveillance Conference + + IEEE + 2008. February 2008 + + + Federal Aviation Administration (2008). National Aviation Research Plan, February 2008. http://nas-architecture.faa.gov/nas/downloads/ + + + + + + MKupfer + + Scheduling Aircraft Landings to Closely Spaced Parallel Runways, Eighth USA/Europe Air Traffic management Research and Development Seminar + Napa, CA + + ATM + 2009. 2009 + + + Kupfer, M. (2009). Scheduling Aircraft Landings to Closely Spaced Parallel Runways, Eighth USA/Europe Air Traffic management Research and Development Seminar (ATM2009). Napa, CA. + + + + + Compilation of DOT and FAA Airport Legal Determinations and Opinion Letters as of December 31, 2020 + 10.17226/26202 + + + + Simultaneous Offset Instrument Approaches -Newark International Airport: An Airport Feasibility Study + + Transportation Research Board + 2001 + + + Magyratis, S., Kopardekar, P., Sacco, N., & Carmen, K. (2001). Simultaneous Offset Instrument Approaches -Newark International Airport: An Airport Feasibility Study, DOT/FAA/CT-TN02/01. + + + + + FAA Air Traffic Controller Selection in the United States + + DMcanulty + + + CZingale + + 10.4324/9781315242538-22 + + + + Staffing the ATM System + + Routledge + 2005 + + + + McAnulty, D., & Zingale, C. (2005) "Pilot-Based Separation and Spacing on Approach to Final: The Effect on Air Traffic Controller Workload and Performance," DOT/FAA/CT-05/14. + + + + + CNS requirements for precision flight in advanced terminal airspace + + MEMiller + + + SDougherty + + + JStella + + + PReddy + + 10.1109/aero.2005.1559471 + + + 2005 IEEE Aerospace Conference + + IEEE + 2005 + + + + Miller, M., Dougherty, S., Stella, J., and Reddy, P. (2005) "CNS Requirements for Precision Flight in Advanced Terminal Airspace," Aerospace, 2005 IEEE Conference, pp. 1-10. + + + + + Co-Operative Air Traffic Management: Concept and Transition + + ThomasPrevot + + + ToddCallantine + + + PaulLee + + + JoeyMercer + + + VernolBattiste + + + EverettPalmer + + + NancySmith + + 10.2514/6.2005-6045 + + + AIAA Guidance, Navigation, and Control Conference and Exhibit + Madeira, Portugal + + American Institute of Aeronautics and Astronautics + 2004 + + + Prevot T., Callantine T., Lee P., Mercer J., Palmer E., & Smith, N. (2004). Rapid Prototyping and Exploration of Advanced Air Traffic Concepts. International Conference on Computational and Engineering Science, Madeira, Portugal. + + + + + Effects of ATC automation on precision approaches to closely spaced parallel runways + + RhondaSlattery + + + KatharineLee + + + BeverlySanford + + 10.2514/6.1995-3367 + + + Guidance, Navigation, and Control Conference + + American Institute of Aeronautics and Astronautics + 1995 + + + Slattery, R., Lee, K., and Sanford,B. (1995). Effects of ATC automation on precision approaches to closely spaced parallel runaways. In Proceedings of the 1995 AIAA Guidance, Navigation, and Control Conference. + + + + + TRB Special Report 314: Federal Aviation Administration's Approach for Determining Future Air Traffic Controller Staffing Needs + 10.17226/18824 + No. DOT/FAA/CT-TN84/24 + + + Federal Aviation Administration + Atlantic City, NJ + + National Academies Press + 1985 + + + Report + Stein, E.S. (1985) "Air Traffic Controller Workload: An Examination of Workload Probe," Report No. DOT/FAA/CT-TN84/24, Federal Aviation Administration, Atlantic City, NJ. + + + + + Comparison of manual and autopilot breakout maneuvers with three closely spaced parallel runway approaches + + SavitaVerma + + + SandraLozito + + + ThomasKozon + + + GordonHardy + + + HerbertResnick + + 10.1109/dasc.2009.5347457 + + + 2009 IEEE/AIAA 28th Digital Avionics Systems Conference + Orlando, FL + + IEEE + 2009 + + + Verma, S.A., Lozito, S., Ballinger, D., Kozon, T., Hardy, G. & Resnick, H. (2009) "Comparison of Manual and Autopilot Breakout Maneuvers for Three Closely Spaced Parallel Runways approaches." Digital Avionics Systems Conference, Orlando, FL. + + + + + Lloyd Doyle launches new inspection concept + + MWaller + + + TDoyle + + + FMcgee + + 10.1108/cw.2000.21726bad.004 + NASA/TM-2000-210091 + + + Circuit World + 0305-6120 + + 26 + 2 + 2000 + Emerald + + + Waller, M., Doyle,T., & McGee,F. (2000). Analysis of the Role of ATC in the AILS Concept, NASA/TM-2000-210091. + + + + + + diff --git a/file760.txt b/file760.txt new file mode 100644 index 0000000000000000000000000000000000000000..e5cd749753c02c47b4811df8fcbf996cb73e4959 --- /dev/null +++ b/file760.txt @@ -0,0 +1,320 @@ + + + + +IntroductionAirports and surface congestion are the biggest bottlenecks in the current national airspace system (NAS) [1,2].In response, technological capabilities are being developed to improve the movement of airport surface traffic.New decision aids are required to integrate other extant technologies in the tower and present it to the users, particularly the tower controllers, in a coherent manner.The tower controllers in the current day rely heavily on the out-the-window view and ASDE-X when available, for aircraft position, and on printed flight strips for flight details.The introduction of new technology to tower controllers is based on the presumption that a similar sized controller team will manage twice the traffic [3].Activities are currently underway at NASA Ames Research Center to implement key elements of the NGATS ATM-Airportal Program.In response, the current study supports selected NGATS ATM-Airportal focus areas, including surface traffic optimization and the management of dynamic airport configuration.In the current study, real-time simulations were developed at NASA Ames Research Center to represent and study a future surface concept in a high fidelity airport tower environment.The aim of this effort was to explore the changes in the procedures, roles and responsibilities of the controllers when they interact with a prototype automation tool.The prototype tool chosen was GoSAFE (Ground Operation, Situation Awareness and Efficiency Flow), which is part of the SOAR 1-4244-1108-4/07/$25.00 ©2007 IEEE.6.B.2-1 concept (Surface Operation and Automation Research) [4].The SOAR concept envisages that future surface operations will involve taxi clearances containing precisely timed taxi routes.This will be achieved through collaboration between tower control tools (e.g., GoSAFE) and advanced automation tools developed for the flight deck.GoSAFE is intended to plan efficient taxi operations [4] with the assumption that flight decks of the future can execute precise taxi commands.Because of timing information embedded in the clearances, it was impractical for these clearances to be input by hand, and thus data link (electronic exchange of information) was required for the procedure.A number of studies have examined tower controllers and their interactions with new tools or procedures.For instance, one study conducted at NASA Ames Research Center explored options for reducing runway incursions at the Los Angeles International Airport (LAX) [5].One option involved a procedural change in the tower, in which two controllers operated one runway each, instead of having one controller manage the two active runways.The study demonstrated that mixed runway operations and runway crossings required significant coordination between the two controllers.The required level of coordination increased the possibility of operational errors that could lead to runway incursions.To minimize such errors, a single controller was assigned to a set of parallel runways.In the current study, an attempt has been made to divide the responsibility for the two parallel runways in DFW without increasing communication and coordination, assisted by the GoSAFE surface automation tool.Another study conducted at NASA Ames Research Center supported the Chicago O'Hare Airport Modernization Program.This study involved a human-in-the-loop simulation of the full build-out Airport Layout Plan [6].During the simulation, it was determined that one of the two controllers responsible for arrival traffic had a significantly higher workload than the other.Reorienting the areas of responsibility from an eastwest to a north-south orientation mitigated the workload imbalance between the two positions.As with the Chicago O'Hare Study [6], the current study also split the areas of responsibility for the two active runways in the north-south direction, but using Dallas Fort-Worth International Airport (DFW) as the test bed.This paper examines the changes in controller roles and responsibilities associated with (1) an increased volume of airport surface traffic, (2) the introduction of the GoSAFE automation technology and (3) changes in controller surface area jurisdiction, while GoSAFE was in operation.GoSAFE was previously introduced in an earlier study (phase-1), and based on the phase-1 results, areas of controller responsibility were modified and implemented in the current study (phase-2).Phase-1 and phase-2 results are discussed and compared. +Summary of Phase-1 ResultsPhase-1, which tested workload with the prototype tool, GoSAFE, found that controller workload was significantly different (i.e., unbalanced) across the four controller positions (F= 130.47, df=3,130 , p<0.001).The Local East#1 (LE1) controller was significantly busier and experienced higher workload on the Workload Assessment Keypad (WAK) scale [7] than the other 3 controllers (see Table 1).The LE1 controller at DFW receives traffic from all directions, requiring management of traffic crossings through active runways for aircraft departing 17C and 13L, and for arrivals going to the terminals.Under phase-1, responsibilities of the GE2 controller had been substantially changed due to the introduction of GoSAFE [8] since this automation tool allowed LE1 to clear the aircraft crossing the active runways on DFW east 17R and 17C all the way to the gate, thereby reducing GE2's level of responsibility.Thus, GE2's tasks were reduced to monitoring the arrivals, while actively managing departures.The change in jurisdiction was identified in this phase. +6.B.2-2 +Current (Phase-2) MethodologyThe current study used the phase-1 results to configure, and later test, new areas of responsibilities for the controllers that were changed due to the introduction of increased surface traffic and GoSAFE.In order to compensate for the uneven distribution of phase-1 workload, changes in the areas of responsibilities for the LE1 and GE2 controllers were introduced under phase-2.The changes in jurisdiction are shown in Figures 1 and2.The most prominent change involved reducing the area of responsibility for the LE1 controller.Specifically, phase-1 LE1 responsibilities of controlling aircraft crossing the south end of the active runways 17R and 17C, on taxiway ER, were transferred to GE2 under phase-2.Thus, the jurisdiction over the active runways was split in the north-south direction, where the north was controlled by LE1 and the south end of the runways was controlled by GE2. +Experimental ConditionsBoth phase-1 and phase-2 of the study used GoSAFE under two experimental conditions: (1) Mixed Communications and (2) Full Datalink, each with three scenarios that were randomly distributed among the runs.The Mixed Communications condition used GoSAFE to deliver the entire taxiway instruction (pre-clearance) via datalink, whereas all routine taxi clearances were issued in segments to the pilots using voice.In the Full Datalink condition, the complete taxi instruction (pre-clearance) and the routine taxi instructions in segments were issued via datalink.A total of 10 runs were conducted, half were mixed communications and the other half were full datalink. +HypothesesThe dependent measures of interest included Workload, Situation Awareness, and Communications.The hypotheses have been categorized under two categories -one that compared phase-1 and -2, and the other that focused on phase-2 only.It was hypothesized that (1) workload among controller positions would become more equally distributed under phase-2, as compared to phase-1, as a result of the DFW jurisdiction changes and the introduction of GoSAFE, (2) situation awareness would remain the same regardless of phase (phase-2 vs. phase-1), and condition (mixed communications vs. full datalink), (3) voice communication loads would be more equally distributed among controllers under phase-2 as compared to phase-1, in terms of number of transmissions and percentage of voice channel occupancy, and (4) under phase-2 jurisdiction, the controllers would experience higher workload in the mixed communications condition than in the full datalink condition. +ParticipantsThe participants in the study were four retired controllers (two local and two ground controllers) who participated in both phase-1 and phase-2 of the study.All participants were experienced tower controllers and one of them had DFW experience.On average, the participants had over 21 years of controller experience, and were retired for approximately six and a half years.In the current6.B.2-3 6.B.2-4study, the participants staffed four tower positions, consisting of two local controller and two ground controller positions.To implement the air traffic control tasks of the simulation, all controllerparticipants were trained on DFW and GoSAFE procedures.The controllers rotated through each of the positions, changing after each study run, to randomize individual effects as much as possible.The simulation also required 5 pseudo pilots who "flew" several aircraft within pre-defined geographical areas of the airport. +Facilities and Simulators UsedThe traffic in the simulations was created using the Airspace Traffic Generator (ATG), a ground and airborne target generator customized for advanced ATC research.The ATG parsed and executed the 4-D taxi commands and emulated the flight deck automation required to operate the concept.The ATG was integrated with GoSAFE using High Level Architecture (HLA).The arrivals were monitored by the local controllers using a JBRITE display, which is an emulation of the FAA's DBRITE.In the current study, all the controllers used the GoSAFE displays to manage traffic and there was no out-the-window view available in phase-2.Time synchronization, data collection and data management details were implemented over the HLA network.Additional details on the software architecture and modules are documented elsewhere [9]. +Traffic and ScenarioExpected future levels of traffic were simulated for DFW.The east side of DFW with a south flow using runways 17R, 17C, 17L, and 13L under clear day conditions were simulated.In general, the traffic count for a 45 minute scenario was 140-160.This is approximately 1.5 times the current level of traffic for the east side only.The three scenarios included an arrival rush, an even flow of arrivals and departures, and a departure rush that morphed into an arrival rush. +ProcedureThe controllers used the GOSAFE displays for management of traffic.The GoSAFE displays are radar like displays on the surface that provide the entire taxi instruction or pre-clearance to the controllers, eliminating the need for high familiarity with the airport.The controllers issued segments of the pre-clearance as commands to pilots in simplified phraseology (by using segment numbers instead of reading the entire timed taxi instruction), which also mitigated the need for high levels of familiarity with the airport. +ResultsThe tools used to collect data included the Workload Assessment Keypad (WAK) [7] and Task Load Index (TLX) [10] scales for measuring workload, and the Situation Awareness Rating Scale (SART) [11] for measuring situation awareness.To assess WAK workload, the participants pressed a key on their workload pad every 5 minutes during the simulation run.This WAK key press represented the participant's assessment of current workload experienced, which ranged from 1 (low workload) to 7 (high workload).In addition, TLX and SART questionnaires were administered to each participant after every simulation run.The researchers also made observation notes and led group discussions with the controllers.All data analysis results from these sources are described in this section. +Workload (TLX and WAK)Table 2 presents summary statistics of the overall data distribution of workload, as measured on the TLX scale, with comparisons between phase-1 of the experiment (initial controller responsibility/ jurisdiction) and phase-2 of the experiment (new responsibility/ jurisdiction).As shown in Table 2, the change in controller jurisdiction had a beneficial effect on most of the controller workload variables, with less mental demand, physical demand, temporal demand and frustration, along with increased performance.However, none of these observed differences were large enough to reach statistical significance. +Table 2. TLX Workload by the PhaseFigure 3 shows a comparison of TLX workload ratings as a function of experiment phase (phase-1: initial jurisdiction, phase-2: new jurisdiction) and condition (mixed, full).Figure 3 shows relatively large differences between phase-1 mixed vs. full conditions, whereas smaller differences are indicated between phase-2 mixed vs. full conditions.Under the phase-1 jurisdiction, statistically significant differences between the mixed and full conditions on the dependent measures of physical demand (F=4.87,df=1,29, p<= 0.05) and temporal demand (F=5.45,df=1,20, p<= 0.05) were realized.Conversely, no statistically significant differences were indicated between mixed and full conditions under phase-2 jurisdiction.Hence, it seems likely that the change in jurisdiction between the two experiment phases had the overall effect of balancing the workload across the mixed and full conditions on the specific variables measured by the TLX scale (i.e., minimized any possible mixed/full effect).The availability of surface automation -GoSAFE aided in keeping the communication and coordination requirements down that would have otherwise increased due to the sharing of active runways with an increased volume of surface traffic.During group discussion, the controllers mentioned that LE1, under phase-2, experienced increased spare mental capacity.This allowed LE1 controllers to issue commands in a timely fashion to the mixed condition pilots, due to the workload redistribution under phase-2.Table 3 presents the differences of the means (absolute value) between the mixed/full conditions for each of the two phases of the experiment, on each of the TLX workload measures: +Table 3. Workload for Condition & Phase(*statistically significant at p < 0.5) Similar results were found with WAK workload, showing a significant interaction effect of phase and condition (F=25.58,df=1,1 , p<=0.05),where the phase-1 mixed/full workload difference of nearly 1 full scale point was virtually eliminated under phase-2.Figures 4 and5 illustrate this effect graphically. +Figure 5. Phase-2 WAK Workload by ConditionTable 4 shows the range of TLX workload means across the 4 controller positions (LE1, LE2, GE1, and GE2) for each of the two experiment phases.Here, range is defined as the difference between the highest mean workload and the smallest mean workload among the four controllers.The range provides a measure of variability of workload across the controller positions.The range values in Table 4 clearly show an overall rebalancing of TLX workload, as a result of the implementation of the new controller jurisdiction (supported by GoSAFE) in phase-2 of the experiment (i.e., the phase-2 range of means is considerably less than the phase-1 range of means in TLX workload measures).Figures 6 and7 illustrate the effects of controller position and experiment phase on the TLX workload ratings.Only those means corresponding to the LE1 and GE2 positions are presented, since there was insufficient phase-1 vs. phase-2 workload variability for the other two positions.This would make sense, since LE1 and GE2 are the only two controller positions directly impacted by the jurisdiction change implemented in phase-2.Figure 6 shows the phase-1 and phase-2 mean responses on each of the TLX workload measures, for position LE1 only.For the most part, responses on all of the workload measures show improvement under the new jurisdiction implemented in phase-2, as compared to the workload responses under the old jurisdiction in phase-1.There was less mental demand, physical demand, temporal demand, effort and frustration, with a slight improvement in performance.It would appear that the implementation of the new controller jurisdiction in phase-2 had the effect of re-distributing workload across the LE1 and GE2 positions.Hence, the workload was spread out more evenly across both positions, with one position experiencing an increase and the other position experiencing a decrease in workload.Figures 6 and7 clearly show that the phase-2 distribution of means are much more similar to each other than the phase-1 distribution of means across the LE1 and GE2 positions, providing further evidence of the observed workload re-distribution.ANOVA statistics of controller position effects on workload for each phase are shown in Tables 5 and6.In general, the phase-1 analysis of position resulted in statistically significant differences on most of the TLX workload dependent measures (Table 5).Conversely, the phase-2 analysis of position on each of the workload measures resulted in only one statistically significant difference (Table 6).This would seem to provide further evidence of the workload re-distribution across positions, resulting from a change in controller jurisdiction.The controllers also mentioned that under phase-2, the GoSAFE technology assisted them with runway crossings, a task traditionally handled by the local controllers.It was further noted that in the absence of the surface technology provided by GoSAFE, this change in areas of responsibility would not have been operationally feasible due to the communication and coordination requirements under phase-2, especially with the increased level of traffic.A position by phase interaction effect of WAK workload (Figure 8) was also realized (F=15.83,df=3,406, p<= 0.05), with a general reduction of workload under phase-2, and similar trends across the controller positions (e.g., LE1 and GE2 workload re-distribution under phase-2). +Situation Awareness (SART)The change in controller jurisdiction implemented in phase-2 had the effect of improving situation awareness on all of the 10 SART measures.As compared to the phase-1 mean responses, the phase-2 mean responses indicated less instability, variability, complexity and division of attention.The phase-2 mean responses also indicated more alertness, spare mental capacity, concentration, information quantity, and familiarity.Statistical significance was achieved on the instability, concentration and familiarity measures.SART means and ANOVA statistics are listed in Table 7. Conversely, SART analysis results, broken down by experiment phase and mixed/full conditions, were less striking.While some marginal differences in the mixed/full condition were observed as a function of experiment phase, neither the main effect of the mixed/full condition, nor the interaction effect of phase by condition, for any of the SART measures, reached statistical significance. +6.B.2-7Figures 9 and 10 illustrate the interaction of experiment phase by controller position on situation awareness.Only the means for those controller positions directly impacted by the change in jurisdiction (i.e., LE1, GE2) are illustrated.Results for positions LE2 and GE1 are not shown, since their phase-1 vs. phase-2 differences were relatively consistent, relative to the LE1 and GE2 differences which were much more striking. +Figure 9. LE1 Situation Awareness by PhaseUnder phase-2 (as compared to phase-1), the LE1 position experienced less instability, variability, complexity, and division of attention.Under phase-2 (as compared to phase-1), the LE1 position also experienced more alertness, spare mental capacity, concentration, information quantity and familiarity, with about the same level of information quality (Figure 9 and Table 8).In group discussions, the controllers also indicated that, under phase-2, they had more spare mental capacity, especially in the mixed condition.So, in general, there was consistent improvement in situation awareness experienced in phase-2 as compared to phase-1 for position LE1.Under the original jurisdiction, the LE1 controller scrolled up and down the map to handle the aircraft crossing both the north and south ends of the active runways.Under the new jurisdiction where LE1 is no longer responsible for the south end of active runways, the LE1 controller scrolled between North 17L and 17R, and taxiways EM, up to the boundary of the new LE1 jurisdiction area.Hence, it would make sense that LE1 had better situation awareness under phase-2, since there was less jurisdiction area under their responsibility.The hypothesis that situation awareness would remain the same between the two phases was not upheld for LE1, since situation awareness improved due to better stability, and spare mental capacity brought about by the changes in controller jurisdiction.Under phase-2 (as compared to phase-1), the GE2 position experienced more instability, variability, complexity, alertness, concentration, division of attention and information quantity.Under phase-2 (as compared to phase-1), the GE2 position also experienced less spare mental capacity, and had about the same level of information quality and familiarity (Figure 10 and Table 8).So, while there was some improvement in situation awareness on several of the SART measures, there was also some degradation or no improvement on most of the measures.This finding is consistent with researchers' observations of phase-2 runs, where GE2 would occasionally ask LE1 if there were more departures on 17R, since aircraft on taxiways ER could not cross 17R until the departure aircraft had left the airport.Therefore, in this instance, GE2 experienced less situation awareness, requiring occasional controller coordination to develop a fuller awareness of surface traffic.However, it should be noted that while some degradation in situation awareness occurred, the level of degradation was usually less than one scale point, and situation awareness generally remained at a relatively high level. +6.B.2-8While Figures 9 and10 clearly show a general level of improvement in situation awareness for the LE1 position, and some degradation in situation awareness for the GE2 position, they also show these effects to be one of situation awareness redistribution across both positions, i.e., the LE1 and GE2 phase-2 curves are much more similar to each other than the LE1 and GE2 phase-1 curves.Situation awareness improvement for LE1 is quite high for some of the SART measures (e.g., instability, variability, complexity) while GE2 situation awareness degradation, if indicated at all, is generally low in magnitude.For instance, spare mental capacity increased for LE1 by 1.3 scale points, whereas GE2's spare mental capacity decreased by only 0.5 scale points.Combined with the relative lack of variability observed from the other two positions, this would point to an overall increased level in situation awareness across all positions.SART means and ANOVA statistics for the phase by position interaction effects are described in Table 8.Again, only cell means for the LE1 and GE2 positions are presented since there is insufficient variability in the phase-1 vs. phase-2 curves for the other two positions to account for any possible significant interaction effect.This would make sense, since again, LE1 and GE2 were the only positions to be directly impacted by the jurisdiction change implemented in phase-2. +Communication +Number of Voice Transmissions by Position & PhaseThe mean number of controller issued voice clearances by position and experiment phase is shown in Figure 11.The change in jurisdiction 6.B.2-9 between the two phases decreased the total number of voice clearances from 95.6 in phase-1 to 80.4 in phase-2 (all positions combined or overall effect of phase).However, this overall effect of phase was not statistically significant.ANOVA results did show an overall significant main effect for position (F=4.13,df=3,68, p<= 0.0094).When the position effect was broken down by phase, ANOVA results showed a significant phase-1 main effect (F=4.15, df=3,68, p <= 0.01), and a non-significant phase-2 effect.This would seem to indicate that the number of voice transmissions was more equally distributed across the positions under phase-2, as compared to phase-1. Figure 11 shows that LE1 experienced a large decrease in the number of voice clearances between the phases, from 201 to 118.Conversely, GE2 experienced a small increase in the number of voice clearances, from 40 to 62. Thus, the change in jurisdiction balanced the number of voice transmissions between the LE1 and GE2 positions, while decreasing the overall number of voice transmissions across the two positions.Under phase-2, the LE1 controller was not required to monitor both the north and south sections of the active runways.Hence, the LE1 controller was able to direct full attention to a smaller jurisdiction area.However, under the old jurisdiction (phase-1), GoSAFE required almost every aircraft on the airport to be handled by the LE1 controller.The graphical user interface (GUI) did not provide an inset map of the south side of the runways, so controllers focused on the north end of the runways, and would sometimes forget to issue clearances to aircraft on the south end of the runways.In phase-2, an inset could have partially mitigated the problem, but the workload would still remain high, thus a split in jurisdiction was warranted.The spilt in jurisdiction and the support provided by GoSAFE's new GUI under phase-2 eliminated this problem.In the absence of the decision support tool, GoSAFE, one would expect an increase in voice transmissions in phase-2 because of the split in authority over active runways.However, by using GoSAFE, the number of communications decreased and the simplification of phraseology developed by the researchers also positively impacted communications. +Voice Channel Occupancy by Position & PhaseVoice channel occupancy is defined as the percentage of radio frequency occupied relative to the total duration of the simulation run. Figure 12 shows the overall distribution of voice channel occupancy by controller position and phase.ANOVA results yielded a significant main effect of controller position (F=4.82,df=3,68, p<=0.01).When the position effect was broken down by phase, ANOVA results showed a significant phase-1 main effect (F=4.42, df=3,68, p <=0.01), and a non significant phase-2 effect.This would seem to indicate that voice channel occupancy was more equally distributed across the positions under phase-2, as compared to phase-1.Further, the means in Table 9 clearly show that LE1 experienced a large decrease, and GE2 experienced a small increase in voice channel occupancy, from phase-1 to phase-2.Thus, the change in jurisdiction balanced and lowered the amount of 6.B.2-10 communication and frequency congestion among the controller positions. +SummaryThis current research effort successfully implemented airport surface configuration changes based on the introduction of new surface automation.The changes directly impacted controller's roles and responsibilities without compromising their workload or increasing frequency congestion.In fact, these changes had the effect of decreasing both workload and frequency congestion while increasing situation awareness of local air traffic controllers.Additionally, some ground controllers experienced a relatively small increase in workload, which may have the beneficial effect of preventing tedium and vigilance decrement, often introduced by automation.Evidence of this was provided from the results of the current study, where the small increase in workload occurred concurrently with an increase in alertness and concentration (under phase-2, as compared to phase-1).All of these beneficial effects were possible due to the introduction of an automation tool (in this case, GoSAFE), allowing controllers to effectively work with an increased volume of surface traffic (relative to those handled by present day airport operations).The aviation community generally recognizes that sharing the management and control of active runways has the effect of increasing coordination and communication among controllers, and may also compromise safety [8].In the controllers' opinion the changes in areas of responsibility would not be acceptable to them in the current day or future operations due to the potential increase in communication load and a potential corresponding decrease in safety.They reported that the surface automation tool, GoSAFE, alleviated some of these concerns.The study also had some limitations.For example, due to practical and technical constraints, it was necessary to implement a randomized ANOVA, under circumstances where possible intercorrelations among the data points may have introduced some bias into the analysis.However, such bias was minimized by the experimental procedures which randomized possible individual effects.To increase the chance of further reducing such bias within the statistical analysis of data, further research with a larger sample size is recommended.Further research which examines other measures such as the controller's ability to deal with anomalies and off-nominal events is also needed to study the impact on safety.Finally, while we would expect that the introduction of GoSAFE in other airports would have similar effects as the current study, it is recommended that further research be conducted using test beds other than DFW, to gain a larger perspective on the generalizability of our findings.In summary, the results clearly indicated that the change in jurisdiction (areas of controller responsibility) and the implementation of GoSAFE had a re-distribution effect for positions LE1 and GE2, and an overall positive impact on the dependent measures of workload, situation awareness and communication, with 1.5 times the current level of traffic.Thus, introduction of new surface automation technology is likely to impact the division of roles and responsibilities between the human and automation.The increase in surface traffic had the effect of increasing the overall level of workload, while the introduction of GoSAFE in phase-1 led to an imbalance in this workload among the four controller positions.In the following phase, the same automation helped redistribute workload by re-designing the areas of responsibility for the controllers.Figure 2 .2Figure 1.Phase-1 Area of Responsibility +Figure 3 .3Figure 3. Workload by Phase & Condition +Figure 4. Phase-1 WAK Workload by Condition +Figure 6 .6Figure 6.Workload by Phase (LE1) Conversely, Figure 7 mostly illustrates just the opposite pattern of mean responses for the GE2 position, showing more phase-2 mental demand, physical demand, temporal demand, effort and frustration. +Figure 7 .7Figure 7. TLX Workload by Phase (GE2) +Figure 8 .8Figure 8. Workload by Phase and Position +Figure 10 .10Figure 10.GE2 Situation Awareness by Phase +Figure 11 .11Figure 11.Mean Number of Voice Transmissions by Position and Phase +Figure 12 .12Figure 12.Mean Percentage of Voice ChannelOccupancy by Position and Phase +Table 1 . WAK Means for Phase 11PositionMeanSDLE14.111.38LE21.841.13GE12.320.86GE21.130.34 +17Mean Responsephase2 mixed6(1=low 7=high)phase2 full54Physical2.91.92.71.2Demand 3Temporal2.81.42.51.2Demand 2Performance 5.21.15.30.9Effort 1LE13.0 LE21.3 GE13.1 GE21.2 ALLFrustration2.51.12.11.1 POSITIONSPhase1 means:Phase2 means:| mixed -full || mixed -full |Mental demand0.90.1Physical demand1.4 *0.3Temporal1.1 *0.4demandPerformance0.30.0Effort0.80.0Frustration0.40.16.B.2-5 +Table 4 . Workload Range across All Positions4TLX measurePhase-1Phase-2RangeRangeMental Demand3.31.6Physical Demand3.40.8Temporal Demand 2.91.6Performance1.51.0Effort2.60.5Frustration1.81.0 +Table 5 . Position Effects on Workload (Phase-1)5F ratio* p<0.5Mental Demand20.8*Physical Demand8.1*Temporal Demand 15.0*Performance1.7Effort11.5*Frustration5.6* +Table 6 . Position Effects on Workload (Phase-2) F ratio * p<0.56Mental Demand10.8*Physical Demand1.2Temporal Demand 3.3Performance2.1Effort0.3Frustration1.4 +Table 7 . Situation Awareness: Phase Effects77Mean Responsephase1 LE16(1=low 7=high)phase2 LE1mean Phase-1mean Phase-2F-ratio p< 0.055Instability3.12.44.4*43Variability3.63.21.82Complexity2.92.60.71Alertness Spare mental capacity Concentration4.8 5.6 4.75.2 5.8 5.41.9 0.3 7.1*i n s t a b i l i t y v a r i a b i l i t y c o m p l e x i t y a l e r t n e s s e n t a l -c a p a c i t y c o n c e n t r a t i o n d i v i s i o n -a t t e n t i o n f o q u a n t i t y f o q u a l i t y f a m i l i a r i t y i n i n s p a r e -mDivision of3.63.30.8attentionInformation5.15.31.2quantityInformation5.15.10.0qualityFamiliarity5.45.96.3* +Table 8 . Situation Awareness: Phase by Position Interaction Effects8Position:LE1LE1GE2GE2F-* p <=Phase:P1P2P1P2ratio0.05Instbty4.72.61.52.55.4*Variabty 5.43.91.93.17.0*Cmplx4.73.31.42.33.4*Alert4.95.54.55.50.8SpMC4.35.66.56.02.7*Conctn5.05.54.05.41.5DivAt5.34.02.23.03.6*InfoQan 5.05.34.65.00.1InfoQal5.15.05.25.40.2Familty5.15.85.65.80.6 +Table 9 . Mean Percentage of Voice Channel Occupancy by Position and Phase9LE1LE2GE1GE2Phase-1 22.78.86.33.5Phase-2 12.87.06.95.8 + + + + +AcknowledgementsWe acknowledge the efforts of our team that made this research successful.Our special thanks go to the developers of the concept and prototype tool Go-SAFE -Shawn Cheng, and Yonggoo Seo from Optimal Synthesis Inc.The SAIC team at NASA Ames Research Center including Ron Lehmer, Carla Ingram, Diana Carpenter, Ramesh Panda, Srba Jovic, John Walker, and Dan Wilkins deserve special acknowledgment for their significant contributions to the simulation effort.6.B.2-11 6.B.2-12 + + + + + + + + + Surface operation automation research for airport tower and flight deck automation + + VH LCheng + + 10.1109/itsc.2004.1398970 + + + Proceedings. The 7th International IEEE Conference on Intelligent Transportation Systems (IEEE Cat. No.04TH8749) + The 7th International IEEE Conference on Intelligent Transportation Systems (IEEE Cat. No.04TH8749)Washington, DC + + IEEE + 2004. October 3-5 + + + Cheng, V. (2004). Surface Operation Automation Research for Airport Tower and Flight Deck Automation. IEEE Intelligent Systems (ITSC), Washington, DC, October 3-5. + + + + + Refractive Index: The Ultimate Tool for Real-Time Monitoring of Solid-Phase Peptide Synthesis. Greening the Process + 10.1021/acs.oprd.1c00051.s001 + + 2005 + American Chemical Society (ACS) + + + COURSE) Project report + Chicago OMP Ultimate-phase Real Time Simulation Evaluation (COURSE) Project report (2005). + + + + + Development of NASA-TLX (Task Load Index): Results of Empirical and Theoretical Research + + SandraGHart + + + LowellEStaveland + + 10.1016/s0166-4115(08)62386-9 + + + Advances in Psychology + + PAHancock + + + NMeshkati + + + Elsevier + 1988 + + + + Hart, S. G. & Staveland, L. (1988). Development of the NASA Task Load ,Index (TLX): Results of empirical and theoretical research. In P. A. Hancock & N. Meshkati (eds.), Human Mental Workload (pp. 139-183). + + + + + Evaluation of Future National Airspace System Architectures + + GeorgeHunter + + + KrisRamamoorthy + + + PaulCobb + + + AlexHuang + + + MattBlake + + + AlexanderKlein + + 10.2514/6.2005-6492 + + + AIAA Modeling and Simulation Technologies Conference and Exhibit + San Francisco, California + + American Institute of Aeronautics and Astronautics + Aug. 15-18, 2005 + + + Hunter, G., Ramamoorthy, K., Cobb, P., Huang, A., Blake, M., & Klein, A. Evaluation of Future National Airspace System Architectures. AIAA Modeling and Simulation Technologies Conference and Exhibit, San Francisco, California, Aug. 15-18, 2005. + + + + + Runway Incursion Studies in NASA's Future Flight Central + + SusanLockwood + + + NancyDorighi + + + BorisRabin + + + MichaelMadson + + 10.2514/6.2002-5811 + FFC-LAXR002 + + + AIAA's Aircraft Technology, Integration, and Operations (ATIO) 2002 Technical Forum + + American Institute of Aeronautics and Astronautics + 2001 + + + Los Angeles International Airport Runway Incursion Study. Phase II Alternative simulation (2001). Future Flight Central, FFC-LAXR002. + + + + + Developing a Decision Support Interface for Surface Domain Air Traffic Controllers + + LynneMartin + + + SavitaVerma + + + DeborahBallinger + + + VictorCheng + + 10.1037/e577822012-013 + + + HFES proceedings + + 2007. October 2007 + American Psychological Association (APA) + Baltimore + + + Martin, L.; Verma, S.; Cheng, V. & Ballinger, D. (2007). Developing a decision support tool interface for surface domain controllers. HFES proceedings, October 2007, Baltimore. + + + + + Distributed System Architecture in VAST-RT for Real-Time Airspace Simulation + + RonaldLehmer + + + ScottMalsom + + 10.2514/6.2004-5436 + version 0.1 + + + AIAA Modeling and Simulation Technologies Conference and Exhibit + + American Institute of Aeronautics and Astronautics + 2005 + + + Northrop Grumman Information Technology (2005). Virtual Airspace Simulation Technology - Real-Time: VAST RT Data processing system requirements and specification. VASTRT DataProc-NASA-DID-P100, version 0.1 + + + + + A good choice: Why the National Aerospace Laboratory NLR in The Netherlands purchased 18 Task ba... + + JWSmeltink + + + MJSoomer + + + PRDe Waal + + + RDVan Der Mei + + 10.2514/6.2023-4013.vid + + 2004 + American Institute of Aeronautics and Astronautics (AIAA) + + + National Aerospace Laboratory NLR + + + Technical Report + Smeltink, J.W. ; Soomer, M.J. ; de Waal, P.R. & van der Mei, R.D. (2004). An Optimisation Model for Airport Taxi Scheduling. Technical Report, National Aerospace Laboratory NLR. + + + + + Scheduling aircraft landings using airlines’ preferences + + MJSoomer + + + GJFranx + + 10.1016/j.ejor.2007.06.017 + + + European Journal of Operational Research + European Journal of Operational Research + 0377-2217 + + 190 + 1 + + 2007 + Elsevier BV + + + Soomer, M.J., & Franx, G.J. (2007) Scheduling Aircraft Landings Using Airlines' Preferences. European Journal of Operational Research, In Press. + + + + + TRB Special Report 314: Federal Aviation Administration's Approach for Determining Future Air Traffic Controller Staffing Needs + 10.17226/18824 + No. DOT/FAA/CT-TN84/24 + + 1985 + National Academies Press + Atlantic City, NJ + + + Report + Stein, E.S. (1985). Air traffic controller workload: An examination of workload probe. (Report No. DOT/FAA/CT-TN84/24). Atlantic City, NJ: Federal Aviation Administration Technical Center. + + + + + Situational Awareness Rating Technique (Sart): The Development of a Tool for Aircrew Systems Design + + RMTaylor + + 10.4324/9781315087924-8 + + + Situational Awareness + + Routledge + 1989. 2007 + 478 + + + + Situational Awareness Rating Technique (SART): the Development of a Tool for Aircrew Systems Design + Taylor, R.M. (1989). Situational Awareness Rating Technique (SART): the Development of a Tool for Aircrew Systems Design. In Proceedings of the AGARD AMP Symposium on Situational Awareness in Aerospace Operations, CP478. Email Addresses Savita Verma: saverma@mail.arc.nasa.gov Thomas Kozon: tkozon@mail.arc.nasa.gov 26th Digital Avionics Systems Conference October 21, 2007 + + + + + + diff --git a/file761.txt b/file761.txt new file mode 100644 index 0000000000000000000000000000000000000000..f48144a9f0a684c4d87d2158eab6fbd364f6d896 --- /dev/null +++ b/file761.txt @@ -0,0 +1,510 @@ + + + + +LIST OF FIGURES +LIST OF TABLES +INTRODUCTIONDemand in the future air-transportation-system concept is expected to double or triple by 2025 (ref.1).Increasing airport arrival rates will help meet the growing demand that could be met with additional runways, but the expansion airports are met with environmental challenges for the surrounding communities when using current standards and procedures.Therefore, changes to airport operations can improve airport capacity without adding runways.Building additional runways between current ones, or moving them closer, is a potential solution to meeting the increasing demand, as addressed by the Terminal Area Capacity Enhancing Concept (TACEC).TACEC requires robust technologies and procedures that need to be tested such that operations are not compromised under instrument meteorological conditions.The reduction of runway spacing for independent simultaneous operations dramatically exacerbates the criticality of wake vortex incursion and the calculation of a safe and proper breakout maneuver.The study presented here developed guidelines for such operations by performing a real-time, human-in-the-loop simulation using precision navigation, autopilot-flown approaches, with the pilot monitoring aircraft spacing and the wake vortex safe zone during the approach. +BACKGROUNDThe Federal Aviation Administration (FAA) has successfully conducted independent approaches to parallel runways for over 40 years using the instrument-landing-system (ILS) navigation and terminal radar monitoring (ref.2).The simultaneous approaches that utilize standard radar are conducted on parallel runways that are separated by at least 4300 ft.It is possible to conduct independent approaches on runways separated by as little as 3000 ft, but it requires a Precision Runway Monitor (PRM) with an update rate of 1 sec.The separation standards between the aircraft on these parallel approaches are 1000-ft vertical separation.Additionally, a 2000-ft-wide "notransgression zone" (NTZ) was placed equidistant from the centerlines of the approach paths on the two parallel runways.Some airports, like San Francisco International Airport (SFO), can support approximately 60 landings per hour on its two parallel runways that are 750 ft apart by using simultaneous offset instrument approaches (SOIA) (ref.3).SOIA approaches require the trailing aircraft in the paired approach to obtain a visual sighting of the lead aircraft with at least a 1200-ft ceiling with 4-nmi visibility.As weather degrades, the current navigation and surveillance system, as well as the existing procedures, lack the accuracy to support SOIA approaches, reducing the landing rate to half the visual-flight-rules (VFR) capacity.Several researchers have investigated alternative procedures for very-closely-spaced parallel runway (VSCPR) operations.Studies have focused on the technologies required to enable the VCSPR operations.Several different requirements have been identified from these studies, such as cockpit displays, collision-prevention systems, and precision navigation, communication, and surveillance systems (refs. 6, 7, and 8).Another critical component that is necessary for the safe execution of VSCPR procedures is the ability to predict the wake vortices for the aircraft nearby and provide wake information to the affected aircraft.Previous research has also evaluated procedures for VCSPR approaches, but most of them have used fast-time simulation to investigate the performance of the procedures.Pritchett & Landry (ref. 6) identified the various parameters related to VCSPR operations, such as separation responsibility and different separation and spacing objectives between the paired aircraft.Few human-in-the-loop studies have been conducted for VCSPR operations.A study to investigate pilot response to VCSPR operations for the Airborne Information for Lateral Separation (AILS) concept is one such example.NASA developed the AILS concept to further examine independent parallel runway operations for runways as close as 2500 ft.The concept requires technologies that enable the use of precise navigation and surveillance data.Automation is presumed to detect blunders or situations that may require the aircraft to perform a breakout maneuver (ref.4).The AILS experiment was designed to study three variables: 1) intruder geometry, 2) runway separation (3400 or 2500 ft), and 3) flight control mode (autopilot versus manual prior to the warning for breakout).The dependent variables were pilot reaction time and miss-distance in off-nominal situations that required the pilot to perform an escape maneuver.The study found that pilot reaction time to detect and perform breakout maneuvers was not affected by runway separation.Across all conditions the average pilot reaction time was 1.11 sec, with a standard deviation of 0.45 sec.The experiment found a statistically significant effect for the flight control mode, with autopilot use prior to the emergency escape maneuver leading to longer reaction times.The current study is different from the AILS experiment because it considers wake, and dynamically generates breakout maneuver.TACEC aims to fly paired approaches on runways that are 750 ft apart in instrument meteorological conditions (ref.5).A ground-based processor will identify aircraft that could be paired approximately 30 minutes from the terminal boundary.The aircraft are selected for pairing based on several parameters, such as aircraft performance, arrival direction, relative timing criteria, and aircraft size-of-wake considerations.The ground-based processor then assigns four-dimensional (4-D) trajectories to the aircraft in the pair.It is assumed that all aircraft will use differential Global Positioning System (GPS)-enabled and high-precision 4-D flight-management-system (FMS) capabilities for the execution of these trajectories.Enhanced cockpit displays that depict both traffic and wake information will also be a requirement for these operations.These operations guarantee a wake-free region by positioning a following aircraft sufficiently close to the lead aircraft on the parallel approach so that the vortex does not have time to spread into the path of the following aircraft (fig.1).When the paired aircraft reach 12 nmi from the airport, their autopilot systems become "coupled" via a longitudinal spacing control mode.In this mode, a speed-control algorithm on the following aircraft uses state data broadcast by the lead aircraft to precisely maintain the separation between the two aircraft until touchdown.The TACEC concept envisions nearly completely automated approaches and landings.Advanced 4-D-capable FMSs execute the assigned 4-D trajectories, and an integrated coupling system maintains safe spacing between the paired aircraft during the final 12 nmi of flight.The pilot is responsible for approving the engagement of coupling and monitoring safe progress using cockpit displays of traffic information (CDTIs) that include display of predicted wake location, alerts for wake hazards, display of ownership and traffic assigned trajectories, and indication of navigation performance relative to assigned trajectories.Little data exist regarding the use of VCSPR technologies and procedures.The objective of the current study was to develop guidelines for the procedures defined by the TACEC using a humanin-the-loop simulation study.Thus the objective of this simulation also included exploring the usefulness and usability of the cockpit displays and procedures associated with this new concept. +HUMAN-IN-THE-LOOP EXPERIMENT +Objectives and HypothesesThe objective of the study was to assess and develop preliminary guidelines for the procedures for the Terminal Area Capacity Enhancing Concept (TACEC) by performing a real-time, human-inthe-loop simulation to conduct human-factors studies of prototype displays and to investigate procedures.This process accelerates development of a far-future (2025) concept through early implementation.The effort required development of a simulated airport with parallel runways 750 ft apart, and designing the airspace around this generic airport to facilitate very-closely-spaced parallel runway (VCSPR) approaches.The concept was implemented in the Advanced Cockpit Flight Simulator (ACFS) by integrating displays that depict wake and traffic information.The participants of the study flew the simulator under various conditions and provided feedback on changes in roles and responsibilities, new procedures, and their opinions about the concept.The report describes all aspects of the study, starting with the approach.Appendices A through D give pilot schedule, demographic, and survey information, appendix E gives observer feedback, appendix F lists questions for group discussion, and appendix G discusses the layout of the airport used in the study. +Approach Test FacilityThe human-in-the-loop study conducted to assess the paired TACEC approaches used the ACFS, a full-mission simulator that resides in the Crew Vehicle Systems Research Facility (CVSRF).The ACFS simulates a generic commercial transport aircraft, and it can be reconfigured to represent future aircraft.Currently, the ACFS can simulate two aerodynamic aircraft models, narrow-body transport aircraft (similar to a Boeing 757) and a C-17 transport.The simulator, as it stands, includes fly-by-wire flight controls, touch controls, touch-sensitive electronic checklists, schematics of aircraft systems, a customizable FMS, and graphical flight displays.The cab is mounted on a six-degree-of-freedom synergistic motion system and uses side stick controllers for aircraft control in the pitch and roll axes.The simulator is run from Silicon Graphics, Inc, (SGI) computers, which provide the simulator flight systems and programmable flight displays.In this study, the CDTI described in the "Background" section was integrated with the flight-display systems in the cockpit.The ACFS motion capabilities were also used for the study.The visual systems in the ACFS offer a 180-degree horizontal and a 40-degree vertical field of view.The ACFS visual databases can depict as many as nine airports (SFO, Los Angeles International Airport (LAX), John F. Kennedy International Airport (JFK), Denver International Airport (DEN), Dallas Fort Worth International Airport (DFW), Sea-Tac Airport (SEA), Hartsfield-Atlanta International Airport (ATL), O'Hare International Airport (ORD), and Boston/General Edward Lawrence Logan International Airport (BOS).For the VCSPR study, the SRT airport visual database was created, which is a modification of the DFW airport, as described in appendix G on the DFW layout. +Airport and Airspace DesignThe airport and airspace used to investigate procedures for the TACEC concept used a fictitious airport that was based on the current DFW's layout and operations.The airport used for the simulation was referred to as "KSRT."The simulation focused on studying TACEC approaches to very closely spaced parallel runways.Since a south air traffic flow was used for the simulation scenarios, the SRT airport utilized only runways 18R, 18L, 17R, and 17C (renamed to 17L).All four runways were assumed to be equipped to a Category IIIB (CAT-IIIB) level.Both 18R and 17L (see figure 2) were moved to within 750 ft of their inboard runways, 18L and 17R, respectively, requiring an adjustment of 464 ft from their current DFW positions.The layout of the DFW is described in detail in appendix G. +TACEC ProceduresThe TACEC concept calls for TACEC-assigned 4-D arrival trajectories for both of the aircraft to be paired at meter fixes located near the edge of the terminal airspace, normally 40-60 nmi from the airport (ref.5).Flights in the simulation began 25 nmi from the airport, assuming they were already paired.Routes to the KSRT airport included approach and departure routes and procedures similar to those for DFW airport.This study focused upon arrivals; no departures were included. +Arrival Traffic FlowSouth flow of traffic was simulated for the generic airport KSRT.All of the four runways (18R, 18L, 17R, 17L) were used for arrival operations.The concept allows for any aircraft arriving from any of the four arrival meter fixes (NE, NW, SE, and SW) to be paired for a simultaneous parallel landing, based on aircraft characteristics and relative timing criteria.Paired aircraft must fly their assigned 4-D trajectories with a high level of accuracy in order to meet timing constraints at the coupling point and ensure wake safety throughout the approach.The 4-D trajectories were carefully designed to provide safe wake-avoiding routes from the arrival meter fixes to the runways.Each route consisted of three segments, and each one of the first segments provided vortex-free 4-D routes extending from the meter fix to the coupling point at 12 nmi from the runway.The second segment began at the coupling point and ended 2 nmi from the runway.During the second segment, one route was straight in, aligned with the runway centerline, while the other was at a 6-degree slew angle from the straight-in route (see fig. 3).At the coupling point, the aircraft were laterally separated by slightly more than 1 nmi.Each of the final segments was aligned with the runway centerlines, extended 2 nmi from the runway threshold, and was about 600 ft above ground level (AGL) in order to provide a straight-in flight path to touchdown.Once the aircraft reached the coupling point, the following aircraft precisely maintained spacing behind the lead aircraft in order to avoid the wake of the lead aircraft.This operation was accomplished by an automated speed-control algorithm on board the following aircraft that maintained the assigned time-based spacing relative to the lead based on state information broadcasted via Automatic Dependent Surveillance -Broadcast (ADS-B) by the lead aircraft.Figure 3 shows the geometry of the final approach portion of the arrivals (i.e., the final 12 nmi before landing. +Coupling Point +Traffic ScenarioThe traffic scenario had two aircraft in the simulation: the following aircraft in the pair, as represented by the ACFS, and the lead aircraft, which was recorded or scripted for this study.The ownship was always the following aircraft, and the recorded one was always the leader aircraft in the closely spaced parallel runway approach.The leader aircraft was a Boeing 747 heavy aircraft representing Japan Airlines (JAL).Based on the wind condition, the ownship was either on the slewed approach landing on runway 18R or on the straight-in approach landing on runway 18L. +Cockpit Display of Traffic and Wake InformationThe primary purpose of the displays used for the TACEC evaluation was to provide the flight crews with information to ensure that adequate separation was being maintained with the lead aircraft and its hazardous wake area.While not evaluated in the present simulation, the displays also provide "breakout" annunciation and guidance if adequate separation is not maintained with the lead aircraft or its wake.The primary flight display (PFD) and the navigation display (ND) are modifications of standard current-generation transport flight displays with added lead-aircraft position and wake information.Figure 4 shows the PFD on the straight-in parallel final at 532-ft radar altitude, while figure 5 shows the ND for the same location.Lateral spacing of the flight paths at this part of the approach was 750 ft.The displays are adaptations of those previously developed by Hardy and Lewis (ref. 8). +Lead-Aircraft PositionThe +Hazardous-Wake-Area DepictionThe shaded white area on the ND and the wake frames on the PFD depict the hazardous wake area.This area was defined as that volume of airspace such that if the apex or center of gravity (cg) of the following aircraft (simulator) remains outside the wake area, no noticeable wake activity would be detected.This area was predicted in real time from aircraft characteristics and onboard sensors of crosswind and atmospheric turbulence.The prediction algorithms were conservative to account for model and sensor errors (ref.9).The shaded area on the ND and the wake frames on the PFD turn amber if the cg of the following aircraft moves to within one wingspan of the hazardous area, and they turn red if its cg penetrates it. +Predictor DotsFive 2-second predictor dots, for a total of 10 seconds, were added to the ND for both aircraft (see slightly to the right of the nominal path for the simulator in fig.5) and also were presented on the PFD (aligned with the position icon of the lead aircraft).These dots show flight path trend information to help the pilot determine the future location of the aircraft. +Longitudinal Situation IndicatorTo maintain the position of the aircraft in the "safe" zone, as shown in figure 4, a longitudinal situation indicator (LSI) was added.The LSI is flagged on the ND and shows the nominal location (in this case 5 seconds behind the lead aircraft) that the auto-throttle is attempting to keep.For this example, the simulator is approximately 400 ft behind its nominal location.The same LSI information is shown on the deviation scale added on the left side of the PFD (fig.4). +Display ScalingA conventional PFD has a field of view of about 40 deg.To be able to see the lead aircraft position and wake information, this field was increased to 80 deg.This increase decreases the resolution of the display, but with future larger display hardware it may not be objectionable.A conventional ND has a maximum zoom-in capability of a 10-mi.range scale.To have adequate resolution for this task, the maximum zoom-in range scale is 0.25 nmi.The display zoomed in increments of 10-, 5-, 2-, 1-, and 0.5-nmi scales. +Experimental Matrix and Independent VariablesThe three variables examined in the study were visibility conditions, direction of the wind, and the distance between the lead and follower aircraft.The visibility conditions were a clear day, or Category IIIB.The study aimed at exploring an adverse cross wind on the follower (ownship), thus the direction of winds was coupled with the follower (ownship) landing on the left or right runway (18L or 18R runways in this study), as shown in figure 5.The approach to runway 18R is referred to as the slewed approach, and the one to 18L is the straight-in approach.The third variable examined in the study was the distance between the lead and follower aircraft at initialization points, which was either 10 or 5 sec.The matrix for the study is shown in table 1, where the gap between the lead and trailing aircraft was 10 sec, and table 2, where the gap between the lead and trailing aircraft was 5 sec.In addition, an additional run led to the consideration of a potential escape maneuver due to the location of the wake and traffic.This situation was observed by the pilots for purposes of preliminary discussion, and the identification of procedures for off-nominal situations. +Dependent VariablesThe dependent variables collected during the study included subjective data on situation awareness, comparison of features provided by the displays, and other subjective questions asked about the usefulness and usability of the displays in a post-interaction survey. +ParticipantsThe participants of the study were three retired pilots from commercial airlines; all of them had experience with glass cockpits and some experience flying SOIA approaches in San Francisco.Their mean age was 65 years, and their mean total years of experience as a pilot was about 40 years.They had an average about 16,500 hours of flying.Their average number of years since retirement was 6.5 years. +Test ScenariosThe study was run for three days, with one pilot participating each day.At the beginning of the day, the pilot was familiarized with the project, the concept, and the new displays in the cockpit.Next, the pilot was taken to the ACFS, where the pilot received a demonstration of the simulator, and more hands-on training on the CDTI and related procedures.The schedule for the study is included in appendix A. The schedule included the eight runs specified earlier (see tables 1 and 2 and appendix A) and an off-nominal escape maneuver run mentioned in the "Observer Notes and discussion" section. +ProtocolThe procedures for VCSPR were being explored in this study, so each pilot flew the ACFS as a captain.The role of the pilot, in general, was to fly in autopilot mode, and monitor the displays to check separation with the lead aircraft and wake.At the coupling point the pilots heard a chime, saw the acknowledgement button light up, and a message on the lower Engine Indicating and Crew Alerting System (EICAS) appeared that read "TACEC Coupling."At this point the pilots pressed the acknowledgement button, and continued to monitor the separation between the two aircraft.The traffic scenario in the next section describes the tow aircraft that were simulated for the study. +Tools Used for Data CollectionSeveral tools were used for collecting subjective data from the pilots.All participants completed a demographic survey before the simulation runs were conducted.It collected information about the pilots such as their age, experience as a pilot, and number of hours flying different aircraft types, any experience with SOIA, and experience using personal computers.Each pilot was asked to complete a Post Interaction Survey at the end of all the runs.It collected information on the pilot-rated usefulness and usability of the displays.Similarly, a feature comparison survey was administered at the end of all of the runs.The pilots had the opportunity to rate the importance of different features in the displays on a scale of 1 to 5, where 1 was equivalent to "very unimportant" and 5 was equivalent to "very important."Pilots also completed the Situational Awareness Rating Technique (SART) (ref.10).The SART gathers a participant's rating of his/her situational awareness (SA) for the preceding period of time on 10 different scales.Each scale has 7 points, with the end points representing the opposite ends of the construct.Participants circled the point on the scale that most closely represented their experienced level of SA.The 10 SART ratings were gathered from every participant at the end of each run-a total of 8 ratings per participant were collected. +RESULTSThis section reports results that focus on the data captured by the tools mentioned in the above paragraph .Results of the post-interaction survey, feature comparison, situation awareness, and observer notes are described in the following section. +Post-Interaction SurveyThe post-interaction survey was administered to each pilot at the end of the eight trial runs.Since the questions were administered after the completion of the simulation, there were no distinctions among the different experimental conditions.The questions instead queried the participants about the general experiences of using VCSPR procedures and tools.Also, because of low statistical power for testing, tests for significance were not conducted.The pilots responded to the question on the overall utility of the displays for VCSPR approaches as highly useful (average of 3, on a scale of 1 to 5).The questions focused on the ease of using the displays to derive information for some of the functions handled by the pilots using the displays.The pilots found that the overall level of ease for extracting information from the displays was very high (M = 5 on a scale of 1 to 5, where 1 was very hard and 5 was very easy).A detailed analyses of ease of deriving information is shown in figure 6.In general, on average the pilots found that the displays provided enough information, and that it was relatively easy to extract for most of the functions.The mean value was greater than or equal to 4 for all functions except flying in low visibility.During the group discussions, the pilots mentioned that they would like to see the tool deployed in clear weather conditions for a period of time to allow the pilots to develop enough trust in the automation before it is used for flying under Category IIIB visibility conditions.They felt that this trust could be improved with more familiarity and use of this type of automation.Also, the pilots mentioned that deriving information about wake characteristics was very easy in this simulation (M = 5).One can infer that the pilots were able to effectively monitor separation of the aircraft from the wake.All the pilots reported that they were able to effectively monitor the lead aircraft, mostly by using the ND.Also, none of the pilots were confused by the interface.On the ability to zoom on the ND, the pilots reported that having a separate zoom capability for the pilot flying and pilot not flying will enable them to maintain both a strategic and tactical view at the same time.The ND zoom capability was handled by a toggle switch on the center console and was available as a function ) for ease of deriving information from the displays.only to the pilot flying.The pilots were asked which aspects of the concept they liked the best, and which aspects they liked the least.The pilots liked the system and the new displays because they will greatly enhance safety in today's air traffic environment.They also agreed that the system will enhance capacity at the airports.In contrast, the pilots repeated that this automation needs to be implemented in good visibility conditions before the pilots will trust the automation for use during instrument meteorological conditions (IMC).They were all concerned about procedures for breakout maneuvers, and definition of standards for proximity.They also wanted more flexibility with maneuvering throttles without disengaging the auto throttles.One pilot also mentioned that all procedures, including airspeed requirements between the coupled aircraft, must be agreed upon by the pilots and controllers involved in the procedures prior to flying.The pilots were also asked to rate some statements regarding the concept and displays (fig.7).They all agreed that automation is required for VCSPR approaches, and that there was little confusion about the displays.They responded with above-average ratings for ease of monitoring separation from the lead aircraft.The participants also found the wake information on the ND and the predictor dots very useful, and they valued being able to visualize the trajectory of the lead aircraft.They rated their level of confidence in the concept as average, and they did not indicate concern in their responses about the role of the pilot in this concept. +Feature ComparisonThe participants were asked to rate the various features on the displays provided to them in the simulator.They rated most features as having above-average importance (ranging from 4 to 4.5 on a scale of 1 to 5) except the lead aircraft and the LSI on the PFD.Those were rated at an average of 3.5 on a scale of 1 to 5, where the higher number indicates higher level of importance.The LSI on the ND was not always visible, and all participants complained about not being able to visually track the LSI because it was hidden under the solid white icon of the aircraft.The LSI on the PFD provided the information about the actual position versus expected position of the simulator in terms of distance, whereas the LSI on the ND provided temporal information as referenced by the 2-sec predictor dots.Despite its poor visibility at certain times, most pilots preferred the LSI on the ND.The predictor dots on the lead aircraft were considered to have an average level of importance, because the pilots always flew the follower aircraft in the approach, and they were concerned with their own trajectory predictions to monitor separation from the lead aircraft and its wake.Similarly, the feature "out-of-the-window visibility" received a 3.5 rating, and the acknowledgement button used for accepting the coupling between the paired aircraft received a 2.6 average rating.During the group discussion, the pilots suggested that pressing the acknowledgment button should arm the coupling of the two aircraft, before they are actually at the coupling point, to keep it consistent with other standard displays.The pilots also mentioned that the flight-mode annunciation should have a visual indicator that is white, depicting that the system is armed before coupling.Eventually it should turn green when actual coupling occurs, at the coupling point.In the present experimental setup, the acknowledgement button changed the flight-management-system (FMS) annunciation to "coupled" and did not give the pilots a chance to "arm."This situation created some confusion and led to the comments the pilots made.Among other concerns and suggestions for improving the design of the system, some pilots had difficulty with interpreting the wake depiction and monitoring the lead aircraft on the PFD.Other pilots felt that when the aircraft starts deviating from its longitudinal position, the procedure should allow for the pilot to adjust the throttles or speed without disengaging the autopilot. +Situational AwarenessThe situational awareness questionnaire, SART was administered to the pilots after every simulation run.They rated 10 SART elements on a scale of 1 to 7, where 1 is "low" and 7 is "high."Thus the data have been analyzed for all the conditions for each of the three pilots.Because of low statistical power for testing, significance tests were not calculated.The SA ratings have been depicted on a line graph to enable better trend comparisons for the conditions.Figure 8 shows that the SA trends for the different sub-elements are the same for the aircraft starting with 10-or 5-sec temporal separation between them.The pilots did not feel that any of these situations were unstable, and level of variability and complexity was similar in the two conditions.In the group discussions, the pilots mentioned that they preferred their aircraft to be ahead rather than behind on the LSI because being behind increased the chances of the aircraft getting into the wake zone and out of the safe zone.Pilots' responses on SA for the simulator flying on the straight-in path (landing on 18L) or on the slewed path (landing on 18R) (fig.9) show similar trends.The pilots considered the slewed path slightly more unstable, variable, and complex, but they also felt that a higher level of concentration and familiarity was required with the situation.The SA responses for the visibility condition (fig.10) showed that the pilots experienced similar levels of awareness in the clear versus poor visibility condition.In general, they felt that the poor visibility condition was slightly more variable, unstable, and complex.The pilots required slightly more alertness, and they had slightly less spare mental capacity in the poor visibility condition as compared to clear visibility condition.The information quality, information quantity, and familiarity with the situation were about the same for both of the visibility conditions.Figure 10.SA responses on clear day vs.Category IIIB visibility. +Observer Notes and Group DiscussionsThe observer data yielded some interesting findings.Comments during and after the simulation runs from the three participants pertained to issues related to the tools and procedures for closely spaced parallel approaches, wake avoidance, and nonnormal events.In addition, many comments were associated with the interface of the concept elements, in particular the alerting and display features.The three pilot participants had several comments about what they perceived were the critical aspects of the closely spaced parallel approach concept as it was represented in this study.Pilots felt the high degree of automation required for the closely spaced tasks was necessary for the precision of the procedure; however, they all expressed the need for some opportunity to intervene or "fine tune" the automation.For example, the ability to manually adjust the speed was recommended by two of the participants.In four of the eight scenarios, pilot participants flew these procedures with visibility at the KSRT airport down to about 600 ft of runway visual range (RVR).Another opinion that had general consensus was that flying these types of closely spaced procedures had a higher risk in these low-visibility surface environments.The comments indicated that although the pilots understood that automation tools would be necessary for navigation guidance and the avoidance of wake vortices, they preferred attaining a visual of the other aircraft to detect any cues that may indicate wake-vortex threat or the threat of a possible unexpected escape maneuver.Pilots also helped identify factors necessary to create and fly an escape maneuver such as traffic, terrain and rest of the airspace.The other four scenarios were in clear weather, and were generally found to be more acceptable conditions for the approaches.The pilot participants had many comments about the display of the wake information.In general, they found the wake depiction and the display locations acceptable.They preferred wake depiction on the ND versus the PFD.One pilot stated that it took him some time to understand wake on the PFD, raising the issue of the limited training the pilots received for this simulation.As the previous comments indicated, there were some concerns about the ability to predict wake responses during low-visibility conditions.In addition, all three pilots stated that they did not fully understand the nature of wake characteristics, and how these characteristics may impact their own aircraft in closely spaced parallel approaches like those flown in our scenarios.They welcomed aircraft automation that provided information on wake behaviors and their impact on these procedures. +CONCLUSIONSThis study investigated a concept that incorporates wake information and new technologies to allow for the use of very-closely-spaced parallel runways in all-weather conditions.The airport and 25 nmi of surrounding airspace were created and simulated as a part of this effort.A highfidelity full-motion simulator with the emulation of a four-dimensional (4-D) flight management system (FMS) was used to implement the concept, and several displays were enhanced to enable simultaneous approaches.The pilots provided feedback through their responses to the questionnaires and debriefings.The three pilots had similar results, and their suggestions were consistent.In general, they were marginally more comfortable with very-closely-spaced parallel runway (VCSPR) approaches and automation in visual meteorological conditions (VMC) rather than Category IIIB visibility conditions, even though their situational awareness (SA) ratings showed similar responses for both conditions.In addition, they indicated that they preferred 10-versus 5-sec spacing between the lead and follower aircraft.The participants felt it was important for them to be able to deploy gear and flaps manually, and influence speed and throttles without disengaging the autopilot.All the pilots were concerned about potential breakout procedures, and they all think automation will play a large role in the determination of the procedures, with direct involvement of the air traffic controller necessary for safe procedures. +FUTURE WORKThe study provides future research ideas and guidelines for developing procedures for VCSPR.Future research efforts by NASA and Raytheon could examine the safety and viability of the procedures and technologies associated with breakout or escape maneuvers under conditions where a simultaneous approach needs to be abandoned.In addition, the representation of more airport traffic and structures are included so that the implications of surrounding constraints could be explored.The possibility of providing more flexibility in the system where pilots could, for example, deploy gears or use throttles for speed control without disengaging the autopilot could also be explored.The TACEC concept calls for TACEC-assigned four-dimensional (4-D) arrival trajectories to begin at meter fixes located near the edge of the terminal airspace, normally 40-60 nmi from the airport.Flights in the VAST TACEC simulation began 25 nmi from the airport.Routes in the SRT airspace were designed to work both with and without TACEC tools and procedures.In order to facilitate a comprehensive design, an effort was made to reuse as much of the existing DFW traffic-flow operations as possible, including: STARs, SIDs, Approach Plates, Arrival Meter Fixes, Departure Meter Fixes, and Standard Operating Practices.Figure 1 .1Figure 1.Wake-safe zone for following aircraft................................................................................ Figure 2. SRT airport diagram........................................................................................................... Figure 3. Final approach geometry for TACEC.................................................................................Figure 4. Primary flight display on straight-in parallel final.............................................................. Figure 5. Navigation display on straight-in parallel final..................................................................Figure 6. Ratings (mean and Standard Error (SE)) for ease of deriving information from the displays............................................................................................................... Figure 7. Pilots' subjective ratings (mean and SE) on statements regarding the concept and displays......................................................................................................... Figure 8. SA responses for 10-vs.5-sec distance between the two aircraft.................................... Figure 9. SA responses for aircraft on straight-in vs. slewed approach........................................... Figure 10.SA responses on clear day vs.Category IIIB visibility..................................................... +Figure 4 .4Figure 1.Wake-safe zone for following aircraft................................................................................ Figure 2. SRT airport diagram........................................................................................................... Figure 3. Final approach geometry for TACEC.................................................................................Figure 4. Primary flight display on straight-in parallel final.............................................................. Figure 5. Navigation display on straight-in parallel final..................................................................Figure 6. Ratings (mean and Standard Error (SE)) for ease of deriving information from the displays............................................................................................................... Figure 7. Pilots' subjective ratings (mean and SE) on statements regarding the concept and displays......................................................................................................... Figure 8. SA responses for 10-vs.5-sec distance between the two aircraft.................................... Figure 9. SA responses for aircraft on straight-in vs. slewed approach........................................... Figure 10.SA responses on clear day vs.Category IIIB visibility..................................................... +Figure 6 .6Figure 1.Wake-safe zone for following aircraft................................................................................ Figure 2. SRT airport diagram........................................................................................................... Figure 3. Final approach geometry for TACEC.................................................................................Figure 4. Primary flight display on straight-in parallel final.............................................................. Figure 5. Navigation display on straight-in parallel final..................................................................Figure 6. Ratings (mean and Standard Error (SE)) for ease of deriving information from the displays............................................................................................................... Figure 7. Pilots' subjective ratings (mean and SE) on statements regarding the concept and displays......................................................................................................... Figure 8. SA responses for 10-vs.5-sec distance between the two aircraft.................................... Figure 9. SA responses for aircraft on straight-in vs. slewed approach........................................... Figure 10.SA responses on clear day vs.Category IIIB visibility..................................................... +Figure G- 1 .1Figure G-1.Current DFW Airport layout........................................................................................... Figure G-2.SRT Airport layout.......................................................................................................... +Figure 1 .1Figure 1.Wake-safe zone for following aircraft. +Figure 2 .2Figure 2. SRT airport diagram. +Figure 3 .3Figure 3. Final approach geometry for TACEC. +position of the simulator was shown on the ND with the conventional triangular icon (solid) at the lower center of the ND.The lead-aircraft position was shown with the open icon at the upper left of the ND.The triangular lead aircraft with the same perspective was shown on the PFD at the left of the display.With augmented GPS navigation, it was assumed that position information was known, with ADS-B to be within a few ft. +Figure 4 .4Figure 4.Primary flight display on straight-in parallel final. +Figure 5 .5Figure 5. Navigation display on straight-in parallel final. +Figure 6 .6Figure 6.Ratings (mean and Standard Error (SE)) for ease of deriving information from the displays. +Figure 7 .7Figure 7. Pilots' subjective ratings (mean and SE) on statements regarding the concept and displays. +e f u l n e s s p r e d i c t o r d o t s c o n c e r n e d a b o u t r o l e o f p i l o t v a l u e l e a d ' s t r a j e c t o r y h a v e c o n f i d e n c e i n f u t u r e o f c o n c e p t +Figure 8 .8Figure 8. SA responses for 10-vs.5-sec distance between the two aircraft. +Figure 9 .9Figure 9. SA responses for aircraft on straight-in vs. slewed approach. +Table 1 .1Matrix where the Distance Between the Lead and Trailing Aircraft is 10 sec ................. +Table 2 .2Distance4-Dfour-dimensionalACFSAdvanced Cockpit Flight SimulatorADS-BAutomatic Dependent Surveillance-BroadcastAGLabove ground levelAILSAirborne Information for Lateral SeparationATLHartsfield-Atlanta International AirportBOSBoston/General Edward Lawrence Logan International Airportcgcenter of gravityCDTIcockpit displays of traffic informationC-LNAVCoupled Lateral NavigationC-SPDCoupled SpeedC-VNAVCoupled Vertical NavigationCVSRFCrew Vehicle Systems Research FacilityDENDenver International AirportDFWDallas Fort Worth International AirportEICASEngine Indicating and Crew Alerting SystemFAAFederal Aviation AdministrationFMSflight management systemGPSGlobal Positioning SystemILSinstrument landing systemJALJapan AirlinesJFKJohn F. Kennedy International AirportKSRTfictitious airport used for the simulation; based on current DFW layout and operationsLAXLos Angeles International AirportLSIlongitudinal situation indicatorNDnavigation displaynminautical milesNTZno-transgression zoneORDO'Hare International AirportPFDprimary flight displaybetween the lead and trailing aircraft is 5 sec .................................................... Table G-1.SRT Runway Positions .................................................................................................... Table G-2.Runway Configurations by Procedure ............................................................................ Table G-3.Runway Configurations by Operation (A = Arrival, D = Departure) ............................ v NOMENCLATURE +TABLE 11. MATRIX WHERE THE DISTANCE BETWEEN THE LEAD AND TRAILING AIRCRAFTIS 10 SECStraight Approach (18L)Slewed Approach (18R)Clear dayRun 1Run 4Low visibility (Category-III b)Run 2Run 3 +TABLE 2 .2DISTANCE BETWEEN THE LEAD AND TRAILING AIRCRAFT IS 5 SECStraight Approach (18L)Slewed Approach (18R)Clear dayRun 5Run 6Low visibility (Category-III b)Run 8Run 7 + Ames Research Center, Moffett Field, CA, 94035-0001. + Raytheon Company, Marlborough, MA 01752. + Science Application International Corporation (SAIC), Ames Research Center, Moffett Field, CA 94035-0001. + Perot Systems, Ames Research Center, Moffett Field, CA 94035-0001. + + + + +AcknowledgmentsFunded by NASA's ARMD, Airspace Systems Program, NextGen-Airportal Project.Team members: Savita Verma, Sandra Lozito, Deborah Ballinger, Greg Trot, Gordon Hardy, Ramesh Panda, Diane Carpenter, Ronald Lehmer, Thomas Kozon. + + + + +Tools Used for Data Collection . + + + +Did the pilot appear to notice the TACEC coupling immediately (e.g., hit "accept")?Did you see any events or hear any comments from the pilot regarding the TACEC alerting?Did you see any events or hear any comments from the pilot regarding situational awareness?Did you see any events or hear any comments from the pilot regarding the display?Did you see any events or hear any comments from the pilot regarding workload?Were there any events that you thought were significant?(Please include description of any malfunctions, display distortions, motion problems, etc.)? +Appendix G -Layout and Design of SRT AirportAs stated previously, the SRT airport is based on the current Dallas/Fort Worth International Airport (DFW) layout.Because the simulation focused on studying TACEC approaches to very-closely-spaced parallel runways, and because of the decision to have a south air traffic flow for the simulation scenarios, the SRT airport utilized only runways 18R, 18L, 17R, and 17C (renamed to 17L).All four runways could be used for arrivals and departures, and all were assumed to be equipped to a Category IIIB level.Both runways 18R and 17L were moved to within 750 ft of their inboard runways, 18L and 17R, respectively, requiring an adjustment of 464 ft from their current DFW position.See table G-1 for the DFW runway position changes made for the simulation.In order to support a variety of scenarios, five runway configurations and associated airspace routes were developed.These configurations are summarized in tables G-2 and G-3.Configuration #2 is the only configuration that the DFW airport currently supports.For this configuration, the airspace design matches the current DFW design.The other four configurations were based on current DFW operations (Standard Terminal Arrivals (STARs), Standard Instrument Departures (SIDs), and Approach Plates), but had to be modified to accommodate both the desired runway configurations and the TACEC concept. + + + + + + + Reduced waste generation technical work plan + 10.2172/5213424 + + Dec. 12, 2004 + Office of Scientific and Technical Information (OSTI) + + + Joint Development and Planning Office: The Next Generation Air Transport System Integration National Plan. Dec. 12, 2004. + + + + + Compilation of DOT and FAA Airport Legal Determinations and Opinion Letters as of December 31, 2020 + 10.17226/26202 + DOT/FAA/CT- TN02/01 + + 2001 + Transportation Research Board + + + Magyratis, S.; Kopardekar, P.; Sacco, N.; and Carmen, K.: Simultaneous Offset Instrument Approaches -Newark International Airport: An Airport Feasibility Study. DOT/FAA/CT- TN02/01, 2001. + + + + + Frequency and severity of deviations during simultaneous independent approaches to parallel runways + + SebastianVMassimini + + + GerryCMcneil + + + NoamTene + + 10.1109/icnsurv.2010.5503313 + MP 06W0000127 + + + 2010 Integrated Communications, Navigation, and Surveillance Conference Proceedings + + IEEE + 2006 + + + Mitre Report + Massimini, V.S.: Simultaneous Independent and Dependent Parallel Instrument Approaches. Assumptions, Analysis, and Rationale. Mitre Report MP 06W0000127, 2006. + + + + + NASA administrator announces resignation + + TSAbbott + + + DMElliot + + 10.1016/s0045-8732(01)80042-5 + Concept. NASA/TP-2001-210665 + + + COSPAR Information Bulletin + COSPAR Information Bulletin + 0045-8732 + + 2001 + 152 + + 2001 + Elsevier BV + + + Abbott, T.S.; and Elliot, D.M.: Simulator Evaluation of Airborne Information for Lateral Spacing (AILS) Concept. NASA/TP-2001-210665, 2001. + + + + + CNS requirements for precision flight in advanced terminal airspace + + MEMiller + + + SDougherty + + + JStella + + + PReddy + + 10.1109/aero.2005.1559471 + + + 2005 IEEE Aerospace Conference + + IEEE + 2005 + + + + Miller, M.E.; Dougherty, S.; Stella, J.; and Reddy, P.: CNS requirements for precision flight in advanced terminal airspace. Aerospace, 2005 IEEE Conference, pp. 1-10, 2005. + + + + + Operational concepts for closely spaced parallel approaches + + ARPritchett + + + SLandry + + 10.1109/dasc.1999.863730 + + + Gateway to the New Millennium. 18th Digital Avionics Systems Conference. Proceedings (Cat. No.99CH37033) + St. Louis, Mo + + IEEE + Oct. 1999 + + + + Pritchett, A.R.; and Landry, S.: Operational Concepts for Closely-Spaced Parallel Approaches. Proc. Digital Avionics Systems Conf. (DASC00), St. Louis, Mo., pp. 4.C.2-1, Oct. 1999. + + + + + Flight Testing WAAS for Use in Closely-Spaced Parallel Approaches + + SHouck + + + ABarrows + + + BParkinson + + + PEnge + + + DPowell + + + + Proc. ION-GPS 1999 Meeting + ION-GPS 1999 MeetingNashville, Tenn + + 1999 + + + Houck, S.; Barrows, A.; Parkinson, B.; Enge, P.; and Powell, D.: Flight Testing WAAS for Use in Closely-Spaced Parallel Approaches. Proc. ION-GPS 1999 Meeting, Nashville, Tenn., 1999. + + + + + Cockpit Display of Traffic and Wake Information for Closely Spaced Parallel Approaches + + GordonHardy + + + EmilyLewis + + 10.2514/6.2004-5106 + AIAA-2004-5106 + + + AIAA Guidance, Navigation, and Control Conference and Exhibit + Providence, R.I. + + American Institute of Aeronautics and Astronautics + Aug. 16-19, 2004 + + + Hardy, G.H.; and Lewis, E.K.: A Cockpit Display of Traffic Information for Closely- Spaced Parallel Approaches. AIAA-2004-5106, AIAA Guidance, Navigation, and Control Conference and Exhibit, Providence, R.I., Aug. 16-19, 2004. + + + + + Models of Wake-Vortex Spreading Mechanisms and Their Estimated Uncertainties + + VernonRossow + + + GordonHardy + + + LarryMeyn + + 10.2514/6.2005-7353 + AIAA-2005-7353 + + + AIAA 5th ATIO and16th Lighter-Than-Air Sys Tech. and Balloon Systems Conferences + Arlington, Va. + + American Institute of Aeronautics and Astronautics + Sept. 26-28, 2005 + + + Rossow, V.J.; Hardy, G.H.; and Meyn, L.A.: Models of Wake-Vortex Spreading Mechan- isms and Their Estimated Uncertainties. AIAA-2005-7353, ATIO Forum, Arlington, Va., Sept. 26-28, 2005. + + + + + Situation Awareness as a Predictor of Performance for En Route Air Traffic Controllers + + FrancisTDurso + + + CarlaAHackworth + + + ToddRTruitt + + + JerryCrutchfield + + + DankoNikolic + + + CarolAManning + + 10.2514/atcq.6.1.1 + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 6 + 1 + + 1998 + American Institute of Aeronautics and Astronautics (AIAA) + + + Appendix F -List of Questions for Group Discussion + Durso, F.T.; Hackworth, C.A.; Truitt, T.R.; Crutchfield, J.; Nikolic, D.; and Manning, C.A.: Situation Awareness as a Predictor of Performance for En Route Air Traffic Con- trollers. Air Traffic Control Quarterly, vol. 6, no. 1, pp. 1-20, 1998. Appendix F -List of Questions for Group Discussion + + + + + Aircraft Safety Information Pamphlet + 10.2307/j.ctvx077dc.13 + + + I Was Waiting to See What You Would Do First + + University of Arkansas Press + + + + + What information would you require from the lead to perform a breakout maneuver? 2. Would you like to see zoom feature change the size of the aircraft icon? + + + + + What Continuing Education Workshops Would You Like to See? + + SharonRaeJenkins + + 10.1037/e387942004-039 + + + American Psychological Association (APA) + + + Would you like to see a forward boundary of safe flying zone on display? + + + + + Creative Problem-Solving – Turning What You Don’t Want into What You Do Want + 10.1017/9781108907712.008 + + + Good Thinking + + Cambridge University Press + + + + + Are we scaling the wake display to the icon size in the way you would like? 6. Do we want predictors (dots) past 10 seconds? + + + + + What Continuing Education Workshops Would You Like to See? + + SharonRaeJenkins + + 10.1037/e387942004-039 + + + American Psychological Association (APA) + + + Would you like to see it differently + Was the datalink message showing the flight trajectory at the start of the simulation to be flown useful? Would you like to see it differently? + + + + + Useful Lists + + BryceButton + + 10.1016/b978-1-578-20096-2.50017-4 + + + Nonlinear Editing + + Elsevier + + + + + Was the acknowledgement button before coupling useful? + + + + + What Do You Like about America? + 10.2307/j.ctvwh8f5f.29 + + + The Woman in the Corner + + University of Pittsburgh Press + + + + + What do you like about the concept? + + + + + 7. Adapting IRAC to ‘discuss’ questions + + SIStrong + + 10.1093/he/9780198811152.003.0007 + + + Oxford University Press + + + Discuss breakout maneuvers based on a scenario as shown 11. Discuss low-visibility problems. Are the displays sufficient? 12. Discuss questions from the post-interactive survey. + + + + + What Practical Things Should I Think About Before I Start My Research? + 10.4135/9781526443144 + + + SAGE Publications Ltd + + + What do you think about the jurisdiction of the tower controller? Where should it start and end? What will be the responsibilities of the controller before and after the coupling point? + + + + + What you doin' about hydraulic system reliability/ques/ + + GMoore + + 10.2514/6.1967-403 + + + Commercial Aircraft Design and Operation Meeting + + American Institute of Aeronautics and Astronautics + + + + What did you think about the display changes that occur after coupling of aircraft? FMS annunciation? Dashed magenta line for lead aircraft? + + + + + “What do you think about having beauty marks on your-Hashek!”: Innovative and Impolite Uses of an Arabic Politeness Formula among French Teenagers + + ChantalTetreault + + 10.1111/jola.12098 + + + Journal of Linguistic Anthropology + Journal of Linguistic Anthropology + 1055-1360 + + 25 + 3 + + + Wiley + + + What do you think about having no manual control over the system? + + + + + What Kind of Animal You Would Be If You Could Be Any Animal + 10.2307/jj.455900.38 + + + In Kind + + University of Iowa Press + + + + + could you be at the coupling as accurately as with automation? What kind of cues will you need + If you are in manual control, could you be at the coupling as accurately as with automation? What kind of cues will you need? + + + + + + diff --git a/file762.txt b/file762.txt new file mode 100644 index 0000000000000000000000000000000000000000..90e2a6c7c11ab286eb0bf2fcbcbf0342b4bbd938 --- /dev/null +++ b/file762.txt @@ -0,0 +1,387 @@ + + + + +IntroductionMany US airports depend on parallel runway operations to meet the growing demand for day-to-day operations.The main objective for increasing simultaneous approaches on parallel runways is to improve the throughput of the airport.Several concepts for simultaneous approaches have focused on achieving Visual Meteorological Conditions (VMC) capacities under Instrument Meteorological Conditions (IMC) because poor weather often reduces the capacity of airport with parallel runways to half.Triple parallel runways have the potential to increase capacity especially when they are 750 ft apart.Some airports such as John F. Kennedy and Atlanta Hartsfield have adequate space between th eir two parallel runways to build a third runway between them such that they are all 750 ft apart.The biggest challenge with closely spaced parallel runways approaches is achieving safe operations.For runways that have greater than 3400 ft separation between them, a No Transgression Zone (NTZ) of 2000 ft between the runways provides a safety net.In the concept investigated for this study, the runways were 750 ft apart and a breakout maneuver was shown on the navigation display in the cockpit.Studies have researched missed approaches using auto pilot or manual procedures for single runway airports [2] and there has been some research to compare the procedures under auto-pilot and manual flight control modes prior to a breakout for two runway operations [3].However, no previous research has been done to compare using auto-pilot and manual flight control modes for flying breakout maneuvers on three parallel runway operations.This paper will compare the procedures for performing breakout maneuvers for triple simultaneous procedures under off nominal conditions using either the auto pilot or manual flight control mode.Several metrics including the workload and situation awareness experienced by the pilots have been compared in this study.The following sections of this paper describe the background research and the experimental approach that was taken to study the effect of manual and autopilot flight control modes on breakout maneuvers.Then, the results and discussion section focuses on separation between the aircraft, accuracy of flying the breakout trajectory and subjective data such as workload and situation awareness. +BackgroundMost of the previous research on very closely spaced parallel approaches has focused on dual runways [4] [5].The research on triple streams of aircraft has been mostly exploratory in nature, such as investigating the effect of adding a third stream of aircraft on capacity.There have been several procedures defined for triple simultaneous approaches, and most of them define a no transgression zone or a safety net to protect against aircraft blundering or deviating from their intended path towards the other aircraft.Previous research [6] described several permutations of Simultaneous Offset Instrument Approach (SOIA) procedures for triple parallel runways.For example -an independent SOIA procedure [7] for triple aircraft arrivals procedure requires an independent monitor for each runway and has a 2000 ft No Transgression Zone established between each pair of simultaneous streams.Breakout procedures were also defined by a concept called Airborne Information for Lateral Separation (AILS) studied by Abbott (2001) [3] at NASA.They explored procedures where the flight control mode prior to the breakout was either autopilot or manual.The breakout procedure was always performed under manual flight control modes.They found that if the pilots were flying under an auto pilot control mode prior to the actual breakout maneuver, they took longer to respond to the breakout.Several researchers have investigated the effect of flight control modes on workload and errors.Casner [2] also explored the effect of flight control mode on different phases of flight that included missed approaches for single runway operations, which are in some ways similar to the breakout maneuver.Among the other variables manipulated in the study were the navigation methods (VOR-Very high frequency Omni Range or GPS -Global Positioning System).He found that overall workload was higher in manual flight mode as compared to the auto pilot for the missed approach phase of flight.The author also found that on the subjective survey items, the pilots indicated a "strongly agrees" to autopilot reducing their workload (4.63 on a scale of 5).They showed preference for using auto-pilot during periods of high workload (4.63 out of 5).Similarly, they also showed mid-level preference (3.38 on a scale of 5) towards using autopilot for missed approaches.The authors of the current study also explored triple runway procedures for breakout maneuvers conducted under manual flight control mode.They found that pilots experienced high workload and a reduction in situation awareness because they had to focus too much attention on flying the breakout maneuver shown on the navigation display using the flight director on the primary flight display [1].The current paper explores the differences in flying the breakout maneuver using the autopilot control mode, as compared to the manual mode with three closely spaced parallel runways. +Experimental Approach +Airport and Airspace DesignThe experiment used a fictitious airport (KSRT) loosely based on the current Dallas/Fort Worth International Airport (DFW) layout and operations except for three parallel runways that were set to be 750 ft apart.Because the simulation focused on TACEC approaches to very closely spaced parallel runways using south flow scenarios, only the west side runways (18L, 18C and 18R) were used.The outside runway (currently 18R) was moved inward to create 18C with a 750 ft separation between the runways and a third 18R was also added at 750 ft from 18C.All three of the runways were assumed to be equipped to a CAT-IIIB level. +TACEC ProceduresTerminal Area Capacity Enhancing Concept [7] (TACEC) allows for any aircraft arriving from any of the four arrival meter fixes (NE, NW, SE, and SW) to be paired for a simultaneous parallel landing, based on aircraft characteristics and relative timing criteria.The three paired aircraft flew their assigned 4D trajectories with a high level of accuracy to meet timing constraints at the coupling point and to ensure wake safety throughout the approach.TACEC assumes augmented Global Positioning System (GPS) and ADS-B (Automatic Dependent Surveillance-B).TACEC calls for the three aircraft to be paired at meter fixes located near the edge of the terminal airspace, normally 40-60 nmi from the airport [8] and given TACEC-assigned 4D arrival trajectories to the runway.Flights in the simulation began 25 nmi from the airport, assuming they were already paired/grouped.Routes to the airport included approach and departure routes, and procedures were defined to be, similar to those at the DFW airport.This study focused on arrivals, and no departures were included in the traffic.The coupling point, which refers to the point at which the speed of the multiple aircraft becomes dependent, or "slaved" to one another, is defined at 12 nmi from the threshold of the runway.From that point onward in the simulation, the center aircraft precisely maintained 12s spacing behind the lead aircraft, and the right aircraft maintained 24s behind the lead aircraft using a speed algorithm to avoid the wake and for safe separation.The approach paths of the two trailing aircraft were at a slewed angle from the center of the runway-six degrees for the aircraft on the center runway and 12 degrees for the aircraft on the right runway, when the aircraft were 25 nmi from the threshold.All three aircraft turned straight and parallel to each other at about 2 nmi from the runway.Onboard automation, based on ADS-B, monitored the three aircraft for potential emergency situations.The automation displayed a predicted hazardous zone for the wake generated by the lead and center aircraft in the cockpits of the second and third planes.ADS-B lateral position and intent information was used to detect and display any deviation from the proposed approach path that would encroach on either of the trailing aircraft.Visual and aural alerts were given to the pilots when the lead-aircraft's blunders or wake presented a dangerous situation to the trailing aircraft.The navigation display depicted a breakout trajectory after the aircraft crossed the coupling point.This breakout trajectory was dynamically generated and considered wake, traffic, buildings and terrain of the airport surroundings.When the breakout was required at different altitudes on the arrival path, different bank angles for the breakout maneuvers were used and the curvature of the breakout trajectory changed on the navigation displays.The pilots flew the breakout trajectory manually using the flight director when they received an aural and visual alert under the manual flight control mode condition.In the auto-pilot condition, they flew the breakout trajectory without disengaging the auto-pilot. +DisplaysThe displays were similar to the displays used for the study of two runway very closely spaced parallel approaches [4] and were based on previous research associated with flight deck displays [8] [9].The Navigation Display (ND) and Primary Flight Display (PFD) are shown in Figures 1 and2.The displays show wake and trajectory information along with standard flight instrument data.Information regarding coupling of aircraft is also shown on PFD.After crossing the coupling point and the pilot's prior acceptance of coupling with the lead aircraft, the flight mode annunciation changes to show that the three aircraft are coupled for speed (C-SPD), coupled for lateral navigation (C-LNAV) and coupled for vertical navigation (C-VNAV).The two trailing aircraft were coupled with the lead aircraft.The autopilot of the trailing aircraft flew the approach; the pilot primarily monitored the aircraft performance and the displays for the remainder of the flight.If the wake of the adjacent aircraft drifted within one wingspan of the own-ship aircraft, the color of the wake hazardous zone of the lead aircraft on the display turned to yellow, and then turned red when the apex of the aircraft was in the wake.Similarly, if the lead aircraft deviated from the planned trajectory towards the following aircraft's path by 60 ft, the lead aircraft symbol turned yellow, and then red when the lead aircraft deviated by 120 ft.The red warnings, accompanied by an aural alert "breakout, climb" required a mandatory breakout, which the pilots flew manually.When the pilots pressed the Take-Off-Go-Around (TOGA) switch, the breakout trajectory, which had been displayed to the pilot in white, became the active route, and was then displayed in magenta for both the flight control modes. +Advanced Concept Flight SimulatorThe human-in-the-loop experiment studied breakout maneuvers using auto-pilot and manual flight control mode for triple TACEC approaches in the Advanced Concepts Flight Simulator (ACFS), which is located at NASA Ames Research Center.The ACFS is a motion-based simulator that represents a generic commercial transport aircraft, enabling it to be reconfigured to represent future aircraft.It has the performance characteristics similar to a Boeing 757 aircraft, but its displays have been modified to study different advanced concepts.In this study, the cockpit displays described in the previous section were integrated with the flight display systems in the cockpit.The visual systems offer a 180 deg horizontal and a 40 deg vertical field of view.This simulator is capable of providing various visibility conditions and was set to IMC for this experiment. +Study DesignFour factors were manipulated in this study on the TACEC concept for triple runways.The primary factor, which is the most pertinent to the focus of this paper, is flight control mode, with two state valuesmanual or autopilot mode under which the breakout maneuver was flown.The second factor was the cause of the breakout maneuver -wind causing the wake of the lead aircraft to drift towards the following (center) aircraft, or the lead aircraft deviating from its original path and towards the trailing aircraft.The third factor was the location of the off-nominal situation, which was above 500 ft, or between 200 ft -500 ft AGL.The fourth factor was the position of the ownship or the simulator which could either be approaching the center or right runway (with the lead aircraft approaching the leftmost runway).All runs had an off-nominal situation that required a breakout maneuver.A total of 16 runs were performed for each participant, 8 of which used the manual flight control mode to fly the breakout maneuver and the rest used the autopilot for the breakout maneuver.The +HypothesesBased on previous research conducted on flight modes [2], we predicted that there would be reduction in workload and improvement in situation awareness under the autopilot control mode, as compared to the manual control mode, for flying the breakout Breakout Trajectory Wake own-ship trajectories.We also predicted increased aircraft separation and improved breakout trajectory accuracy with autopilot, as compared to the manual breakout procedures, due to the precise nature of the autopilot mode. +ParticipantsThe participants were three recently retired pilots from commercial airlines; all were male and they all had experience with glass cockpits.Their average experience as a pilot was about 38 years.Their average number of years since retirement was less than two. +Experimental ProcedureThe study ran in two parts: the first part collected data on manual flight control mode with three pilots and the second part had the same pilots who participated in the auto-pilot flight control mode fly the breakout maneuvers.At the beginning of the experimental run for both manual and auto-pilot set of conditions, the pilots were familiarized with the project, the concept, and the new displays in the cockpit.The pilot received a demonstration of the ACFS and hands-on training on the flight deck displays and related procedures.Since procedures for triple Very Closely Spaced Parallel Runways (VCSPR) were being explored in this study, each pilot flew the ACFS in the left seat (as captain) along with a confederate who acted as the first officer for both the flight control modes -manual and auto pilot.Prior to flying the breakout maneuver, the role of the pilot was to fly in auto pilot mode and monitor the displays to check separation with the lead aircraft and with wake.Prior to the coupling point the pilots heard a chime, saw the acknowledgement button light up, and received a "TACEC Coupling" message on the lower Engine Indicating and Crew Alerting System (EICAS) display.At this point the pilots pressed the accept button.They flew as the center or as the trailing aircraft and both of those aircraft were coupled with the leader aircraft on the left most runway.They were coupled with the leader's speed and continued to monitor the separation between the three aircraft.The flight mode annunciation also changed to show that the two aircraft were coupled for speed (C-SPD), coupled for Lateral navigation (C-LNAV) and coupled for Vertical navigation (C-VNAV).If the pilots received a visual and aural alert from the displays they were required to perform a breakout maneuver.Under the manual flight control mode for flying the breakout maneuver, the pilot would press the TOGA switch, disengage the autopilot, leave the auto throttle on, and fly the breakout trajectory shown on the ND.Pressing the TOGA switch would capture the breakout trajectory, and the pilots used the flight director to fly the trajectory.They flew different breakout trajectories at different altitudes, with the breakout above 500 ft altitude requiring an initial bank angle of 30 deg, and the breakout at altitudes between 200-500ft requiring an initial bank angle of 10-deg.They had an initial heading change of 20-deg if they were the center aircraft on 18C and a heading change of 40-deg if they were the trailing aircraft on 18R.In all the above cases, the aircraft had to climb to 3,000 ft as a part of the break out procedure.The pilots then followed the 'S' shaped breakout trajectory displayed on the ND.The trajectory was 'S' shaped so that the final leg of the trajectory became parallel to the runways.The final leg of the breakout trajectory was 1.5 nmi abeam for 18C and 3 nmi for 18R.Under the auto-pilot flight control mode, the only difference from the above procedures was that the pilot pressed the TOGA button to execute the breakout maneuver, and did not disengage the auto-pilot.Rest of the procedures for the auto-pilot mode were the same as that for the manual mode. +Traffic ScenarioThe traffic scenario had three aircraft: (1) The ACFS (B757) was always one of the two following aircraft (center or trailing) in the triplet, and the other two aircraft were scripted, depending upon the experimental condition, and (2) the leader aircraft was a Boeing 747-400, which was prerecorded and scripted for this study and landed on 18L under nominal conditions.The pilot who flew the ACFS simulator always landed on either 18R or 18C or performed the breakout, depending upon the simulator position for the particular data collection run.Operationally, the trailing aircraft should be upwind of the cross wind, but this is not always possible so all scenarios included adverse crosswind.It should also be noted that larger aircraft would ideally be the trailing aircraft (from an intra-echelon perspective); a leading 'heavy' aircraft in the upwind position represents the worst-case scenario for this concept. +Tools used for Data CollectionSeveral tools were used for collecting subjective data from the pilots.All participants completed a demographic survey before the simulation runs were conducted.The survey collected information about the pilots such as their age, experience and number of hours flying different aircraft types, any experience with SOIA approaches, and experience using personal computers.All pilots were asked to complete a Post Interaction Survey at the end of all the runs.This survey allowed them to rate the information content and the usability of the displays.The participants completed the NASA Task Load Index (TLX) rating scales [10] after each simulation run but did not complete the pair-wise scale comparison included as part of the TLX, so the six scales were analyzed separately.Pilots also completed the Situation Awareness Rating Tool (SART) [11].The SART gathers a participant's rating of situation awareness (SA) for the preceding period of time on ten different scales.Each scale has 7 points, with the end points representing opposite ends of the construct.Participants circled the point on the scale that most closely represented their experienced level of SA.The ten SART ratings together with the TLX ratings were gathered from every participant at the end of each simulation run.In addition to the assessment instruments described above, the flight simulator's digital data collection system was used.A host of objective flight data from each of the simulation runs was collected on the variables pertinent to the hypotheses of the experiment.All collected data were indexed with a common timestamp, which was used as the basis of time synchronization as it updates in real-time while the simulation run advances.All digital data were collected at a rate of 30 Hz. +Results and DiscussionStatistical analysis of the study-data focused on three areas: (1) the flight simulator's digital data collection outputs, (2) the pilot participants' workload and situation awareness assessments, and (3) verbal feedback provided by the pilot participants at the end of the simulation runs.As a means of controlling for the possible confounding influence of variables that could impact the results pertinent to the current investigation, other factors were built into the statistical analysis paradigm.More specifically, autopilot vs. manual breakout differences were analyzed along with 3 other independent variables using a 4-way repeated measures analysis of variance (ANOVA) procedure.These three additional variables -cause of breakout, location of breakout, and position of ownship were analyzed in a previous study [1], as they were pertinent to that investigation, and results on these factors were fully addressed and reported.However, since the focus of the current paper is on autopilot vs. manual breakout differences, only the results on this factor will be reported. +Aircraft Separation from Breakout through 30 Seconds Past BreakoutThe dependent measure of aircraft separation is defined as slant range, also known as the displacement distance between two aircraft.Analysis of aircraft separation as it changes over time from breakout point was implemented, to determine if there were any instances of unsafe separation between aircraft during the most critical phase of the breakout maneuver, i.e., the time span that immediately follows breakout point, defined as breakout time through 30 seconds past breakout time.Separate analyses were performed in comparing (1) Leading and center aircraft separation, and (2) Center and trailing aircraft separation.Table 2 shows summary statistics of the combined autopilot and manual breakout data as they changed over time originating from breakout point.As indicated in Table 2, there is a clear trend towards increased separation between each of the two pairs of aircraft analyzed, with some overall increase 15 seconds past breakout, and a larger increased separation at 30 seconds past breakout.Figures 3 through 6 show the same aircraft separation data displayed in Table 1, broken down by manual vs. autopilot breakout conditions.Generally, Figures 3 through 6 show similar distribution patterns of separation data between autopilot and manual conditions for each aircraft pair analyzed.Also, the behavior of the separation data as it changes over time, broken down by autopilot and manual conditions, is very similar to the overall separation distribution shown in Table 1 with a clear trend towards increased separation between each of the two pairs of aircraft analyzed, with some overall increase 15 seconds past breakout, and a larger increased separation at 30 seconds past breakout.The only apparent exception to this trend, shown on several individual time-series in Figures 5 and6, indicate a relatively small decrease in separation 15 seconds past breakout, prior to increased separation 30 seconds after breakout.This behavior occurred only occasionally, and only with the center/trailing aircraft pair.These data suggest that this particular trend, which occurred under both manual and autopilot conditions, reflects the complex geometry of the breakout maneuvers.Specifically, the center aircraft needs to separate itself from the leader aircraft towards the trailing aircraft, which may initially decrease separation for a very short period of time.Even so, during this critical window of time, there were no cases where the slant range between either of the aircraft pairs was less than 2400 ft, indicating zero instances of unsafe separation.These data compare favorably with the data collected by the FAA's MPAP study [12], which defined a test criterion violation (TCV) as 500 ft of separation between aircraft.Using the same definition, a TCV was not observed at any time, at or beyond breakout.Clearly, the objective evidence shows significantly larger separation than the TCV value indicated in the MPAP studies.Further, possible differences between autopilot and manual conditions were assessed on the dependent measure of slant range separation 15 seconds past breakout.ANOVA results comparing the two study conditions indicate some increased separation between the paired aircraft when the autopilot was used.These results seem to indicate that either the autopilot or manually flow breakout procedures could provide a basis for safe breakout maneuvers for the concept under study.In addition, ANOVA results indicated some added safety benefit, in terms of increased aircraft separation, provided by flying the VCSPR breakout using autopilot. +Leader &Center Separation +Mean +Accuracy of Breakout Trajectory: Cross Track and Track Angle ErrorTrajectory accuracy is measured by the actual ownship/simulator position against the breakout trajectory generated by the system and displayed on navigation display in the cockpit (see Figure 1) averaged across time.Two measures of ownship trajectory particularly sensitive to breakout maneuvers include cross track error and track angle error.For each flight simulation run, cross track error and track angle error were averaged across time from the breakout point to the end of the flight.A two-way repeated measures ANOVA yielded a main effect of condition (autopilot vs. manual) on each of the two dependent measures.Both of these results are consistent with respect to the directionality of the means across both track angle and cross track error.Less cross track error and less track angle error were observed under the autopilot breakout condition, as compared to the manual breakout condition.ANOVA summary statistics from this analysis are listed in Table 4.The pilots flew the breakout trajectories with higher precision under the autopilot condition, which would make sense, due to the increased level of automation accuracy that the autopilot provides during breakout.This result is consistent with the results that Casner [2] found where the use of auto-pilot for missed approaches led to a smaller average number of errors.Also, since the autopilot was used to fly the breakout procedure, the pilots would be able to focus more attention on the information provided by the displays, rather than manually flying the breakout, which would necessarily have the effect of increasing pilot situation awareness.This dynamic will be discussed in greater detail later in this paper. +Mean +WorkloadParticipants completed the NASA TLX workload questionnaire after every run.Data were collected on each of the six TLX workload measures, and a variable measuring overall workload combining all six of these measures was derived, for a total of 7 workload measures.This overall workload variable, also known as the "composite" measure, once derived, was then scaled down to match the 1 to 7 scale for direct comparison with the other six measures.Also, the "performance" measure was analyzed on an inverse scale, so a higher score would actually mean less performance.Results on all 7 of these measures, comparing autopilot vs. manual breakout results, are summarized in Figure 7.Results shown in Figure 7 indicate that pilot workload was consistently lower in autopilot breakout runs as compared to manual breakout runs in all of the 6 workload measures, as well as the overall workload composite measure.This was expected, since manual breakout procedures require the pilots to manually fly the ownship according to the breakout trajectory while also monitoring the displays.Under autopilot breakout, the pilots were mostly concerned with monitoring the displays, thereby decreasing workload and enhancing situation awareness, since the actual flying of the breakout maneuver was taken over by the automation.In particular, the physical workload and effort decreased for the auto-pilot condition, which is also consistent with results from the Casner study [2].It should also be noted that workload measured across all scales and conditions was found to be manageable, at low to moderate levels (Figure 7).Hence, workload seems to be low enough to be reasonable, but high enough to prevent tedium and vigilance decrement based on criteria established by previous research [13]. +Situation AwarenessThe SART scale, mentioned earlier, measures situation awareness on ten scales [11].Participants provided ratings on each of these ten scales after every simulation run.All collected SART data were then used to derive three broader categories concerned with a) the demands of the situation b) the 'supply' or personal resources that the participants have to bring to the situation and c) situational provision that the situation provides in the form of information through displays.The first broad category combines the three SART scales -instability, variability and complexity of the situation.The second broad category of personal resources combines the SART scales on alertness, spare mental capacity, concentration, and division of attention.The third broad category, situation provision, combines the three SART scales on information quantity, information quality, and familiarity.After all data were collected and the three broader categories were derived, results were then scaled down to range from 1 (very low) to 7 (very high).Figure 8 shows situation awareness results on the three derived variables, comparing autopilot and manual breakout conditions.It was found that the situation demands of the autopilot breakout runs were lower than those of the manual breakout runs.This result is consistent with our result of lower pilot workload levels in the autopilot condition, since workload correlates with the three SART subscales which from the broader variable of situation demands.Again, the manual breakout condition requires that pilots safely maneuver the aircraft by following the breakout trajectory and maintain adequate situation awareness, which equates to more situation demands than those of the autopilot breakout condition.Results on personal resources indicate almost no difference between the two breakout conditions.This may be due to the anticipation of a breakout anytime, which required equal levels of alertness and concentration across both conditions.Likewise, there was almost no difference between the two breakout conditions in situation provision, suggesting equal amounts of information quantity, information quality and familiarity throughout the course of the simulation runs.Finally, relative to the possible range of values for each of the three broader situation awareness measures, Figure 8 indicates high levels of personal resources and situation provision, with moderately low levels of situation demands across both breakout conditions, suggesting that situation awareness was maintained throughout the course of the current investigation, providing additional support for the efficacy of the TACEC concept. +Summary and ConclusionsTriplet aircraft procedures were investigated in a high fidelity human-in-the-loop simulation incorporating new tools and technologies involving very closely spaced parallel runway operations under both autopilot and manually flown breakout procedures.The results indicated that the autopilot breakout procedures were flown with greater accuracy and better separation than the manually flown breakout procedures.Also, the pilot participants maintained higher levels of situation awareness and lower levels of workload in the autopilot condition as compared to the manual condition.Also, data analysis comparing both study conditions resulted in additional improvement on all of the dependent measures of interest under the autopilot breakout condition.An analysis of aircraft separation during breakout, depicted that, the observed slant range between aircraft never fell below 2400 ft., which is well above the FAA's MPAP test criterion violation threshold between aircraft [12].A statistically significant result was also observed, indicating increased separation under the autopilot breakout, as compared to manual breakout procedures, thus upholding our hypotheses.Analysis of cross track and track angle error indicated statistically significant results between the autopilot and manual conditions, indicating increased trajectory accuracy under the autopilot breakout procedures as compared to manual breakout.The pilots experienced lower workload and situational demands placed on them during autopilot breakout as compared to manual breakout.While realizing these differences, the results also indicate that workload was manageable, and an adequate level of situation awareness was maintained across both conditions.Overall, our hypothesis regarding autopilot breakout procedures decreasing workload and increasing situation awareness, and also showing increased separation as compared to the manually flown breakout procedures were upheld.While more research is still necessary especially with trajectory errors and uncertainties that were not considered in the paper, these results attest to the potential promise of this concept for possible integration into the future NextGen operational environment.Figure 1 :Figure 2 :12Figure 1: Navigation Display (ND) during final approach +Figure 7 .7Figure 7. Effects of Autopilot & Manual Breakout on Pilot Workload Measures (error bars represent ± 1 standard error) +Figure 8 .8Figure 8. Effects of Autopilot & Manual Breakout on Pilot Situation Awareness Measures (error bars represent ± 1 standard error) +Wake Breakout Cause: Aircraft Deviation Breakout Location: > than 500 ft Center/Trailing Ownship Center/Trailing Ownship Breakout Location: 200 ft -500 ft Center/Trailing Ownship Center/Trailing OwnshipTable 1 : Test matrix repeated for flight control mode.1table shows the test matrix, which was repeated to get 8 autopilot and 8 manual flight control mode runs.Repeated runs were made for each breakout cause, breakout location, and position of the aircraft under each flight control mode.BreakoutCause: +Table 2 . Aircraft Separation Following Breakout (combined manual and automated condition) Figures 3-4. Aircraft Separation Following Breakout: Leader & Center Slant Range Under Autopilot and Manual Conditions (each time-series represents one simulation "flight") N=8 Figures 5-6. Aircraft Separation Following Breakout: Center & Trailing Slant Range Under Autopilot and Manual Conditions (each time-series represents one simulation "flight") N=82SDMaxMin(ft)(ft)(ft)(ft)Breakout2551982674 2447Point15 Seconds28591503172 2534Past Breakout30 Seconds36543084106 3038Past BreakoutCenter &TrailingSeparationBreakout2855612968 2796Point15 Seconds28881333192 2552Past Breakout30 Seconds37345464509 3094Past Breakout +Center & Trailing: (F=20.63; df=1,2) Autopilot 2902 110 Manual 2875 154Table 3 . Autopilot vs. Manual Breakout Effect on Aircraft Separation 15 s Past Breakout (* p<0.05)3Table 3 provides summary statistics and ANOVA results pertinent to this finding.Aircraft PairCondition Mean(ft) SD (ft)Leader & Center:Autopilot2867139Manual2849163* +88 deg 0.52 deg Manual 2.27 deg 1.98 deg Table 4. Autopilot vs. Manual Breakout Effect on Ownship Cross Track and Track Angle Error (* p<0.05)SD* Cross Track Error:(F=72.30 df=1,2)Autopilot28.44 ft19.36 ftManual74.49 ft78.91 ft*Track Angle Error(F=28.80;df=1,2)Autopilot0. + + + + +AcknowledgementsWe would like to thank members of Simlab team at NASA Ames Research Center -Ramesh Panda, Darrell Wooten, Diane Carpenter, Ron Lehmer, John Walker and Dan Wilkins, without their effort this simulation would not have been possible. + + + + + + + + + Procedures for off-nominal cases: Very closely spaced parallel runway operations + + SavitaVeram + + + SandyLozito + + + ThomasKozon + + + DeborahBallinger + + + HerbertResnick + + 10.1109/dasc.2008.4702783 + + + 2008 IEEE/AIAA 27th Digital Avionics Systems Conference + + IEEE + 2009 + + + Verma, S., Kozon, T., Lozito, S. Ballinger, D. Herb Resnick, Ramesh Panda, Diane Carpenter, Darrell Wooten, Gordon Hardy, 2009, Evaluation of Triple Closely Spaced Parallel Runway Procedures for Off-nominal Cases, In Press. + + + + + Perceived vs. measured effects of advanced cockpit systems on pilot workload and error: Are pilots' beliefs misaligned with reality? + + StephenMCasner + + 10.1016/j.apergo.2008.10.002 + + + Applied Ergonomics + Applied Ergonomics + 0003-6870 + + 40 + 3 + + 2009. 2009 + Elsevier BV + + + Casner, S., 2009, Perceived vs. measured effects of advance cockpit systems on pilot workload and error: Are pilots' beliefs misaligned with reality? Applied Ergonomics 40 (2009), pg 448-456. + + + + + NASA administrator announces resignation + + TAbbott + + + DElliot + + 10.1016/s0045-8732(01)80042-5 + Concept. NASA/TP-2001-210665 + + + COSPAR Information Bulletin + COSPAR Information Bulletin + 0045-8732 + + 2001 + 152 + + 2001 + Elsevier BV + + + Abbott, T. & Elliot, D. (2001). Simulator Evaluation of Airborne Information for lateral Spacing (AILS) Concept. NASA/TP-2001-210665. + + + + + Procedures for off-nominal cases: Very closely spaced parallel runway operations + + SavitaVeram + + + SandyLozito + + + ThomasKozon + + + DeborahBallinger + + + HerbertResnick + + 10.1109/dasc.2008.4702783 + + + 2008 IEEE/AIAA 27th Digital Avionics Systems Conference + St. Paul, MN + + IEEE + 2008 + + + Verma, S., S. Lozito, T. Kozon, D. Ballinger, and H. Resnick, 2008, Procedures for off-nominal cases: Very closely spaced parallel runways approaches, Digital Avionics System Conference, St. Paul, MN. + + + + + Preliminary Guidelines on Flight Deck procedures for Very Closely Spaced Parallel Approaches + + SVerma + + + SLozito + + + GTrot + + + 2008 + International Council for Aeronautics (ICAS) + Anchorage, Alaska + + + Verma, S., Lozito, S., G. Trot, 2008, Preliminary Guidelines on Flight Deck procedures for Very Closely Spaced Parallel Approaches, International Council for Aeronautics (ICAS) Anchorage, Alaska. + + + + + + + CGladstone + + + GGlover + + + PMassimini + + + CShiotsuki + + + BSimmons + + + 2000 + + + Analysis of Triple arrivals to Hartsfield Atlanta International Airport, MTR00W0000023 Mitre Technical Report + Gladstone, C., G. Glover, P. Massimini, C. Shiotsuki, and B. Simmons, 2000, Analysis of Triple arrivals to Hartsfield Atlanta International Airport, MTR00W0000023 Mitre Technical Report. + + + + + Closely Spaced Parallel Approaches in Terminal Airspace + + AbdulmalikAl-Bulushi + + + NarinChau + + + RobertEftekari + + + RyanGraziano + + + AmonTarakemeh + + 10.1109/sieds.2006.278674 + + + 2006 IEEE Systems and Information Engineering Design Symposium + + IEEE + 2006 + + + Al-Bilushi, A., Chau, N., Eftekari, R. & Tarakemeh, A. (2006). Closely Spaced Parallel Approaches in Terminal Airspace. Proceedings of the 2006 Systems and Information Engineering Design Symposium. + + + + + Cockpit Display of Traffic and Wake Information for Closely Spaced Parallel Approaches + + GordonHardy + + + EmilyLewis + + 10.2514/6.2004-5106 + + + AIAA Guidance, Navigation, and Control Conference and Exhibit + Providence, RI4 + + American Institute of Aeronautics and Astronautics + 2004 + + + Hardy, G. H. and E. K. Lewis, 2004, A Cockpit Display of Traffic Information for Closely Spaced Parallel Approaches, AIAA-2004-5106, AIAA Guidance, Navigation, and Control Conference and Exhibit, Providence, RI4. + + + + + Models of Wake-Vortex Spreading Mechanisms and Their Estimated Uncertainties + + VernonRossow + + + GordonHardy + + + LarryMeyn + + 10.2514/6.2005-7353 + AIAA-2005-7353 + + + AIAA 5th ATIO and16th Lighter-Than-Air Sys Tech. and Balloon Systems Conferences + Arlington, VA + + American Institute of Aeronautics and Astronautics + 2005 + + + Rossow, V. J., G. H. Hardy, and L.A. Meyn, 2005, Models of Wake-Vortex Spreading Mechanisms and Their Estimated Uncertainties, AIAA-2005-7353, ATIO Forum, Arlington, VA. + + + + + Development of NASA-TLX (Task Load Index): Results of Empirical and Theoretical Research + + SandraGHart + + + LowellEStaveland + + 10.1016/s0166-4115(08)62386-9 + + + Advances in Psychology + + PHancock + + + NMeshkati + + Amsterdam, The Netherlands + + Elsevier + 1988 + + + + Hart, S., & L. Staveland, 1988, Development of a Multi-Dimensional Workload Rating Scale: Results of Empirical and Theoretical Research, In P. Hancock & N. Meshkati (Eds.), Human Mental Workload, Amsterdam, The Netherlands, Elsevier, pp. 139-183. + + + + + Situation Awareness as a Predictor of Performance for En Route Air Traffic Controllers + + FrancisTDurso + + + CarlaAHackworth + + + ToddRTruitt + + + JerryCrutchfield + + + DankoNikolic + + + CarolAManning + + 10.2514/atcq.6.1.1 + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 6 + 1 + + 1998 + American Institute of Aeronautics and Astronautics (AIAA) + + + Durso, F., C Hackworth, R Truitt, J Cruthfield, D Nikolic, C Manning, 1998, Situation Awareness as a Predictor of Performance for En Route Air Traffic Controllers, Air Traffic Control Quarterly, 6(1), pp.1-20. + + + + + Training Pilots Prior to Simultaneous Instrument Landing System (ILS) Precision Runway Monitor (PRM) Approaches to Closely-Spaced Parallel Runways + + SMagyarits + + + ROzmore + + 10.4271/arp5453 + + + May, 2002 + SAE International + + + S. Magyarits & R. Ozmore, Evaluation of triple Independent Instrument Landing System Approaches to Runway Spaced 4000 ft and 5300 ft apart using a Precision Runway Monitor System, DOT/FAA/CT-TN02/16. May, 2002. + + + + + Automation and Human Performance: Theory and Applications + + RParasuraman + + + MustaphaMouloua + + + 1996 + Lawrence Erlbaum Associates + + + Parasuraman, R. and Mustapha Mouloua, 1996, Automation and Human Performance: Theory and Applications, Lawrence Erlbaum Associates. + + + + + + diff --git a/file763.txt b/file763.txt new file mode 100644 index 0000000000000000000000000000000000000000..e5e21b4676b2acb21d6620c008afdff3c664f3ba --- /dev/null +++ b/file763.txt @@ -0,0 +1,371 @@ + + + + +INTRODUCTIONThe biggest challenge airports must address with closely spaced parallel runways is that their capacity is reduced when visual approaches are not possible due to poor visibility [1].The FAA's Nextgen and Eurocontrol's Single European Sky ATM Research (SESAR) [2] have a common goal to maintain visual capacities under all weather conditions at airports with closely spaced parallel runways.Previous concepts investigated safety issues related to parallel runway operations but did not examine the information and procedures for pairing aircraft.Studies investigated the safety issues associated with parallel approaches that may require aircraft to perform breakout maneuvers due to hazardous conditions [3,4] such as wake of lead aircraft drifting towards the follower or the lead aircraft blundering towards the follower.In addition, the role of controllers in aircraft pairing for simultaneous approaches was explored [5], including the examination of controller responsibilities and communication tasks.A gap exists regarding research examining the integration of the flight deck and ground procedures and tools necessary for pairing aircraft.This human-in-the-loop study investigates the dynamic role of controllers and pilots for pairing aircraft to parallel runways for simultaneous approaches.The objective of this investigation was to evaluate the integrated procedures, information requirements, and automation for pilots and controllers when pairing aircraft for closely spaced approaches. +II. BACKGROUNDThe FAA has successfully conducted independent approaches to parallel runways for over forty years using the Instrument Landing System (ILS) navigation and terminal radar monitoring [1].Some airports, like San Francisco International, can support approximately 60 landings per hour on two parallel runways that are 750 ft apart using visual approaches, and Simultaneous Offset Instrument Approaches (SOIA) under limited cloud ceiling visual meteorological conditions (VMC).As visibility degrades, the current navigation and surveillance system, as well as the existing procedures, cannot support SOIA approaches, dramatically reducing the landing rate.Several researchers have investigated alternative procedures for Very Closely Spaced Parallel Runway (VSCPR) operations.A number of requirements were identified from these studies, such as cockpit displays, collision prevention systems, precision navigation, communication, surveillance systems and wake information [6] [7].In addition, Pritchett & Landry [8] explored the parameters and procedures related to VCSPR operations such as separation responsibility and spacing objectives.These studies provided important insight into necessary technologies, information, and procedures for VSCPR implementations.There have been a number of human-inthe-loop studies that have explored VCSPR operations.The Airborne Information for Lateral Spacing (AILS) concept is an example of an investigation of pilot response to VCSPR operations [9].The concept requires technologies that enable the use of precise navigation and surveillance data, as well as technology for the detection of blunders.Further simulations have been conducted by NASA to examine pilot procedures for paired approaches on runways that are 750 ft apart in instrument meteorological conditions [3].Enhanced cockpit displays that depict both traffic and wake information were provided to the flight crew for these operations.The results from these investigations revealed that even in the blunder cases, pilot workload was manageable, and an adequate level of situation awareness (SA) was maintained.There are some data regarding the role of the controller in parallel runway operations.Under SOIA, the controller has positive control over the aircraft until the pilot breaks through the clouds and the follower aircraft has visual contact with the leading aircraft.Under AILS, the final approach controller has positive control over the aircraft pair until the trailing aircraft is given a clearance for the AILS approach [10].The Terminal Area Capacity Enhancing Concept (TACEC) [11] was collaboratively developed by Raytheon and NASA Ames Research Center.TACEC is a technique that can be used for conducting simultaneous instrument approaches to two or even three closely-spaced parallel runways that are 750 ft apart.The concept requires a leader and follower aircraft in a staggered configuration, with a safe zone behind the leader where the trailing aircraft is protected from the wake of the leader.The suggested safe trailing distance for the following aircraft is from 5s to 25s, with 15s representing the optimal temporal distance and +/-10s representing the tolerance.This recommended distance was derived from analytic assessments and previous human-in-the loop studies, and is intended to allow for a safe wake distance and provide for distance for a potential break-out maneuver [3].Pilot procedures and information requirements for TACEC were explored in several studies, and controller procedures were examined in a separate investigation [5].In the pairing concept explored in this research, the controllers determine and suggest an appropriate aircraft pair.The flight crew of the trailing aircraft is responsible for accepting and maneuvering the aircraft into a position, where it is dependent upon the lead aircraft.Prototypic automation tools are provided to both controllers and pilots to assist in the pairing tasks.Although controllers have responsibility for maintaining separation between aircraft pairs, the crew of the trailing aircraft is responsible for maintaining the 5-25 s distance behind the lead aircraft.The remainder of the paper will discuss the methods and procedures used by both the pilot and controller participants, as well as the results and a summary of the findings from this simulation. +III. METHODSThe participants were six glass-cockpit qualified flight crews and three controller teams.Each controller team consisted of three controllers.All participants had at least 10 years of experience in their respective fields.The study was run for two days per flight crew with each ATC team participating for four days.The pairing procedures were developed with the assistance of pilot and air traffic control subject matter experts.The scenarios were based upon airspace around San Francisco Airport.Both teams were briefed and trained on the pairing concept, the new displays, and their automation tools.Nine scenarios with a VFR level of traffic arriving on the approach routes were used.The scenarios were scripted to simulate an upstream scheduler that metered traffic into the terminal area.Each participating flight crew flew a motion-based flight simulator in these nine scenarios.Pseudo-pilots controlled other aircraft targets in the scenarios to add realism.There was always an opportunity to pair with another aircraft, with the simulator always representing the following aircraft.All participants completed questionnaires and took part in a debrief at the end of the study. +A. Flight Crew Tools and ProceduresThe study used the Advanced Concepts Flight Simulator (ACFS) located at NASA Ames Research Center.The ACFS is a motion-based simulator that represents a generic commercial transport aircraft.The displays were modified to study the pairing concept. +1) Flight Deck Display ConditionsThe position of the simulator was shown on the navigation display (ND) in the ACFS with the conventional white triangular icon.The lead aircraft position was shown by an open chevron icon on the ND.With augmented GPS (Global Positioning System) navigation, it was assumed that ADS-B (Automatic Dependent Surveillance-Broadcast) position information was accurate within a few feet.The study varied two sets of displays for the pilot participants to help them monitor their pairing conformance.Each of the two display conditions provided information on both the primary flight display (PFD) and the ND for the captain and first officer.The Display 1 (Position Display) condition provided data about the distance error for the aircraft to the coupling point with the use of conformance bars (Figure 1).The coupling point, which is about 12 nmi from the runway threshold, is where the automation systems of the two aircraft would be linked with each other for the rest of the approach.The trailing plane used flight deck automation to control speed and maintain precise spacing of 15s in trail behind the leader.This distance error was relative to a desired position on the aircraft's profile.Display 2 (Prediction Display) offered an estimated time of arrival (ETA) prediction based upon the aircraft's current ground speed.The features for both display conditions included conformance bars that indicated the spacing window behind the leader on the ND and markers for the spacing on the PFD to help the crews manage conformance.The bars and markers would turn yellow if the aircraft was outside of conformance parameters (5-25s window).Another display feature associated with The Position Display included a Longitudinal Situation Indicator (LSI) for showing the ideal location of the aircraft.The Predictive Display is similar to the green arc currently used by the flight crews in glass cockpits, whereas the position display was a new display and is similar to the one used by the controllers for conformance monitoring. +2) Flight Deck Automation ConditionsIn addition to the display variable, the pilot participants were also presented with two automation conditions (see Table 1).An auto speed control flight deck automation tool was developed to assist the crew in the task of maintaining the required spacing behind the lead aircraft.The Airborne Spacing for Terminal Arrival Routes (ASTAR) was originally developed at NASA Langley Research Center for merging and spacing operations [12].For this study, ASTAR was modified to manage the speed of the simulator (as the following aircraft) to maintain 5 -25s behind the lead aircraft on a parallel runway.In the conditions where the flight crew participants did not have the automation that offered auto speed control, they needed to manage their own aircraft speed using current day flight deck automation (manual automation) (e.g., FMS input or mode control panel [MCP] input). +3) Flight Crew ProceduresThe flight crew pairing procedures involved the use of the display and automation features.When the flight crew received a data link message with the pairing instruction from air traffic control, the Secondary Flight Display (SFD) presented textual information about the relevant pairing parameters.These data included the call sign and aircraft type of the lead aircraft, its current speed, its planned approach speed, the ETA of the lead and following aircraft at the coupling point and the current spacing error and the coupling status.Using the data provided, the crew decided to "accept" or "reject" the clearance.They had the option to use the pairing procedures or to not engage in pairing.The flight crews were also told that if they decided not to engage in pairing, they could cancel the pairing operations; however, they did need to inform the controller.If they did not pair, then they would make the approach as a single aircraft.On receiving the initial pairing clearance, one of the two display conditions was presented to the crew.In addition, sometimes the auto speed control automation was available and, after the pairing was accepted, the flight crew could select to engage it.In the cases where the auto speed control automation was present and used, the automation managed the speed to maintain the required separation.The crews did not make speed adjustments manually unless they decided to discontinue use of the automation.The crews were informed that they could turn the automation off whenever they felt it was necessary.In cases where aircraft were early or late, the controller could cancel the pair.Pilots could also cancel the pairing at any time. +B. Air Traffic Control Tools and ProceduresControllers were able to pair aircraft from any of the five arrival streams; however, two aircraft from the same stream could not be paired.Speed changes were the only adjustments allowed to maintain pairing and spacing.The goal of the pairing procedure was to have the trailing aircraft reach the coupling point between 5 and 25s behind the lead aircraft as this represented the safe zone for the trailing aircraft.The 5s lag allows for the lead aircraft to execute an escape maneuver across the path of the trailing aircraft and the 25s cutoff point protects the trailer from the spreading wake of the leader.The three air traffic controller positions used were an Area Coordinator, Boulder Sector Controller and Niles Sector Controller.The coordinator position was responsible for the creation of pairs and overlays two sectors-Niles and Boulder.The sector controllers were responsible for maintaining the pairs to the coupling point with the desired intra-pair spacing of 15s; and maintaining the current day spacing of 3nmi between consecutive pairs of aircraft.Based on previous research [5], a level of automation was selected for the pairing tool, in which the automation suggested pairs of aircraft, and the controllers could manually override the suggested pairs.The main goal for the coordinator was to evaluate pairs to ensure the two aircraft were capable of landing between 5 and 25s of each other.Each of the three controllers had a pairing table (which listed all pairs in order of their ETA, a continuallyupdated timeline (configured to show the ETAs of the aircraft to the two parallel runways), and a conformance monitoring tool, which displayed two bars to show the leading and trailing edge of the 5-25s conformance envelope.To finalize a pair, the coordinator evaluated the pair suggested by the automation against the timeline.If the pair was acceptable, the coordinator sent a data link message to the two aircraft.When the pilots acknowledged the pairing, the aircraft call signs turned green in the pairing table.Both aircraft in the pair were given an approach clearance electronically by the sector controller who owned the trailing aircraft in the pair.The approach clearance also implicitly delegated separation authority to the aircraft.Aircraft pairs that were out of conformance could not be given approach clearances.If a pair lost conformance, controllers either re-paired aircraft after making speed adjustments (if possible), landed the planes as singles, or vectored them away and returned them back to the flow upstream. +IV. RESULTS AND DISCUSSIONThe study goal was to explore the feasibility of aircraft pairing on arrival.The key metrics were the spacing of the aircraft relative to each other by the beginning of the coupled approach, and the participant's subjective ability to complete his/her tasks, including workload and situation awareness.The spacing metric between the aircraft pair helps determine the safety of the operations. +A. Number of Aircraft Pairs Created/ Deleted and Number of Single AircraftFigure 3 shows the mean number of aircraft pairs created/deleted and the number of aircraft that arrived as singles for all the test conditions.As indicated in this figure, the controller participants, on average, paired most aircraft in each scenario (N=14.7 pairs or 29.4 aircraft/run), canceled very few pairs (N < 1 pair/run) and left a relatively small number of aircraft (N=5.5 aircraft/run) to land as singles.These statistics seem to provide some evidence of controllers' ability to use the pairing tool, suggesting a high level of usability.Although the objective of the controller was to land as many pairs as possible, having a small number of singles helps with efficiency, particularly in cases when an aircraft was vectored or had a go-around and had to be reintegrated into the flow. +B. In trail Spacing between aircraft 1) Spacing from the flight crew perspective.The leader-trailer spacing for the aircraft pair including the ACFS was of particular interest.Out of 47 runs, in 32 (61.5%) the ACFS flew over its coupling point close to its ideal position, that is between 10s and 20s behind the leader, and 44 runs (93.6%) were inside the 20s window (5 to 25s in-trail).Figure 4 shows the 47 runs in the order of the simulator's spacing behind its leader (the ideal 15s behind is the 0 on the yaxis) not in chronological order.Above the top green line the ACFS was more than 25s behind the leader, i.e., late.Below the bottom green line it was less than 5s behind the leader, i.e., early.In 47 runs, only twice was the ACFS spaced outside the 20s window (4.2%)once it was early (in front of the window, run 1) and once it was late (behind the window, run 47).A third run was on the borderline for being early.Based on crew feedback, both outliers and the borderline run seem to be the result of pilots testing the system to see how long they could wait before they intervened.At the end of each run, pilots provided feedback about conformance at coupling.For the two outlier runs both crews were aware that they may have not achieved their conformance window.All four pilots commented at some level that they were watching their "current spacing error" indicator vacillate between 10s (which is "on-time") and 11s (which is late or early). +2) In Trail Spacing Between Leader and Follower Aircraft at the Coupling Point (Controller's Perspective)This metric was defined as the difference between the time the leading aircraft arrived at the coupling point and the time the trailing aircraft arrived at the coupling point.It was used as a measure of how well the aircraft achieved precise 15s in trail spacing between the leader and the follower within an aircraft pair.Results showing the distribution of this metric across all simulation runs for all aircraft are presented in Figure 5.While Figure 5 indicates a mean value very close to the optimum separation of 15s and well within the preferred range of in trail spacing values, it is also quite clear that a fairly wide range of values exist within the overall distribution.Still, most of these values fall within the goal of 15s (+/-10 s) temporal separation, which suggests that most aircraft met the conditions required by the concept under study. +3) Spacing DiscussionNearly all aircraft pairs in all runs crossed the coupling point within the specified spacing window, suggesting that the concept is feasible.The flight deck crew were aware on the runs when the ACFS did not meet its window, suggesting that procedures need to be more carefully defined (rather than improving the display of in/out of conformance information). +C. Operator WorkloadThe ATWIT (Air Traffic Workload Input Technique) [13] was used to collect both pilot and controller opinions of their workload during the scenarios they worked.The seven-point ATWIT scale was built into the controllers' workstations and was available on a keypad placed in front of the flight crew during each run.Every five minutes, all participants were asked to rate their overall workload level at that moment from 1 = "very low workload" to 7 = "very high workload." +1) Pilot WorkloadWith the runs lasting around 20mins, crews rated their workload about 4 times per run.The initial analysis below considers only the mean ratings, which combine these four responses.Across all runs crews reported "a little" workload (M= 2.7), which is encouraging as this was not a full mission simulation.Although most ratings were at the low end of the scale, there were a few individual 6 and 7 ratings, which are important to note because this means some pilots thought at certain points they could not cope with any more load/tasks (interruptions, landing preparation, etc.).To explore what might affect workload, means were aggregated by the four study conditions.Figure 6 compares these mean values.Differences between the mean ratings are smallless than 0.5 of a scale-point separates the lowest mean of study condition C from the highest mean of study condition D.Figure 6 shows the interaction between the two variables over the four conditions -auto speed control automation is linked to both the lowest and the highest mean workload ratings depending upon the type of display crews used.Participants rated the Prediction Display as causing them a higher workload but only in the auto speed control automation condition.A Friedman test was applied on these workload ratings across the four conditions and showed the differences are significant ( 2 (3)=9.423,p=.024).A series of pairwise Wilcoxon tests show the only difference at the p<.05 level was between conditions C and D (Position Display and Prediction Display under auto speed control automationthe two points on the right in Figure 6), where ratings showed participants thought the workload was higher in the D condition (p=.024, meanC = 2.5, meanD = 2.8).The difference between the Position Display mean workloads under the two automation conditions (A & C) only approached significance (p=.054).These results are surprising given that condition A (Position Display) was rated on average to incur a higher workload than condition B (Prediction Display) when participant feedback suggested the Prediction Display was harder to use.The crew experienced lower workload in the Prediction Display used under current automation, possibly because its prediction function is similar to that of the green arc used for altitude prediction in the current glass cockpit.Also, the Position Display was found to be easier to use in the auto-speed (future) condition. +2) Controller WorkloadMean workload scores indicated very low workload across all three ATC positions (Table 2) and an Analysis of Variance did not yield a statistically significant effect of controller position on workload.The means and standard deviations would seem to suggest that overall workload was low; however, the ranges suggest that it was high at various times.Finally, while the area coordinator had a slightly higher workload compared to the other two positions, the range of the scale means across all positions is less than 0.2, supporting the lack of statistical significance. +Position +3) Workload DiscussionStudy results suggest that overall workload for both the pilots and controllers was low enough to be manageable.However, workload ratings varied, suggesting that workload was occasionally high enough to require the necessary attention to maintain performance.Crew workload ratings suggest that the style of information presentation in the Position Display coupled with the auto speed control automation's speed management (condition C) incurred the lowest workload. +D. Situation AwarenessAt the end of each simulation run, study participants completed three subscales of the Situation Awareness Rating Technique (SART) [14].The three questions were answered on a 7point scale from "very low" to "very high" and were always answered in the same order.The questions query the respondent's understanding of, demand from, and supply of attention available to complete their task.Also, an overall SART scale measure was obtained by combining the three subscales in accordance with established practice [14].The supply subscale rating was subtracted from demand and then this result was subtracted from the understanding rating.Therefore, the SART scale has potentially a 19 point range from -5 (extremely low SA) to 13 (very high SA). +1) Pilot Situation AwarenessOverall, crews rated their situation awareness as "good" (mean overall SART=7.8).On the subscales, crews rated their understanding and supply of attention as "high" while the demand was rated as "medium," suggesting they felt they had enough resources to meet the demand.Pilots' ratings for the three questions (and the overall SART) were compared by the two display conditions.The mean values for this comparison are shown in Figure 7.The displays' mean ratings are very close for the SART and its components, with pilots estimating a slightly higher level of understanding of the Position Display with a lower demand but also a lower attention supply for this display set.A Wilcoxon Signed-Rank test shows that pilots found the Prediction Display was significantly more demanding than the Position Display (Z=-2.115,p=.034, meanD1=4.0,meanD2=4.1).Given the verbal feedback crews gave about the displays, this is a meaningful difference.Neither the SART nor the other two questions were significantly statistically different between display conditions.The SART ratings were also evaluated by the two levels of automation conditions.Again, there are no clear differences between the SART ratings by the two automation levels (meanCurrent = 7.6, meanAuto speed control=8.0).Comparing the means for the three questions indicates that pilots found the current automation more demanding (p=.025, meanCurrent =4.2, meanAuto speed control=3.9).Again, although this is a small difference, it is meaningful.While not statistically significant, the other means may suggest that the pilots had less understanding of the situation and slightly more attention capacity when using the current automation.Figure 8 shows that demand on attention was moderate, situation understanding was very high and supply of attention was also high.These results indicate attention demand was low enough to be manageable, but high enough to prevent tedium and vigilance decrement.Results also suggest the controller participants understood the scenarios quite well and that they were not overwhelmed by the task at hand (supply of attention).Collectively, these results would suggest that overall level of controller situation awareness was high.No statistically significant results were indicated between the three ATC positions on all of the three measures.However, the directionality of means indicates coordinator attention demand was somewhat higher relative to the other two positions, which would reflect the area coordinator's greater responsibilities overseeing an area encompassing multiple sectors, pairing the aircraft in different sectors, and monitoring the pairs and the flow.Although there is no statistical significance,, the trend shows that the area coordinator is required to perform a higher level of multi-tasking relative to the other two positions. +3) Situation Awareness DiscussionResults suggest that the overall level of pilot and controller situation awareness was high.Demand on attention was low enough to be manageable, but high enough to prevent tedium.Participants reported good understanding of the situation and a high supply of attention indicating that they were not overwhelmed by the task at hand.Crew situation awareness ratings support a preference for condition C as this condition was reported to be less demanding than the others. +E. Participant Opinions 1) Feedback from the pilotsFlight deck crews had a number of opportunities to comment generally on the concept and other aspects of the study.They raised concerns over procedures; for example one or two pilots noted that they had to try to fit in with the controller managing the speed of the lead aircraft suggesting that controller-pilot roles need clarification.In addition, they also indicated confusion over the way current speed restriction procedures were related to the pairing procedures.Situation awareness (SA) was a concern among some pilots.There were SA concerns on approaches where the leader originated in a different sector, which meant crews could not hear ATC communications with the leader.There were also problems with, or omissions on, the displays that crews said made awareness hard to maintain.A number of crews requested more information about the lead aircraft, which some said would be necessary to increase their comfort with reducing current standard separation during pairing.One pilot had a general concern with SA in the concept, stating that it required too much heads-down time, thus losing outside reference and traffic avoidance.Few comments were collected regarding about workload.Only one crew commented that the mental workload for the pairing task was high.Another pilot noted that workload approaching the coupling point was high.This is potentially problematic because at the coupling point, crews will be busy with tasks to prepare for landing.High workload earlier on the arrival would be more acceptable.There was also some concern that since landing was not required in the study, the full workload of this phase of flight was not represented.The majority of the general comments were suggestions for display modifications.Overall, pilots preferred the Position Display over the Prediction Display, referencing issues of confusability and apparent instability of the timing parameters related to the pure error calculations.It is interesting to note that the feedback provided on the post-run questionnaire did not indicate any of the confusion about the Prediction Display that was reflected in the debrief or the post-simulation questionnaire.Since the post-run questionnaires were administered after each scenario, the researchers feel that the pilots completed these questionnaires quickly and perhaps with less care so they could move onto the next simulation scenario or take a break.This may have prompted them to rush through their responses and may therefore account for the inconsistency with the other data.The pilots did suggest that dampening the variability in the error parameters would help for the usability of the timing data.In addition, three common requests were for more information about the lead aircraft, repeating the key conformance data on the PFD and the navigation displays, and modifying the depiction of some of the data. +2) Feedback from the controllersAir traffic controllers had several opportunities to provide feedback on the pairing procedures and the distribution of roles and responsibilities between the controllers and pilots.The controllers mentioned that once aircraft are paired, they are not inclined to break the pair unless the flight deck informs them of their inability to stay in the pair.The controllers seem to feel that if there were concerns about the pairs, they preferred to keep the traffic flow stable.The controllers often gave speed commands to the lead aircraft to indirectly manage the following aircraft and keep it inside its conformance bars.The controllers also expressed frustration at their inability to control the following aircraft directly, since automation on the flight deck was managing its speed to ensure that it landed in the safe window of 5-25s behind the lead aircraft.If the lead and following aircraft were in different sectors, the two sector controllers' collaboration and communication was increased.The controllers were responsible for inter-pair spacing and the automation handled intra-pair spacing and this procedure also impacted communication workload between the two sector controllers.Different sources of information were used to draw the conformance monitoring graphics for the flight deck and controllers.This sometimes presented different information to the air and ground and became a source of confusion.There is need to have not only clarity in the division of roles and responsibilities but also have clarity on authority and level of hierarchy. +V. SUMMARYThe objective of this study was to explore the procedures and information requirements for pairing aircraft for VCSPR.This study focused on three key metrics: the spacing of the aircraft relative to each other by the beginning of the VCSPR and operators' reports of their workload and situation awareness.All three metrics, when taken at a study-wide level, indicated the feasibility of the concept; ATC and the flight deck crews were able to maneuver aircraft into a paired approach and to cross the coupling point in their specified window using the automation options and flight deck display features in all conditions.The average workload for both ATC and crews was manageable and the average situation awareness was adequate.There were some issues with procedures and information requirements.For example, flight crews were hesitant to cancel their pairing when they were close to coupling if their aircraft was just outside the +/-10s window.The crews' behavior and feedback indicate that they felt they were "close enough" to the spacing parameters when they were in that range.Procedural parameters need to be specified at a greater level of detail to avoid this in future studies.While the pilots and controllers were able to complete their respective pairing tasks with the information provided, the presentation format of key information seemed critical to their performance.Scan patterns and heads-down time were concerns expressed by the pilots.They suggested key information should be redundantly presented on their focal displays (PFD, ND) and should be filtered to indicate when they need to act.However, this study has offered an optimistic start by investigating the integrated dynamic role of controllers and pilots and clarifying where controller-pilot-automation interaction confusions exist.The results of this simulation have indentified the need for additional investigation.. Future research is needed to define information requirements for pilots and controllers when conducting pairing operations for parallel runways.Further study is also necessary to determine when the cancellation of pairing may be required, and the impact cancellations may have on arrival procedures.Figure 1 :1Figure 1: Display conditions 1 and 2 for the navigation display (ND).Triangles represent the following aircraft for both displays. +Figure 2 :2Figure 2: Partial view of the finalized pairs in the controller's pairs table.Leading and trailing aircraft are suggested within the list for each category. +Figure 3 .3Figure 3. Number of aircraft pairs created/deleted and number of aircraft flown as singles (Note: The error bars on all figures, where indicated, represent 1 standard error above and below the mean). +Figure 4 :4Figure 4: Distance of ACFS-follower behind its leader at the coupling point +Figure 5 .5Figure 5.In trail temporal separation between leader and follower aircraft within all aircraft pairs at the Coupling Point. +Figure 6 :6Figure 6: Mean workload by study condition.This figure shows only a portion of the scale. +Figure 7 :7Figure 7: Mean SART ratings under two display conditions 2) Controller Situation Awareness +Figure 88Figure8presents the results on these three measures for each controller position. +Figure 8 .8Figure 8. Situation awareness by controller position. +Table 1 .1Flight Crew Experimental Conditions.PositionPredictionDisplayDisplayCurrentCondition ACondition BAutomationAuto SpeedCondition CCondition DControl +Table 2 .2Controller workload statistics for 47 runs.MeanSDRangeBoulder1.40.484.00Coordinator1.50.525.00Niles1.30.444.00 + + + +FIRST AUTHOR BIOGRAPHYSavita Verma has an M.S. (Human factors) from San Jose State University and has been working with NASA for the last ten years.Her research areas include human performance modeling, datalink, surface and terminal operations. + + + + + + + Frequency and severity of deviations during simultaneous independent approaches to parallel runways + + SebastianVMassimini + + + GerryCMcneil + + + NoamTene + + 10.1109/icnsurv.2010.5503313 + + + 2010 Integrated Communications, Navigation, and Surveillance Conference Proceedings + Mclean, VA + + IEEE + 2006 + + + + Massimini, S. (2006) Simultaneous independent and dependent parallel instrument approaches: Assumptions, analysis, and rationale. Mitre Product: MP 06W0000127, Case number 06-0984. Mclean, VA. + + + + + + Eurocontrol + + + The European Air Traffic Master Plan + + 2010. December 15. 2010 + + + EUROCONTROL (2010).The European Air Traffic Master Plan. December 15,2010. + + + + + Comparison of manual and autopilot breakout maneuvers with three closely spaced parallel runway approaches + + SavitaVerma + + + SandraLozito + + + ThomasKozon + + + GordonHardy + + + HerbertResnick + + 10.1109/dasc.2009.5347457 + + + 2009 IEEE/AIAA 28th Digital Avionics Systems Conference + Orlando, FL + + IEEE + 2009 + + + Verma, S.A., Lozito, S., Ballinger, D., Kozon, T., Hardy, G. & Resnick, H. (2009) "Comparison of Manual and Autopilot Breakout Maneuvers for Three Closely Spaced Parallel Runways approaches." Digital Avionics Systems Conference, Orlando, FL. + + + + + Case Study of Paired Approach Procedure to Closely Spaced Parallel Runways + + JonathanHammer + + 10.2514/atcq.8.3.223 + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 8 + 3 + + 2000 + American Institute of Aeronautics and Astronautics (AIAA) + + + Hammer, J. (2000). Case study of paired approach procedure to closely spaced parallel runways. Air Traffic Control Quarterly, 8 (3), 223-252. + + + + + Preliminary Guidelines on Controller's Procedures for Pairing Aircraft for Simultaneous Approaches under Different Levels of Automation + + SavitaVerma + + + ThomasKozon + + + DeborahBallinger + + 10.1201/ebk1439834916-c52 + + + Advances in Human Factors and Ergonomics Series + Orlando + + CRC Press + 2010. July 2010 + + + + Verma, S, Kozon, T., & Ballinger, D (2010). Preliminary Guidelines on Controller's Procedures for Pairing Aircraft for Simultaneous Approaches under Different Levels of Automation. Applied Human factors and Ergonomics, Orlando, July 2010. + + + + + Pair approach: a closely-spaced parallel runway approach concept + + TRBone + + + OOlmos + + + AMundra + + + + International Symposium on Aviation Psychology + Columbus, OH + + 2001 + + + Bone, T. R., Olmos, O. and Mundra, A. (2001) Pair approach: a closely-spaced parallel runway approach concept. Presented at the International Symposium on Aviation Psychology, Columbus, OH + + + + + The Frankfurt Capacity Program: Wake Vortex Considerations for paired approaches + + DeutscheFlugsicherung + + + 1999 + + + Deutsche Flugsicherung, "The Frankfurt Capacity Program: Wake Vortex Considerations for paired approaches". 1999. + + + + + Displaying procedural vs. real-time information for paired approaches [ATC] + + SJLandry + + + ARPritchett + + 10.1109/dasc.2002.1052973 + + + Proceedings. The 21st Digital Avionics Systems Conference + The 21st Digital Avionics Systems ConferenceCambridge, MA + + IEEE + 2002. 2002 + + + Landry, S and Pritchett., A. (2002). Examining assumptions about pilot behavior in paired approaches. In Proceedings of HCI Aero 2002. Cambridge, MA. + + + + + NASA administrator announces resignation + + TAbbott + + + DElliot + + 10.1016/s0045-8732(01)80042-5 + Concept. NASA/TP-2001-210665 + + + COSPAR Information Bulletin + COSPAR Information Bulletin + 0045-8732 + + 2001 + 152 + + 2001 + Elsevier BV + + + Abbott, T. & Elliot, D. (2001). Simulator Evaluation of Airborne Information for lateral Spacing (AILS) Concept. NASA/TP-2001-210665. + + + + + + MWaller + + + TDoyle + + + FMcgee + + Analysis of the Role of ATC in the AILS Concept, NASA/TM-2000-210091.Symposium on Aviation Psychology + Columbus, OH + + 2000 + + + Waller, M., Doyle,T., & McGee,F. (2000). Analysis of the Role of ATC in the AILS Concept, NASA/TM- 2000-210091.Symposium on Aviation Psychology, Columbus, OH. + + + + + CNS requirements for precision flight in advanced terminal airspace + + MEMiller + + + SDougherty + + + JStella + + + PReddy + + 10.1109/aero.2005.1559471 + + + 2005 IEEE Aerospace Conference + + IEEE + 2005 + + + + Miller, M., Dougherty, S., Stella, J., and Reddy, P. (2005) "CNS Requirements for Precision Flight in Advanced Terminal Airspace," Aerospace, 2005 IEEE Conference, pp. 1-10. + + + + + Evaluation of an airborne spacing concept to support continuous descent arrivals + + JLMurdoch + + + BEBarmore + + + BTBaxley + + + TSAbbott + + + WRCapron + + + + Eighth USA/Europe Air Traffic management Research and Development Seminar + Napa, CA + + ATM + 2009. 2009 + + + Murdoch, J.L. Barmore, B.E., Baxley, B. T., Abbott, T. S., and Capron, W. R. (2009) Evaluation of an airborne spacing concept to support continuous descent arrivals. Eighth USA/Europe Air Traffic management Research and Development Seminar (ATM2009). Napa, CA + + + + + Air Traffic Controller Workload: An examination of workload probe (FAA/CT-TN90/60) + + ESStein + + + 1985 + FAA + Atlantic City, New Jersey + + + Stein, E.S. (1985). Air Traffic Controller Workload: An examination of workload probe (FAA/CT- TN90/60). Atlantic City, New Jersey: FAA. + + + + + Evaluation of the Situational Awareness Rating Technique (SART) as a tool for aircrew systems design + + SJSe1con + + + RMTaylor + + + + Proceedings of the AGARD AMP Symposium on Situational Awareness in Aerospace Operations + the AGARD AMP Symposium on Situational Awareness in Aerospace OperationsSeuilly-sur Seine, France + + NATO AGARD + 1989 + 478 + + + Se1con, S.J. & Taylor, R.M. (1989). Evaluation of the Situational Awareness Rating Technique (SART) as a tool for aircrew systems design. Proceedings of the AGARD AMP Symposium on Situational Awareness in Aerospace Operations, CP478. Seuilly-sur Seine, France: NATO AGARD. + + + + + + diff --git a/file764.txt b/file764.txt new file mode 100644 index 0000000000000000000000000000000000000000..28ae6440d7108a90a5910e547b0c11dea0728f26 --- /dev/null +++ b/file764.txt @@ -0,0 +1,368 @@ + + + + +INTRODUCTIONThe FAA allows simultaneous instrument approaches on two and three runways spaced 4300 ft or more apart, as well as Precision Runway Monitor (PRM) approaches on runways 3000 ft apart, at all but three of the 35 busiest domestic airports.When two parallel runways cannot be used simultaneously, arrival capacities halve, which cause delays during instrument meteorological and marginal visual conditions.The runway pairs can only be used for simultaneous arrivals when pilots can provide visual separation.Mundra et al. show that twenty four percent of all delays are caused when the arrival airport is unable to conduct visual approaches due to reduced visibility, lowered ceiling, haze, or fog.Sixty-six percent of these delays are associated with airports that have at least one pair of runways that are closely spaced (i.e., less than 3000 ft) [1].Procedures for triple parallel runways have been examined for runways spaced by at least 2500 ft and for situations in which two, but not all three, aircraft are dependent on each other [5].This paper pushes the limit by examining a new concept and procedures for three simultaneous arriving aircraft that are dependent on each other when the three parallel runways are spaced by 750 ft.The biggest concern with simultaneous landings on runways closer than 1200 ft are breakout maneuvers due to off nominal conditions.This paper describes an advanced concept and procedures used to address off-nominal situations caused by wake and aircraft-blunders, for aircraft flying three simultaneous closely spaced runway approaches during IMC.The paper will describe the experiment, discuss the results on separation between the leader and follower aircraft, the accuracy of flying the breakout trajectory, workload, and situation awareness for the pilots under these off-nominal situations. +A. BackgroundMost of the previous research on very closely spaced parallel approaches has focused on dual runways [2] [3].Some previous research has focused on modeling capacity gains for closely spaced parallel runways using different procedures [4].The research on triple streams of aircraft has been mostly exploratory in nature, investigating the effect of adding a third stream of aircraft on capacity.There have been several procedures defined for triple simultaneous approaches, and each one of them defines a nontransgression zone or a safety net to protect against aircraft blundering or deviating from their intended path towards the other aircraft.In 2000, Hartsfield Atlanta International airport (ATL) commenced a research study to increase capacity [5] that would consider adding a third stream of aircraft to one of the pairs of runways, thus creating triple simultaneous approaches to increase capacity.Gladstone et. al. (2000) [5] led a research effort that explored several procedures for ATL such as Simultaneous Offset Instrument Approach (SOIA) and Along-Track Spacing (ATS).These procedures were then adapted and further investigated for triple approaches for ATL.The SOIA procedure has been successfully implemented in several airports with two runways including St. Louis International and San Francisco International.Since the runways are spaced closer than 3000 ft, one approach path is aligned to the runway and the other approach path is offset at a distance greater than 3000 ft.The aircraft continue on the approach path until they reach visual conditions, also referred to as the clear of clouds point.The weather minimums for SOIA at SFO are 2100ft/4nm visibility at break of clouds.The trailing aircraft on the offset approach must report visual contact with the leading aircraft prior to the Missed Approach Point (MAP) and then air traffic control will issue a visual approach clearance.Once the trailing aircraft receives the clearance, it executes a visual sidestep maneuver to align itself to the runway and land.If the trailing aircraft is unable to make visual contact, it executes a missed approach that takes the aircraft away from the runways.Each pair of aircraft maintains standard wake separation to avoid wake turbulence.Along-Track Spacing (ATS) is another concept for dual closely spaced runways that involves the use of straight-in approaches to each runway.Deutsche Flugsicherung (DFS, 1999a) [6] proposed using 1.5 nmi diagonal spacing between the leading aircraft on one runway and the trailing aircraft on the other runway.Wake turbulence protection is provided by displacing the threshold of the trailing aircraft by 1500 m (4921 ft) and effectively raising the trailing aircraft's flight path by 80 m (262 ft).The concept of relying on a displaced threshold for vertical separation is called the High Approach Landing System (HALS).Similar to SOIA, wake protection between pairs is provided by standard separations.Also, ATS does not require a side step maneuver as is required in SOIA.Thus HALS provides a capability to fly dual streams of aircraft to closely spaced runways that have displaced ILS installed, which permit landings up to Category I minima, potentially allowing for more capacity benefits than SOIA.Gladstone et. al (2000) [5] described four procedures with three approach streams developed for ATL with two sets of parallel runways.The procedures include Independent SOIA Triples, Dependent SOIA Triples, Angled SOIA Triples, and Triples using Along-Track Spacing (ATS).The Independent SOIA for triples procedure requires an independent monitor for each runway and has a 2000 ft No Transgression Zone (NTZ) established between each pair of simultaneous streams.The Dependent SOIA triples build on the current standard that allows dependent approaches to two parallel runways with a minimum of 2500 ft between runways and a diagonal separation of at least 1.5 nmi between aircraft, with one NTZ established between the north and south pairs of runways.So, on ATL's north runways there is one aircraft on 26R, whereas on south runways, the in-board runway has a straight in aircraft, the out-board runway has the aircraft on a SOIA, and the trailing aircraft follows the lead aircraft by 1.5 nmi.The Angled SOIA triple approach is similar to the Independent approach with NTZs except the aircraft on the SOIA is offset by 3 degrees.The ATS triple approach was similar to the Dependent SOIA approach in that there was one approach stream on the outboard runway for the north set of runways and there were two straight-in approaches to the south runways.The two aircraft on the south runways are separated by 1.5 nmi diagonally.The procedures described above still require considerable testing and development, since standards don't exist for most of them, especially the ATS concept.All the procedures for triple runways described in the previous paragraphs have runway spacing of at least 2500 ft apart.Only two aircraft out of the triples are in a formation or dependent on each other, so the third aircraft is not dependent, which limits the capacity benefits.The breakout procedures for the triples in Gladstone's study were similar to those used with SOIA.The Multiple Parallel Approach (MPAP, 2002) [7] team performed a real time simulation to evaluate simultaneous instrument landing system approaches to three parallel runways spaced 4000 ft and 5300 ft apart.The MPAP team introduced blunders to test the concept's ability to maintain adequate separation between aircraft on final approach.Here, the controllers were responsible for separation between aircraft on a triple approach, and they used Precision Runway Monitor with a 1.0 sec update rate to help with the task.The concept investigated in this paper has three runways spaced at 750 ft from each other, and the three aircraft flew in an echelon formation (see Figure 1).The pilots were responsible for separation and were provided new tools and procedures to achieve the same.The following section describes the concept, and the procedures for the triples using Terminal Area Capacity Enhancing Concept (TACEC) [8]. +A. Airport and Airspace DesignThe experiment used a fictitious airport (KSRT) loosely based on the current Dallas/Fort Worth International Airport (DFW) layout and operations except for three parallel runways that were set to be 750 ft apart.Because the simulation focused on TACEC approaches to very closely spaced parallel runways using south flow scenarios, only the west side runways (18L, 18C and 18R) were used.The outside runway (currently 18R) was moved inward to create 18C with a 750 ft separation between the runways and a third 18R was also added at 750 ft.All three of the runways were assumed to be equipped to a CAT-IIIB level. +B. TACEC ProceduresTACEC calls for the three aircraft to be paired at meter fixes located near the edge of the terminal airspace, normally 40-60 nmi from the airport [8] and given TACECassigned 4D arrival trajectories to the runway.Flights in the simulation began 25 nmi from the airport, assuming they were already paired.Routes to the airport included approach and departure routes and procedures similar to those for DFW airport.This study focused upon arrivals, and no departures were studied.TACEC [8] allows for any aircraft arriving from any of the four arrival meter fixes (NE, NW, SE, and SW) to be paired for a simultaneous parallel landing, based on aircraft characteristics and relative timing criteria.The three paired aircraft flew their assigned 4D trajectories with a high level of accuracy to meet timing constraints at the coupling point and to ensure wake safety throughout the approach.TACEC assumes augmented Global Positioning System (GPS) and ADS-B (Automatic Dependent Surveillance-B).The coupling point, which refers to the point at which the speed of the multiple aircraft becomes dependent, or "slaved" to one another, is defined at 12 nmi from the threshold of the runway.From that point onward in the simulation, the center aircraft precisely maintained 12 s spacing behind the lead aircraft, and the right aircraft maintained 24 s behind the lead aircraft using a speed algorithm to avoid the wake and for safe separation.The approach paths of the two trailing aircraft were at a slewed angle from the center of the runway-six degrees for the aircraft on the center runway and 12 degrees for the aircraft on the right runway, when the aircraft were 25 nmi from the threshold.All three aircraft turned straight and parallel to each other at about 2 nmi from the runway.Onboard automation based on ADS-B monitored the three aircraft for potential emergency situations.The automation displayed a predicted hazardous zone for the wake generated by the lead and center aircraft in the cockpits of the second and third planes.ADS-B lateral position and intent information was used to detect and display any deviation from the proposed approach path that would encroach on either of the trailing aircraft.Visual and aural alerts were given to the pilots when the lead-aircraft's blunders or wake presented a dangerous situation to the trailing aircraft.The navigation display depicted a breakout trajectory after the aircraft crossed the coupling point.This breakout trajectory was dynamically generated and considered wake, traffic, buildings and terrain of the airport surroundings.When the breakout was required at different altitudes on the arrival path different bank angles for the breakout maneuvers were used and the curvature of the breakout trajectory changed on the navigation displays.The pilots flew the breakout trajectory manually using the flight director when they received an aural and visual alert. +C. DisplaysThe displays were similar to the displays used for the study of two runway very closely spaced parallel approaches [2] and were based on previous research associated with flight deck displays [9] [10].The Navigation Display (ND) and Primary Flight Display (PFD) are shown in Figures 2 and3.The displays show wake and trajectory information along with standard flight instrument data.After crossing the coupling point and the pilot's prior acceptance of coupling with the lead aircraft, the flight mode annunciation changes to show that the three aircraft are coupled for speed (C-SPD), coupled for lateral navigation (C-LNAV) and coupled for vertical navigation (C-VNAV).The two trailing aircraft were coupled with the lead aircraft.The autopilot flew the approach, the pilot primarily monitored the aircraft performance and the displays for the remainder of the flight.If the wake of the adjacent aircraft drifted within one wingspan of the ownship aircraft, the color of the wake hazardous zone on the display turned to yellow, and then turned red when the apex of the aircraft was in the wake.Similarly, if the lead aircraft deviated from the planned trajectory towards the following aircraft's path by 60 ft, the lead aircraft symbol turned yellow, and then red when the lead aircraft deviated by 120 ft.The red warnings, accompanied by an aural alert "breakout, climb" required a mandatory breakout, which the pilots flew manually.When the pilots pressed the Take-Off-Go-Around (TOGA) switch, the breakout trajectory, which had been displayed to the pilot in white, became the active route, and was then displayed in magenta. +D. Advanced Concept Flight SimulatorThe human-in-the-loop experiment studied breakout maneuvers for triple TACEC approaches in the Advanced Concepts Flight Simulator (ACFS) located at NASA Ames Research Center.The ACFS is a motion-based simulator that represents a generic commercial transport aircraft, enabling it to be reconfigured to represent future aircraft.It has the performance characteristics similar to a Boeing 757 aircraft, but its displays have been modified to study different advanced concepts.In this study, the cockpit displays described in the previous section were integrated with the flight display systems in the cockpit.The visual systems offer a 180 deg horizontal and a 40 deg vertical field of view.This simulator is capable of providing various visibility conditions and was set to IMC for this experiment. +E. VariablesFour independent variables were examined in this study of the TACEC concept for triple runways.First was the presence or absence of an off-nominal situation that may warrant a breakout maneuver.The second independent variable was the cause of the breakout maneuver -wind causing the wake of the lead aircraft to drift towards the following (center) aircraft, or the lead aircraft deviating from its original path and towards the trailing aircraft.The third independent variable was the location of the offnominal situation, which was above 500 ft, or between 200 ft -500 ft above the ground.The fourth independent variable under study was the position of the own-ship or the simulator which could either be approaching the center or right runway.A total of 24 runs were performed for each participant in which 8 were normal and 16 had off-nominal situations.In the runs that required a breakout maneuver, repeated runs were made for each breakout cause, breakout location, and position of the aircraft. +F. HypothesesIn the absence of previous research on triple-runway closely-spaced approaches, the researchers predicted that the location of the off-nominal situation or the nature of the off-nominal situation or the position of the ownship (center versus right runway) would not affect pilots' behavior on the following parameters: Separation from lead at breakout point  Accuracy of flying trajectory  Workload  Situation awareness However, it was expected that there would be differences in situation awareness and workload experienced by the pilots in the runs that have the off-nominal situation versus the runs that do not. +G. ParticipantsThe participants were eight recently retired pilots from commercial airlines; all were male and all of them had experience with glass cockpits.Their average experience as a pilot was about 38 years.Their average number of years since retirement was less than two. +H. Experimental ProcedureThe study ran for eight days with one pilot participating each day.At the beginning of the day, the pilot was familiarized with the project, the concept, and the new displays in the cockpit.The pilot received a demonstration of the ACFS and hands-on training on the flight deck displays and related procedures.Since procedures for triple Very Closely Spaced Parallel Runways (VCSPR) were being explored in this study, each pilot flew the ACFS in the left seat (as captain) along with a confederate who acted as the first officer.The role of the pilot was to fly in auto pilot mode and monitor the displays to check separation with the lead aircraft and with wake.Prior to the coupling point the pilots heard a chime, saw the acknowledgement button light up, and received a "TACEC Coupling" message on the lower Engine Indicating and Crew Alerting System (EICAS) display.At this point the pilots pressed the accept button.They flew as the center or as the trailing aircraft and both of those aircraft were coupled with the leader aircraft on the left most runway.They were coupled with the leader's speed and continued to monitor the separation between the three aircraft.The flight mode annunciation also changed to show that the two aircraft were coupled for speed (C-SPD), coupled for Lateral navigation (C-LNAV) and coupled for Vertical navigation (C-VNAV).If the pilots received a visual and aural alert from the displays they were required to perform a breakout maneuver.To fly the breakout maneuver, the pilot would press the TOGA switch, disengage the autopilot, leave the auto throttle on, and fly the breakout trajectory shown on the ND.Pressing the TOGA switch would capture the breakout trajectory, and the pilots used the flight director to fly the trajectory.They flew different breakout trajectories at different altitudes, with the breakout above 500 ft altitude requiring an initial bank angle of 30 deg, and the breakout Breakout Trajectory Wake at altitudes between 200-500ft requiring an initial bank angle of 10-deg.They had an initial heading change of 20deg if they were the center aircraft on 18C and a heading change of 40-deg if they were the trailing aircraft on 18R.In all the above cases, the aircraft had to climb to 30,00ft as a part of the break out procedure.The pilots then followed the 'S' shaped breakout trajectory displayed on the ND.The trajectory was 'S' shaped so that the final leg of the trajectory became parallel to the runways.The final leg of the breakout trajectory was 1.5 nmi abeam for 18C and 3.0 nmi for 18R. +I. Traffic ScenarioThe traffic scenario had three aircraft: (1) The ACFS (B757) was always one of the two following aircraft (center or trailing) in the triplet, and the other two aircraft were scripted, depending upon the experimental condition, and (2) the leader aircraft was a Boeing 747-400, which was prerecorded and scripted for this study and landed on 18L under nominal conditions.The pilot who flew the ACFS simulator always landed on either 18R or 18C or performed the breakout, depending upon the simulator position for the particular data collection run.Operationally, the trailing aircraft should be upwind of the cross wind, but this is not always possible so scenarios included adverse crosswind. +J. Tools used for Data CollectionSeveral tools were used for collecting subjective data from the pilots.All participants completed a demographic survey before the simulation runs were conducted.The survey collected information about the pilots such as their age, experience, and number of hours flying different aircraft types, any experience with SOIA approaches, and experience using personal computers.All pilots were asked to complete a Post Interaction Survey at the end of all the runs.This survey allowed them to rate the information content and the usability of the displays.The participants completed the NASA Task Load Index (TLX) rating scales [12] after each simulation run but did not complete the pair-wise scale comparison included as part of the TLX, so the six scales were analyzed separately.Pilots also completed the Situation Awareness Rating Tool (SART) [11].The SART gathers a participant's rating of situation awareness (SA) for the preceding period of time on ten different scales.Each scale has 7 points, with the end points representing the opposite ends of the construct.Participants circled the point on the scale that most closely represented their experienced level of SA.The ten SART ratings together with TLX were gathered from every participant at the end of each run -a total of 16 ratings per participant were collected.In addition to the assessment instruments described above, the flight simulator's digital data collection system was used.A host of objective flight data for each of the simulation runs was collected on some of the variables pertinent to the hypotheses of the experiment.All collected data were indexed with a common timestamp, which was used as the basis of time synchronization as it updates in real-time while the simulation run advances.All digital data were collected at a rate of 30 Hz. +III. RESULTS & DISCUSSIONStatistical analysis of the study data focused on three areas: (1) the flight simulator's digital data collection outputs, (2) the pilot participants' workload and situation awareness assessments, and (3) verbal feedback provided by the pilot participants at the end of the simulation runs. +Aircraft Separation from Breakout through 30 Seconds Past BreakoutThe dependent measure of aircraft separation is defined as slant range, or straight-line displacement distance between two aircraft.Analysis of aircraft separation as it changes in time from breakout point was implemented, to determine if there were any instances of unsafe separation between aircraft during the most critical phase of the breakout maneuver, i.e., the time span that immediately follows breakout point, defined as breakout time through 30 seconds past breakout time.Separate analyses were performed in comparing (1) Leading and center aircraft separation, and (2) Center and trailing aircraft separation.Table 1 shows summary statistics associated with these data and Figures 4 and5 show the aircraft separation for all flights as it changed over time originating from breakout point: As indicated in Table 1, and Figures 4 and5, there is a clear trend towards increased separation between each of the two pairs of aircraft analyzed, with some overall increase 15 seconds past breakout, and a larger increased separation at 30 seconds past breakout.The only apparent exception to this trend is shown in Figure 5, showing the separation between the center and trailing aircraft, where some runs show a relatively small decrease in separation 15 seconds past breakout, prior to increased separation 30 seconds after breakout.It is suggested that this trend in the data reflects the complex geometry of the breakout maneuvers in the case of the center aircraft which needs to separate itself from the leader aircraft towards the trailing aircraft, which may initially decrease separation for a very short period of time.Even so, during this critical window of time, there were no cases where the slant range between the center and trailing aircraft was less than 2500 ft, indicating zero instances of unsafe separation (Figure 5).Furthermore, no instances of unsafe separation between the leader and center aircraft were observed (Figure 4).These data compare with the data collected by FAA's MPAP [7], where they defined a test criterion violation (TCV) as 500ft of separation between the aircraft.Using the same definition, no TCV occurred between the lead and center aircraft or between the center and the right most (trailing) aircraft. +Leader/Center Separation +MeanClearly, the objective evidence shows no single instance of unsafe separation during the critical 30 second time period past breakout.In addition, inferential findings comparing study conditions were uncovered which augment this result.Tables 2 and3 A statistically significant main effect of breakout cause was observed (F=89.87,df=1,7, p<0.0001) on the dependent measure of slant range between the leader and center aircraft.Aircraft separation was greater under the wake condition, as compared to the aircraft blunder condition.This effect may be due to the relative uncertainty of wake behavior, resulting in the pilots attempting to achieve more separation when the wake drifts to allow for the unpredictable performance of wake phenomenon.A statistically significant main effect of breakout location was observed in comparing the slant range separation between and leader/center aircraft (F=20.45,df=1,7, p<0.01), and also between the center/trailing aircraft (F=44.73,df=1,7, p<0.001),where separation was greater at the higher altitude breakout.This effect reflects the different post-breakout geometries between the aircraft, where breakout procedures require an initial 30 degree bank angle at higher altitudes, and only a 10 degree bank angle at lower altitudes.While a main effect of center/trailing ownship was not realized on the dependent measure of aircraft separation between the leader and the center aircraft at breakout, it was realized in comparing the aircraft separation between the center and trailing aircraft at breakout (F=40.39;df=1,7; p<0.001).Separation was greater when the ownship was the center aircraft, as compared to the trailing aircraft.This might be indicative of the unique position of the center aircraft, where separation involving the ownship and two other aircraft are necessary to maintain safety, whereas the trailing aircraft needs to maintain separation with only one other aircraft.Due to this unique physical position between the other two aircraft, special vigilance may have been exercised by the pilot, resulting in increased separation. +Accuracy of Breakout Trajectory: Cross Track and Track Angle ErrorTrajectory accuracy is measured by the actual ownship/simulator position against the breakout trajectory generated by the system and displayed on Navigation Display averaged across time.Two measures of ownship trajectory particularly sensitive to breakout maneuvers include cross track error and track angle error.For each flight simulation run, cross track error and track angle error were averaged across time from the breakout point to the end of the flight.Two-way repeated measures ANOVA yielded main effects of breakout cause, breakout location, and center/trailing ownship on each of the two dependent measures.All of these results are consistent with respect to the directionality of the means across both track angle and cross track error.More cross track error and more track angle error were observed (1) when the cause of breakout was wake, (2) at breakout locations above 500 ft as compared to breakout locations at or below 500 ft., and (3) when the ownship was the trailing aircraft.ANOVA summary statistics on the significant results from this analysis are listed in Tables 4 &5.The pilots flew the breakout trajectories with higher precision under the condition where aircraft deviation led to a breakout.It is possible that the uncertainties and unpredictable nature of aircraft deviations and a faster developing hazardous situation might have led the pilots to precisely follow the breakout trajectory generated by the automation. +Mean (ft) SD (ft)CauseThe effect of breakout location may have occurred due to the perceived immediacy of the response at an altitude of below 500 ft, since airspace is highly congested close to major airports at lower altitudes, requiring increased vigilance of flight crews at this stage of approach.Lower approach altitudes introduce special concerns, with possible pilot errors creating an increased chance of dangerous consequences.Pilots are also keenly aware of other possible factors, such as low altitude wind shear, which could have the effect of complicating an already dangerous situation.Hence, perceived immediacy of the response, combined with increased vigilance, may have contributed to this effect.From an operational perspective, this may suggest that the pilot participants are inherently assessing the need for a very accurate response to a dangerous situation during flight times that may have other immediate and critical issues.Also, the location effect may in part reflect breakout procedures, where the maneuver below 500 ft has an initial bank angle of 10 deg, which is fairly easy to execute with the side-stick control used in the ACFS, allowing the pilots to fly the breakout trajectory projected on the ND more accurately.Therefore, this result should be interpreted for its relativity to the other independent variables and as providing trend information.It seems likely that the effect of center/trailing ownship on track angle and cross track error is related to the perception that the center aircraft may be the most vulnerable to possible unsafe separation or wake, due to its dual proximity to both the leading and the trailing aircraft.Hence, it is speculated that a potentially dangerous situation requiring breakout might generate more psychological discomfort when in the center position, motivating a greater degree of vigilance in conducting the breakout maneuver.On the other hand, when the ownship is the trailing (right most) aircraft, the most immediate concern involves possible unsafe separation/wake with only one aircraft, causing less initial discomfort and vigilance required to escape the potentially dangerous situation. +Mean +WorkloadParticipants completed the NASA TLX workload questionnaire after every run.Data were collected on each of the six TLX workload measures, and a variable measuring overall workload combining all six of these measures was derived.Data analysis comparing breakout vs. non-breakout runs on each of the workload measures was implemented.ANOVA results indicated that pilot workload was significantly higher in breakout runs as compared to nonbreakout runs in 5 of the 7 workload measures, as well as the overall workload composite measure (p<0.05).This was expected, since breakout procedures require pilots to manually fly the ownship according to the breakout trajectory rather than monitor the displays in the normal approach procedures.Figure 6 shows the mean score of each of the 6 workload measures, as well as the overall composite score, broken down by run category (breakout / nonbreakout).Looking at the data collected from the breakout runs only, it was found that the pilots' overall workload was reasonably manageable with a mean composite score of 21.98 (sd = 5.83), where 6 indicates low workload and 42 indicates high workload.A significant main effect of breakout location was observed on this composite score (F=6.97;df=1,7; p<0.05),where the higher altitude breakout generated higher workload (mean=23.15,sd=6.13)than the lower altitude breakout (mean=20.79,sd=5.31).It seems likely that this effect is due to the differing breakout procedures, where the higher altitude breakout requires a bank angle maneuver of 30 degrees while the breakout at lower altitudes requiring only a 10 degree bank angle, making one maneuver more aggressive than the other.In addition to this main effect of breakout location, a significant interaction effect of breakout cause by center/trailing ownship on composite workload was also observed (F=11.07;df=1,7; p<0.05).Under the aircraft blunder condition, higher workload was observed when the ownship was the center aircraft, and under the wake condition, higher workload was observed when the ownship was the trailing aircraft.Figure 7 shows this interaction graphically.Again, since blundering aircraft have an unpredictable nature, it makes sense that the pilots of the center ownship would have more workload than pilots of the trailing ownship under the aircraft blunder condition, since the center ownship has two neighboring aircraft it needs to maintain safe separation with, and the trailing aircraft only needs to maintain separation with one.However, wake could seem to pose a greater concern with the trailing aircraft, since the trailing aircraft may perceive the need to avoid the wake generated by the two aircraft in front of it whereas the center aircraft needs to avoid the wake from the leader aircraft only. +Situation AwarenessThe SART scale, mentioned earlier, measures situation awareness on ten scales.Participants provided ratings on each of these ten scales after every simulation run.All collected SART data were then used to derive three broader categories [11] concerned with a) the demands of the situation b) the 'supply' or personal resources that the participants have to bring to the situation and c) situational provision that the situation provides in the form of information through displays.The first broad category combines the three SART scales -instability, variability and complexity of the situation, where the values can range from 3 to 21.The second broad category of personal resources combines the SART scales on alertness, spare mental capacity, concentration, and division of attention, where the resultant scores can range from 4 to 28.The third broad category, situation provision combines the three SART scales on information quantity, information quality, and familiarity, and the resultant value can range from 3 to 21.Statistical analysis comparing nominal and breakout conditions on situation awareness of the pilot participants yielded a significant difference on the scale of situational demands (F=25.46,df=2,6, p<.01).The situation demands of the breakout runs were higher than the nominal runs.This result is consistent with the result of higher pilot workload levels in the off-nominal (i.e., breakout) condition, which correlate with higher levels of instability and variability, as compared to the nominal condition.This would be expected, since the off-nominal condition requires that pilots safely maneuver the aircraft by following the breakout trajectory, rather than implement normal approach procedures.Results on personal resources indicate almost no difference between nominal and breakout runs.This may be due to the anticipation of a breakout anytime, which required equal levels of alertness and concentration across nominal and breakout runs.Likewise, there was almost no difference between the nominal and breakout runs in situation provision, suggesting equal amounts of information quantity, information quality, and familiarity, providing some support for the efficacy of the TACEC concept.The means and standard deviations of the three situation awareness variables across both conditions are graphically depicted in Figure 8.Further analyses of the SART data within the breakout condition revealed no meaningful significant effects.Hence, the pilots experienced similar levels of situation awareness irrespective of the cause of breakout, the location of the breakout, or center/trailing ownship.Finally, relative to the possible range of values for each of the three composite situation awareness measures, Figure 8 indicates high levels of personal resources and situation provision, with moderately low levels of situation demands, suggesting that situation awareness was maintained throughout the course of the current investigation, providing support for the TACEC concept. +IV. CONCLUSIONSTriplet aircraft procedures were investigated in a high fidelity human-in-the-loop simulation incorporating new tools and technologies involving very closely spaced parallel runway operations.Scenarios included nominal and off-nominal cases.Statistically significant differences were observed.The results indicated that overall, pilots successfully "flew" the simulator through all of the study scenarios, both accurately and safely within and across all conditions.An analysis of aircraft separation between leader/center aircraft and between center/trailing aircraft during breakout indicated zero instances of unsafe separation.During breakout, the minimum observed slant range between all aircraft across all conditions was 2437 ft., which is well above the FAA's MPAP test criterion violation threshold of 500 ft separation between aircraft.Further analysis of aircraft separation during breakout indicated statistically significant differences of cause and location of breakout, as well as center vs. trailing ownship, suggesting that pilots may be more inclined to fully trust the automation to guide them along the breakout trajectory when confronted with an aircraft blunder, and separation may vary as a result of differences in breakout procedures at different altitudes, as well as the unique position of the ownship among the triplet aircraft.Analysis of cross track and track angle error indicated that overall, the breakout trajectory was flown quite accurately across all conditions.The pilots experienced higher workload and situational demands placed on them during breakout as compared to the normal approach procedure.While realizing these differences, the results also indicate that workload was manageable, and an adequate level of situational awareness was maintained across all conditions.Overall, the data indicate that very closely spaced triplet parallel runway approach procedures can increase efficiency of flight operations, while maintaining an adequate level of safety.While more research is necessary, these results attest to the potential promise of the current concept under investigation.Figure 1 .1Figure 1.The line formation for triples +Figure 2 :Figure 3 :23Figure 2: Navigation Display during final approach +Figure 5 .5Figure 4. Aircraft Separation Immediately Following Breakout: Leader/Center Slant Range (each time-series represents one simulation "flight") +Figure 6 .6Figure 6.Effects of Breakout on Pilot Workload Measures (* indicates p<0.05; error bars represent ± 1 standard deviation; composite score on a scale of 1-7) +Figure 7 .7Figure 7. Significant Interaction Effect of Breakout Cause by Center/Trailing Ownship on Overall Pilot Workload +Figure 8 .8Figure 8. Effects of Breakout on Pilot Situation Awareness Measures (* indicates p<0.05; error bars represent ± 1 standard deviation) +Table 1 . Aircraft Separation Following Breakout1SDMaxMin(ft)(ft)(ft)(ft)Breakout Point2550962675243715 Seconds Past286315832182534Breakout30 Seconds Past285432243522987BreakoutCenter/TrailingSeparationBreakout Point2854592979279415 Seconds Past287213431912545Breakout30 Seconds Past365150245592918Breakout +Table 2 . Aircraft Separation 15 s past breakout: leader/center slant range CENTER/TRAILING AIRCRAFT SEPARATION Mean (ft) SD (ft) Location: F=44.73 df=1,7 p<0.001 Breakout Location > 500 ft 2924 169 Breakout Location ≤ 500 ft 2828 892provide ANOVA statistics (F values) on the dependent measure of aircraft separation 15 seconds past breakout.LEADER / CENTERMeanSDAIRCRAFT(ft)(ft)SEPARATIONCause: F=89.87 df=1,7p<0.0001Aircraft Blunder279194Wake2958253Location: F=20.45 df=1,7p<0.01Breakout Location > 500 ft 2949268Breakout Location ≤ 500 ft 280062Location: F=40.39 df=1,7p<0.001Center Ownship2962128Trailing Ownship2791100 +Table 3 . Aircraft Separation 15 s past breakout: center/trailing slant range3 +Table 4 . Significant Main Effects on Ownship Cross Track Error4: F=10.37 df=1,7p<0.05Aircraft Blunder7379Wake103102Location: F=48.09 df=1,7p<0.001Breakout Location>130110500 ftBreakout Location≤4638500 ftCenter/Trailing Aircraft:F=5.20 df=1,7 p=0.05Ownship Center7391Ownship Trailing10491 +Table 5 . Significant Main Effects on Ownship Track Angle Error5SD(deg)(deg)Cause: F=10.50df=1,7 p<0.05Aircraft Blunder2.281.62Wake2.872.59Location: F=58.75df=1,7 p<0.001Breakout Location3.652.46>500 ftBreakout Location ≤1.501.08500 ftCenter/TrailingAircraft: F=10.09df=1,7 p<0.05Ownship Center1.962.02Ownship Trailing3.202.16 + + + +AUTHOR BIOGRAPHY + + + + + + + Potential benefits of a Paired Approach procedure to closely spaced parallel runways in instrument and marginal visual conditions + + AnandDMundra + + + WayneWCooper + + + ArthurPSmith + + + LaurenceFAudenaerd + + + ClarkRLunsford + + 10.1109/dasc.2008.4702807 + + + 2008 IEEE/AIAA 27th Digital Avionics Systems Conference + St. Paul, MN + + IEEE + 2008. October 2008 + + + A. Mundra, W. Cooper, A. Smith, L. Audenaerd, and C. Lunsford (2008). "Potential benefits of a paired approach procedure to closley spaced parallel runways in instrument and marginal visual conditions," Digital Avionics System Conference, St. Paul, MN, October 2008. + + + + + Procedures for off-nominal cases: Very closely spaced parallel runway operations + + SavitaVeram + + + SandyLozito + + + ThomasKozon + + + DeborahBallinger + + + HerbertResnick + + 10.1109/dasc.2008.4702783 + + + 2008 IEEE/AIAA 27th Digital Avionics Systems Conference + St. Paul, MN + + IEEE + October 2008 + + + S. Verma, S. Lozito, T. Kozon, D. Ballinger, and H. Resnick, "Procedures for off-nominal cases: very closely spaced parallel runways approcahes," Digital Avionics System Conference, St. Paul, MN, October 2008. + + + + + Procedures for off-nominal cases: Very closely spaced parallel runway operations + + SavitaVeram + + + SandyLozito + + + ThomasKozon + + + DeborahBallinger + + + HerbertResnick + + 10.1109/dasc.2008.4702783 + + + 2008 IEEE/AIAA 27th Digital Avionics Systems Conference + Anchorage, Alaska + + IEEE + 2008 + + + S. Verma, S. Lozito, G. Trot, "Preliminary Guidleines on Flight Deck procedures for very closely spaced parallel approaches." International Council for Aeronautics (ICAS) Anchorage, Alaska, 2008. + + + + + Modelling the capacity of closely-spaced parallel runways using innovative approach procedures + + MilanJanic + + 10.1016/j.trc.2008.01.003 + + + Transportation Research Part C: Emerging Technologies + Transportation Research Part C: Emerging Technologies + 0968-090X + + 16 + 6 + + December, 2008 + Elsevier BV + + + M. Janic, "Modelling the capacity of closley spaced parallel runwayss using innovative approach procedures," Transportation Research Part C:Emerging Technologies. Volume 16, Issue 6, December, 2008, pg 704-730. + + + + + Analysis of Triple arrivals to Hartsfield Atlanta International Airport + + CGladstone + + + GGlover + + + PMassimini + + + CShiotsuki + + + BSimmons + + + + MTR00W0000023 Mitre Technical Report + + 2000 + + + C. Gladstone, G. Glover, P. Massimini, C. Shiotsuki, and B. Simmons, "Analysis of Triple arrivals to Hartsfield Atlanta International Airport." MTR00W0000023 Mitre Technical Report, 2000.. + + + + + The Frankfurt Capacity Program: Wake Vortex Considerations for paired approaches + + DeutscheFlugsicherung + + + 1999 + + + Deutsche Flugsicherung, "The Frankfurt Capacity Program: Wake Vortex Considerations for paired approaches." 1999 + + + + + Training Pilots Prior to Simultaneous Instrument Landing System (ILS) Precision Runway Monitor (PRM) Approaches to Closely-Spaced Parallel Runways + + SMagyarits + + + ROzmore + + 10.4271/arp5453 + + + May, 2002 + SAE International + + + S. Magyarits & R. Ozmore, "Evaluation of triple Indpendent Instrument Landing System Approaches to Runway Spaced 4000 ft and 5300 ft apart using a Precision Runway Monitor System." DOT/FAA/CT-TN02/16. May, 2002. + + + + + CNS requirements for precision flight in advanced terminal airspace + + MEMiller + + + SDougherty + + + JStella + + + PReddy + + 10.1109/aero.2005.1559471 + + + 2005 IEEE Aerospace Conference + + IEEE + + + + + M. Miller, S. Dougherty, J. Stella, and P. Reddy, "CNS Requirements for Precision Flight in Advanced Terminal Airspace," Aerospace, 2005 IEEE Conference, pp. 1-10. + + + + + Cockpit Display of Traffic and Wake Information for Closely Spaced Parallel Approaches + + GordonHardy + + + EmilyLewis + + 10.2514/6.2004-5106 + + + AIAA Guidance, Navigation, and Control Conference and Exhibit + Providence, RI4 + + American Institute of Aeronautics and Astronautics + 2004 + + + Hardy, G. H., & Lewis, E. K., 2004, A Cockpit Display of Traffic Information for Closely Spaced Parallel Approaches, AIAA-2004- 5106, AIAA Guidance, Navigation, and Control Conference and Exhibit, Providence, RI4. + + + + + Models of Wake-Vortex Spreading Mechanisms and Their Estimated Uncertainties + + VernonRossow + + + GordonHardy + + + LarryMeyn + + 10.2514/6.2005-7353 + AIAA-2005-7353 + + + AIAA 5th ATIO and16th Lighter-Than-Air Sys Tech. and Balloon Systems Conferences + Arlington, VA + + American Institute of Aeronautics and Astronautics + 2005 + + + Rossow, V. J., Hardy, G. H., and Meyn, L. A., 2005, Models of Wake-Vortex Spreading Mechanisms and Their Estimated Uncertainties, AIAA-2005-7353, ATIO Forum, Arlington, VA. + + + + + Situation Awareness as a Predictor of Performance for En Route Air Traffic Controllers + + FrancisTDurso + + + CarlaAHackworth + + + ToddRTruitt + + + JerryCrutchfield + + + DankoNikolic + + + CarolAManning + + 10.2514/atcq.6.1.1 + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 6 + 1 + + 1998 + American Institute of Aeronautics and Astronautics (AIAA) + + + Durso, F. T., Hackworth, C. A., Truitt, T. R., Crutchfield, J., Nikolic, D., & Manning, C.A.,1998, Situation Awareness as a Predictor of Performance for En Route Air Traffic Controllers, Air Traffic Control Quarterly, 6(1), pp. 1-20. + + + + + Development of NASA-TLX (Task Load Index): Results of Empirical and Theoretical Research + + SandraGHart + + + LowellEStaveland + + 10.1016/s0166-4115(08)62386-9 + + + Advances in Psychology + + PHancock + + + NMeshkati + + Amsterdam, The Netherlands + + Elsevier + 1988 + + + + Hart, S., & Staveland, L.,1988, Development of a Multi-Dimensional Workload Rating Scale: Results of Empirical and Theoretical Research, In P. Hancock & N. Meshkati (Eds.), Human Mental Workload, Amsterdam, The Netherlands, Elsevier, pp. 139-183. + + + + + + diff --git a/file765.txt b/file765.txt new file mode 100644 index 0000000000000000000000000000000000000000..73a5c1e89e9d7c23f9e4b0d2714781dd7e674c98 --- /dev/null +++ b/file765.txt @@ -0,0 +1,311 @@ + + + + +IntroductionThe NextGen air transportation system is being designed with the expectation that the volume of the traffic will double or triple by 2025 [1].Many air transportation forecasts expect a significant growth for air travel demand.To meet this demand, parallel runways operations are a potential solution to increasing the throughput of an airport.Several airports like Chicago's O'Hare, Dallas Fort Worth and Denver International depend on parallel runways operations to meet growing demand.The FAA has successfully conducted independent approaches to parallel runways for over 40 years using the Instrument Landing System (ILS) navigation and terminal radar monitoring [2].The simultaneous approaches that use standard radar are conducted on parallel runways that are at least 4300 ft apart.To conduct parallel approaches on runways that have 3000 ft spacing between them requires the use of Precision Radar Monitor (PRM) with an update of rate of 1.0 s. [2] Some airports, like San Francisco International airport, can support approximately 60 landings per hour using both of the parallel runways that are 750 ft apart by using the Simultaneous Offset Instrument Approach (SOIA) [3].SOIA approaches require the trailing aircraft in the paired approach to obtain a visual sighting of the lead aircraft, and at least a 2100 ft ceiling and 3 nm visibility.As weather degrades, the current navigation and surveillance systems, and existing procedures, do not provide the accuracy necessary to support SOIA approaches.This reduces the landing rate to half the Visual Flight Rules (VFR) capacity.In the SOIA procedures, air traffic control is responsible for pairing the aircraft, detecting any blunders and commanding breakout maneuvers, if required.Independent simultaneous approaches, down to 2500 ft spacing, were examined by Airborne Information for Lateral Spacing.In that investigation autopilot-flown approaches with onboard warnings were provided to the pilot when a breakout needed to be performed due to an aircraft blunder [4].To achieve significant capacity gains during both good and inclement conditions, runways closer than 2500 ft need to be explored.Building additional runways between current ones, or moving them closer, is a potential solution to meeting the increasing demand.The Raytheon Corporation, working with NASA developed the concept called Terminal Area Capacity Enhancing Concept (TACEC) [5].The concept requires robust technologies and procedures that need to be ©2008 IEEE.Personal use of this material is permitted.However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redistribution to servers or lists, or to reuse any copyrighted component of this work in other works must be obtained from the IEEE.developed and evaluated such that operations are not compromised under instrument meteorological conditions.The reduction of runway spacing for independent simultaneous operations dramatically exacerbates the likelihood of wake vortex incursion and requires the calculation of a safe and proper breakout maneuver.The study presented here investigated procedures for breakout maneuvers due to off-nominal situations such as the blundering of the lead aircraft or its wake drifting towards the trailing aircraft.A real-time, human-in-the-loop simulation studied procedures using precision navigation, autopilot-flown approaches, with the pilot monitoring aircraft spacing and the wake vortex safe zone during the approach.There were aural and visual alerts provided to the pilots to manually perform the breakout maneuvers. +BackgroundTo explore operations on runways closer than 3000 ft, NASA explored a new concept called Airborne Information Lateral Spacing (AILS).NASA developed the AILS concept to further examine independent parallel runway operations on runways as close as 2500 ft.The concept requires technologies that enable the use of precise navigation and surveillance data.Automation is presumed to detect blunders or situations that may require the aircraft to perform a break-out maneuver.The AILS experiment was designed to study three variables-intruder geometry, runway separation (3400 ft or 2500 ft), and flight control mode (auto-pilot versus manual prior to the warning for breakout).The dependent variables were pilot reaction time and miss-distance in off-nominal situations that required the pilot to perform an escape maneuver.The study found that pilot reaction time to detect and perform break out maneuvers was not affected by runway separation.Across all conditions the average pilot reaction time was 1.11 s, with a standard deviation of 0.45 s.The experiment found a statistically significant effect for the flight control mode, with auto-pilot use prior to the emergency escape maneuver leading to longer reaction times.TACEC would allow paired approaches on runways that are 750 ft apart in instrument meteorological conditions [5].The concept includes a ground-based processor which identifies aircraft that could be paired approximately 30 minutes from the terminal airspace boundary.The aircraft are selected for pairing based on several parameters such as relative aircraft performance, arrival direction, and the size of aircraft's wake.The ground based processor then assigns 4-Dimensional (4D) trajectories to the aircraft in the pair.It is assumed that all aircraft will use differential GPSenabled, high precision 4-D flight management system capabilities for the execution of these trajectories.Enhanced cockpit displays that depict both traffic and wake information will also be a requirement for these operations.The current study is different from the AILS experiment in that the algorithms and displays consider wake data, breakout maneuvers are dynamically generated, and the runways are only 750 ft apart. +Breakout ManeuversThe TACEC operational concept necessitates an understanding of unusual events where the approach path of one aircraft might intrude into the approach path of another aircraft.Although these events should be rare, such off-nominal events must be considered to insure the safety of the tools and procedures.In the ILS/ PRM approaches earlier described, there are two approach controllers that monitor each runway.A non-transgression zone (NTZ) with a width of 2000 ft between the two parallel approach paths is defined.The PRM controller detects and initiates breakout when aircraft penetrates the NTZ ,and the pilots have to manually fly the breakout maneuver.SOIA approaches have a similar procedure: the controllers monitor the SOIA flights using the PRM and other standard ATC equipment.Blunders are detected and breakout maneuvers are initiated by the controllers, similar to the ILS/PRM approaches.Breakout instructions that are provided by the ATC are usually long.It is interesting to note that an NTZ exists until the Missed Approach Point (MAP), and that the approach courses are separated by 3000 ft until that point.The trailing aircraft is always on the ILS offset.After exiting the Clear of Clouds (CC) point (shown in Figure 1), the trailing aircraft has about 25 s to obtain visual contact with the lead aircraft, before reaching the missed approach point.If visual sighting is not obtained, then the aircraft has to execute a missed approach. +Figure 1 SOIA ApproachesThe AILS experiment [4] also made provisions for breakout maneuvers.The on-board system detected potential conflicts between the lead and trailing aircraft.Separation responsibility was delegated to the flight crews.AILS defined the breakout maneuver as an Emergency Escape Maneuver (EEM).It required the aircraft to immediately climb and turn 45 deg away from the intruding aircraft.The navigation display showed an escape bug placed at 45 deg, but wake turbulence issues were addressed by existing separation standards.The TACEC study examined breakout maneuvers that require a less extreme turn when compared to the AILS maneuvers.This paper investigates breakout maneuvers for TACEC operations that propose very closely spaced parallel runways.The procedures are defined in the Experimental approach section and results describing the pilots' responses to the maneuvers are described in the Results and Discussion section. +Experimental Approach +Airport and Airspace DesignThe experiment used a fictitious airport (KSRT) loosely based on the current Dallas/Fort Worth International Airport (DFW) layout and operations except for runways that were set be 750 ft apart as shown in Figure 2. Because the simulation focused on TACEC approaches to very closely spaced parallel runways using south flow scenarios, only the west side runways (18R and 18L) were used.The outside runway was moved inward to create a 750 ft separation between the runways.Both the runways were assumed to be equipped to a CAT-IIIB level. +Figure 2 Final Approach geometry for TACEC +TACEC ProceduresThe TACEC concept calls for TACEC-assigned 4D arrival trajectories for both aircraft to be paired at meter fixes located near the edge of the terminal airspace, normally 40-60 nmi from the airport [5].Flights in the simulation began 25 nmi from the airport, assuming they were already paired.Routes to the airport included approach and departure routes and procedures similar to those for DFW airport.This study focused upon arrivals, and no departures were included.The TACEC concept allows for any aircraft arriving from any of the four arrival meter fixes (NE, NW, SE, and SW) to be paired for a simultaneous parallel landing, based on aircraft characteristics and relative timing criteria.Paired aircraft flew their assigned 4D trajectories with a high level of accuracy to meet timing constraints at the coupling point and ensure wake safety throughout the approach.A coupling point is defined at 12 nmi from the runway.From that point onward, the following aircraft precisely maintained spacing behind the lead aircraft to avoid wake using a speed control algorithm.The paths of the trailing aircraft were at a slewed angle when the aircraft was 25 nmi from threshold, then became parallel at about 2 nmi from the runway.Onboard automation monitored the paired aircraft for potential conflicts.Automation also displayed predicted safe zone from the wake generated by the lead aircraft.Visual and aural alerts are used to alert pilots to lead aircraft blunders or wake drifting towards the trailing aircraft.The navigation display depicted the breakout trajectory after crossing the coupling point.This breakout trajectory was dynamically generated considering wake, traffic, buildings and terrain of the airport surroundings.The locations of the breakout on the arrival path require different breakout maneuvers, which change the angle of the escape trajectory on the navigation displays.The pilots flew the breakout trajectory manually using the flight director when they received an aural and visual alert. +DisplaysThe displays were similar to displays used for the preliminary study of very closely spaced parallel approaches [9] and were based on previous research associated with flight deck displays [6] [7].The Navigation Display (ND) and Primary Flight Display (PFD) are shown in Figure 3 and4.The displays show both wake and trajectory information as well as standard flight instrument data.After crossing the coupling point, and the pilot's prior acceptance of the coupling, the flight mode annunciation changes to show that the two aircraft are coupled for speed (C-SPD), coupled for lateral navigation (C-LNAV) and coupled for vertical navigation (C-VNAV).Since the autopilot flew the approach, the pilot primarily monitored the aircraft performance and the displays for the remainder of the flight.If the wake of the lead aircraft drifted within one wingspan of the trailing aircraft, the color of the wake on the display turned to yellow, and then turned red when the apex of the aircraft was in the wake.Similarly, if the lead aircraft deviated from the planned trajectory towards the following aircraft's path by 60 ft, the outline of the lead aircraft symbol turned yellow, and then red when the lead aircraft deviated by at least 120 ft.The red warnings require a mandatory breakout, which the pilots flew manually.Once the pilots pressed the TOGA switch, the breakout trajectory, which had been displayed to the pilot in white, became the active route, and was then displayed in magenta. +Advanced Concept Flight Simulator (ACFS)The human-in-the-loop experiment studied breakout maneuvers for paired TACEC approaches in the Advanced Cockpit Flight Simulator (ACFS) located at NASA Ames Research Center.The ACFS is a motion-based simulator that represents a generic commercial transport aircraft, enabling it to be reconfigured to represent future aircraft.It has the performance characteristics similar to a Boeing 757 aircraft, but its displays have been modified to Breakout Trajectory Wake study different advanced concepts.In this study, the cockpit displays described in the previous section were integrated with the flight display systems in the cockpit.The visual systems offer a 180 deg horizontal and a 40 deg vertical field of view. +VariablesThree variables were examined in this study to examine the TACEC concept.First was the presence or absence of an off-nominal situation that may warrant a breakout maneuver.The second variable was the cause of the breakout maneuverwind causing the wake of the lead aircraft to drift towards the trailing aircraft, or the lead aircraft deviating from its original path and towards the trailing aircraft.The third variable being studied was the location of the off-nominal situation, which was above 500 ft, or between 200 ft -500 ft above the ground.A total of 16 runs were performed in which 8 were normal and rest had off-nominal situations.In the runs that required a breakout maneuver, repeated runs were made for each cause of the breakout and location of the off-nominal situation. +HypothesisIn the absence of previous research, the researchers predicted that the location of the offnominal situation or the nature of the off-nominal situation would not affect pilots' behavior on the following parameters.Any differences observed will guide the formalization of procedures.• Early breakouts • Breakout response time • Separation from lead at breakout point • Accuracy of flying trajectory • Workload • Situation awareness However, it is expected that there will be differences in situation awareness and workload experienced by the pilots in the runs that have the off-nominal situation versus the runs that do not. +ParticipantsThe participants were nine recently retired pilots from commercial airlines; all were male and all of them had experience with glass cockpits.Their average experience as a pilot was about 38 years.Their average number of years since retirement was less than two. +Experimental ProcedureThe study ran for nine days with one pilot participating each day.At the beginning of the day, the pilot was familiarized with the project, the concept, and the new displays in the cockpit.The pilot received a demonstration of the ACFS, and hands-on training on the flight deck displays and related procedures.Since procedures for Very Closely Spaced Parallel Runways (VCSPR) were being explored in this study, each pilot flew the ACFS in the left seat (as captain) along with a confederate who acted as the first officer.The role of the pilot was to fly in auto pilot mode, and monitor the displays to check separation with the lead aircraft and wake.At the coupling point the pilots heard a chime, saw the acknowledgement button light up, and received a "TACEC Coupling" message on the lower Engine Indicating and Crew Alerting System (EICAS) display.At this point the pilots pressed the accept button.They were coupled with the leader's speed, and continued to monitor the separation between the two aircraft.The flight mode annunciation also changed to show that the two aircraft were coupled for speed (C-SPD), coupled for Lateral navigation (C-LNAV) and coupled for Vertical navigation (C-VNAV).If the pilots received a visual and aural alert from the displays they had to perform a breakout maneuver.To fly the breakout maneuver, the pilot had to press the Take-Off-Go-Around (TOGA) switch, disengage the autopilot, leave the auto throttle on, and fly the breakout trajectory shown on the ND.Pressing the TOGA switch would capture the breakout trajectory, and the pilots used the flight director to fly the trajectory.They flew different breakout trajectories at different altitudes, The breakout performed above 500 ft altitude required an initial bank angle of 30 deg, and the breakout at altitude between 200-500ft required an initial bank angle of 10-deg.The pilots then followed the 'S' shaped breakout trajectory displayed on the ND. +Traffic ScenarioThe traffic scenario had two aircraft: (1) The following aircraft in the pair, as represented by the ACFS, and (2) A Boeing 747-400, which was prerecorded and scripted for this study.The pilot who flew the ACFS simulator always landed on 18L.The recorded/scripted aircraft was the leader aircraft that always landed on 18R in the closely spaced parallel runway approach. +Tools used for Data CollectionSeveral tools were used for collecting subjective data from the pilots.All participants completed a demographic survey before the simulation runs were conducted.The survey collected information about the pilots such as their age, experience, and number of hours flying different aircraft types, any experience with SOIA approaches, and experience using personal computers.All pilots were asked to complete a Post Interaction Survey at the end of all the runs.This survey allowed them to rate the information content and the usability of the displays.The participants completed the NASA Task Load Index (TLX) rating scales [10] after each simulation run but did not complete the pair-wise scale comparison that is part of the measure, so the six scales were analyzed separately.Pilots also completed the Situation Awareness Rating Tool (SART) [8].The SART gathers a participant's rating of situation awareness (SA) for the preceding period of time on ten different scales.Each scale has 7 points, with the end points representing the opposite ends of the construct.Participants circled the point on the scale that most closely represented their experienced level of SA.The ten SART ratings were gathered from every participant at the end of each run -a total of 16 ratings per participant were collected.In addition to the assessment instruments described above, the flight simulator's digital data collection system was used.A host of objective flight data for each of the simulation runs was collected on some of the variables pertinent to the hypotheses of the experiment.All collected data were indexed with a common timestamp, which was used as the basis of time synchronization as it updates in real-time while the simulation run advances.All digital data were collected at a rate of 30 Hz. +Results & DiscussionStatistical analysis of the study data focused on three areas: (1) the flight simulator's digital data collection outputs, (2) the pilot participants' workload and situation awareness assessments, and(3) open-ended feedback provided by the pilot participants at the end of the simulation runs.Inferential statistical analysis techniques such as repeated measures Analysis of Variance (ANOVA) and binary logistic regression were employed to address the primary research questions of interest, and descriptive statistics were also reported to augment the results. +Early Breakout AssessmentDuring the course of the breakout runs, the traffic symbol color (aircraft deviation condition) or the traffic wake color (wake condition) would transition from white (nominal) to amber (warning), to red (breakout required).Under these conditions, the pilot participant's initiation of breakout should occur only when the color on the display transitions to red.However, it was noted that pilot participants would sometimes initiate a breakout when the traffic display transitioned to amber, resulting in a somewhat less than optimal breakout maneuver.A binary logistic regression analysis was implemented to assess potential differences across the study conditions, on the incidents of early breakout across the study conditions.The regression model included both levels of each of the independent variables as covariates, and the Wald statistic was computed to assess the significance of the model.Cause of breakout (aircraft deviation vs. wake) was found to be significant in the model (Wald = 4.459, df=1, p< 0.05) whereas location of breakout was not.Thus the hypothesis that there would be no difference in early breakouts due to the cause of breakout was not upheld, whereas location of breakout was upheld.Frequencies and percentages of early breakout response incidents are listed in Table 1. +Location +Table 1: Frequencies and Percentages of Early Breakout by Location and CauseAs indicated in Table 1, 72.2% of the early breakout cases were observed in the wake condition, as compared to 27.8% cases in the aircraft deviation condition, suggesting that the salience of the wake situation might inspire a greater sense of immediacy to maneuver away from the cause of potential danger, even prior to the required breakout response.This may have occurred for a number of reasons.Wake behavior is relatively hard to predict, so the uncertainty of its characteristics may lead to more caution on the part of the pilot, even though the pilots were told that the predicted wake danger area displayed was calculated conservatively.Also, the wake display is large relative to the traffic symbol display.That is, the wake display shows the physical size of the nearby wake vortex, which tends to expand as the lead aircraft moves closer to the ownship.The traffic symbol display, on the other hand, changes color (as does the wake display), but remains static in size.It may be possible that the increased frequency of early breakout response under the wake condition may have occurred as a result of the relative "largeness" of the display, which on some level, might have signaled a situation that was perceived as more critical than it was, leading to a premature response.This may reflect a need for some adaptation of the displays to minimize this effect. +Breakout Response TimeBreakout response is defined as the difference between the time at which the wake or traffic symbol display transitions to the color red, which is the same time an aural alert occurs on the flight deck, and the time when the pilot initiates the breakout response.A two-way repeated measures ANOVA was used to test the hypothesis that there will be no significant main effects or interactions on the dependent variable of breakout response, with cause of breakout and location of breakout as the two independent variables.A significant main effect of breakout location was observed (F=4.86;df=1,8 , p .05), with breakouts occurring above 500 ft AGL showing a larger (i.e., slower) response time than breakouts occurring below 500 ft.No other significant effects were yielded from this analysis.Means and standard deviations associated with the significant main effect are shown in Table 2.A breakout response time of less than 2s should be interpreted with caution.It is unusually low compared to what would be anticipated in the real world, where the novelty and non-expectancy of the situation might make it impossible to act this quickly.In the study the pilots expected an offnominal situation and were ready to breakout, in some cases they even performed early breakouts, which explains the unusually low breakout response time.In actual operations, these off-nominal events should be rare, and the pilots would likely need more time due to their infrequency and unexpectedness.Thus, these times should be viewed as providing trend and relative information only. +Table 2. Significant Main Effect of Breakout Location on Blunder Response TimeThe null hypothesis that there would be no difference in breakout response time due to the cause of breakout was upheld, but it was not upheld for the location of breakout.This effect may have occurred as a result of the perceived immediacy of the response at an altitude of less than 500 ft, since airspace is highly congested close to major airports at lower altitudes, coupled with the proximity to the ground and the terminals requiring increased vigilance of flight crews at this stage of approach.Breakouts at lower altitudes introduce special concerns, because pilot errors carry an increased risk of dangerous consequences.Pilots are also keenly aware of other possible factors, such as low altitude wind shear, which could have the effect of complicating an already dangerous situation.Hence, the perceived immediacy of the response, combined with increased vigilance, may have contributed to the faster breakout response time.Operationally, this may suggest that the pilot participants are inherently and correctly assessing the need for a faster response to a dangerous situation, during flight times that may have other immediate and critical issues +Separation from Lead at Breakout PointThe dependent measure of aircraft separation at breakout is defined as slant range, or straight-line distance, between the leading aircraft causing the breakout and the ownship.Again, the effects of the two independent variables of breakout cause and breakout location on the dependent measure were tested in this analysis.A significant main effect of breakout cause on the dependent measure was observed (F=37.21,df=1,8, p< 0.001), with greater aircraft separation under the wake condition than under the aircraft deviation condition.No other significant main or interaction effects were observed from this analysis.The hypothesis that there would be no differences for cause of breakout was not upheld, but it was upheld for the location of breakout.Means and standard deviations describing the details of the significant main effect are listed in Table 3.As a check on the reasonableness of the results reported in Table 3, a Kruskal-Wallis one-way ANOVA by ranks was implemented on the aircraft separation data, due to a possible violation of the variance homogeneity assumption.Consistent with results shown in Table 3, a significant main effect of breakout cause on aircraft slant range at breakout was observed (Kruskal-Wallis Test Statistic = 47.52,df=1, p<0.0001).Again, it seems that the off-nominal situation caused by wake has special characteristics that might help to explain a greater degree of aircraft separation at breakout time.The pilots during the group discussion mentioned that the uncertainty regarding wake characteristics prompted them to make responses more quickly. +F=37Operational considerations might include adapting the aircraft deviation and wake displays to account for differences in which pilots react to the onset of situations that might evolve into blunders (e.g., premature maneuvering, possible lack of vigilance in the case of inadequate display format, etc.) +Accuracy of Trajectory: Cross Track and Track Angle ErrorTrajectory accuracy is measured by the actual ownship/simulator position against the breakout trajectory generated by the system and displayed on ND averaged across time.Two measures of ownship trajectory particularly sensitive to breakout maneuvers include cross track error and track angle error.For each flight simulation run, cross track error and track angle error was averaged across time from the breakout point to the end of the flight.A two-way repeated measures ANOVA yielded a main effect of breakout location on each of the two dependent measures.Both of these results are consistent with respect to the directionality of the means.More cross track error and more track angle error were observed at breakout locations above 500 ft as compared to breakout locations at or below 500 ft.No other main or interaction effects were observed.ANOVA summary statistics on the significant results from this analysis are listed in Tables 4 &5 Also, the maneuver below 500 ft has an initial bank angle of 10 deg, which is fairly easy to execute with the side-stick control used in the ACFS.Most pilots complained about the stick shift, but did like the 10-deg bank angle at the lower altitude, since it allowed them to fly the breakout trajectory projected on the ND more accurately.Thus the cross track error and track angle error shown in the Tables 4 and5 should be interpreted for its relativity to the different independent variables and as providing trend information. +WorkloadParticipants completed the NASA TLX workload questionnaire after every run.In general the pilot's workload was quite manageable and was below average.A statistically significant difference was observed between the breakout condition and the nominal condition for the dependent variable of overall workload (F=6.17,df = 1,8, p<=.05), with higher workload experienced in the breakout runs as compared to normal runs.Further analyses depicted significant differences between the normal and breakout conditions on the sub elements of workload such as effort (F=10.81 ; df=1,8 ; p<0.05) and frustration (F=7.16,df=1,8,p<0.05).Marginally significant differences were also observed on mental demand (F= 4.77, df=1,8, p=0.06), and temporal demand (F=4.53,df=1,8, p=0.06).Means and standard deviations of all workload sub-scale assessments, comparing nominal vs. breakout conditions, are graphically depicted in Figure 5. Analysis of workload assessment within the breakout condition was also done.There were no significant differences in the workload experienced by the pilots as a result of location or cause of breakout. +Situation AwarenessParticipants rated the ten SART scales after every simulation run.Each scale has seven points, where 1 represents 'little' or 'no' and 7 represents 'a lot' or 'very.'These ten scales were combined to three broader categories concerned with the a) demands of the situation b) the 'supply' or personal resources that the participants has to bring to the situation and c) situational provision that the situation provides in the form of information through displays.The first broad category combines the three SART scales -instability, variability and complexity of the situation, where the values can range from 3 to 21.The second broad category of personal resources combines the SART scales on alertness, spare mental capacity, concentration, and division of attention, where the resultant scores can range from 4 to 28.The third broad category, situation provision combines the three SART scales on information quantity, information quality, and familiarity, and the resultant value can range from 3 to 21.Statistical analysis comparing normal and breakout conditions on situation awareness of the pilot participants yielded a significant difference on the subscale of situational demands (F=15.42,df=1,8, p<.01).Also higher pilot workload levels were experienced in the off-nominal (i.e., breakout) condition, which correlate with higher levels of instability, variability, and complexity, as compared to the nominal condition.This would be expected, since the off-nominal condition requires that pilots safely maneuver the aircraft by following the breakout trajectory, rather than implement normal approach procedures.Less striking differences were observed between the nominal and breakout runs on the other two situation awareness variables of personal resources and situation provision.This may be due to the anticipation of a breakout anytime, which required equal levels of alertness and concentration.It is interesting to note that between the nominal and breakout scenarios the pilots experienced equally high levels of information quantity, and quality, and familiarity.The means and standard deviations of the three situation awareness variables across both conditions are graphically depicted in Figure 6.Further analyses of the SART data within the breakout condition revealed no significant difference as a result of the location or the cause of the off-nominal situation.The pilots experienced similar levels of situation awareness irrespective of the cause of the breakout (wake or aircraft deviation) or the location of the breakout. +SummaryThe TACEC procedures were investigated in a human-in-the-loop simulation incorporating new tools and technologies.Scenarios included nominal and off-nominal cases.Statistically significant differences were observed in this current investigation using the analyzed digital data collection variables and some of the subjective variables.However, it is also interesting, and reassuring to note that the pilot participants successfully "flew" the simulator through all of the study scenarios, both accurately and safely within and across all conditions.While early breakouts are not entirely consistent with the concept, the breakout maneuvers were successfully "flown," and safety was not compromised when they did occur.Wake, possibly due to its salience in the displays did cause more early breakouts than the blundering of the lead aircraft.During group discussion, pilots indicated that the warnings associated with aircraft blundering were not clear and visible.The overall breakout aircraft slant range separation mean was over 2500 ft and the breakout trajectory was also quite accurately flown across all conditions.The location of the off-nominal situation did impact the slant range between the lead and trailing aircraft, and also the accuracy with which the breakout trajectory was flown.The pilots in general preferred the initial 10 deg bank angle they flew on breakout trajectories initiated between 200 ft and 500 ft and provided feedback that it was easier to fly than the more aggressive 30-deg initial bank angle used for breakouts at higher altitudes.The pilots also provided the feedback that the ability to see the trajectory on the ND aided them in flying the trajectory accurately.The pilots experienced higher workload and situational demands placed on them during breakout as compared to the normal landings.While realizing these differences, the results also indicate that workload was manageable, and an adequate level of situational awareness was maintained across all conditions.Overall, the data provide support for the contention that very closely spaced parallel runway approach procedures, when implemented wisely, can increase efficiency of flight operations, while maintaining an adequate level of safety.Hence, the results attest to the potential promise of the current concept under investigation.Figure 3 :Figure 4 :34Figure 3: Navigation Display during final approach +F=4 +Figure 5 .5Figure 5. Effects of Breakout on Pilot Workload Measures (error bars represent ± 1 standard deviation) +Figure 6 .6Figure 6.Effects of Breakout on Pilot Situation Awareness Measures (error bars represent ± 1 standard deviation) +Table 3 . Significant Main Effect for Cause of Breakout on Aircraft Separation3.21Mean (ft)Standarddf=1,8Deviationp<0.001Aircraft2820.45174.48DeviationWake2994.5414.83 +Table 4 . Significant Main Effect of Breakout Location on Ownship Cross Track Error F=157.58 df=1,8 p<0.0001 Mean (deg) SD (deg)4.F=45.08 df=1,8Mean (ft)SD ( ft)p<0.001Breakout Location73.0825.23> 500 ftBreakout Loaction39.4327.42500 ftBreakout Location3.410.95> 500 ftBreakout Location1.420.74500 ft +Table 5 . Significant Main Effect of Breakout Location on Ownship Track Angle Error5 + + + + + + + + + Joint Planning and Development Office (JPDO) Unmanned Aircraft Systems (UAS) + + YogiShah + + 10.1109/icnsurv.2013.6548689 + + + 2013 Integrated Communications, Navigation and Surveillance Conference (ICNS) + + IEEE + 2004 + + + JPDO Report + Joint Development and Planning Office, 2004, The Next Generation Air Transport System Integrated National Plan, JPDO Report. + + + + + Frequency and severity of deviations during simultaneous independent approaches to parallel runways + + SebastianVMassimini + + + GerryCMcneil + + + NoamTene + + 10.1109/icnsurv.2010.5503313 + MP 06W0000127 + + + 2010 Integrated Communications, Navigation, and Surveillance Conference Proceedings + + IEEE + 2006 + + + Mitre Report + Massimini, V.S., 2006, Simultaneous Independent and Dependent Parallel Instrument Approaches. Assumptions, Analysis, and Rationale, Mitre Report MP 06W0000127. + + + + + Compilation of DOT and FAA Airport Legal Determinations and Opinion Letters as of December 31, 2020 + 10.17226/26202 + + + + Simultaneous Offset Instrument Approaches -Newark International Airport: An Airport Feasibility Study + + Transportation Research Board + 2001 + + + Magyratis, S., Kopardekar, P., Sacco, N., & Carmen, K.,2001, Simultaneous Offset Instrument Approaches -Newark International Airport: An Airport Feasibility Study, DOT/FAA/CT-TN02/01. + + + + + NASA administrator announces resignation + + TSAbbott + + + DMElliot + + 10.1016/s0045-8732(01)80042-5 + NASA/TP-2001-210665 + + + COSPAR Information Bulletin + COSPAR Information Bulletin + 0045-8732 + + 2001 + 152 + + 2001 + Elsevier BV + + + Abbott, T. S., & Elliot, D. M., 2001, Simulator Evaluation of Airborne Information for Lateral Spacing (AILS) Concept, NASA/TP-2001-210665. + + + + + CNS requirements for precision flight in advanced terminal airspace + + MEMiller + + + SDougherty + + + JStella + + + PReddy + + 10.1109/aero.2005.1559471 + + + 2005 IEEE Aerospace Conference + + IEEE + 2005 + + + + Miller, M.E., Dougherty, S., Stella, J., & Reddy, P.,2005, CNS Requirements for Precision Flight in Advanced Terminal Airspace, Aerospace, 2005 IEEE Conference, pp. 1-10. + + + + + Cockpit Display of Traffic and Wake Information for Closely Spaced Parallel Approaches + + GordonHardy + + + EmilyLewis + + 10.2514/6.2004-5106 + + + AIAA Guidance, Navigation, and Control Conference and Exhibit + Providence, RI4 + + American Institute of Aeronautics and Astronautics + 2004 + + + Hardy, G. H., & Lewis, E. K., 2004, A Cockpit Display of Traffic Information for Closely Spaced Parallel Approaches, AIAA-2004-5106, AIAA Guidance, Navigation, and Control Conference and Exhibit, Providence, RI4. + + + + + Models of Wake-Vortex Spreading Mechanisms and Their Estimated Uncertainties + + VernonRossow + + + GordonHardy + + + LarryMeyn + + 10.2514/6.2005-7353 + AIAA-2005-7353 + + + AIAA 5th ATIO and16th Lighter-Than-Air Sys Tech. and Balloon Systems Conferences + Arlington, VA + + American Institute of Aeronautics and Astronautics + 2005 + + + Rossow, V. J., Hardy, G. H., and Meyn, L. A., 2005, Models of Wake-Vortex Spreading Mechanisms and Their Estimated Uncertainties, AIAA-2005-7353, ATIO Forum, Arlington, VA. + + + + + Situation Awareness as a Predictor of Performance for En Route Air Traffic Controllers + + FrancisTDurso + + + CarlaAHackworth + + + ToddRTruitt + + + JerryCrutchfield + + + DankoNikolic + + + CarolAManning + + 10.2514/atcq.6.1.1 + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 6 + 1 + + 1998 + American Institute of Aeronautics and Astronautics (AIAA) + + + Durso, F. T., Hackworth, C. A., Truitt, T. R., Crutchfield, J., Nikolic, D., & Manning, C.A.,1998, Situation Awareness as a Predictor of Performance for En Route Air Traffic Controllers, Air Traffic Control Quarterly, 6(1), pp. 1-20. + + + + + Procedures for off-nominal cases: Very closely spaced parallel runway operations + + SavitaVeram + + + SandyLozito + + + ThomasKozon + + + DeborahBallinger + + + HerbertResnick + + 10.1109/dasc.2008.4702783 + + + 2008 IEEE/AIAA 27th Digital Avionics Systems Conference + Anchorage, Alaska + + IEEE + 2008 + + + Verma, S. A., Lozito, S., & Trot, G.,2008, Preliminary Guidelines on Flight Deck Procedures for Very Closely Spaced Parallel Approaches, International Council of Aeronautical Sciences, Anchorage, Alaska. + + + + + Development of NASA-TLX (Task Load Index): Results of Empirical and Theoretical Research + + SandraGHart + + + LowellEStaveland + + 10.1016/s0166-4115(08)62386-9 + + + Advances in Psychology + + PHancock + + + NMeshkati + + Amsterdam, The Netherlands + + Elsevier + 1988 + + + + Hart, S., & Staveland, L.,1988, Development of a Multi-Dimensional Workload Rating Scale: Results of Empirical and Theoretical Research, In P. Hancock & N. Meshkati (Eds.), Human Mental Workload, Amsterdam, The Netherlands, Elsevier, pp. 139-183. + + + + + + diff --git a/file766.txt b/file766.txt new file mode 100644 index 0000000000000000000000000000000000000000..67abd00ede23044e387f7258e64011ffcb85b845 --- /dev/null +++ b/file766.txt @@ -0,0 +1,340 @@ + + + + +IntroductionDemand in the future air transportation system concept is expected to double or triple by 2025 [1].Increasing airport arrival rates will help meet the growing demand that could be met with additional runways.But the expansion of the airport is often met with environmental challenges for the surrounding communities when using current standards and procedures.Independent simultaneous operations are allowed today with 4300 ft spacing or down to 3000 ft if special radar is used [2].Simultaneous Offset Instrument Approaches (SOIA) are allowed for closer runway spacing (down to 750 ft) by offsetting the aircraft longitudinally and requiring visual separation, which reduces arrival rate during poor weather [3].Independent simultaneous approaches down to 2500 ft spacing were examined by Airborne Information for Lateral Spacing (AILS) through autopilot-flown approaches with on-board warnings provided to the pilot when a breakout needed to be performed due to an aircraft blunder [3].To achieve capacity gains, runways closer than 2500 ft need to be explored.Building additional runways between current ones, or moving them closer, is a potential solution to meeting the increasing demand, as addressed by the terminal area capacity enhancing concept (TACEC).The concept requires robust technologies and procedures that need to be tested such that operations are not compromised under instrument meteorological conditions.The reduction of runway spacing for independent simultaneous operations dramatically exacerbates the criticality of wake vortex incursion and the calculation of a safe and proper breakout maneuver.The study presented here developed guidelines for such operations by performing a real-time, human-inthe-loop simulation using precision navigation, autopilot-flown approaches, with the pilot monitoring aircraft spacing and the wake vortex safe zone during the approach. +BackgroundThe FAA has successfully conducted independent approaches to parallel runways for over forty years using the Instrument Landing System (ILS) navigation and terminal radar monitoring [2].The simultaneous approaches that utilize standard radar are conducted on parallel runways that are separated by at least 4300 ft apart.It is possible to conduct independent approaches on runways separated by as little as 3000 ft, but it requires a Precision Runway Monitor (PRM) with an update rate of 1 s.The separation standards between the aircraft on these parallel approaches are 1000 ft vertical separation.Additionally, there is a 2000 ft wide "no-transgression zone" (NTZ) that was placed equidistant from the centerlines of the approach paths on the two parallel runways.Some airports like San Francisco International Airport can support approximately 60 landings per hour on its two parallel runways that are 750 ft apart by using SOIA [3].SOIA approaches require the trailing aircraft in the paired approach to obtain a visual sighting of the lead aircraft with at least a 1200 ft ceiling with 4nm visibility.As weather degrades, the current navigation and surveillance system, as well as the existing procedures, lack the accuracy to support SOIA approaches, reducing the landing rate to half the VFR capacity.Several researchers have investigated alternative procedures for Very Closely Spaced Parallel Runway (VSCPR) operations.Studies have focused on the technologies required to enable the VCSPR operations.Several different requirements have been identified from these studies, such as cockpit displays, collision prevention systems, and precision navigation, communication, and surveillance systems [6] [7] [8].Another critical component that is necessary for the safe execution of VSCPR procedures is the ability to predict the wake vortices for the aircraft nearby and provide wake information to the affected aircraft.Previous research has also evaluated procedures for VCSPR approaches, but most of them have used fast-time simulation to investigate the performance of the procedures.Pritchett & Landry [6] identified the various parameters related to VCSPR operations such as separation responsibility and different separation and spacing objectives between the paired aircraft.Few human-in-the-loop studies have been conducted for VCSPR operations.The study to investigate pilot response towards the VCSPR operations for the AILS concept is one such example [4].NASA developed the AILS concept to further examine independent parallel runway operations to runways as a close as 2500 ft.The concept requires technologies that enable the use of precise navigation and surveillance data.Automation is presumed to detect blunders or situations that may require the aircraft to perform a break-out maneuver.The AILS experiment was designed to study three variables-intruder geometry, runway separation (3400 ft or 2500 ft), and flight control mode (auto-pilot versus manual prior to the warning for breakout).The dependent variables were pilot reaction time and miss-distance in off-nominal situations that required the pilot to perform an escape maneuver.The study found that pilot reaction time to detect and perform break out maneuvers was not affected by runway separation.Across all conditions the average pilot reaction time was 1.11 s, with a standard deviation of 0.45 s.The experiment found a statistically significant effect for the flight control mode, with auto pilot use prior to the emergency escape maneuver leading to longer reaction times.TACEC aims to fly paired approaches on runways that are 750 ft apart in instrument meteorological conditions [5].A ground-based processor will identify aircraft that could be paired approximately 30 minutes from the terminal boundary.The aircraft are selected for pairing based on several parameters such as aircraft performance, arrival direction, relative timing criteria, and aircraft size of wake considerations.The ground based processor then assigns 4D trajectories to the aircraft in the pair.It is assumed that all aircraft will use differential GPS-enabled, and high precision 4-Dimensional (4-D) flight management system capabilities for the execution of these trajectories.Enhanced cockpit displays that depict both traffic and wake information will also be a requirement for these operations.The current study is different from the AILS experiment because it considers wake, and dynamically generates break-out maneuver.Little data exist regarding the use of VCSPR technologies and procedures.The objective of the current study is to develop guidelines for the procedures defined by the TACEC using a human-in-the-loop simulation study.The goal of this simulation is to explore the usefulness and usability of the cockpit displays and procedures associated with this new concept. +Experimental Approach 2.1 Airport and Airspace DesignThe airport and airspace used to investigate procedures for the TACEC concept used a fictitious airport that was based on the current Dallas/Fort Worth International Airport (DFW) layout and operations.The airport used for the simulation was referred to as "KSRT."Because the simulation was focused on studying TACEC approaches to very closely spaced parallel runways, and because of the decision to have a south air traffic flow for the simulation scenarios, the SRT airport only utilized runways 18R, 18L, 17R, and 17C (re-named to 17L).All four runways were assumed to be equipped to a CAT-IIIB level.Both 18R and 17L were moved to within 750 ft of their inboard runways, 18L and 17R respectively.This required an adjustment of 464 feet from their current DFW positions. +TACEC proceduresThe TACEC concept calls for TACECassigned 4D arrival trajectories for both the aircraft to be paired at meter fixes located near the edge of the terminal airspace, normally 40-60 nautical miles from the airport [5].Flights in the simulation began 25nmi from the airport, assuming they were already paired.Routes to the KSRT airport included approach and departure routes and procedures similar to DFW airport.This study focused upon arrivals and no departures were included. +Arrival Traffic FlowSouth flow of traffic was simulated for the generic airport KSRT.All the four runways (18R, 18L, 17R, 17L) were used for arrival operations.The concept allows for any aircraft arriving from any of the four arrival meter fixes (NE, NW, SE, and SW) to be paired for a simultaneous parallel landing, based on aircraft characteristics and relative timing criteria.Paired aircraft must fly their assigned 4D trajectories with a high level of accuracy in order to meet timing constraints at the coupling point and ensure wake safety throughout the approach.The 4D trajectories were carefully designed to provide safe wake-avoiding routes from the arrival meter fixes to the runways.Each route consisted of three segments, and each one of the first segments provided vortexfree 4D routes extending from the meter fix to the coupling point at 12 nmi from the runway.The second segment began at the coupling point and ended 2 nmi from the runway.During the second segment, one route was straight in, aligned with the runway centerline, while the other was at a 6-degree slew angle from the straight-in route (see Figure 1).At the coupling point, the aircraft were laterally separated by slightly more than 1 nmi.Each of the final segments were aligned with the runway centerlines and extended 2 nmi from the runway threshold and were about 600 ft Above Ground Level (AGL) in order to provide a straight-in flight path to touch down.Once the aircraft reached the coupling point, the following aircraft precisely maintained spacing behind the lead aircraft in order to avoid the lead's wake.This was accomplished by an automated speed control algorithm on-board the following aircraft that maintained the assigned time-based spacing relative to the lead based on state information broadcasted via ADS-B by the lead aircraft.Figure 1 shows the geometry of the final approach portion of the arrivals (i.e. the final 12nmi before landing). +Cockpit Display of Traffic and Wake InformationThe primary purpose of the displays used for the TACEC evaluation was to provide the flight crews with information to ensure that adequate separation was being maintained with the lead aircraft and its hazardous wake area.While not evaluated in the present simulation, the displays also provide "breakout" annunciation and guidance if adequate separation is not maintained with the lead aircraft or its wake.The Primary Flight Display (PFD) and the Navigation Display (ND) are modifications of standard current generation transport flight displays with added lead aircraft position and wake information.Figure 2 shows the PFD on the straight-in parallel final at 532 ft radar altitude while Figure 3 shows the ND for the same location.Lateral spacing of the flight paths at this part of the approach was 750 ft.The displays are adaptations of those previously developed by Hardy and Lewis (2004) [8]. +Lead aircraft positionThe position of the simulator was shown on the ND with the conventional triangular icon (solid) at the lower center of the ND.The lead aircraft position was shown with the open icon at the upper left of the ND.The same perspective triangular lead aircraft position was shown on the PFD at the left of the display.With augmented GPS navigation, it was assumed that position information was known with ADS-B to be within a few feet. +Hazardous wake area depictionThe shaded white area on the ND and the wake frames on the PFD depict the hazardous wake area.This was defined as that volume of airspace such that if the simulator's apex or center of gravity (cg), remains outside the wake area, no noticeable wake activity would be detected.This area was predicted in real time from aircraft characteristics and on-board sensors of crosswind and atmospheric turbulence.The prediction algorithms were conservative to account for model and sensor errors [9].The shaded area on the ND and the wake frames on the PFD turns amber if the simulator's cg moves to within one wingspan of the hazardous area, and it turns red if its cg penetrates it. +Predictor dotsFive two-second predictor dots, for a total of ten seconds were added to the ND for both aircraft (see slightly to the right of the nominal path for the simulator in Figure 3) and also were presented on the PFD (aligned with the lead aircraft's position icon).These show fight path trend information to help the pilot determine the future location of the aircraft. +Longitudinal Situation IndicatorTo maintain the aircraft's position in the "Safe" zone, as shown in Figure 3, a Longitudinal Situation Indicator (LSI) was added.The LSI is flagged on the ND and shows the nominal location (in this case five seconds behind the lead aircraft) that the auto-throttle is attempting to keep.For this example, the simulator is approximately 400 feet behind its nominal location.The same LSI information is shown on the deviation scale added on the left side of the PFD (Figure 2). +Display scalingA conventional PFD has a field of view of about 40 degrees.To be able to see the lead aircraft position and wake information this was increased to 80 degrees.This decreases the resolution of the display but with future larger display hardware, it may not be objectionable.A conventional ND has a maximum zoom-in capability of a ten-mile range scale.To have adequate resolution for this task, the maximum zoom-in range scale is 0.25 nmi.The display zoomed in increments of 10, 5, 2, 1, and 0.5 nmi scales. +MethodologyThe objective of the study was to explore new procedures called paired approaches that are intended for very closely spaced parallel runways (750 ft apart in this study).Retired commercial airline pilots participated and flew a series of scenarios using a flight simulator of a glass cockpit aircraft that included new tools and procedures. +Advanced Concept Flight Simulator (ACFS)The human-in-the-loop study conducted to assess the paired TACEC approaches used the Advanced Cockpit Flight Simulator (ACFS) located at NASA Ames.The ACFS is a motionbased simulator that represents a generic commercial transport aircraft, enabling it to be reconfigured to represent future aircraft.It has the performance characteristics of Boeing 757 aircraft, but its displays have been modified to study different advanced concepts.In this study, the cockpit display described in Section 2 was integrated with the flight display systems in the cockpit.The visual systems offer a 180-degree horizontal and a 40-degree vertical field of view. +Experimental matrixThe three variables that were examined in the study were visibility conditions, direction of the wind, and the distance between the lead and follower aircraft.The visibility conditions were a clear day or Category-IIIB.The study aimed at exploring an adverse cross wind on the follower (simulator), thus, the direction of winds was coupled with the follower (simulator) landing on the left or right runway (18L or 18R runways in this study).The approach to 18R is referred to as the slewed approach and the one to 18L is the straight in approach.The third variable examined in the study was the distance between the lead and follower aircraft at initialization points, which was either 10s or 5s. +ParticipantsThe participants of the study were three retired pilots from commercial airlines; all of them had experience with glass cockpits and some experience flying SOIA approaches in San Francisco.Their mean total years of experience as a pilot was about 40 years.They had on an average about 16,500 hours of flying.Their average number of years since retirement was 6.5 years.The study was run for three days with one pilot participating each day.At the beginning of the day, the pilot was familiarized with the project, the concept, and the new displays in the cockpit.Next, the pilot was taken to the ACFS, where he received a demonstration of the simulator, and more hands-on training on the CDTI and related procedures. +ProcedureThe procedures for VCSPR were being explored in this study, so each pilot flew the ACFS as a captain along with a confederate who acted as the first officer.The role of the pilot, in general, was to fly in auto pilot mode, and monitor the displays to check separation with the lead aircraft and wake.At the coupling point the pilots heard a chime, saw the acknowledgement button light up, and a message on the lower engine indicating and crew alerting system (EICAS) appeared that read "TACEC Coupling."At this point the pilots pressed the acknowledgement button, and continued to monitor the separation between the two aircraft.The flight mode annunciation also changed to show that the two aircraft were coupled for speed (C-SPD), coupled for Lateral navigation (C-LNAV) and coupled for Vertical navigation(C-VNAV).Since the autopilot flew the approach, the pilot primarily monitored the aircraft performance and the displays for the remainder of the flight. +Traffic ScenarioThe traffic scenario had two aircraft: the following aircraft in the pair, as represented by the ACFS, and another aircraft, which was recorded or scripted for this study.The simulator was always the following aircraft and the recorded one was always the leader aircraft in the closely spaced parallel runway approach.The leader aircraft was a simulated Boeing 747-400.Based on the wind condition, the simulator was either on the slewed approach landing on runway 18R or on the straight in approach landing on runway 18L. +Tools used for data collectionSeveral tools were used for collecting subjective data from the pilots.All participants completed a demographic survey before the simulation runs were conducted.It collected information about the pilots such as their age, experience as a pilot, and number of hours flying different aircraft types, any experience with SOIA, and experience using personal computers.Each pilot was asked to complete a Post Interaction Survey at the end of all the runs.It collected information on the pilot-rated usefulness and usability of the displays.Similarly a feature comparison survey was administered at the end of all of the runs.The pilots had the opportunity to rate the importance of different features in the displays on a scale of 1 to 5, where 1 was equivalent to "very unimportant" and 5 was equivalent to "very important."Pilots also completed the Situation Awareness Rating Tool (SART) [10].The SART gathers a participant's rating of his/her situation awareness (SA) for the preceding period of time on ten different scales.Each scale has 7 points, with the end points representing the opposite ends of the construct.Participants circled the point on the scale that most closely represented their experienced level of SA.The ten SART ratings were gathered from every participant at the end of each run -a total of 8 ratings per participant were collected. +Results and DiscussionThis section reports results that focus on the data captured by the tools mentioned in section 3.6.Results of the post interaction survey, feature comparison, situation awareness, and observer notes are described in the following section. +Post Interaction SurveyThe post interaction survey was administered to each pilot at the end of the eight trial runs.Since the questions administered after the simulation was complete, there were no distinctions among the different experimental conditions, but instead queried the participants about the general experiences of using VCSPR procedures and tools.Also, due to low statistical power for testing, tests for significance were not conducted.The pilots responded to the question on the overall utility of the displays for VCSPR approaches as highly useful (average of 3, on a scale of 1 to 5).The questions focused on the ease of using the displays to derive information for some of the functions handled by the pilots using the displays.The pilots found that the overall level of ease for extracting information from the displays was very high (M=5 on a scale of 1 to 5, where 1 was very hard and 5 was very easy).In general, on average the pilots found that the displays provided enough information, and that it was relatively easy to extract the information for most of the functions.The mean value was greater than or equal to 4 for all functions except flying in low visibility.During the group discussions, the pilots mentioned that they would like to see the tool deployed in clear weather conditions for a period of time to allow the pilots to develop enough trust in the automation before it is used for flying under Category-IIIB visibility conditions.They felt that this trust could be improved with more familiarity and use of this type of automation.Also, the pilots mentioned that deriving information about wake characteristics was very easy in this simulation (M=5).One can infer that they were able to effectively monitor separation of the aircraft from the wake.Ratings for ease of deriving information from the displays All the pilots reported that they were able to effectively monitor the lead aircraft.Also, none of the pilots were confused by the interface.On the ability to zoom on the navigation display, the pilots reported that having a separate zoom capability for the pilot flying and pilot non-flying will enable them to maintain both a strategic and tactical view at the same time.The navigation display zoom capability was handled by a toggle switch on the center console and was available as a function only to the pilot flying.The pilots were asked which aspects of the concept they liked the best, and which aspects they liked the least.The pilots also said that the system and the new displays will greatly enhance safety in today's air traffic environment.They also agreed that the system will enhance capacity at the airports.In contrast, the pilots repeated that this automation needs to be implemented in good visibility conditions before the pilots will trust the automation for use during IMC.They were all concerned about procedures for breakout maneuvers, and definition of standards for proximity.They also wanted more flexibility with maneuvering throttles without disengaging the auto throttles.One pilot also mentioned that all procedures, including airspeed requirements between the coupled aircraft, must be agreed upon by the pilots and controllers involved in the procedures prior to flying it.The pilots were also asked to rate some statements regarding the concept and displays (Figure 5).They all agreed that automation is required for VCSPR approaches, and that there was little confusion about the displays.They responded with above average ratings for ease of monitoring separation from the lead aircraft.The participants also found the wake information on the navigation display and the predictor dots very useful, and they valued being able to visualize the lead aircraft's trajectory.They rated their level of confidence in the concept as average, and they did not indicate concern in their responses about the role of the pilot in this concept. +Feature ComparisonThe participants were asked to rate the various features on the displays provided to them in the simulator.They rated most features as having above average importance (ranging from 4 to 4.5 on a scale of 1 to 5) except the lead aircraft and the LSI on the PFD.Those were rated at an average of 3.5 on a scale of 1 to 5, where the higher number indicates higher level of importance.The LSI on the ND was not always visible and several participants complained about not being able to visually track the LSI because it was hidden under the aircraft's solid white icon.The LSI on the PFD provided the information about the simulator's actual position versus expected position in terms of distance, whereas the LSI on the ND provided temporal information as referenced by the 2s predictor dots.Despite its poor visibility at certain times, the LSI on the ND was preferred by most pilots.The lead aircraft's predictor dots were considered to have average level of importance, because the pilots always flew the follower aircraft in the approach, and they were concerned with their own trajectory predictions to monitor separation from the lead aircraft and its wake.Similarly, the feature-out of the window visibility received a 3.5 rating and the acknowledgement button used for accepting the coupling between the paired aircraft, received a 2.6 average rating.During the group discussion, the pilots suggested that pressing the acknowledgment button should arm the coupling of the two aircraft, before they are actually at the coupling point to keep it consistent with other standard displays.The pilots also mentioned that the flight mode annunciation should have a visual indicator that is white in color, depicting that the system is armed before coupling.Eventually it should turn green when actual coupling occurs, at the coupling point.In the present experimental setup, the acknowledgement button changed the FMS annunciation to "coupled" and did not give the pilots a chance to "arm."This created some confusion and led to the comments made by the pilots.Among other concerns and suggestions for improving the design of the system, some pilots had difficulty with interpreting the wake depiction and monitoring the lead aircraft on the PFD.Other pilots felt that when the aircraft starts deviating from its longitudinal position, the procedure should allow for pilot to adjust the throttles or speed without disengaging the autopilot. +Situational AwarenessThe situation awareness questionnaire, SART was administered to the pilots after every simulation run.They rated 10 SART elements on a scale of 1 to 7, where 1 is 'low' and 7 is 'high.'Thus the data has been analyzed for the all the conditions for each of the three pilots.Due to low statistical power for testing, significance tests were not calculated.The situation awareness ratings have been depicted on a line graph to enable better trend comparisons for the conditions.Figure 6 shows that the SA trends for the different sub-elements are the same for the aircraft starting with 10s or 5s temporal separation between them.The pilots did not feel that any of these situations were unstable, and level of variability and complexity was similar in the two conditions.In the group discussions, the pilots mentioned that they preferred their aircraft to be ahead rather than behind on the LSI because that increased the chances of the aircraft getting into the wake zone and out of the safe zone.Pilot's responses on situation awareness for the simulator flying on the straight-in path (landing on 18L) or on the slewed path (landing on 18R) (Figure 7) show similar trends.The slewed path was considered slightly more unstable, variable, and complex by the pilots, but they also felt that higher level of concentration and familiarity was required with the situation.The situation awareness responses for the visibility condition (Figure 8) showed that the pilots experienced similar levels of awareness in the clear versus poor visibility condition.In general, they felt that the poor visibility condition was slightly more variable, unstable, and complex.The pilots required slightly more alertness, and they had slightly less spare mental capacity in the poor visibility condition as compared to clear visibility condition.The information quality, information quantity, and familiarity with the situation were about the same for both of the visibility conditions +Observer Notes and Group DiscussionsThe observer data yielded some interesting findings.Comments during and after the simulation runs from the three participants pertained to issues related to the tools and procedures for closely spaced parallel approaches, wake avoidance, and non-normal events.In addition, many comments were provided that were associated with the interface of the concept elements, in particular the alerting and display features.The three pilot participants had several comments about what they perceived were the critical aspects of the closely spaced parallel approach concept as it was represented in this study.Pilots stated that they felt that the high degree of automation required for the closelyspaced tasks was necessary for the precision of the procedure; however, they all expressed the need for some opportunity to intervene or "fine tune" the automation.For example, the ability to manually adjust the speed was recommended by two of the participants.In four of the eight scenarios, pilot participants flew these procedures with visibility at the KSRT airport down to about 600 feet of Runway Visibility Range (RVR).Another opinion that had general consensus was that flying these types of closely spaced procedures had a higher risk in these low-visibility surface environments.The comments indicated that although the pilots understood that automation tools would be necessary for navigation guidance and the avoidance of wake vortices, they preferred attaining a visual of the other aircraft to detect any cues that may indicate wake vortex threat or the threat of a possible unexpected maneuver.The other four scenarios were in clear weather, and were generally found to be more acceptable conditions for the approaches.The pilot participants had many comments about the display of the wake information.In general, they found the wake depiction and the display locations acceptable.They preferred wake depiction on the ND versus the PFD.One pilot had stated that it took him some time to understand wake on the PFD, raising the issue of the limited training the pilots received for this simulation.As the previous comments indicated, there were some concerns about the ability to predict wake responses during low visibility conditions.In addition, all three pilots stated that they did not fully understand the nature of wake characteristics, and how they may impact their own aircraft in closely spaced parallel approaches like those flown in our scenarios.They welcomed aircraft automation that provided information on wake behaviors and their impact on these procedures. +SummaryThis study investigated a concept that incorporates wake information and new technologies to allow for the use of very closely spaced parallel runways in all-weather conditions.The airport and 25 nmi of surrounding airspace were created and simulated as a part of this effort.A high-fidelity simulator with the emulation of a 4-D FMS was used to implement the concept, and several displays were enhanced to enable simultaneous approaches.The pilots provided feedback through their responses to the questionnaires and debriefings.The three pilots had similar results and their suggestions were consistent.In general, they were marginally more comfortable with VCSPR approaches and automation in VMC rather than CAT-IIIB visibility conditions, even though their situational awareness ratings showed similar responses for both conditions.In addition, they indicated that they preferred 10 s versus 5 s spacing between the lead and follower aircraft.The participants stated that they felt it was important for them to be able to deploy gear and flaps manually, and influence speed and throttles without disengaging autopilot.All the pilots were concerned about potential breakout procedures, and think automation will play a large role in the determination of the procedures, with direct involvement of the air traffic controller necessary for safe procedures. +Future WorkThe study provides future research ideas, and guidelines for developing procedures for VCSPR.Current research efforts by NASA and Raytheon are examining the safety and viability of the procedures and technologies associated with escape maneuvers.In addition, the representation of more airport traffic and structures are included so that the implications of surrounding constraints could be explored.The possibility of providing more flexibility in the system where pilots could, for example, deploy gears or use throttles for speed control without disengaging autopilot could also be explored.Figure 11Figure 1 Final approach geometry for TACEC +Figure 22Figure 2 Primary flight display on straight-in parallel final +Figure 33Figure 3 Navigation Display on straight-in parallel final +Figure 44Figure 4Ratings for ease of deriving information from the displays +Figure 55Figure 5 Pilots' subjective ratings on statements regarding the concept and displays +Figure 66Figure 6 Situation awareness responses 10s v/s 5s distance between the two aircrafts +Figure 7 Figure 878Figure 7 SA responses for aircraft on straight in v/s slewed approach + + + +Copyright StatementThe authors confirm that they, and/or their company or institution, hold copyright on all of the original material included in their paper.They also confirm they have obtained permission, from the copyright holder of any third party material included in their paper, to publish it as part of their paper.The authors grant full permission for the publication and distribution of their paper as part of the ICAS2008 proceedings or as individual off-prints from the proceedings. + + + + + + + Reduced waste generation technical work plan + 10.2172/5213424 + + December 12, 2004 + Office of Scientific and Technical Information (OSTI) + + + Joint Development and Planning Office. The Next Generation Air Transport System Integration National Plan. December 12, 2004. + + + + + Compilation of DOT and FAA Airport Legal Determinations and Opinion Letters as of December 31, 2020 + 10.17226/26202 + DOT/FAA/CT-TN02/01 + + 2001 + Transportation Research Board + + + K + Magyratis, S.; Kopardekar, P.; Sacco, N. & Carmen (2001), K. Simultaneous Offset Instrument Approaches -Newark International Airport: An airport feasibility study. DOT/FAA/CT-TN02/01. + + + + + Frequency and severity of deviations during simultaneous independent approaches to parallel runways + + SebastianVMassimini + + + GerryCMcneil + + + NoamTene + + 10.1109/icnsurv.2010.5503313 + MP 06W0000127 + + + 2010 Integrated Communications, Navigation, and Surveillance Conference Proceedings + + IEEE + 2006 + + + Mitre Report + Massimini, V.S. (2006). Simultaneous Independent and Dependent parallel Instrument Approaches. Assumptions, Analysis, and rationale. Mitre Report MP 06W0000127. + + + + + NASA administrator announces resignation + + TSAbbott + + + DMElliot + + 10.1016/s0045-8732(01)80042-5 + Concept. NASA/TP-2001-210665 + + + COSPAR Information Bulletin + COSPAR Information Bulletin + 0045-8732 + + 2001 + 152 + + 2001 + Elsevier BV + + + Abbott, T. S. & Elliot, D. M. (2001). Simulator Evaluation of Airborne Information for lateral Spacing (AILS) Concept. NASA/TP-2001-210665. + + + + + CNS requirements for precision flight in advanced terminal airspace + + MEMiller + + + SDougherty + + + JStella + + + PReddy + + 10.1109/aero.2005.1559471 + + + 2005 IEEE Aerospace Conference + + IEEE + 2005. 2005 + + + + Miller, M.E.; Dougherty, S.; Stella, J. & Reddy, P. (2005). CNS requirements for precision flight in advanced terminal airspace. Aerospace, 2005 IEEE Conference, pp 1-10. + + + + + Operational concepts for closely spaced parallel approaches + + ARPritchett + + + SLandry + + 10.1109/dasc.1999.863730 + + + Gateway to the New Millennium. 18th Digital Avionics Systems Conference. Proceedings (Cat. No.99CH37033) + St. Louis MO + + IEEE + October 1999 + + + + Pritchett, A. R. and Landry, S., Operational Concepts for closely spaced parallel approaches, Proceedings of the Digital Avionics Systems Conference (DASC00), St. Louis MO, October 1999, pp. 4.C.2-1. + + + + + Flight Testing WAAS for use in closely spaced parallel approaches + + SHouck + + + ABarrows + + + BParkinson + + + PEnge + + + DPowell + + + + Proceedings of the ION-GPS 1999 + the ION-GPS 1999 + + 1999 + + + Houck, S.; Barrows, A.; Parkinson, B.; Enge, P.; Powell, D. (1999) Flight Testing WAAS for use in closely spaced parallel approaches. Proceedings of the ION-GPS 1999 + + + + + + + Meeting + + + Nashville, TN + + + Meeting, Nashville, TN. + + + + + Cockpit Display of Traffic and Wake Information for Closely Spaced Parallel Approaches + + GordonHardy + + + EmilyLewis + + 10.2514/6.2004-5106 + AIAA-2004-5106 + + + AIAA Guidance, Navigation, and Control Conference and Exhibit + Providence, RI + + American Institute of Aeronautics and Astronautics + 2004. August 16-19, 2004 + + + Hardy, G. H. and Lewis, E. K., (2004). "A Cockpit Display of Traffic Information for Closely Spaced Parallel Approaches", AIAA-2004-5106, AIAA Guidance, Navigation, and Control Conference and Exhibit, Providence, RI, August 16-19, 2004. + + + + + Models of Wake-Vortex Spreading Mechanisms and Their Estimated Uncertainties + + VernonRossow + + + GordonHardy + + + LarryMeyn + + 10.2514/6.2005-7353 + + + AIAA 5th ATIO and16th Lighter-Than-Air Sys Tech. and Balloon Systems Conferences + + American Institute of Aeronautics and Astronautics + 2005 + + + Rossow, V. J., Hardy, G. H., and Meyn, L. A. (2005). "Models of Wake-Vortex Spreading Mechanisms and Their Estimated + + + + + Models of Wake-Vortex Spreading Mechanisms and Their Estimated Uncertainties + + VernonRossow + + ATIO Forum + + + + GordonHardy + + ATIO Forum + + + + LarryMeyn + + ATIO Forum + + + 10.2514/6.2005-7353 + AIAA-2005-7353 + + + AIAA 5th ATIO and16th Lighter-Than-Air Sys Tech. and Balloon Systems Conferences + Arlington, VA + + American Institute of Aeronautics and Astronautics + September 26-28, 2005 + + + Uncertainties", AIAA-2005-7353, ATIO Forum, Arlington, VA, September 26-28, 2005. + + + + + Situation Awareness as a Predictor of Performance for En Route Air Traffic Controllers + + FrancisTDurso + + + CarlaAHackworth + + + ToddRTruitt + + + JerryCrutchfield + + + DankoNikolic + + + CarolAManning + + 10.2514/atcq.6.1.1 + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 6 + 1 + + 1998 + American Institute of Aeronautics and Astronautics (AIAA) + + + Durso, F. T., Hackworth, C. A., Truitt, T. R., Crutchfield, J., Nikolic, D., & Manning, C.A. (1998). Situation Awareness as a predictor of performance for en route air traffic controllers. Air Traffic Control Quarterly, 6(1), 1-20. + + + + + + diff --git a/file767.txt b/file767.txt new file mode 100644 index 0000000000000000000000000000000000000000..f5dfbdd597e89ce58055fee6ec5f9b6adee36f09 --- /dev/null +++ b/file767.txt @@ -0,0 +1,421 @@ + + + + +I. INTRODUCTIONUrban Air Mobility (UAM) is gaining interest as the need for On Demand Mobility in today's congested road traffic is increasing in metropolitan areas [1,2].UAM is envisioned as the concept to transport passengers and cargo safely and efficiently using innovative aircraft in these urban areas [3].It is expected to improve mobility for the general public, decongest road traffic, reduce transport time, and reduce strain on existing public transport networks [3].There exist several challenges to UAM, such as integration of procedures with airspace and the airport, maintaining operations within acceptable noise levels, gaining public acceptance, developing a path for vehicle certification, and more.Integration of UAM operations in the National Airspace System (NAS) has been the focus of the research conducted at NASA Ames Research Center under the Air Traffic Management -eXploration (ATM-X) UAM sub-project.Previous research on UAM operations focused on understanding the capabilities and limitations of helicopter operations in the current day environment that might be applicable to UAM [4].The research found that current day requirements for helicopter operations, similar to UAM operations, for obtaining verbal clearances to Class Bravo airspace was the biggest limiting factor.Digitizing communication would not be a feasible solution for such clearances for UAM operations since it would simply substitute the verbal communications, changing the nature of the workload but not reducing the workload for the air traffic controllers.To make the UAM operations scalable, the traffic management for the small Unmanned Aircraft Systems (sUAS) was proposed as a solution.The UAS Traffic Management (UTM) paradigm [5,6], describes a service-oriented architecture with a focus on third party services.The study reported in this paper explores the application of the UTM architecture to UAM operations.A simulation performed in collaboration with Uber Technologies investigated if the performance of the UTM architecture and its implementation from UTM's Technical Capability Level 4 (TCL-4) [6,7] were extensible for UAM operations, and if the data exchange between multiple operators as planned under UTM were adequate for UAM operations in shared airspace.The main objectives of this engineering evaluation were to explore information and data exchange requirements, identify key questions regarding access of controlled airspace from the Air Navigation Service Provider (ANSP) perspective, and to explore digital airspace integration procedures.In the next section, Section II, different concepts of operations that exist for UAM operations will be described.In Section III, a description of the method that includes a highlevel overview of the system configuration, airspace definitions, traffic scenarios, and engineering evaluation use cases will be provided.Section IV will describe the results from analyzing the post-simulation data and also capture some of the lessons learned.Section V will include a summary and proposed next steps. +II. BACKGROUNDThe approach to airspace management for UAM operations simulated in this study is an extension of the concept for UTM to enable UAM operations, with special consideration given to ensuring the extensions are interoperable with the existing Air Traffic Management (ATM) system.Details of the UTM operational concept are included in the Federal Aviation Administration's (FAA's) UTM Concept of Operations (ConOps) document [12].An integral aspect of this airspace management approach is the idea that third parties will provide airspace management services to aircraft.The UTM architecture described in [5,6] refers the service providers for UAS as UAS Service Suppliers (USS).It is likely that service providers for UAM operators will be differentiated in the future, but for now we refer to them as USS in this paper.The tests described in this paper were designed to highlight key challenges related to extending the UTM concept to UAM and establish the simulation and testing infrastructure necessary to develop more detailed concepts and procedures for UAM operations in the NAS.The study assumed the following operating assumptions: aircraft operations simulated in this study were conducted under today's visual flight rules (VFR), with a qualified pilot in command onboard the aircraft.Each all-electric, vertical takeoff and landing (eVTOL) aircraft is expected to have a capacity of transporting up to four passengers and the pilot.Operations will depart from and arrive at vertiports that have been constructed to the relevant guidelines and approved for this UAM use case; they will not operate out of unprepared sites.Aircraft will fly along highly structured routes that are known and acceptable to the local air navigation service provider and communities.These flights will be relatively short because they are conducted within a metropolitan area, normally under twenty minutes.An important consideration for accessing Class B airspace is the manner in which "airspace authorization" is achieved.Airspace authorization is the means by which an operation is approved to operate in a particular airspace.Today, authorization may be granted through submission of a flight plan and verbal clearance from an air traffic controller (e.g., for Instrument Flight Rules [IFR] operations), or be allowed without explicit clearances (e.g., for VFR operations in uncontrolled airspace).Authorization to conduct a particular operation will require that aircraft are appropriately certified, and pilots are licensed for the types of operations they will conduct, that maintenance schedules are being followed, that weather conditions are appropriate, and other criteria are met.It is expected that systems and procedures will be developed to support automated, digital authorization of an operation that enables aircraft to access terminal airspaces (i.e.airspace classes B, C, and D) without first receiving a verbal clearance or making verbal contact with air traffic control.This authorization would occur using the UTM architecture after some modifications have been made to it.Within busy controlled airspace there will likely need to be specially designated "airspaces" in which UAM aircraft are allowed to operate without verbal clearances or contact with air traffic controllers.These "UAM-authorized airspaces" are envisioned as discrete subspaces of the greater Class B, C, or D airspace that would be geographically static over weeks or months, though their accessibility for UAM operations could be relatively dynamic, based on current air traffic flows, weather, airport configurations, and other factors.Similar to helicopter routes, these "airspaces" would be located in areas not typically used by traditional aviation and would avoid noise-sensitive areas or geographies with other incompatible land uses.The airspace used for UAM operations would continue to be classified according to the surrounding airspace, just as the special flight rules area (SFRA) [14 CFR Part 93 Subpart G)] above Los Angeles International Airport (LAX) continues to be designated Class B, but permits access using a different authorization method.Importantly, as envisioned, the UAM airspace/routes would not be associated with a particular UAM operator, and any appropriately qualified and authorized UAM aircraft would be permitted to operate in them.The key operational benefit of the UAM airspace/routes are that operations in Class B, C, or D airspace would not require traditional air traffic control services provided by ANSPs and utilize instead services provided by a third party (e.g., a USS) in conjunction with pilot responsibilities.In effect, the traditional ANSP services necessary to ensure the safety and efficiency of operations in these otherwise complex environments would be replaced by highly structured routes, common aircraft operating characteristics and procedures, as well as information exchanges between participating aircraft facilitated by their respective USS.Significant work remains for the UAM community to define the requirements for these services and the technologies that will deliver them. +III. METHODS +A. Extension of the UTM paradigmThe UTM architecture delineates the functions provided by the ANSP, third party service providers or the USS Network.This service-oriented architecture includes connection to the regulatory authority via the Flight Information Management System (FIMS).FIMS is a central, cloud-based component that enables information exchange between ANSP (e.g.FAA) systems and the USS Network.The USS enables UAM operators to communicate with each other and with the ANSP about both planned and ongoing operations.The UTM system developed for TCL-4 [6,7] field tests were used as the starting point for UAM operations in this research.Compared to the current UTM operations that serve the sUAS that fly below 400 AGL, UAM operations pose a higher safety risk compared to UTM operations as they are envisioned to transport people flying between 500 AGL and 3,000 AGL, with more interactions with other traffic.Given such differences, this research aimed to understand the applicability of the UTM paradigm and architecture for coordinating operations within the UAM routes that are shown in the Airspace section.Within the UTM paradigm, USS submit operations with their intended flight path as operational volumes.The operational volumes can be either Transit Based Operational Volumes (TBOV) or Area Based Operational Volumes (ABOV).TBOVs are based on a known route or flight profile, where lateral and vertical boundaries are built around a centerline.The TBOV includes any geographical buffer required to account for the UAS' ability to maintain flight along the centerline (navigation performance capabilities, environmental factors, etc.).ABOVs represent a larger block of airspace encompassing a mission profile because the projected route may be too complex (e.g., survey operation) or dynamic (e.g., search and rescue mission) to describe, or if the UAS has minimal navigation capabilities and can only perform visual line of operations.TBOVs were selected for this study since they could be adapted for routes that were planned for UAM flights shown in Fig. 1.The design and size of the TBOVs is expected to be managed by the UAM operator.This study contrasted two approaches to TBOV design; NASA and Uber each designed their TBOVs with different parameters.In both designs the maximum length of TBOV was set to 100,000 feet to prevent any operator from blocking large portions of airspace for their operations and to allow several TBOVs to be built for the experimental UAM routes.This can be compared to the 6000 feet maximum length of TBOV allowed in UTM.Fig. 1 shows a notional representation of the TBOVs, and how they are stitched together to represent a UAM path between two vertiports.NASA planned for each TBOV's maximum traversal time to be 60 sec in length.The dimensions of the NASA-defined volumes were 500 ft vertical by 1500 ft lateral between planned, adjacent routes that traverse opposite directions.The spacing between the routes and the height/width of the volume were calculated based on the following assumptions:1. Flights will have required navigational accuracy of 0.1 nmi, which is approximately 600 ft on either side (laterally) of the route; 2. A 150 ft buffer is added to the lateral distance, required by additional services such as a conformance monitoring/ Separation Service used by NASA and also by the target generator that was used to fly the UAM flights; 3. A 250 ft buffer was added above and below the vehicle.Uber designed its volumes such that each TBOV connects two consecutive waypoints of the route.1. TBOVs are set at 75 ft either side laterally of the vehicle and 100 ft each side vertically making the rectangular cross section of the volume 150 ft wide and 200 ft tall.2. For the Uber routes that intersected with NASA routes, TBOVs provided by Uber started at the crossing fixes.Uber's projections of air traffic assume airspace utilization with many routes connecting vertiports in cities.These engineering evaluations allowed investigation of the given volumetric dimensions around the routes that assumed far-term future Required Navigation Performance (RNP) levels. +B. AirspaceThe technical approach taken by NASA involved constructing the initial UAM airspace system by connecting UTM TCL4 and Testbed [12] at NASA Ames Research Center.Several services were developed for UAM operations and connected to Testbed to investigate information requirements for multiple UAM operator coordination.The two UAM operators (i.e.NASA and Uber) operated in the same airspace using the UTM architecture to share operational information.The test focused on the basic services originating from the UTM paradigm, that would be required for UAM routes operating in Classes B/D/E/G airspaces.NASA also planned to test additional services, such as scheduling and conflict detection and resolution.For the purposes of the engineering evaluation, the interaction of these flights with Air Traffic Control (ATC) was not considered.Although the concept of operations is still being finalized, it is likely that for UAM operations, routes will be pre-defined and published and shared with all operators, which was not a requirement for UTM operations.Routes were planned in Dallas-Fort Worth airspace.There were some routes that connected the large airports like Dallas-Fort Worth (DFW), Dallas-Love Field (DAL) and Addison (ADS).Both DFW and DAL are in Class B airspace whereas ADS is in Class D airspace.Most of the routes in Class B airspace were designed to fly UAM vehicles at 500 or 1000 ft AGL.There were some routes that were designed and developed in Class E/G airspace to study the interaction between the different operators.Interaction with sUAS was not considered when these routes were developed and should be considered in future studies.The routes in Class G and E airspace had routes flowing in the opposite directions that were 1500 ft apart.Fig. 2 shows the UAM routes planned in the DFW area and their colors show the altitude at which those flights were planned. +C. Traffic ScenariosThere were two traffic scenarios used for the engineering evaluation that defined the level of simulated UAM traffic.However, the results of only one scenario is described in this paper; this scenario operated about 110 NASA operations on all the UAM routes tested.The number of Uber operations planned for the entire simulation were about 85 for their traffic scenario.This led to a total of 200 operations planned for the 40 min period.UTM TCL-4 performance limits were considerably expanded to achieve the 200 simultaneous operations since TCL-4 implementation did not test with more than 40 operations. +D. Use CasesThe Use Cases were developed to investigate non overlapping resources between multiple operators, as a reasonable approximation of actual implementation of operations, before more complex conditions were considered.The rationale was to allow development of services that start by providing situational awareness to the two UAM operators and expand to scheduling and separation functions in both the strategic and tactical timeframes.Thus the use cases were built with increasing complexity with the goal of allowing future operations where routes would be shared among operators. +1) Use Case A: Two UAM Operators, Different Resources, Shared AirspaceUse Case A was developed to evaluate two UAM operators that manage different resources in a shared airspace.NASA and Uber provided their own separate sets of pre-defined UAM routes which were shared as adaptation data prior to running the simulation (Fig. 3 and4).In this use case, Uber and NASA did not share any routes or vertiports, but were aware of each other's routes and vertiports.Executing the UTM paradigm, the USSs shared position information with each other and operated within the same operational area.Each operator was responsible for the management and scheduling of their flights on their own set of UAM routes.All UAM flights were assumed to be equipped with ADS-B out, and their real-time positions were available.NASA and Uber agreed to allow 4D volumes to overlap among their operations, disabling the negotiation function, in order to leverage other separation services.Uber's UAM flights operated with some of their basic services such as route generation and flight trajectory management.NASA operated with some of its additional services for scheduling and separation in some of the runs. +2) Use Case B: Multiple UAM Operators, Constrained Resources, Shared AirspaceUse Case B was focused on multiple UAM operators in shared airspace with constrained resources (see Fig. 5).The crossing fixes between the NASA and Uber flights served as the constrained resources.The following process was used for Use Case B: NASA submitted its TBOV for any given flight to the NASA USS, followed by Uber's submission of their flight's TBOV.It was expected that Uber would submit its TBOVs overlapping the crossing fixes, six minutes prior to the NASA flight's scheduled/ desired departure time, and it was used by NASA's scheduling service to condition the traffic flows, when applicable.At that time, the scheduling service was provided with scheduled departure times and Estimated Times of Arrival (ETAs) to any constraint points along the route and at the destination vertiport.Thus, NASA flights saw Uber flights as fixed constraints and would be re-scheduled to avoid the Uber flights.NASA services continued to deconflict NASA flights with each other using a set of separation minima described in Section F, and also de-conflict with Uber's TBOVs. +3) Use Case C: Multiple Operators, Shared Airspace, Managed ContingencyThe objective of Use Case C was for NASA and Uber UAM operators to share the airspace and manage a contingency that needed to be resolved in a tactical time frame.The initial steps involved in NASA's and Uber's pre-departure and pre-planning activities were the same as in Use Case A.At some point during the flight, the Uber UAM flight detected a battery issue.The Uber USS determined the nearest vertiport was VP5 as depicted in Fig. 7.The new Uber operational volume V3, overlapped their own volume V1, enroute at the contingency point, and also overlapped NASA's flight volume V2, destined to vertiport VP4.The Uber USS submitted an operational plan change request to volume V3, and updated the ETA and the operation's end time, which was then accepted by the system.The Uber UAM aircraft received the approval and adjusted its flight maneuvers to stay within V3.The Uber contingency aircraft was given the highest priority in the system and was treated as a "priority or exempt flight" (i.e., hard constraint) by NASA Scheduling and Separation Services in the runs where those services were utilized.Uber elevated that operation to an emergency severity.NASA Separation Services detected conflicts and issued resolution maneuvers to the NASA aircraft to maintain separation between Uber's contingency aircraft (with its new operational volume) and the NASA UAM aircraft and its volume. +E. Uber Services and ParametersUber has developed a set of core airspace management services tailored to controlled airspace access.They include a Route Generation Service which handles the nominal route creation that ingests and converts waypoint-based routes to ARINC429-inspired route structure.The Operation Management Service handles the lifecycle of each operation including their creation, activation, publication and termination.An Operation Lifecycle Manager is responsible for managing the state of each operation from creation/scheduling to landing/closing.In addition, the service conforms to the UTM specification and provides the UTM operational states (proposed, accepted, activated, nonconforming, rogue, closed) for each of Uber's operations.State transitions processed here are triggers to other services; i.e., sending closure announcements to other USSs when an operation closes.An Operation Volume Service generates TBOVs as part of the operation creation process.The TBOV Generation configuration was described above.A Flight Monitoring Service continuously performs trajectory conformance monitoring by comparing the flown trajectory with the predicted trajectory.Any deviations that are detected trigger appropriate state transitions in the Operation Management Service.A Vehicle Integration Service (VIS) handles commands to the vehicles (e.g.takeoff, landing), and uses those commands as triggers to call the other services.When a contingency plan is commanded, VIS immediately calls the Operation Management Service to generate the new set of TBOVs and update them. +F. NASA Services and AssumptionsNASA services had similar functions as Uber's Basic Services.The trajectories generated for NASA UAM aircraft used a Target Generator (TG) prior to departure, which incorporated the simulated UAM aircraft model characteristics such as speed and flight profiles during climb and descent.The Fleet Operator then generates the TBOVs that encompass the trajectory, which is then submitted by the USS as part of the operation plan.The operation volumes were built to be up to 60s temporal length for every operation.The Fleet Operator was responsible for the states and lifecycle of all NASA's operations, connects NASA airspace management services with NASA's simulation environment (Testbed) and with the NASA USS (NUSS).Like other USS, NUSS was responsible for trajectory conformance monitoring and announced the operations as nonconforming when they departed their given operational volumes.To evaluate the impact of strategic scheduling and tactical separation services on the UAM operations, NASA also utilized additional services -Scheduling Service and Separation Service that were tested in selected runs.These services were used to manage scheduling and separation of UAM operations in addition to the strategic de-confliction by operation volumes under the UTM paradigm.NASA's Scheduling Service provided the strategic de-confliction function and conditioned the flows with pre-defined in-trail, crossing, and merge points spacing requirements for the NASA flights and also took into consideration Uber's flights and their volumes that overlapped with NASA flight's volume at the crossing points.NASA's Scheduling Service, Network Scheduler (NS), is based on a scheduler developed for NASA's Airspace Technology Demonstrations 1 and 2 (ATD-1 and ATD-2) and has a tested and proven outcome and stability [10].The NS manages the schedule by scheduling arrivals at nodes on a route network.Given that TBOVs are used in UAM operations, the scheduling algorithm is modified to accept the ETAs from the NASA flights and TBOVs from other operators using a node blocking functionality that separates flights from the beginning and end of the TBOV.The result is that TBOVs may overlap, but the vehicles are separated by the relevant spacing requirement.NASA's Scheduling Service assumed that the capacity at each vertiport was one arrival and one departure per minute, thus the departure demand built into the traffic ranged between 40s and 90s.The Scheduling Service worked towards 45s in-trail separation and 15s separation at the crossing points.The crossing point restriction was applied inside the controlled airspace only; it was assumed that the Separation Service will have the flexibility to maneuver outside controlled airspace.Uber sent 15-20s volumes with a buffer of 45s on either side for flights that shared the crossing point with NASA's flights, and the scheduling service applied 15s on either side of Uber's TBOV, making the constraint volume approximately 140s long centered on the crossing points.The other service that was developed and tested in some runs was NASA's Separation Service (Auto Resolver-AR) [11], which provided the tactical de-confliction function.The Target Generator was used for generating trajectories prior to departure, but the Separation Service (AR) algorithm also constantly predicts trajectories when the UAM operation is airborne.This service generated commands to the vehicle for maneuvering them for conflicts, which are then passed to the Target Generator via the Fleet Operator and commands the targets/vehicles.The look ahead time for conflicts was set to 5 minutes by the service.The separation minima used by this service was 400ft of vertical separation and 1200ft lateral separation between two NASA UAM aircraft.The same separation minima that was applied between UAM and other VFR aircraft.The separation minima that was applied between NASA and Uber was 750ft lateral and 250ft vertical.At the vertiports, arrival compression was taken into account and the vertiports were assumed to have a "bubble" around the vertiport where the separation criteria would not be applied.The radius of this "bubble" was set to 1 nmi.This simplification allowed the services to focus on the enroute separation requirements. +G. Experimental MatrixThe experimental conditions included the three Use Cases (A, B, and C) as detailed earlier, under four different conditions including:• Baseline (Basic services only),• Use of Scheduling Service (NS) only • Use of Separation Service or AutoResolver (AR) only • Use of both Scheduling and Separation Services (NS +AR)The next section discusses results and for brevity, mostly focuses on Use Case B where both NASA and Uber shared resources in the airspace (crossing points) in a strategic time frame.Some results in Use Case C where there was tactical deconfliction required, due to an emergency case, are also shared.These metrics described below point to some key differences from the way operations are defined for UTM. +IV. RESULTS AND LESSONS LEARNEDThis section focuses on results contrasting the NASA and Uber services under the three use cases, and examines the number of volumes per operation, the number of TBOVs per operation per route, and the position messages per operation.Recall that the nature and type of missions envisioned for UAM is different than UTM in many ways, and researchers wished to explore potential system limits that might be encountered with the UAM concept.The number of volumes per operation is expected to depend upon the complexity of the intended operations, airspace constraints, and route changes.Increases in number of volumes per operation is likely to stress the system by adding latency between publish and response time to messages required for those operations.Increased TBOVs also create more messaging requirements as USSs provide more position updates, which can also stress the system. +A. Number of Volumes per Operation 1) Distribution of number of TBOVs for Baseline ScenarioThe number of volumes generated for each operation in each of the three Use Cases are depicted in Fig. 9, 10 and 11 for the Baseline scenario where basic services were utilized in each of the three simulation runs.In the histogram plots, the horizontal axis shows the number of operational volumes in a single operation and the vertical axis shows the number of operations with a given number of volumes for both NASA and Uber operations.Use Case A has fewer operations in total, since there were fewer routes flown, and they were not shared between NASA and Uber.Use Cases B and C have higher numbers of operations, and those operations have higher numbers of operational volumes due to the length of the new routes that were added for Use Cases B and C.Uber flights utilized basic services in all their operations and generated the same number of routes and operations, which explains why their number of volumes per operations is consistent across Use Cases (10 or 15 volumes for most operations).In Use Case C, there are several Uber operations with six volumes per operation because those flights were exercised as emergency or high priority flights and updated their operations with few volumes while they intersected the NASA flights.For the NASA flights, the number of volumes per operation is similar for Use Cases B and C, which is about 20 or 25 volumes per operation, whereas Use Case A shows that most of the NASA operations have 20 volumes per operation.NASA generated volumes that were kept under or equal to 60 sec as temporal length for its operations whereas Uber generated volumes of different sizes, which could at times overlap among its operations.This data shows that different operators in the operational world could design TBOVs with different parameters that could signify UAM vehicles with different performance characteristics.However, the smaller the lateral, vertical or temporal length of the volume could place greater demands on conformance monitoring.Larger number of volumes per operation were possible because the UTM TCL4 performance limits were expanded. +2) Distribution of Number of TBOVs with Advanced ServicesThis subsection describes the effect of the strategic Scheduling Service or the Separation Service on the distribution of number of TBOVs per operation for NASA operations only.The following histogram plots (Fig. 12 and 13) show the distribution of TBOVs for Use Case B only.Fig. 12 shows that when both strategic scheduling and separation services are utilized, the flow of traffic is conditioned and highest number of volumes per operation is shifted to the right.There are about 30 volumes for about 58 operations and there are a few flights (under 10) that have 35 or 40 volumes per operation as well.The higher number of TBOVs are generated because the advanced services tend to modify trajectories leading to extra waypoints, around which extra volumes may be generated.This effect is particularly pronounced when the strategic Scheduling Service (NS) is utilized (Fig. 13).The service tends to use Trajectory Generator to delay flights on the ground when it detects conflicts with Uber flights at the crossing points, and those trajectory generations provide additional waypoints.These additional waypoints add new volumes keeping the design of the TBOVs in mind.Thus, advanced services should be designed keeping the design for the generation of TBOVs in mind, since they impact the number of TBOVs per operation. +B. Number of TBOVs per Operation per RouteThe number of TBOVs per operation per route is shown in Fig. 14 and 15 for NASA and Uber flights respectively for Use Case B, Baseline condition, where only basic services were utilized by both NASA and Uber.The number of volumes per operation per route is proportional to the length of the route for Uber flights except for one outlier (Fig. 15), whereas it is not always proportional to the length of the route for NASA flights but has a linear trend with a few outliers (Fig. 14).One of the reasons that the number of volumes is scattered for the NASA flights (Fig. 14) vs Uber flights is that flight route length depends on the route structure; a route with more turns is likely to have a larger number of waypoints and thus larger number of TBOVs generated around those waypoints, especially when the design restricted the size of the NASA volumes to be under 60 sec.In general, when total number of operations increase in the airspace, a higher number of volumes generally means that the volumes sizes are smaller and that conformance to the volumes can be a challenge for the overall system. +C. Number of POSITION Messages per Operation 1) Distribution of Number of Position Messages for Baseline ScenarioThe average number of position messages per nmi that were exchanged during the course of the simulation for each of the three Use Cases are compared for the Baseline condition where basic services were utilized.The largest fraction of the messages between the different operations for NASA and Uber flights were position messages, which an aircraft sends every second while it is airborne; the distribution of position messages is listed in Table 1.As expected, the data show that the average number of position messages were directly proportional to the length of the routes, that ranged from 13.5 to 33 miles for NASA flights and from 18 nmi to 26 nmi for the Uber flights.Uber had fewer operations due to fewer routes planned in all Use Cases and that is reflected the lower number of Position messages per nmi of the route.For NASA flights, it's observed that Use Case A has the fewest routes and produced the fewest Position messages.Use Cases B and C have additional Position messages shared due to additional routes and crossing points between NASA and Uber flights.This increase in the Position messages using the UTM TCL-4 implementation required the addition of several machines to archive the position data since they were recorded at 1 hz.Operators may decide to share Position data among themselves at all times to allow for tactical separation and collision avoidance for UAM operators, which was not considered a need for UTM operations, where strategic deconfliction of operations was sufficient to enable safe operations.Table 2 lists the Total number of Position messages produced during the entire simulation run and shows that the number of Position messages were in the range of 100,000-120,000 for NASA operations and in the range of 21,000-27,000 for Uber flights.This means that although only basic services were utilized during the Baseline condition, the number of Position messages shared was similar across the different Use Cases within each Operator's flights. +2) Distribution of Number of Position Messages with Operational Services ProvidedIn this subsection, for NASA operations only, the effect of having the Scheduling Service (NS) or the Separation Service (AR) is compared with the distribution of total number of Position messages exchanged in Use Case B. Table 3 shows the highest total number of position messages were seen for Baseline conditions, when the traffic flow is not conditioned.It was also observed that the Total number of Position messages exchanged were reduced when both the NASA advanced services were utilized.Table 3 shows that when Scheduling Service is utilized by itself, it has the least total number of Position messages.This is the case because the service delays the flight pre-departure on the ground and Position messages are shared only when the flight is airborne.The total number of Position messages exchanged when both services are utilized is higher than Scheduling Service acting alone.This is because the Scheduling service imposes ground delay, but Separation Service may increase the position messages due to maneuvers or resolutions that it sends to de-conflict the operations, it may also delay the flight pre-departure.In contrast, the Separation Service has relatively fewer Position messages exchanged than Baseline since it also imposes a small amount of ground delay for operations that may have an imminent conflict at the time of departure +D. DiscussionThe results show that generation of TBOVs or operational volumes is an essential service to the UTM paradigm that was adopted and tested for UAM operations.However the design and implementation of this service has an effect on the size and the number of the volumes per operation.It was essential to remove the maximum TBOV spatial length imposed by UTM for UAM operations because UAM vehicles followed a specific route that could be several miles long.It was observed that if the length of the route took the flight through several grid cells, it led to higher numbers of discovery or position messages to be exchanged.Also, if the route had many turns or merge points, this affected the number of TBOVs that were generated for those extra waypoints, even if the route length were kept the same.Most aircraft followed one of a few routes.Thus for operations on the same routes, the TBOVs that were submitted had the same geometric boundaries with temporal overlapping.It may be more efficient to only have a pointer to the 3D volumes and operation-specific time thresholds for each volume.The size of the operational volume was a design consideration but a large volume size, especially on the crossing points, led to a reduction in density of flights passing through the crossing points.This was because the separation and scheduling service was based on avoiding the TBOVs of the other operator, whereas the usage of scheduled time of arrival with some buffers would have considerably improved the density of operations over the crossing points.It was also observed that the TBOV around merging points tend to be small, generally due to close waypoints associated with routes that have two-way operations for multiple operators.The trajectory predictions made pre-flight were generally not found to be very precise due to uncertainties in the system, and simulation capabilities led to the actual operational volumes around crossing points to be larger than expected.NASA operations expected Uber operations to provide their TBOVs at the crossing and merge points around six minutes prior to departure of the flight.From discussions between NASA and Uber, it was realized that providing operational data more than six minutes prior to departure would require larger buffers at the merge points to accommodate for uncertainties in the system,.In this study, trajectory predictions with uncertainties such as winds were not included and would be expected to have an additional impact on the size of the TBOVs.In the operational world, operators may design TBOVs keeping the performance characteristics of the vehicle in mind.The smaller the buffers used in the design of the TBOVs, the higher are needs placed on conformance monitoring.It is likely that the UAM concept of operations will require that UAM routes and airspace are pre-published and shared with all operators in a given region.This is significantly different from UTM operations where flight plans and strategic deconfliction is performed at the time an operation is submitted.In contrast, UAM routes and airspace will need to be strategically deconflicted from traditional traffic, so that they could operate with minimal or no ATC monitoring.In this research, strategic de-confliction was provided by NASA's Scheduling Service for UAM operations.This service deconflicted NASA flights from the Uber flights while keeping the Uber flights as a constraint to condition the flows.An alternative to this pre-departure scheduling would be the use of scheduled times of arrival at merge points and crossing points that could also be achieved tactically.Similarly, NASA's tactical Separation Service was used for de-confliction of UAM operations, which can help with inflight re-routing or maneuvering, but the time required to authorize the new operational volume could be on the order of several seconds, which can have an impact on the conformance of the flight to the new route.Thus, there is a need for a look ahead time that takes into account the time required for the system to authorize the new TBOV.The number of position messages increased when NASA's advanced services-Separation Service was used to de-conflict UAM operations, which can stress the performance of the system when the total number of operations increase.Position updates were made at 1hz and it was observed that UTM TCL4 system did not test at such a high tempo of operations; future work will investigate alternate protocols for handling high number of Position messages. +V. CONCLUSIONNASA and Uber utilized the UTM paradigm for UAM operations with different use cases, in a study where resources between the two sets of operations were shared in terms of crossing points.Strategic deconfliction the crossing points was explored and found to require a balance between when the operation details could be shared by the size of the volumes assigned to the crossing points.It was observed that NASA's tactical separation service was able to manage de-confliction with a high priority flight and results showed that the lookahead time must also incorporate the time required to obtain approval of the new route, in allowing conformance to the new route.Increases in number of volumes per operation, or position messages being exchanged, are likely to stress the system and add to the latency between publish and response message time, requiring higher buffer sizes that may be a potential source of inefficiencies.This study showed that overall, the UTM architecture can be successfully applied for UAM operations and that the implementation of services can have a considerable impact on the efficiency of the system.Future work will focus on improving the implementation of the advanced services and also investigating sharing routes between multiple operators.Figure 1 .1Figure 1.Notional diagram representing a series of TBOVs along a route between origin and destination pair. +Figure 2 .2Figure 2. Routes planned for Engineering Evaluation in DFW Area. +Figure 3 .3Figure 3. Use Case A. Multiple operators-different resources in shared space. +Figure 4 .4Figure 4. Use Case A. NASA and Uber routes in DFW Area. +Figure 5 .5Figure 5. Use Case B. Multiple operators sharing resources (crossing points) in shared airspace. +Fig. 55Fig. 5 and 6 show the interaction between NASA and Uber routes, and examples of the crossing fixes that required predeparture scheduling in the strategic time frame.Fig. 6 also shows that a new NASA route was added in Class G airspace to create the constrained and shared resource (crossing fixes) between the NASA and Uber operators. +Figure 6 .6Figure 6.Use Case B. Crossing fixes between Uber and NASA routes +Figure 7 .7Figure 7. Use Case C. Operators share airspace and manage a contingency. +Figure 8 .8Figure 8. Use Case C. Change in Uber flight due to tactical situation +Figure 1 .Figure 2 .Figure 3 .123Figure 1.Number of volumes per operation, Use Case A, Baseline Condition for NASA and Uber flights +Figure 4 .Figure 5 .45Figure 4. Number of Volumes per operation for basic services and advanced service for NASA operations +Figure 6 .6Figure 6.Number of Volumes per operation per route for NASA flights. +Figure 7 .7Figure 7. Number of Volumes per operation per route for Uber flights. +Table 1 . Average number of Position messages per nmi for Use Case A, B and C, Baseline condition for both NASA and Uber flights1Use CaseNASAUberA621552B700669C654492 +Table 2 . Total number of Position messages for Use Case A, B and C, Baseline condition for both NASA and Uber flights2Use CaseNASAUberA118,03426,728B100,59426,283C115,52921,898 +Table 3 . Average number of Position messages per nmi for Use Case B, Baseline compared to Advanced services (scheduling and separation) for NASA flights3ScenarioTotal PositionmessagesBaseline137,602Scheduling and Separation services86,247Scheduling Service Only76,919Separation Service Only114,210 + + + + + + + + + Enabling Airspace Integration for High-Density On-Demand Mobility Operations + + EricRMueller + + + ParmialHKopardekar + + + KennethHGoodrich + + 10.2514/6.2017-3086 + AIAA 2017-3086 + + + 17th AIAA Aviation Technology, Integration, and Operations Conference + Denver, Colorado + + American Institute of Aeronautics and Astronautics + 2017. June 2017 + 5 + + + Mueller, E.; Kopardekar, P. & Goodrich, K. (2017). Enabling Airspace Integration for High-Density On-Demand Mobility Operations. 17th AIAA Aviation Technology, Integration, and Operations Conference. 5- 9 June 2017, Denver, Colorado. AIAA 2017-3086 + + + + + Fast-Forwarding to a Future of On-Demand Urban Air Transportation + + UberElevate + + 10.3030/774086 + + + 2017 + Publications Office of the European Union + + + Uber Elevate, "Fast-Forwarding to a Future of On-Demand Urban Air Transportation," 2017, https://www.uber.com/elevate.pdf. + + + + + Urban Air Mobility Airspace Integration Concepts and Considerations + + DavidPThipphavong + + + RafaelApaza + + + BryanBarmore + + + VernolBattiste + + + BarbaraBurian + + + QuangDao + + + MichaelFeary + + + SusieGo + + + KennethHGoodrich + + + JeffreyHomola + + + HusniRIdris + + + ParimalHKopardekar + + + JoelBLachter + + + NatashaANeogi + + + HokKwanNg + + + RosaMOseguera-Lohr + + + MichaelDPatterson + + + SavitaAVerma + + 10.2514/6.2018-3676 + + + 2018 Aviation Technology, Integration, and Operations Conference + Atlanta, Georgia + + American Institute of Aeronautics and Astronautics + 2018. 25 th -29 th June 2018 + + + Thipphavong, D.P. et.al. (2018) Urban Air Mobility Airspace Integration Concepts and Considerations. 18th AIAA Aviation Technology, Integration, and Operations Conference. 25 th -29 th June 2018, Atlanta, Georgia. + + + + + Exploration of Near term Potential Routes and Procedures for Urban Air Mobility + + SavitaVerma + + + JillianKeeler + + + TamsynEEdwards + + + VictoriaDulchinos + + 10.2514/6.2019-3624 + + + AIAA Aviation 2019 Forum + Denver, Dallas, Texas + + American Institute of Aeronautics and Astronautics + 2019. 17 th June 2019 + + + Verma S.; Keeler, J.; & Edwards, T. (2019). Exploration of Near-term Potential Routes and Procedures for Urban Air Mobility. 19th AIAA Aviation Technology, Integration, and Operations Conference. 17 th June 2019, Denver, Dallas, Texas. + + + + + + PKopardekar + + + JRios + + + T;Prevot + + + MJohnson + + + JJung + + + IiiRobinson + + + JE + + Unmanned Aircraft System Traffic Management (UTM) Concept of Operations. 16 th AIAA Aviation Technology, Integration, and Operations Conference + Washintong D. C + + 2016. 17 th June 2016 + + + Kopardekar, P.; Rios, J.; Prevot, T; Johnson, M.; Jung, J.; & Robinson III, J. E. (2016). Unmanned Aircraft System Traffic Management (UTM) Concept of Operations. 16 th AIAA Aviation Technology, Integration, and Operations Conference, 17 th June 2016, Washintong D. C. + + + + + Standard Specification for UAS Traffic Management (UTM) UAS Service Supplier (USS) Interoperability + + JLRios + + + ISSmith + + + PVenkatesen + + + JRHomola + + + MAJohnson + + + JJung + + 10.1520/f3548 + NASA/TM-2019- 220376 + + + UAS Service Supplier Specification + + 2019 + ASTM International + + + Rios, J L.; Smith I. S.; Venkatesen P.; Homola J. R.; Johnson M. A.; & Jung J. (2019). UAS Service Supplier Specification. NASA/TM-2019- 220376. + + + + + Standard Specification for UAS Traffic Management (UTM) UAS Service Supplier (USS) Interoperability + + JLRios + + + ISSmith + + + PVenkatesan + + + DRSmith + + + VBaskaran + + 10.1520/f3548 + Level 4 NASA/TM-2018-220050 + + 2018 + ASTM International + + + Rios, J. L.; Smith, I. S.; Venkatesan, P.; Smith, D. R.; & Baskaran, V (2018). UTM UAS Service Supplier Development -Sprint 2 Toward Technical Capability Level 4 NASA/TM-2018-220050. + + + + + Flight Demonstration of Unmanned Aircraft System (UAS) Traffic Management (UTM) at Technical Capa... + 10.2514/6.2020-2851.vid + + + Federal Aviation Administration + Washington, D.C. + + American Institute of Aeronautics and Astronautics (AIAA) + 2020 + + + "Concept of Operations v2.0 Unmanned Aircraft System (UAS) Traffic Management (UTM)." Federal Aviation Administration, Washington, D.C., 2020. + + + + + UAS Traffic Management (UTM) Concept of Operations to Safely Enable Low Altitude Flight Operations + + PKopardekar + + + JRios + + + TPrevot + + + MJohnson + + + JJung + + + JERobinson + + 10.2514/6.2016-3292 + + + 16th AIAA Aviation Technology, Integration, and Operations Conference, AIAA AVIATION Forum + + AIAA + 2016-3292, Jun. 2016 + + + Kopardekar, P., Rios, J., Prevot, T., Johnson, M., Jung, J., and Robinson, J. E., "UAS Traffic Management (UTM) Concept of Operations to Safely Enable Low Altitude Flight Operations," 16th AIAA Aviation Technology, Integration, and Operations Conference, AIAA AVIATION Forum, AIAA 2016-3292, Jun. 2016. doi: 10.2514/6.2016-3292. + + + + + NASA's ATM technology demonstration 1 (ATD-1): Integrated concept of arrival operations + + BTBaxley + + + HSwenson + + + TPrevot + + + TCallantine + + 10.1109/dasc.2012.6382981 + ATD- 1 + + + 2012 IEEE/AIAA 31st Digital Avionics Systems Conference (DASC) + + IEEE + + + + B. T. Baxley, W. C. Johnson, J. Scardina, and R. F. Shay, "Air Traffic Management Technology Demonstration-1 Concept of Operations (ATD- 1 + + + + + + + Conops + + Version 3.0 + + 101 + + + ConOps), Version 3.0," p. 101. + + + + + Simulation Evaluations of an Autonomous Urban Air Mobility Network Management and Separation Service + + ChristabelleBosson + + + ToddALauderdale + + 10.2514/6.2018-3365 + + + 2018 Aviation Technology, Integration, and Operations Conference + + American Institute of Aeronautics and Astronautics + 2018 + 3365 + + + Bosson, Christabelle, and Todd A. Lauderdale. "Simulation evaluations of an autonomous urban air mobility network management and separation service." In 2018 Aviation Technology, Integration, and Operations Conference, p. 3365. 2018. + + + + + Develeopment of a Highfidelity simulation environment for shadow-mode assessments for air traffic concepts + + JERobinson + + + ALee + + + CFLai + + + + Royal Aeronautical Society: Modeling and Simulation in Air Traffic Management Conference + London, UK + + 2017. November 2017 + + + Robinson, J.E.; Lee, A. & Lai, C.F. (2017). "Develeopment of a High- fidelity simulation environment for shadow-mode assessments for air traffic concepts." Royal Aeronautical Society: Modeling and Simulation in Air Traffic Management Conference, London, UK, November 2017 + + + + + + diff --git a/file768.txt b/file768.txt new file mode 100644 index 0000000000000000000000000000000000000000..82520a9845e2ed53d0d1447e57e80d4e5d8dd406 --- /dev/null +++ b/file768.txt @@ -0,0 +1,538 @@ + + + + +INTRODUCTIONManaging terminal area traffic is challenging due to the density of traffic and complexity of trajectories and separation standards.Conflict Alert is a short time-horizon conflict detection tool currently in operational use in both the en route and terminal area in the U.S. National Airspace, but it is often inhibited or desensitized in the terminal area because it generates a high number of false alerts.Similarly, The Short Term Conflict Alert (STCA) in the Belgian Military airspace, Semmerzake (see Appendix D. of [1]) has rendered STCA ineffective due to high number of nuisance alerts.These tools help maintain safety in the current system, sometimes at the expense of capacity, because controller workload is often considered as a limitation to capacity [2].Conflict Alert uses only dead reckoning to determine when aircraft are in dangerous proximity to each other as compared to alerting for losses of separation.Tang et al. [3] augmented the legacy dead reckoning approach with flight trajectory intent information to create a short time-horizon tool called Terminalarea Tactical Separation Assured Flight Environment (T-TSAFE).The tool was developed to address the inadequacies of Conflict Alert (CA), which is used in the current air traffic control milieu.A comparison of T-TSAFE against a model of Conflict Alert, in a fast-time environment, found that T-TSAFE reduced the false alert rate from 20 per hour to 2 per hour when altitude intent was available and provided an average alert leadtime of 38 seconds [3].There are several conflict detection tools in the current National Airspace such as Conflict Alert (CA), Minimum Safe Altitude Warning (MSAW) system and Automated Terminal Proximity Alert (ATPA) in the final approach that do not communicate with each other.It would be ideal to have a single tool throughout the terminal airspace including the final approach.This research aims to test the T-TSAFE algorithm for depicting alerts in the data block and providing for a visual graphic for showing compression errors in the final approach.The objective of the current work was to test the T-TSAFE algorithm driving the visual graphic similar to ATPA in the final approach phase, and also vary the presence or absence of aircraft in the traffic flying under visual flight rules.ATPA shows the final approach controller a visual graphic (cones) for separating two aircraft flying straight in on final, and provides warnings when separation is predicted to be lost.The paper describes previous research in the field followed by a section that compared the T-TSAFE cones with the ATPA cones.It continues to delineate the experimental conditions and the relevant results are followed by discussion and conclusions. +II. BACKGROUNDThe Federal Aviation Administration (FAA) has forecasted an increase in air traffic demand that may see traffic more than double by the year 2025 [4] [5].Increases in air traffic will burden the air traffic management system, and higher levels of efficiency will be required.Maintaining current levels of safety will be more difficult in a more constrained and crowded terminal airspace.Thus, automation is proposed to aid the terminal area controllers with the task of assuring separation.Terminal airspace has proven to be difficult for tactical conflict detection automation.The factors that contribute to this difficulty include dense traffic, frequent large turns made by aircraft, imprecise flight plans, a complex set of separation standards and the fact that the aircraft operate close to the minimum separation standards leading to compression errors (horizontal separation violation) on approaches [3].In the current-day environment, CA, a legacy system, was shown to be inaccurate at times, and have a high rate of false alerts [6].An analysis of Conflict Alert showed that in the terminal area, controllers respond to alerts 56% of the time [7], which suggests a high false or high nuisance alert rate.Eurocontrol has been developing guidelines for a Short Term Conflict Alert (STCA) [1].The guidelines define environment data and parameters that should be used for conflict detection such as type of flight, wake category, Reduced Vertical Separation Minima (RVSM) status, cleared or block flight levels, and manually entered flight levels.The STCA guidelines propose using the linear prediction filter (dead reckoning) with a look-ahead time along with cleared flight levels entered by the controller, when available.The new algorithm for tactical conflict detection (T-TSAFE) developed by Tang et al. [3] aims to address the inadequacies of Conflict Alert in terminal airspace and incorporates some of the recommendations made by Friedman-Berg et al. [9], e.g., using a single analytic trajectory that takes into account both flight intent information and the current state of the aircraft.It also follows several of the Eurocontrol"s STCA guidelines such as taking into consideration different route structures, using flight intent information from the airspace definitions, speed restrictions and Area Navigation (RNAV) departure routes, segments of nominal TRACON routes and manually entered altitude clearances by controllers.Tang et al. [3] compared the T-TSAFE algorithm with a model of the Conflict Alert algorithm using recorded data from Dallas/Fort Worth TRACON that included 70 operational errors between January 2007 and April 2009.An analysis of fast-time simulation data showed that T-TSAFE would have prevented most of these operational errors, and that T-TSAFE also yielded a false alert rate of 2 per hour with 38 seconds of lead alert time, giving the controller more time to address conflict situations before they became critical.This finding addresses another Eurocontrol STCA guideline, which is to provide an alert with ample time to conflict so that the controller has enough time to de-conflict the two aircraft.In addition, when the algorithm has information about where and which aircraft will level off, fast time analyses showed further significant reductions in false alerts.The potential benefit from additional altitude intent information was the rationale for asking controllers to enter some commanded altitudes in the current investigation.This is similar to Eurocontrol"s guidelines for Cleared Flight Levels (CFL) for the STCA [1], where controllers are expected to enter the assigned altitudes manually.Subsequent to T-TSAFE"s fast-time study [3], several HITL studies also tested T-TSAFE in the terminal airspace [8][ [9].Results provide additional evidence for T-TSAFE"s efficacy, and the possible usefulness of altitude entries made by the controller.The initial HITL study [9] examined controllers" usage of the ATPA cones, which were used to automatically depict minimum separation between the aircraft on final approach, and the next HITL study [8] compared cones driven by the ATPA algorithm to T-TSAFE alerts in the data block (Fig. 1).Either the ATPA cones were shown to provide warnings or the T-TSAFE alerts were shown in the data block of the aircraft on final approach.The Federal Aviation Administration (FAA) recently fielded ATPA during the final approach phase of flight [10].The tool provides controllers with visual warnings if the minimum separation criteria is being exceeded or has the potential for being exceeded.The controllers give commands to pilots to make necessary maneuvers.ATPA may also allow the controllers to achieve better arrival rates by maintaining precise separation between aircraft.The tool shows the controller a cone, whose narrow end is placed on the aircraft icon and its length is based on the required separation between the two aircraft flying in-line (Fig. 3).Currently there are several conflict detection tools in the terminal area such as CA, MSAW, ATPA and each of these tools acts independently.Allendoerfer & Friedman-Berg (2012) discuss that these tools are not integrated, so they do not exchange inputs or outputs, and sometimes provide contradictory information to the user [11].They also found that tools developed in a non-integrative manner interfere with each other and require additional effort for controllers to manage the system.T-TSAFE was extended to include cones on final approach to integrate some of these tools.T-TSAFE attempts to improve upon, and bridge any gaps or missing functions in CA and ATPA. +III. EXPERIMENTAL TOOL-T-TSAFEThe current investigation examined using the T-TSAFE algorithm to drive ATPA-like cones and compared them with the original implementation of the ATPA cones.The objective of the study was to examine final approach operations using either ATPA cones or T-TSAFE cones under different operating conditions.The main tool used in the experiment was T-TSAFE.The alerts were shown in the data block and also as cones.There were different levels of alerts and they are all described in this section.T-TSAFE uses 1000 feet minimum vertical separation, wake turbulence lateral separation standards, and a look-ahead time of 120 sec to calculate conflicting trajectories.T-TSAFE alerts the controllers to a conflict by placing the number of seconds to predicted Loss of Separation (LoS) at the end of the first line in the data block, and the call sign of the conflicting aircraft in the third line of the data block, both in yellow or red (Fig. 1).There are two levels of alerts annunciated in the data block-yellow alerts when the time to predicted loss of separation is greater than 45 sec, and red alerts when the time to LoS is below 45 sec.If more than one other aircraft is involved in a conflict, the third line shows the call sign of the aircraft closest to a LoS.The controller can also roll the cursor over any aircraft showing a conflict, causing the data blocks of all other conflicting aircraft on the display to turn yellow for five seconds.The controllers also entered assigned altitude to the system and were provided resolution advisories that were either altitude or speed advisories.Both Speed and Altitude resolution advisories are shown in magenta color to the controller, in the second line of the data block.The assigned altitude entered by the controllers via keyboard is shown in green color in Fig. 2. The controllers were not required to make any entries for speed commands they issued to the controllers.As soon as the T-TSAFE algorithm receives altitude intent from the controller"s input, it no longer detects the conflict and removes it from the data block, thus decreasing false alerts.The green assigned altitude stays until the aircraft ascends or descends 300 ft above or below the assigned altitude.The current research compared ATPA cones (Fig. 3) with T-TSAFE cones (Fig. 4) in the final approach area.The T-TSAFE cones have the same visual graphic as the ATPA cones, but they also have added time to LoS in the data block.Other differences between the ATPA and T-TSAFE cones are described in Table I.As mentioned in Table I, ATPA cones appear on the aircraft only after the aircraft are established on the localizer, and the cones provide alerts in the form of yellow and orange cones for compression errors only.The warning yellow and orange ATPA cones did not also display T-TSAFE alerts to avoid confusion and clutter.This phase required controllers to enter the level-off altitudes (issued to the pilots as verbal commands) via keyboard command for the purpose of reducing the number of false alerts.These altitude entries had no effect on the ATPA cones.The final approach controller made keyboard entries for aircraft that were put under visual approach to runway 24R in the Mixed operating conditions.They pushed the keys "4R" on the keyboard to assign the aircraft to a visual approach on 24R.The scratch pad entries in the data block can distinguish between an aircraft on ILS approach and visual approach.The aircraft on ILS approach to 24R shows "I4R" and to 25L shows "I5L".The aircraft on visual approach to 24R shows "24R" (Fig. 5) and to 25L shows "25L" in the scratch pad of the data block.Another new tool explored in the study was controller lookahead time.The controllers were provided a tool that allowed them to set the look-ahead time on the display to control the alerts they wanted to see on their scopes.They were allowed to select from a drop-down menu -45, 60, 90 or 120 sec as their look-ahead time.For example, a controller could select 60 sec as the look-ahead time, and all alerts within the 60 sec of predicted time to conflict were shown on the display.This allowed controllers to control how early and how many conflicts they chose to see on their scope. +ATPA Cones T-TSAFE cones +Cones show compression error +IV. METHODOLOGYThe experimental approach was a Human-In-The-Loop (HITL) evaluation of the T-TSAFE tools under current-day operational conditions. +A. Experiment Matrix.This study phase tested four conditions: T-TSAFE cones under Instrument Flight Rules (IFR) and Mixed operating conditions, and ATPA Cones under IFR and Mixed operating conditions.Four traffic scenarios were exercised with each of the conditions for a total of 16 runs.The two variables being manipulated are cone type-T-TSAFE and ATPA, and operations type-IFR and Mixed. +B. Traffic ScenariosThe study simulated five arrival streams to and one departure stream from Los Angeles International Airport (LAX) using current airspace and procedures within the TRACON.The scenarios were designed to create situations that would result in a LoS between aircraft unless a controller intervened.Occasionally, the controllers were able to successfully avoid conflicts for extended periods due to early interventions.It was therefore necessary to add conflicts to the scenarios using observers collaborating with pseudo pilots.Each scenario involved heavy current-day traffic, with all LAX traffic under Instrument Landing System (ILS) simultaneous approaches on runways 24R and 25L for Conditions A and B. In the Mixed conditions (Condition C and D), some aircraft approaching runway 24R managed by the Stadium controller operated under visual flight rules, while the rest of the traffic to runway 24R and all approaches to runway 25L operated under Instrument Flight Rules (IFR). +C. Experiment ProceduresThe study was conducted over a two-week period with two teams of controllers, each participating for a week.Eight recently retired controllers served as participants, each controlling simulated traffic in the Southern California Terminal Radar Approach Control (TRACON) for approximately eight hours total.Each controller team consisted of four controllers that had retired less than two years ago from Southern California TRACON.Both controller teams were briefed on the T-TSAFE concept, the T-TSAFE interface, and the conditions of the study.During each week, the controller team completed 16 runs, four runs in each of the four different conditions, rotating through four different traffic scenarios.Controllers also rotated between sector positions after each run.Pseudo-pilots flew all the aircraft in the scenarios.The controllers worked the East Feeder and Zuma, feeder sectors, and Downe and Stadium, approach sectors in the Southern California TRACON, rotating positions after each run.All controllers completed questionnaires after every run and took part in a debrief session at the end of the study.Both conflict detection performance and controller subjective feedback were collected and analyzed, to characterize the performance of the T-TSAFE prototype. +V. RESULTS & DISCUSSIONBoth the digital and questionnaire data were collected and analyzed for the two independent variables: Operating conditions (IFR and Mixed), and Cone types (ATPA and T-TSAFE).Only the Stadium controller experienced the Mixed condition, therefore, the analysis of the Mixed condition pertains to the Stadium controller only.This is compared with the IFR condition analysis pertaining to all four controller positions.Also, the cones were displayed to only the final approach positions-Stadium and Downe, thus their data are considered in the data analysis comparing ATPA and T-TSAFE under IFR operating condition only. +A. Total Number of Alerts Generated & DisplayedThe total number of alerts generated by the T-TSAFE algorithm versus the alerts that were displayed is shown in Fig. 6 for IFR and the Mixed operating conditions.The total number of alerts generated by T-TSAFE is based on a lookahead time set inside the T-TSAFE algorithm whereas the alerts displayed in the data tag to the controller are managed by several factors, some of which are controlled of the controller, such as aircraft under visual approach, controller look-ahead time, and presence of cones.The alerts displayed to the controller impacts the controller"s workload therefore it is compared with the Total number of alerts generated.As depicted in Fig. 6, T-TSAFE generated significantly larger number of alerts under the Mixed condition (average of 17) than IFR condition (average of 10) (F (1,126)=15.97,p<0.001).The higher number of alerts generated for the Mixed operating condition is due to the fact that aircraft under visual separation come very close to other aircraft because they have no standard separation requirement.It is interesting to note that the total number of T-TSAFE alerts displayed is considerably smaller than the number of alerts generated.On average 8 alerts were displayed during the course of a 40 minute run on each controller station.The reason for the low number of alerts being displayed is because alerts are suppressed when aircraft are on visual approach, have ATPA cone or are outside a controller"s look-ahead time (discussed in the next subsection).All these factors contribute to the low number of alerts displayed on the aircraft. +B. Controller Look-ahead TimeThe controller manipulated look-ahead time was a new concept in the study.The look-ahead time inside the T-TSAFE algorithm was set at 120 sec.The controller was allowed to manipulate the "controller look-ahead," as it impacted the display which allowed the controller to decide how far in advance they wanted to display the alerts.Controller participants could, for example, choose 45 sec as the lookahead, for which they would see alerts with time to predicted LoS below 45 sec.However, the participants often confused this to mean that they would get every alert at 45 sec to conflict.This means they will have to be trained well to accurately use the controller look-ahead time.The different analyses show that controllers prefer 60 sec as the look-ahead time for getting alerts in their respective scopes.Any alerts generated by T-TSAFE with time to predicted LoS greater than the controller look-ahead time were suppressed by the system and not displayed to the controller.There were no significant differences found on a chi square test, where the expected frequency of a certain look-ahead time was compared with the actual frequency of that lookahead time between IFR and the Mixed conditions.There was a similar tendency for the controllers to select 60 sec most frequently irrespective of the type of cone -ATPA or T-TSAFE (Fig. 8).The look-ahead time can interact with the Time-to-Predicted-LoS provided by the algorithm.Similar to the T-TSAFE alerts, T-TSAFE cones were also impacted by controller"s look-ahead time as controllers did not see any warnings for compression errors on T-TSAFE cones unless the time-to predicted-LoS was below the controller"s look-ahead time.In the case of ATPA cones, since the warning alerts on the cones are based on hard-coded numbers i.e. time-to- predicted-LoS being 45 sec invoked a yellow cone, and it being 24 sec or less invoked an orange cone, controllers" lookahead time had no impact on the ATPA warning cones.Since controller look-ahead time does not impact the display of alerts on ATPA cones, this presents integration and training issues when ATPA cones and T-TSAFE alerts are integrated. +D. Keyboard EntriesControllers made altitude inputs, which are similar to cleared flight levels or assigned altitudes, into the system.The number of these altitude entries is compared across the conditions.It was found that controllers make about two altitude entries per run per controller station and there was no difference between the IFR and Mixed conditions (Fig. 11).The number of altitude entries may be small because the T-TSAFE algorithm derives flight intent information from the published nominal approaches that aircraft are expected to fly.The altitude entries made by the controller in one station are shared with the other controller station, providing shared situational awareness.Similarly, no difference between the altitude entries was observed between the ATPA and T-TSAFE cones condition (Fig. 12).The visual operations under different conditions are also analyzed for visual entries.These are the "4R" entries made via keyboard on aircraft on visual approach (Fig. 5).There is a significant difference in the entries made for visual approach under the Mixed condition as compared to IFR condition (Fig. 11) as indicated by the ANOVA (F (1,126) = 630.2,p<0.001).This result is expected since visual approach entries were allowed only under the Mixed condition.The difference in the cones did not impact the visual approach entries (Fig. 12). +E. Workload & Situational AwarenessParticipants completed the NASA TLX workload questionnaire [12] after every run.Data were collected on each of the six TLX workload measures.In addition, a seventh variable measuring overall workload combining all six of these measures was derived.The overall workload variable, also known as the "composite" measure, once derived, was then scaled down to match the 1-to-5 range for direct comparison with the other six measures (1=very low, 5=very high).Also, the "performance" measure was analyzed on an inverse scale, so a higher score indicates lower performance.Participants also completed an abbreviated version of the Situational Awareness Rating Technique (SART) scale at the end of each simulation run [13].The participants responded to questions on the demands the situation posed on them and their understanding of the situation offered to the user by the displays and procedures.They also responded to a question on attention capacity that refers to the user"s skills and attention needs.The situational awareness scales ranged from 1 to 7, i.e., from very low to very high situational awareness.The results on workload (Fig. 13) and Situational Awareness (SA) (Fig. 14) show that there were no significant differences yielded on the ANOVA when IFR conditions were compared with the Mixed.There were subtle trends observed, controllers reported somewhat decreased workload under mixed operating conditions.Visual operations usually decrease workload on the controller because the responsibility for separation is delegated to the flight deck.It is interesting to note in Fig. 6 that visual aircraft did not get many alerts displayed, as they do with the legacy system CA, which can potentially increase workload.In fact they received less than 1 alert per run, which was caused by general aviation aircraft under visual flight rules.All other alerts on visual aircraft were suppressed.Similar trends are seen with Situational Awareness (SA) data; trends show that the demand on attention is lower for the Mixed condition.There were no differences on the rest of SA variables.The workload reported by the controller participants was compared for the T-TSAFE cones and ATPA cones under the IFR operating conditions.These were considered because the only controller positions that used the cones were the final approach controllers.The Mixed operating conditions were not included in the analysis because between the two Final Approach controllers, only the Stadium controller was allowed to have visual separation between some aircraft.Analysis of Variance (ANOVA) did not yield any statistical difference between the ATPA and T-TSAFE under IFR operating conditions on the workload and SA measures.It is worth noting that despite the differences between the ATPA and T-TSAFE cones, the controllers did not perceive them differently in terms of workload or SA.ATPA is currently fielded in some of the US TRACONs.The similarities in the workload and SA reported for the ATPA cones and TTSAFE cones conditions will make bringing T-TSAFE cones in the air traffic controllers" work environment easier in future.The controllers were not confused about whether they were working with T-TSAFE cones or ATPA because they reported that the T-TSAFE cones appeared on aircraft much sooner than the ATPA cones.Thus there is anecdotal evidence that T-TSAFE cones did provide the controllers better situational awareness than the ATPA, but no subjective measures to support it. +F. Safety & ControlabilityThe controller participants were asked to rate operations for safety, controllability and acceptability on a scale of 1 to 5, where 1 referred to the lower end of the scale and 5, referred to the higher end of the scale.The controller participants on an average reported high levels on safety, controllability and acceptability of the operations.On the questions asked about the safety, controllability and acceptability of IFR operations as compared to the Mixed operations, no difference in the ratings were reported by the controller participants.Similarly, no significant differences were reported between the operations for ATPA cones vs. T-TSAFE cones for the Stadium & Downe controllers. +G. Comparison of Features between ATPA cones & T-TSAFE conesControllers were also asked rate their preference over the features of ATPA and T-TSAFE cones (Fig. 15).These features have been described in Table I.Forty percent of the controllers preferred the ATPA cones warning for showing compression error, whereas sixty percent of the controllers prefer that the T-TSAFE cones show LoS, which means that altitude separation between aircraft do not result in warning cones under T-TSAFE cones, even if lateral separation is less than the minimal requirement.The controller"s preference for T-TSAFE and ATPA on LoS vs. compression error was similar, resulting in a non-significant binomial test.Among all the other features, The T-TSAFE cone features such as their warnings, their early appearance, the impact altitude entries have on the cones, and alerts shown for merging aircraft were preferred by almost 100% of the participants over the ATPA features.All other features yielded significantly favorable results for T-TSAFE cones with significance achieved at 99% probability level.Only 10% of the controllers did not want the additional alerts provided by the T-TSAFE, whereas 90% of the participants preferred the additional alert information such as the time-to-predicted LoS shown on the T-TSAFE cones that was missing in the ATPA cones.This result on additional alert information achieved marginal significance with p<0.07 on a binomial test. +H. Complacency in AutomationA Complacency Potential Rating Scale was used to collect data on automation-induced complacency [14].Wiener [15] defined complacency as "a psychological state characterized by a low index of suspicion."Automation is often identified as a significant factor that induces complacency.Procedures, roles and responsibilities are also potential factors that induce complacency.According to Wickens [16], reliability in automation engenders excessive trust and over-reliance in pilots.Singh et.al. [14] identified four factors that may be related to over-trust or complacency in automation.These are confidence, reliance, trust, and safety in automation.Some examples of scale items that measure different complacency constructs are shown in Table III.The Complacency Potential Rating Scale was adapted and used to collect data for all controller positions across all conditions.The scale uses a 5-point Likert scale that ranges from "strongly disagree" to "strongly agree."Some of the questions in the rating scale were reversed to ensure reliability in the responses.The scale was adapted to ask questions about the IFR, Mixed, ATPA cones, and T-TSAFE cones.Fig. 16 shows the controller complacency ratings, comparing IFR operating conditions to The Mixed operating conditions and comparing the T-TSAFE cones with ATPA cones.Statistically significant ANOVA differences were found for several constructs of complacency and they have been outlined in Table IV.The results show that there are significant differences in the level of confidence, reliance, trust and safety for operations type.IFR operations have consistently higher levels of confidence, reliance, safety and trust than the Mixed operations.Similarly, there are significant differences in the level of confidence and trust for the T-TSAFE cones showing that participants were beginning to rely on the T-TSAFE cones much more than the ATPA cones.Previous research has examined automation complacency [17] [18] and the extent to which over-reliance on automation can lead to operational errors (e.g., in the case of occasional automation failures).This might suggest that very high or very low levels of automation complacency are not desirable, but rather, optimum levels of automation complacency would be somewhere between these two extremes.Thus neither mistrust nor over-reliance in the automation is desirable.The high scores on T-TSAFE indicate over reliance on the tool and warrants extra caution during deployment. +I. Usability FunctionsThe controller participants also rated on a scale of 1 to 5 their subjective responses towards several questions related to procedures, information requirements, and usability.Significant results were found on the ease of locating aircraft with potential LoS, awareness to potential LoS and timeliness of alerts.The controllers reported that it was easier to locate an aircraft with potential LoS with T-TSAFE cones (Fig. 17) as compared to ATPA cones (F(1,7)=5.727,p<0.05).This result makes sense because the T-TSAFE cones appear sooner than the ATPA cones.The controllers also had an easier time maintaining awareness of potential loss of separation than the ATPA cones (F (1,7)=7.0,p<0.05).This may be due to the fact that the T-TSAFE cones have warnings that have additional information such as time to predicted LoS shown in the data, block.The consistency between the alerts and the warning cones also makes for better alerts with T-TSAFE cones than ATPA cones.It was also found that T-TSAFE cones performed significantly better than the ATPA cones on timeliness of alerts (F (1,7)=7, p<0.05).Again the features of T-TSAFE such as cones appearing on the aircraft 30 sec before the aircraft is established on localizer, and providing warning for aircraft that are ahead of the schedule of the current aircraft whereas ATPA provides warnings with aircraft that are physically ahead of it, and not necessarily ahead in the schedule.Overall, T-TSAFE cones performed similar to or better than the ATPA cones.There were no significant differences in the ease of locating potential LoS between IFR and Mixed operations.Similarly, there are no significant differences between IFR and the Mixed operating conditions for the responses towards maintaining awareness of LoS or on the timeliness of alerts.Thus, the operating condition did not impact the nature of alerts or any functions related to conflict detection.Figure 17.Subjective responses on ease of locating aircraft with potential loss of separation +VI. CONCLUSIONSThe study investigated a tactical conflict detection tool (T-TSAFE) in the terminal area.This tool combines dead reckoning with flight intent where possible and has fulfilled many of the requirements set by Eurocontrol"s guidelines for a STCA.Initial and follow-up human-in-the-loop air traffic control simulation experiments were conducted, to test the human factors of this new conflict detection and resolution tool, T-TSAFE.Whereas the initial investigation verified T-TSAFE as an improvement over the legacy conflict detection system currently used in the field, it also revealed aspects of the T-TSAFE system that require further investigation [5].To address this need, a follow-up experiment was conducted, testing T-TSAFE under four experiment conditions.The first two conditions varied the operating conditions-IFR vs. the Mixed conditions where aircraft were allowed to have visual separation in the latter.The second and third conditions varied the software driving the cones -Automatic Terminal Proximity Alert (ATPA) and T-TSAFE.The cones are a graphical tool used for monitoring final approach and providing a warning for possible compression errors in the final approach phase of flight.The test bed for the investigation was simulated-Southern California TRACON.Workload and Situational differences were not found for either cone type or operating type.But trends showed that the Mixed operating condition was easier on the controller with respect to attentional capacity and workload, because for visual operations the controllers delegate responsibility for separation to the flight deck.The controller participants preferred to see alerts within 60 sec of conflict, even though they had the choice of 45 sec, 90 sec and 120 sec as controller look-ahead time.They were provided with two levels of alerts -red and yellow and used the color of the alert to prioritize their cognitive resources.Controller participants reported preferring most of the T-TSAFE cones" features to the ATPA cones" features.They particularly liked the fact that T-TSAFE cones appeared 30 sec before getting established on the localizer making it easy to locate an aircraft having potential LoS and to maintain awareness of the LoS.They found the alerts provided by T-TSAFE as very timely.There was also higher consistency between the alerts and cones in case of T-TSAFE cones, when compared to ATPA cones.For the subjective responses on complacency towards automation, T-TSAFE cones were being relied upon more than ATPA cones.The Mixed operating procedures are keeping the controllers in the optimum zone for complacency for automation.The comparison between IFR and Mixed operations for T-TSAFE has shown that Mixed operations do not increase the workload or decrease situational awareness, suggesting that the tool is not adding any unnecessary alerts in the Mixed condition.Thus, this investigation suggests that having one system (T-TSAFE) that shows the alerts in the data blocks and also drives the cones in the final approach will circumvent the need for integrating different systems.Previous results have shown T-TSAFE an improvement over the legacy system.The results of this study show similarities between the ATPA cones and T-TSAFE cones will make it easier to migrate the air traffic control system from ATPA cones to T-TSAFE cones.The study also shows that T-TSAFE alerts and ATPA cones do to increase undue workload or confusion, thus the transition of T-TSAFE alerts with ATPA in the final approach holds much potential.Figure 1 .1Figure 1.T-TSAFE Data Tag (yellow and red alerts) +Figure 2 .2Figure 2. T-TSAFE Altitude Resolution and Altitude Entries +Cones show Loss of Separation (LoS) Cones appear after established on localizer Cones appear 30 sec before being established on localizer Altitude Intent has no impact Altitude intent reduces false alerts Yellow warning cones when predicted time to LoS is 45 sec Yellow warning cones when predicted time to LoS is greater than 45 sec Orange warning cones when predicted time to LoS is 24 sec Red warning cones when predicted time to LoS is less than 45 sec T-TSAFE compression alerts are suppressed on aircraft with ATPA cones Only time to LoS is shown in data block on aircraft with T-TSAFE cones Conflict predictions between aircraft are made with the aircraft physically ahead Conflict predictions between aircraft are based on the schedule to the runway The T-TSAFE cones evaluate required separation, thus sufficient altitude separation prevents the T-TSAFE cones from showing yellow or red alert status.ATPA cones, on the other hand, turn yellow or orange if there is a compression error despite sufficient altitude separation between two aircraft.The cones for T-TSAFE appear 30 sec before aircraft are established on the localizer.T-TSAFE builds predictions based on scheduled arrival at the runway, allowing for the accurate display of compression alerts even when two aircraft are not yet physically in line.ATPA cones are built for the trailing aircraft only when it is physically behind another aircraft.At face value, the ATPA and T-TSAFE cones appear the same with the exception of additional time-to-LoS shown in the data tag for the T-TSAFE cones for compression errors (Fig. 3 & Fig.4).The color of the T-TSAFE warning cones matches the color of the T-TSAFE alerts in the data block (Fig.3).Other similarities between the ATPA cones and T-TSAFE cones include: The blue cone shows no LoS or compression error; the number displayed inside the cone is the distance between the leading aircraft and current aircraft; the size of the cone for both ATPA and T-TSAFE is based on wake requirements given the aircraft types for both the aircraft.The experiment examined T-TSAFE cones with ATPA cones under IFR and Mixed operating conditions.The Mixed operating conditions involved having part of the traffic on visual approach. +Figure 3 .3Figure 3. Automated Terminal Proximity Alert +Figure 4 .4Figure 4. T-TSAFE cones (Time to LOS shown in data tag) +Figure 5 .5Figure 5. Data tag showing visual entry to 24R +Figure 6 .6Figure 6.Total Number of alerts generated vs. displayed (IFR vs. Mixed) An ANOVA revealed a significant difference in total number of alerts displayed for the visual aircraft ( one on an average) versus none for IFR (F (1,126)=17.38,p<0.001).In general, most alerts on aircraft under visual separation were suppressed, except any alerts with general aviation aircraft flying under visual flight rules (VFR aircraft).Fig. 7 shows the Total Number of alerts generated versus displayed for T-TSAFE cones vs. ATPA cones condition for the two final controllers -Stadium & Downe.Both the ATPA and the T-TSAFE cone conditions are provided with different cones in the final approach, but T-TSAFE alerts show in the data tags for rest of the traffic in the airspace.It was found hat the cone condition did not impact the number of alerts generated by the T-TSAFE algorithm and no significant differences were yielded on ANOVA.As mentioned earlier, the number of T-TSAFE alerts generated for ATPA and T-TSAFE cone condition has the same T-TSAFE algorithm working in background and thus the number is quite similar (about 15 alerts per run).Fig. 7 also shows that the number of alerts displayed in the two conditions is somewhat similar.About 12 T-TSAFE alerts were shown in the ATPA condition and 15 alerts were shown in the data block of aircraft in T-TSAFE cones condition reaching marginal significance on ANOVA (F(1,44)=3.4,p<0.07).The reason for this difference lies in the fact that T-TSAFE alerts for compression errors are suppressed when ATPA cones are depicted on the aircraft, +Figure 7 .7Figure 7.Total Number of alerts generated vs. displayed (ATPA vs. T-TSAFE cone conditions) +Figure 8 .8Figure 8. Frequency of different Controller Look-ahead times ATPA cones vs. T-TSAFE cones (Downe & Stadium Controller)C.Duration of Displayed AlertsThe duration of the alerts displayed for the IFR conditions is compared with the Mixed Condition in Fig. 9.The duration of the alerts is longer in the IFR conditions (61 sec compared to 44 sec in the Mixed) because in the Mixed condition, only the alerts with VFR aircraft are shown.There is no standard separation required with VFR aircraft, but the algorithm uses a 1.5 nmi lateral and 500 ft vertical separation to generate alerts between IFR and VFR aircraft.The alert durations did yield significant differences between the IFR and the Mixed conditions (F(1,1201)=4.98,p<0.05).The Duration for the alerts displayed under the ATPA and T-TSAFE condition is shown in Fig. 10.The T-TSAFE alerts with ATPA are suppressed as soon as the cones appear, which explains the lower duration of the T-TSAFE alerts on the ATPA cones condition.There was marginal significance between the duration of T-TSAFE alerts shown in the ATPA cones and T-TSAFE cones conditions (F(1,650)=3.27,p<0.07).The duration of the T-TSAFE alerts on the ATPA cones is still 60 sec giving the controller enough time to deal with them. +Figure 9 .9Figure 9. Duration of Alerts IFR (All positions) vs. Mixed (Stadium only) +Figure 11 .11Figure 11.Altitude and Visual entries for IFR (all positions) vs. Mixed (Stadium) +Figure 12 .12Figure 12.Altitude and Visual entries for ATPA cones vs. T-TSAFE cones (Downe & Stadium) +Figure 13 .Figure 14 .1314Figure 13.Workload IFR (all positions) vs. Mixed (Stadium only) +Figure 15 .15Figure 15.Controller preference for features of T-TSAFE cones and ATPA +TABLE I .IDIFFERENCES BETWEEN ATPA & T-TSAFE CONES +TABLE IIII.EXAMPLES OF STATEMENTS USED TO MEASURECOMPLACENCY IN AUTOMATIONConfidenceT-TSAFE makes air traffic in the terminal environment saferunder IFR conditions.RelianceT-TSAFE cones have made the controller"s job easier.TrustT-TSAFE is more likely to be correct than manual conflictdetection under visual conditions.SafetyI feel safer having ATPA driven cones than relying on manualconflict detection. +TABLE III .IIISIGNIFICANT RESULTS ON VARIOUS CONSTRUCTS OFPOTENTIAL FOR COMPLACENCYConstructF statisticSignificance(df values)ConfidenceF(1,7) = 15.91 Significant for operation type at p<0.005ConfidenceF(1,7) = 4.82Marginally Significant for cone type atp<0.06RelianceF(1,7) = 5.90Significant for operation type at p<0.05TrustF(1,7) = 14.97 Significant for operation type at p<0.05TrustF(1,7) = 3.94Marginally Significant for cone type atp<0.08SafetyF(1,7) = 6.67Significant for operation type at p<0.05Figure 16. Potential for complacency compared across conditions. + + + + + + + + + + + An fMRI investigation of trust-related decision-making associated with the use of short term conflict alert (STCA) + + JimmyYZhong + 0000-0003-3879-278X + + 10.31224/2470 + + May19, 2009 + Open Engineering Inc + + + Eurocontrol Guidance Material for Short Term Conflict Alert. Released Issue, Edition no. 2, May19, 2009. + + + + + Estimation of European Airspace Capacity from a Model of Controller Workload + + ArnabMajumdar + + + WashingtonOchieng + + + JohnPolak + + 10.1017/s037346330200190x + + + Journal of Navigation + J. Navigation + 0373-4633 + 1469-7785 + + 55 + 3 + + 2002 + Cambridge University Press (CUP) + + + Majumdar, A., Ochieng, W. & Polak, J. (2002). Estimation of European Airspace Capacity from a Model of Controller Workload. The Journal of Navigation. 55. 381-403 + + + + + Tactical Conflict Detection in Terminal Airspace + + HuabinTang + + + JohnERobinson + + + DallasGDenery + + 10.2514/1.51898 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 34 + 2 + + 2011 + American Institute of Aeronautics and Astronautics (AIAA) + + + Tang, H., Robinson, J.E. & Denery, D.G. (2011). Tactical Conflict Detection in Terminal Airspace. Journal of Guidance, Control, and Dynamics. Vol 34, No. 2, 403-413. + + + + + The Next Generation of Air Supply Units + + JanvierEBeaver + + + KarlHering + + 10.1115/rtd2004-66002 + + + Joint Rail + + ASMEDC + 2004. December 12, 2004 + + + Joint Development and Planning Office (2004). The Next Generation Air Transport System Integration National Plan. December 12, 2004. + + + + + FAA Aviation Forecasts: Fiscal Years 1981-1992. Federal Aviation Administration, U.S. Department of Transportation, Washington, D.C. 20591. 1980. 69p + 10.1177/004728758102000159 + + + Journal of Travel Research + Journal of Travel Research + 0047-2875 + 1552-6763 + + 20 + 1 + + 2009 + SAGE Publications + Washington DC + + + Federal Aviation Administration (2009). FAA Aerospace Forecast Fiscal Years 2009-2025. Washington DC, 31. + + + + + Tactical Conflict Alerting Aid for Air Traffic Controllers + + RussellAPaielli + + + HeinzErzberger + + + DannyChiu + + + KarenRHeere + + 10.2514/1.36449 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 32 + 1 + + 2009. January-February 2009 + American Institute of Aeronautics and Astronautics (AIAA) + + + Paielli, R.A., Erzberger, H., Chiu, D. & Heere, K. (2009). Tactical Conflict Alerting Aid for Air Traffic Controllers. Journal of Guidance, Control, and Dynamics, Vol. 32, No. 1, January-February 2009. + + + + + Nuisance Alerts in Operational ATC Environments: Classification and Frequencies + + FerneFriedman-Berg + + + KennethAllendoerfer + + + ShantanuPai + + 10.1177/154193120805200123 + + + Proceedings of the Human Factors and Ergonomics Society Annual Meeting + Proceedings of the Human Factors and Ergonomics Society Annual Meeting + 2169-5067 + 1071-1813 + + 52 + 1 + + 2008 + SAGE Publications + + + Friedman-Berg, F., Allendoerfer, K. & Pai, S. (2008). Nuisance Alerts in Operational ATC Environments: Classification and Frequencies. Proceedings of the Human Factors and Ergonomics Society 52 nd Annual Meeting. + + + + + Phase-2 evaluation of a Tactical conflict detection tool in the Terminal area + + TKozon + + + AFarrahi + + + SVerma + + + Huabin Tang + + + DBallinger + + 10.1109/dasc.2012.6382337 + + + 2012 IEEE/AIAA 31st Digital Avionics Systems Conference (DASC) + Willamsburg, VA + + IEEE + 2012. October, 2012 + + + Kozon, T.; Verma, S.; Farrahi, A.; Tang, H. & Ballinger, D. (2012). Phase-2 Evalautaion of a Tactical Conflict Detection Tool in Terminal Area. 31 st Digital Avionics System Conference, Willamsburg, VA, October, 2012. + + + + + Initial Evaluation of a Conflict Detection Tool in the Terminal Area + + SVerma + + + HTang + + + TKozon + + + AFarrahi + + + DBallinger + + + + 4th International Conference on Applied Human Factors and Ergonomics + + 2012. July, 2012 + + + Verma, S., Tang, H., Kozon, T., Farrahi, A. & Ballinger, D. (2012). Initial Evaluation of a Conflict Detection Tool in the Terminal Area. 4th International Conference on Applied Human Factors and Ergonomics, July, 2012. + + + + + Making the Case for NextGen Concepts—Using the FAA Benefit-Cost Analysis Methodology + + AmitLagu + + + SuzanneAkkoush + + + AnnieCheng + + + BillDunlay + + 10.2514/6.2013-4226 + + + + 2013 Aviation Technology, Integration, and Operations Conference + Washington, D.C. + + American Institute of Aeronautics and Astronautics + 2011 + + + Federal Aviation Administration (2011). FAA"s NextGen Implementation Plan. NextGen Integration and Implementation Office. Washington, D.C., www.faa.gov/nextgen. + + + + + Methods & Tools + + KRAllendoerfer + + + FJFriedman-Berg + + 10.1201/b11933-50 + + + Contemporary Ergonomics and Human Factors 2012 + San Francisco, CA + + CRC Press + 2012. July 2012 + + + Allendoerfer, K. R. & Friedman-Berg, F.J. (2012). When Tools Collide: Human Factors Effects of Using Multiple ATC Tools in Combination. Applied Human Factors & Ergonomics Conference, San Francisco, CA, July 2012. + + + + + Development of NASA-TLX (Task Load Index): Results of Empirical and Theoretical Research + + SandraGHart + + + LowellEStaveland + + 10.1016/s0166-4115(08)62386-9 + + + Advances in Psychology + + PHancock + + + NMeshkati + + Amsterdam, The Netherlands + + Elsevier + 1988 + + + + Hart, S. & Staveland, L. (1988). Development of a Multi-Dimensional Workload Rating Scale: Results of Empirical and Theoretical Research, In P. Hancock & N. Meshkati (Eds.), Human Mental Workload, Amsterdam, The Netherlands, Elsevier, 139-183. + + + + + Situation Awareness as a Predictor of Performance for En Route Air Traffic Controllers + + FrancisTDurso + + + CarlaAHackworth + + + ToddRTruitt + + + JerryCrutchfield + + + DankoNikolic + + + CarolAManning + + 10.2514/atcq.6.1.1 + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 6 + 1 + + 1998 + American Institute of Aeronautics and Astronautics (AIAA) + + + Durso, F.T., Hackworth, C.A., Truitt, T.R., Crutchfield, J., Nikolic, D. & Manning, C.A. (1998). Situation Awareness as a Predictor of Performance for En Route Air Traffic Controllers, Air Traffic Control Quarterly, 6(1), pp 1-20. + + + + + Automation- Induced "Complacency": Development of the Complacency-Potential Rating Scale + + IndramaniLSingh + + + RobertMolloy + + + RajaParasuraman + + 10.1207/s15327108ijap0302_2 + + + The International Journal of Aviation Psychology + The International Journal of Aviation Psychology + 1050-8414 + 1532-7108 + + 3 + 2 + + 1993 + Informa UK Limited + + + Singh, I., Molloy, R. and Parasuraman, R. (1993). Automation-Induced "Complacency": Development of the Complacency-Potential Rating Scale. The International Journal of Aviation Psychology, 3(2), 111-122. + + + + + Complacency: Is the term useful for air safety? + + ELWiener + + + + Proceedings of the 26 th Corporate Aviation Safety Seminar + the 26 th Corporate Aviation Safety SeminarDenver, CO + + Flight Safety Foundation, Inc + 1981 + + + + Wiener, E.L. (1981). Complacency: Is the term useful for air safety? In Proceedings of the 26 th Corporate Aviation Safety Seminar (pp.116- 125). Denver, CO: Flight Safety Foundation, Inc. + + + + + Engineering Psychology and Human Performance + + ChristopherDWickens + + + JustinGHollands + + + SimonBanbury + + + RajaParasuraman + + 10.4324/9781315665177 + + 1992 + Psychology Press + New York + + + Wickens, C.D. (1992). Engineering psychology and human performance. New York: Harper Collins. + + + + + Examination of Automation-Induced Complacency and Individual Difference Variates. NASA Technical Memorandum + + LPrinzel + + + HDevries + + + FFreeman + + + PMikulka + + TM-2001-211413 + + 2001 + + + Prinzel, L., DeVries, H., Freeman, F. & Mikulka, P. (2001). Examination of Automation-Induced Complacency and Individual Difference Variates. NASA Technical Memorandum TM-2001-211413. + + + + + Automation- Induced "Complacency": Development of the Complacency-Potential Rating Scale + + IndramaniLSingh + + + RobertMolloy + + + RajaParasuraman + + 10.1207/s15327108ijap0302_2 + + + The International Journal of Aviation Psychology + The International Journal of Aviation Psychology + 1050-8414 + 1532-7108 + + 3 + 2 + + 2009 + Informa UK Limited + + + Singh, A., Tiwari, T. & Singh, I. (2009). Effects of Automation Reliability and Training on Automation-Induced Complacency and Perceived Mental Workload. Journal of the Indian Academy of Applied Psycholgy. Vol 35, 9-22. + + + + + + diff --git a/file769.txt b/file769.txt new file mode 100644 index 0000000000000000000000000000000000000000..a70665e1dd76dfa8964db9488750859b91cde2b4 --- /dev/null +++ b/file769.txt @@ -0,0 +1,309 @@ + + + + +INTRODUCTIONManaging terminal area traffic is challenging due to the density of traffic and complexity of trajectories and separation standards.Conflict Alert is a short time horizon conflict detection tool currently in operational use in both the en route and terminal area, but it is often inhibited or desensitized in the terminal area because it generates a high number of false alerts.Conflict Alert uses only dead reckoning to determine when aircraft are in dangerous proximity to each other.Safety is maintained in the current system but at the expense of capacity, since controller workload is often considered as a limitation to capacity (Majumdar et al., 2002).The objective of the current work is to develop a reliable and effective conflict alerting system.Tang et al. (2011) augmented the legacy dead reckoning approach with flight trajectory intent information to create a short time-horizon tool called Terminal Tactical Separation Assurance Flight Environment (T-TSAFE).The tool was developed to address the inadequacies of the operational Conflict Alert (CA).A comparison of T-TSAFE against a model of Conflict Alert found that T-TSAFE reduced the false alert rate and provided an average alert lead time of 38 seconds (Tang et al., 2011).Previous Human-In-The-Loop (HITL) research on the predecessor TSAFE tool was conducted in the en route phase of flight (Homola et al., 2009).The current experiment tests the tool for the first time in the terminal environment with current day operations and technology.The experiment varied levels of altitude intent made available to the algorithm and assessed conflict detection performance as well as controller subjective feedback.Eight recently retired controllers served as participants, each controlling simulated traffic in the Southern California TRACON over approximately eight hours.Objective and subjective data are presented that characterize the performance of the Terminal TSAFE prototype. +BACKGROUNDThe Federal Aviation Administration (FAA) has forecasted an increase in air traffic demand that may see traffic more than double by the year 2025 (JPDO, 2004) (FAA, 2009).Increases in air traffic will burden the air traffic management system, and higher levels of safety and efficiency will be required.Maintaining current levels of safety will be more difficult in a more constrained and crowded terminal air space.Thus, automation is proposed to aid the terminal area controllers with the task of assuring separation.Terminal airspace has proven to be difficult for tactical conflict detection automation.The factors that contribute to this difficulty include dense traffic, frequent large turns made by aircraft, imprecise flight plans, a complex set of separation standards and the fact that the aircraft operate close to the minimum separation standards due to compression errors on approaches (Tang et al., 2011).In the current-day environment, Conflict Alert, a legacy system, was shown to be inadequate (Paeilli & Erzberger, 2007).Analysis of Conflict Alert by Friedman-Berg et al. (2008) shows that in the terminal area, controllers respond to CA alerts 56% of the time.Also, they analyzed the duration of the alerts, defined as the time between alert onset and the time when the conflict is resolved.Analyses of Conflict Alert durations showed that 36% of the alerts lasted less than 15 seconds.The short duration of CA alerts does not provide adequate time for the controller to respond and have any effect on the situation.Friedman-Berg et al. (2008) estimated that about 81% of all alerts in the terminal area are false, or nuisance, alerts.If automation can take into account aircraft intent, nuisance alerts are expected to decrease in number, improving the controller's trust in the credibility of the alert.En-route conflict detection automation incorporates some of this intent functionality already, incorporating information about routes and interim altitudes when calculating conflicts, but current terminal area conflict detection automation either lacks this functionality or controllers do not commonly use it.Most of the research on the parent TSAFE tactical tool has focused on en-route airspace.En-route prototypes have been developed and HITL studies were performed at NASA Ames.These studies compared conflict detection and resolution done manually by the controller to conflict detection and resolution performed by TSAFE.The concept of operations in these studies required new technologies such as Automatic Dependent Surveillance-Broadcast (ADS-B) and a Required Navigation Performance (RNP) of 1.0, making it a mid-to far-term concept.All the aircraft in the test airspace were capable of performing trajectorybased operations via datalink communications.For the flights that maintained their trajectory, TSAFE was responsible for the detection and resolution of strategic conflicts.Trajectory changes to resolve conflicts were uplinked directly to the aircraft without the controller's involvement.Overall results showed that using TSAFE resulted in better resolution of tactical conflicts and fewer separation violations than without TSAFE.These studies investigated TSAFE in the en-route airspace with far-term concepts of operations and technologies.Adapting TSAFE to the terminal environment and using it with a near term or current day concept of operations became the focus of this study.A new algorithm for tactical conflict detection (T-TSAFE) developed by Tang et al. (2011) aims to address the inadequacies of Conflict Alert and incorporates some of the recommendations made by Friedman-Berg et al. (2008).The T-TSAFE algorithm uses a single analytic trajectory that takes into account both flight intent information and the current state of the aircraft.In addition to the flight plan, it takes into consideration Area Navigation (RNAV) departure routes, segments of nominal TRACON routes, speed restrictions, and altitude clearances inferred from the recorded track data.Tang et al. (2011) compared the T-TSAFE algorithm with a model of the Conflict Alert algorithm using recorded data from Dallas/Fort Worth TRACON that included 70 operational errors.An analysis of fast-time simulation data showed that T-TSAFE yielded a false alert rate of 2 per hour with 38 seconds of lead alert time, giving the controller adequate time to address conflict situations before they became critical.When the algorithm has information about where and which aircraft will level off, fast-time analyses showed further significant reductions in false alerts.The potential benefit from additional altitude intent information was the rationale for asking controllers to enter some commanded altitudes in the current investigation.Although Terminal-TSAFE has undergone considerable fast-time testing, this study was the first to test terminal TSAFE with humans-in-the-loop, and also to evaluate the algorithm under current-day operational conditions.This study also investigates the feasibility of having controllers enter the level-off altitudes, into the automation, which they also gave the pilots verbally for the purpose of reducing the number of false alerts. +METHODSThis study compared T-TSAFE's alerting performance with an emulation of Conflict Alert in a human-in-the-loop test.As noted earlier, TRACON controllers currently do not provide input into Conflict Alert with commanded altitudes, as is current practice for en-route controllers.Prior studies (Tang et. al, 2011) have shown the rate of false alerts drops dramatically if altitude intent information can be provided to conflict detection automation.This is because, with altitude information, the algorithm knows where aircraft will level off and will not predict conflicts based on a presumption of continued descent.This study included one condition that required the controllers to enter the altitudes at which they verbally commanded aircraft to level off, and another condition where the commanded altitude information was provided via short-term intent broadcast over ADS-B from the flight deck. +Experiment matrixThe three conditions in the experiment described in this paper are shown in Table 1.The experiment included a baseline condition where Conflict Alert was emulated, but those objective results are not described in detail in this paper.The Conflict Alert emulation used in this study was simply represented with only 1000 ft vertical separation and 3 nmi lateral separation requirements between aircraft, and it did not provide meaningful alert data.However, the subjective results comparing Conflict Alert and T-TSAFE still have value because controllers related their experience with the fielded version of Conflict Alert to the T-TSAFE tool.Thus, Conflict Alert data are only discussed in the subjective results.The three T-TSAFE conditions described in the results were: T-TSAFE with no additional altitude information (Condition A); T-TSAFE with controllers entering verbally commanded level-off altitude, if the altitude was for conflict resolution (Condition B), and T-TSAFE where the controller assigned an altitude for conflict resolution by voice and an ADS-B broadcast returned the altitude entered by the pilot to T-TSAFE (Condition C).In conditions B and C, the assigned altitude was shown in green in the second line of the data tag with 'A' prefixed to the three-digit altitude.Four traffic scenarios were exercised with each of the conditions for a total of 12 runs. +Air Traffic Control Tools and ProceduresThe study simulated five arrival streams and one departure stream into Los Angeles International Airport (LAX) using current airspace and procedures within the TRACON.The scenarios were designed to create many situations that would result in a loss of separation between aircraft unless a controller intervened.Controllers in the study occasionally were able to successfully avoid conflicts for extended periods due to early intervention.It was therefore necessary to add conflicts to the scenarios using observers collaborating with pseudo pilots.Each scenario contained the traffic equivalent of heavy current-day traffic, with all LAX traffic under Instrument Landing System (ILS) simultaneous rules.The controller's goal was to avoid losses of separation.The controllers worked the East Feeder and Zuma feeder sectors, and Downe and Stadium approach sectors in the Southern California TRACON, rotating positions after each run. +Figure 1. T-TSAFE data tagsT-TSAFE used 1000 feet minimum vertical separation, wake turbulence lateral separation standards, and a look-ahead time of 120 seconds to calculate conflicting trajectories.T-TSAFE predicted wake encounters as well as physical losses of separation, and factored in flight plan information and standard procedures, as well as using dead reckoning to predict conflicts.T-TSAFE alerts the controllers to a conflict by placing the number of seconds to predicted Loss of Separation (LoS) on the end of the first line in the data block, and the call sign of the conflicting aircraft in the third line of the data block, both in red.If more than one other aircraft is involved in a conflict, the third line shows the call sign of the aircraft closest to a loss of separation (Figure 1).The controller could also roll the cursor over any aircraft showing a conflict, and the data blocks of all other conflicting aircraft on the display would turn yellow for 5 seconds, e.g., UAL842 and MXA902 as other conflicting aircraft with the conflict shown in Figure 1. +Experiment ProceduresThe study was conducted over a two-week period with two teams of controllers, each participating for a week.Each controller team consisted of four controllers that had retired recently from Southern California TRACON.Both controller teams were briefed on the T-TSAFE concept, the T-TSAFE interface, and the conditions of the study.During each week, the controller team completed twelve runs, four runs each of the three different conditions, rotating through four different traffic scenarios.Controllers rotated between sector positions after each run.Pseudo-pilots flew all the aircraft in the scenarios.All controllers completed questionnaires after every run and took part in a debrief session at the end of the study. +RESULTS & DISCUSSIONThe metrics discussed in the paper include the following: total number of alerts, total number of false alerts, workload and usability of the T-TSAFE interface. +Total Number of T-TSAFE Alerts and Total Number of False AlertsFigure 2 shows that the total number of alerts is similar between the three conditions.Each controller dealt with about 14 conflicts on average every run.There was no statistical difference in the number of alerts per condition.The false alerts were defined as a condition where an alert was provided, but the two aircraft did not lose separation, even though the controller did not intervene from 60 sec before the alert through the predicted loss of separation time.As shown in Figure 2, the number of false alerts was about 1 false alert per run, which were not statistically different between the different conditions.Contrary to the expectation, the altitude entries did not reduce the number of false alerts.This was possibly because controllers were able to enter commanded altitudes for conflict resolution only once or twice each run.It is likely that there was insufficient data to evaluate the impact of altitude entries on the number of false alerts. +Duration of the Alert and Response Time to AlertThe duration of alert was defined as the time between the alert onset and the time when the conflict was resolved, which is dependent on several factors.One factor was the look-ahead time that the T-TSAFE algorithm used to predict conflicts.The average duration of an alert across different conditions was about 110 sec (Figure 3).The mean alert durations were slightly shorter in the two conditions where altitude clearances were provided via keyboard or ADS-B, as compared to the condition with no altitude entries.However, these differences did not reach statistical significance.The response time to an alert was measured as the difference between the onset of the alert and the time the pseudo pilot responded to the controller's commands.Figure 3 shows that the time to respond to an alert was 22 sec on average.It seems the duration of the alerts was more than adequate for the controller to see the conflict, determine a resolution, inform the pilots, and for the pilots to initiate the commands issued by the controllers.The controllers also mentioned that they used the "time to predicted loss of separation" provided in the data block to prioritize their tasks.The feedback provided by the controllers made it clear that in the TRACON environment they often address other high-priority emerging situations first and act to resolve a conflict in the last 25-30 sec prior to the predicted loss of separation. +Subjective data -WorkloadParticipants completed the NASA TLX workload questionnaire (Hart et al., 1988) after every run.As mentioned earlier, the subjective data include a baseline condition where the Conflict Alert model was simulated.Data were collected on each of the six TLX workload measures, and a variable measuring overall workload combining all six of these measures was derived, for a total of seven workload measures.The overall workload variable, also known as the "composite" measure, once derived, was then scaled down to match the 1-to-5 scale for direct comparison with the other six measures.Also, the "performance" measure was analyzed on an inverse scale, so a higher score would actually mean lower performance and a lower score is indicative of better performance.Results on all 7 of these measures, comparing the four experimental conditions, are summarized in Figure 4.The directionality of the mean values for most of the workload categories shows that controllers reported higher workload when using the Conflict Alert model and T-TSAFE runs where commanded altitude entries were made via keyboard, as compared to the other two conditions.TRACON controllers are not used to making data entries, and they often reported that this task increased their workload.Likewise, the controllers expressed many negative opinions about the Conflict Alert model, so it is not surprising that workload ratings were mostly higher for this condition.Conversely, for most of the workload categories, workload was rated lower during T-TSAFE runs where no commanded altitude entries were required and T-TSAFE runs where commanded altitudes were received via ADS-B.It seems likely that the lower workload ratings were due to the absence of required commanded altitude entries.However, it should be noted that these trends should be viewed with some caution, since the mean differences did not reach statistical significance. +Subjective data -ProceduresThe controllers were asked questions about T-TSAFE and Conflict Alert procedures, which they rated on a scale of 1-to-5.These data were then analyzed using Analysis of Variance.Table 2 shows the mean ratings of the controllers' ability to maintain awareness of potential conflicts.Results indicated that the controllers found it significantly easier to maintain awareness of potential conflicts with T-TSAFE as compared to the Conflict Alert model (F=7.99, df=[1,7], p<0.05).These data were validated in discussions between researchers and controllers, where controllers indicated that having the Time-to-Conflict in the first line of the data block provided them with information that helped them prioritize their tasks.Similarly, T-TSAFE was rated easier to use than Conflict Alert, possibly due to controllers' experience with the high number of false or nuisance alerts that are routine for Conflict Alert as it is currently implemented in the field.There was a significant difference in the acceptability of the T-TSAFE procedures as compared to Conflict Alert (F=5.73,df=[1,7], p<0.05).The explanation of the interface and the logic behind T-TSAFE given to the controllers might explain their levels of higher acceptability of T-TSAFE.The controllers also perceived that T-TSAFE alerts were more timely and useful than Conflict Alerts.This result is consistent with controller feedback, indicating that the duration of T-TSAFE alerts was adequate for the controllers to respond in time to prevent a loss of separation.A statistically significant difference was found for the perceived number of false alerts between df=[1,7], p<0.05).Again, the controllers were drawing on past experience with Conflict Alert in the field. +CONCLUSIONSTerminal T-SAFE is being developed by NASA as a conflict detection tool to address the inadequacies of Conflict Alert as it is currently used in the field.The selection of the independent variables in this study was guided by previous research on T-TSAFE showing that making altitude intent information available led to a reduction in false alerts.T-TSAFE received commanded altitudes either from controllers entering commanded altitudes issued for conflict resolution or as shortterm intent over ADS-B.Results showed that there was no difference in the rate of false alerts across conditions.This could possibly be explained by the fact that controllers do not make too many altitude entries that were expected to impact the rate of false alerts.Also, the duration of the alerts was not affected by the conditions, and neither was the controllers' response time to the alerts.Subjective data included results comparing a simplified model of Conflict Alert to the three T-TSAFE conditions.The controllers reported that they experienced similar levels of workload, with a slight increase in the physical component of the workload in the condition that required the controllers to make commanded altitude entries via the keyboard.The subjective data on the procedures were favorable towards T-TSAFE over Conflict Alert.The controllers found that they maintained better awareness of potential conflicts and had adequate time to act on the alerts with T-TSAFE.They also reported the T-TSAFE procedures as more acceptable and easier than those with Conflict Alert.Further human in the loop testing of the tool with longer run durations is necessary to better understand the impact of altitude entries on the overall performance of the conflict detection and resolution tool.Figure 2 .2Figure 2. Total Number of T-TSAFE Alerts Compared with False Alerts (Mean Value Per Controller Per Run) +Figure 3 .3Figure 3.Comparison of Duration of Alert and Response Time to an Alert +Figure 4 .4Figure 4. Workload Across the Experimental Conditions +Table 1 . Experiment Matrix Altitude Intent Condition1No Altitude EntryCondition AAltitude Entry via KeyboardCondition BAltitude Entry via ADS-BCondition C +Table 2 . Mean Subjective Ratings on Procedures (*p<0.05)2T-SEConflictSETSAFEAlert + + + + + + + + + Steam--electric plant air and water quality control data. Summary report for the year ended December 31, 1974 + + HErzberger + + 10.2172/6593530 + + + Proceedings of the Fourth USA/Europe Air Traffic management R&D Seminar + the Fourth USA/Europe Air Traffic management R&D SeminarSanta Fe, New Mexico, USA; Washington DC + + Office of Scientific and Technical Information (OSTI) + 2001. December 2001. 2009 + 31 + + + FAA Aerospace Forecast Fiscal Years 2009-2025 + Erzberger, H. (2001). The Automated Airspace Concept, Proceedings of the Fourth USA/Europe Air Traffic management R&D Seminar, Santa Fe, New Mexico, USA, December 2001. Federal Aviation Administration (2009). FAA Aerospace Forecast Fiscal Years 2009-2025. Washington DC, 31. + + + + + Nuisance Alerts in Operational ATC Environments: Classification and Frequencies + + FerneFriedman-Berg + + + KennethAllendoerfer + + + ShantanuPai + + 10.1177/154193120805200123 + + + Proceedings of the Human Factors and Ergonomics Society Annual Meeting + Proceedings of the Human Factors and Ergonomics Society Annual Meeting + 2169-5067 + 1071-1813 + + 52 + 1 + + 2008 + SAGE Publications + + + Friedman-Berg, F., Allendoerfer, K. & Pai, S. (2008). Nuisance Alerts in Operational ATC Environments: Classification and Frequencies. Proceedings of the Human Factors and Ergonomics Society 52 nd Annual Meeting. + + + + + Development of NASA-TLX (Task Load Index): Results of Empirical and Theoretical Research + + SandraGHart + + + LowellEStaveland + + 10.1016/s0166-4115(08)62386-9 + + + Advances in Psychology + + PHancock + + + NMeshkati + + Amsterdam, The Netherlands + + Elsevier + 1988 + + + + Hart, S. & Staveland, L. (1988). Development of a Multi-Dimensional Workload Rating Scale: Results of Empirical and Theoretical Research, In P. Hancock & N. Meshkati (Eds.), Human Mental Workload, Amsterdam, The Netherlands, Elsevier, 139-183. + + + + + Human/automation response strategies in tactical conflict situations + + JeffreyHomola + + + ThomasPrevot + + + JoeyMercer + + + MatthewMainini + + + ChristopherCabrall + + + SanJose + + 10.1109/dasc.2009.5347468 + + + 2009 IEEE/AIAA 28th Digital Avionics Systems Conference + + IEEE + 2009. October, 2009 + + + Homola, J., Prevot, T., Mercer, J., Manini, M. & Cabrall, C. (2009). Human/Automation Response Strategies in Tactical Conflict Situations. 28 th Digital Avionics System Conference, October, 2009. + + + + + The Next Generation of Air Supply Units + + JanvierEBeaver + + + KarlHering + + 10.1115/rtd2004-66002 + + + Joint Rail + + ASMEDC + 2004. December 12, 2004 + + + Joint Development and Planning Office (2004). The Next Generation Air Transport System Integration National Plan. December 12, 2004. + + + + + A review of conflict detection and resolution modeling methods + + JKKuchar + + + LCYang + + 10.1109/6979.898217 + + + IEEE Transactions on Intelligent Transportation Systems + IEEE Trans. Intell. Transport. Syst. + 1524-9050 + + 1 + 4 + + 2002. December 2000 + Institute of Electrical and Electronics Engineers (IEEE) + + + Kuchar, J. & Yang, L. (2002). A Review of Conflict Detection and Resolution Modeling Methods. IEEE Transactions on Intelligent Transportation Systems. Vol. 1, No. 4, December 2000, 179-189. + + + + + Estimation of European Airspace Capacity from a Model of Controller Workload + + ArnabMajumdar + + + WashingtonOchieng + + + JohnPolak + + 10.1017/s037346330200190x + + + Journal of Navigation + J. Navigation + 0373-4633 + 1469-7785 + + 55 + 3 + + 2002 + Cambridge University Press (CUP) + + + Majumdar, A., Ochieng, W. & Polak, J. (2002). Estimation of European Airspace Capacity from a Model of Controller Workload. The Journal of Navigation. 55. 381-403. + + + + + Tactical Conflict Alerting Aid for Air Traffic Controllers + + RussellAPaielli + + + HeinzErzberger + + + DannyChiu + + + KarenRHeere + + 10.2514/1.36449 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 32 + 1 + + 2007 + American Institute of Aeronautics and Astronautics (AIAA) + + + Paeilli, R.A. & Erzberger, H. (2007). Design and Testing of Tactical Conflict Alerting Aid for Air Traffic Controllers. + + + + + Human-in-the-Loop Evaluation of Ground-Based Automated Separation Assurance for NextGen + + ThomasPrevot + + + JeffreyHomola + + + JoeyMercer + + 10.2514/6.2008-8885 + + + The 26th Congress of ICAS and 8th AIAA ATIO + Napa Valley, CA, USA + + American Institute of Aeronautics and Astronautics + 2009. 2009. June 2009 + + + Prevot, T., Homola, J., Mercer, J., Manini, M. & Cabrall, C. (2009). Initial Evaluation of NextGen Air/Ground Operations with Ground-Based Automated Separation Assurance. Eight USA/Europe Air Traffic management Research and Development Seminar (ATM 2009). June 2009, Napa Valley, CA, USA. + + + + + Tactical Conflict Detection in Terminal Airspace + + HuabinTang + + + JohnERobinson + + + DallasGDenery + + 10.2514/1.51898 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 34 + 2 + + 2011 + American Institute of Aeronautics and Astronautics (AIAA) + + + Tang, H., Robinson, J.E. & Denery, D.G. (2011). Tactical Conflict Detection in Terminal Airspace. Journal of Guidance, Control, and Dynamics. Vol 34, No. 2, 403-413. + + + + + + diff --git a/file770.txt b/file770.txt new file mode 100644 index 0000000000000000000000000000000000000000..12a06bc63e27493efb9979a4f5d8c7669d239587 --- /dev/null +++ b/file770.txt @@ -0,0 +1,248 @@ + + + + +will continue to require ground-based planning and control.When aircraft transition from en route to terminal airspace, their trajectories often must merge subject to intrail separation or time-based flow constraints necessary for meeting airport capacity limitations.The unconstrained nature of free flight complicates these groundbased tasks even more.Hence, providing controllers with effective decision support tools that not only support freeIlight, but also support the transition of t+rec-tlight aircraft into the terminal airspace, is essential.A logical first step would be to achieve as many freel]ight benefits as possible without significantly altering current ATC procedures or requiring expensive airborne equipment modifications.These benefits could be achieved by a ground-based system that is highly responsive to user preference in en route airspace while also providing ['or an orderly transition to terminal areas.User preference is the general term for any aircraft operation that is explicitly requested or assumed to be advisories for all aircraft, with a locus on supporting controllers in analyzing and resolving complex, highly constrained traffic situations.When combined, the integrated AT/ST system supports user preference in any air route traffic control center (ARTCC or Center) sector.This system should also he useful in evaluating more advanced free-Ilight concepts by providing a test bed for future research.This document presents an overview of the design concept, explains its anticipated benefits, and recommends a development strategy that leads to a deployable system.The AT and ST are described in detail, and a new ATM position, the "airspace coordinator," is defined.Examples of conflict resolution for typical conflict scenarios are also given. +Functional DesignThe Automatic conflict detection at the ST level will initially be implemented to display only conflicts between:• any pair of aircraft "owned" by that sector controller (independent of whether the conflict resides within the controller's sector);• any "unowned" aircraft within a controller-defined distance (or time) l'rom the sector boundary and any other such unowned aircraft or any owned aircraft; or • any owned aircraft within a controller-defined distance (or time) from its next sector and any aircraft within that next sector or any aircraft within a specified distance (or time) from and due to arrive in that next sector.To include probing the effects of other aircraft, the controller can manually identify the desired aircraft to be probed.Once an unowned aircraft has been identified, it is probed until the ST determines that it can no longer impact aircraft owned by the controller.Display of potential conflicts beyond sector boundaries to all sector controllers involved will facilitate the solution of multisector conflicts.The ST can aid a sector controller to quickly create and evaluate a provisional ("what-iff') plan through the use of provisional clearance feedback, which allows the controller to determine the effect of issuing a clearance without affecting other controllers or other parts of the automation system.In addition to provisional clearance feedback, controllers are expected to use the ST to provide information for maneuver feedback.For example, a controller may absorb required delay by turning an aircraft away from its intended metering fix, while observing a countdown in remaining delay to be absorbed.After the required amount of delay has been absorbed, the controller would turn the aircraft back on a path to the metering fix.The ST also helps the controller to monitor aircraft spacing by displaying separation information at points along predicted trajectories.In-trail spacing can be shown at a defined location (e.g., at a Center or terminal airspace boundary) for selected streams of aircraft (e.g., aircraft exiting into a specified, adjacent Center or assigned to a defined meter fix All display features will ultimately be integrated into an advanced display (such as the display system replacement, or DSR), but some features may appear in early development phases on an auxiliary display interface, as shown in figure 3, where an example of controller interaction with the ST is shown.In the figure, flight UAL001 must be delayed to meet a desired crossing time (13:01) at the TOMSN metering fix while avoiding a conflict with overtlight UAL0(12.Through interacting with the ST's provisional planning tools, the controller has determined that a horizontal path stretch with turnback directly to TOMSN is a workable solution.In this case, the controller issues the path stretch clearances to UAL001 at the turnout and turnback points, followed by a descent clearance approximately 30 n. mi.prior to the TOD point.The AT-detected conflict between UALI00 and UAL200 is also shown with supporting information.The AT resolution advisory of a 20-knot indicated air speed (KIAS) reduction is shown on the fourth line of the data tag for UAL200.(Note that the Iburth line is used for illustrative purposes only.)The AT and ST advisories will be dcsigned so the controller can easily distinguish between them, possibly through colorcoding or blinking.The figure should not be considered a final display interface; significant effort will be devoted to make the interface as effective and easy to use as possible. +Integrated AT/ST ToolsThe needs of any en route traffic environment can be met through integration of the AT and ST.In a completely +Conflict Resolution ScenariosSince much of the concept relies on providing increased flexibility (e.g., routing) for users, considerable attention has been given to determining how the integrated tools will help resolve conflicts.In this section, the anticipated operation of the tools is described in some detail tot several conflict sccnarios.The expected benefits of the integrated tools in each traffic environment arc also discussed.Four typical separation conflict scenarios are identified in figure 4. Consider the numbered areas to be sectors within a Center.In the figure, the sectors are shown to be horizontally adjacent, but the scenarios also hold for vertically adjacent sectors.They even hold for adjacent sectors in different Centers, although the coordination between facilities will be more complicated.For convenicncc in the discussions that follow, all the conflicts are shown to occur in sector 2. When an aircraft crosses a controller's awareness boundary, the ST displays infbrmation to that controller about conflicts between the aircraft and any other aircraft in the downstream sector or any aircraft that has also crossed that awareness boundary.In figure 5, aircraft A is predicted to have a conflict in the downstream sector.When aircraft A crosses the awareness boundary for the upstream controller (SC1), the ST displays the conflict information to SC 1 ; when the aircraft crosses the awareness boundary for the downstream controller (SC2), the conflict information is displayed to SC2.If the ST detects a conflict, it displays the conflict inlbrmation to SC4, even though the conflict occurs outside sector 4. Recall that the conflict is displayed to SC4 even if one or both aircraft are outside SC4's awareness boundary since the upstream controller's awareness boundary is for displaying conflicts with aircraft in scctor 2 (conflicts between aircraft owned by the same controller are always detected by the ST, independent of the sector in which the conflict exists).SC4 thcn uses the provisional planning capability of the ST to resolve the conflict and issue a clearance to the aircraft.The ST also displays the conflict to SC2 if both aircraft are within SC2's awareness boundary (i.e., the downstream controller's awareness boundary).The display of conflict intbrmation to the downstream sector allows for SC2 to request an early handoff if desired.If SC2 does not request and obtain handoffs for aircraft A and B, the resolution is pcrlbrmed by SC4 as described.In general, for conflict resolution involving more than one controller, there is a potential to lower workload through the SC/ST display logic discussed previously.For typical choke-point sectors (such as a low-altitude sector containing a feeder fix), a benefit in providing an upstream controller with an opportunity to resolve the conflict is also possible, thereby redistributing the overall workload.In addition, if an upstream solution is not desired, SC2has the option to resolve the conflict by requesting early handoffs from SC4.These workload benefits are expected to apply to many traffic situations, so the controllers will have a strong incentive to take full advantage of the integrated tools. +External Intruder ConflictFor the external intruder conflict case (fig.6(c)), assume that the AT has chosen to modify the trajectory of aircraft A in order to resolve the conflict predicted to occur in sector 2. When the ST receives the conflict resolution packet that contains the advisory for aircraft A, it checks the display logic for SC2.If SC2 is configured to accept AT resolutions, the ST probes for conflicts.Finding none (or if SC2 is accepting AT resolutions with ST conflicts),the ST displays the conflict information and resolution to SC2.If SC4 has also been configured to accept display of AT advisories, the conflict information is displayed with a message that a resolution is pending in sector 2. This message is important to alert SC4 that any clearance issued to aircraft B at this time may cause the AT resolution to be invalidated and that sector 2 should be notified if verbal negotiation is desired.SC4 will also be aware that, if SC2 rejects the advisory, the AT may resolve the conflict by modifying the trajectory for aircraft B. Again, all situational awareness is accomplished without need for voice communication between controllers.If the conflict has not already been resolved by the AT, the ST detects the conflict and displays the conflict information to SC2 and SC4 after aircraft B crosses the awareness boundary and when the conflict is within the decision-time horizon for each controller.The resolution may be performed by SC2 for aircraft A, by SC4 for aircraft B, or for both aircraft in a negotiated solution.In early implementations, the decision of who will resolve the conflict is made verbally (or procedurally) between the two affected controllers.ST provisional planning aids are expected to be used by the controllers for resolution.Display of conflict information to both controllers may also facilitate an early handoff of aircraft B to SC2 if desired. +Early implementationsof the integrated tools may require the AC to perform the role of selecting which aircraft receives the AT advisory.In this example, the AC may know that the traffic situation in sector 2 makes aircraft B more appropriate than aircraft A for receiving a resolution advisory.By placing the integrated tools in operation with functions such as this performed manually, data can be generated that will serve as a basis for a heuristics-based set of aircraft selection criteria, which could later be automated.In addition, the external intruder scenario would benefit greatly from automatic conflict resolution in the ST, which will also be developed for an advanced implementation.ST automatic resolution logic would parallel the logic of the AT; it would remove much of the need lor negotiation between controllers for scenarios that require resolutions based upon controller intent. +Intersector ConflictAgain assume that aircraft A is chosen by the AT to resolve the conflict.wishes to resolve the conflict, SC2 may ask for early handoffs of both aircraft.Because the interscctor scenario occurs when aircraft in separate sectors have a conflict predicted at a point outside either sector, it is the most complicated situation for sector controller coordination; therefore, it offers the greatest potential for increased efficiency and reduced workload.As in the external intruder scenario, large benefits are expected by using the AT to coordinate conflict resolutions that involve two or more controllers.The manual negotiations currently required for solving multisector conflicts can be reduced significantly, so controllers should have an incentive to use the tool.Early implementation of the integrated tools may require the AC to perform the coordination role tbr AT advisories.As in the previous scenario, ST automatic conflict resolution (to be implemented in an advanced development) is expected to further reduce the need for verbal controller coordination. +Development StrategyA phased development approach is proposed that focuses on obtaining benefits as early as possible, validating the concept under real-world conditions, and using operational experience to expand tool capabilities. +Phase 1In the first phase, the AT and the ST will be developed and evaluated as independent decision support aids.All integration between the AT and the ST is performed manually.A strong emphasis will be placed on development and evaluation of the core capabilities of each tool, and on validation of the fundamental concepts.In addition to laboratory development and evaluation, the tools will undergo operational evaluation in a limited area of en route airspace, involving a few representative sectors (representing both traffic environment extremes) that are chosen based on development and evaluation goals.All display interfaces will be developed only to a level that permits concept evaluation and human-factorsrelated research.The AT will provide automated detection of potential conflicts and the probabilities associated with these predictions for en route aircraft in the Center airspace.The dynamic conflict display will be used to provide this information to the AC, who may then notify the appropriate area supervisor of projected areas of high congestion through voice communication.The sector controllers will use the ST to probe for predicted spacing and conflicts between specified aircraft (manual or limited automatic conflict detection), resolve predicted conflicts through provisional planning, and support aircraft in making efficient descents.An advanced radar tracker will be used to make accurate conflict predictions up to 20 minutes in advance.Research for Phase I will concentrate on validating concept feasibility.It will be designed to answer fundamental questions regarding benefits to controllers and users.The research will also focus on defining needed operational procedures (e.g., intersector coordination) and the key elements of a mature display interface through controller evaluations.Some of the concept feasibility issues to be explored are:• the effectiveness of the dynamic conllict display in assisting the AC in managing the airspace;• the extent of the assistance the ST provides to the controller in devising and executing traffic plans, especially for managing transitioning aircraft;the appropriate sector controller decision-making time horizon for aircraft trajectories with differing traffic management constraints;whether the controller considers the benefits received from the ST to outweigh the additional workload required to interface with the tool; the expected time horizon tor AT cost-effective resolutions based on advanced radar trackcr data, and the timeliness of these resolutions in accommodating user preferences and not infringing upon controller intentions; the sensitivity of AT cost-effective resolutions to sector controller issuance timing and its effect on resolution effectiveness; and the appropriate probability threshold for display of an ST-detected conflict to a controller. +Phase 2In the second phase, most of the capabilities of the fully developed integrated tools will be achieved by allowing some of the tasks to be performed manually.The AT will provide cost-effective resolutions to the AC, who will then use experience-based judgment to determine whether to accept the solutions or modify them using provisional planning techniques.The AC will then request the AT to send the resolutions to the ST for display to the controller through a fully developed display interface.The controller will provide an input to the ST to notify the AC (through the AT display interlace) whether the advisory is accepted or rejected.The ST will provide full automatic conflict detection in addition to the tools provided in Phase I. Automatic detection should allow the controller to devote more attention to other tasks required in this phase, such as acceptance or rejection of AT advisories.Extensive human factors development is expected during this phase.A limited operational deployment could possibly be achieved after Phase 2; if so, the deployment is expected to be limited to a set of sectors chosen on the basis of benefits and cost.Phases 1 and 2 will require the use of an auxiliary display and an additional controller to be located at each sector position.To maintain all current radar controller operations, the additional controller will interact with the ST and then will transfer advisory clearances to the radar position (R-side).The display should provide a plan-view graphical interface and a keyboard lor input.In addition to the proposed Phase 2 functions, this workstation should have all capabilities currently used to perform sector controller duties, such as accepting handoffs, displaying trend vectors, and providing tools for aiding separation maintenance.Sector-certified radar controllers will probably be required to interface with the auxiliary display.The handoff position could be responsible for monitoring the additional display for sectors with heavy traffic, and the interphone or flight-data (D-side) position could assume this responsibility for sectors with light traffic.When the ST is fully developed and approved for direct use by both the R-and D-side sector controllers, its functions will be In en route environments, the integrated tool anticipates and facilitates user preferences, while providing advisory aids to help the controller solve complex traffic management problems.• An operational system can be placed in the field quickly, where it can serve as a testbed for new technology.Planned technology exploration includes trajectory negotiation with airborne flight management systems, integration with traffic schedulers such as TMA, and free-flight concepts that transfer responsibility for maintaining separation to the user.• Sector controllers will have an incentive to use the proposed system because it will improve their capability and reduce their workload.The system is not intended to be a replacement for controllers, but an aid to increase productivity. +•The design leads to a logical and systematic evolution.It will be implemented as a series of new functions that will gradually increase system capability.Manual tasks will be automated based on the semi-automatic operation of early deployments, thereby freeing controllers to give attention to more advanced tasks as the design ew)lves. +•The system is not dependent on planned hardware upgrades, such as DSR, to be successful.It will, however, take advantage of such upgrades.Although development of this system faces many challenges, no unresolvable implementation issues are anticipated.desired by the airline or pilot.Preferred aircraft operations can vary from non-ATe-preferred mutings to the use of airborne vertical navigation (VNAV) automation during descents into terminal airspace.In a system that requires positive ATe control, user preferences are facilitated (or enabled) through verbal or procedural ATC clearances.Developing decision support tools for controllers that identify user preferences and their effects on the current traffic situation would enable controllers to quickly assess the effort of incorporating user preferences into the current traffic plan.Furthermore, by designing automation to determine the minimum change to the user preference required for incorporation into the traffic plan, a large step toward free-flight benefits would he achieved.To facilitate user preference in all en route environments, a system concept based on an extension of the Center/ TRACON Automation System (CTAS) has been developed.(See ref. 4.) It consists of two integ,'atcd components.An airspace tool (AT) focuses on unconstrained en route aircraft (e.g., not transitioning to the terminal airspace), taking advantage of the unconstrained nature of their trajectories and using long-range trajectory prediction to maximize user efficiency by providing cost-effective conflict resolution advisories to sector controllers.A sector tool (ST) generates efficient +In addition to providing more advance time for traffic planning, automatic detection should enable the controller to resolve conflicts more efficiently.The increased confidence provided by automatic conflict detection may also be useful in reducing the number of conservative clearances that controllers currently issue to ensure separation when the5' are unsure of whether a conflict will occur. +the resolutions must be available to the sector controller before the conflict is within the controller's decisionmaking time horizon.Work based on current prediction accuracy suggests that these cost-effective resolutions would generate advisories approximately 10 to 14 minutes in advance of the conflict (ref.5), a time that is expected to be within the prediction horizon needed for the ST.However, reference 5 indicates that increasing trajectory prediction accuracy increases the advance time of the minimum-cost point.One prediction error source is the existing FAA radar tracking algorithm of the Center Host computer.Using an advanced radar tracking system (ref.10) is expected to improve prediction accuracy, thereby enabling minimum-cost resolutions 20 or more minutes in advance.The system is now in place for testing at the Denver ARTCC, and its accuracy is being verified through analysis of flight-test data. +Figure 4 (4Figure 4(a) shows an example of an "intrasector" conflict, the situation where both aircraft and the predicted point of conflict (i.e., initial loss of minimum separation requiremcnts) are within a single sector.This scenario should yield the most straightforward resolution since only one controllcr is involved.A somewhat more complicated scenario, an "external" conflict, is shown in figure 4(b): both aircraft are in one sector, and the point of conflict is in anothcr.Figure 4(c) shows an "external intruder" scenario: one aircraft and the predicted conflict point are in one sector, and the other aircraft is in another sector. Figure 4(d) shows an "intersector" conflict, where the two aircraft are in different sectors and the predicted conflict point is in a third sector.The latter two scenarios generally require the greatest amount of coordination between controllers. +Figure 4 .4Figure 4. Typical conflict scenarios in a multisoctor environment. +Figure 6 .6Figure 6.Concluded. +Table I .IInitial AT/ST conflict probing characteristicsAirspace toolSector toolAircraft probedUnconstraineden route traffic onlyAll en route trafficTrajectory constraintsNoneTraffic managementDetection responsibilityEntire Center airspaceFor each sector, all aircraft within sector and individual aircraft in neighboringsectorsConflict displayed toAC a for resolutionSC(s)bConflict resolutionAC resolves conflicts with AT and sends to STSC manual resolution (aided by ST(AC can negotiate resolutions with sectorthrough provisional planning aids)controllers)SC issues clearancesResolution typeCost-effectivetrajectory with provisionalST provisional planning aidsplanning aidsaAC = Airspace CoordinatorbSC = Sector ControllerIn order for the cost-effectiveAT resolutions to have animpact in an environmentin which the AT and ST areworking together to facilitate user preferencesfor amixture of both unconstrainedand constrained aircraft, +Table 2 .2Expected conflict resolution behavior of the AT/ST systemThe positions of +When the ST receives the ATadvisory for aircraft A (fig. 6(d)), it checks the displaylogic for SCI, and ifSClis configured to accept, the STdisplays the resolution advisory to SC 1. If the resolutionis acceptable, SC1 then issues the clearance. If SC4 isconfigured to display AT advisories, the conflict infor-mation is displayed with a message that a resolution thatinvolves aircraft B is pending in sector I. Again, SC4knows that any clearance issued to aircraft B at this timemay invalidate the AT advisory. In addition, SC2 may beinformed about the conflict situation by display of theconflict and a message that a resolution involving aircraftA and B is pending in sector 1. SC2 is informed of theconflict if both aircraft are within their respective aware-ness boundaries, and SC2 has all informationneeded toknow the AT resolution plan. If the AT advisory inter-fetes with SC2 traffic planning, SC2 can ask SC 1 to rejectthe AT advisory and/or negotiate with both controllers forearly handoffs.Assume that the ST detects a conflict between aircraft Aand B. SCI and SC4 have the conflict displayed whenboth aircraft have crossed the awareness boundarieswith respect to sector 2 and are within the controllers"decision-makingtime horizon. The two controllers thenuse the ST provisional planning aids and work togetherto negotiate a solution. SC2 also sees the conflict andaircraft information if SC2's display is configured toshow all potential conflicts within sector 2. Again, if SC2 +Table 33with many of the tool functions performed manually.Both simulation experiments and field testing are anintegral part of Phases 1 and 2 development.Phase 3provides the fully developed integrated system describedin this document, including its use as a research platlbrmfor advanced concepts. Simulations and field evaluationssummarizesa three-phase developmentstrategy. Phase 1will be used during this phase to automate many of theconcentrateson demonstrationof the core capabilitiesmanual functions developed in earlier phases. Withof the individual AT and ST tools and develops theadequate staffing, the developmentcould be completedfunctionalitynecessary to perlbrm concept validation.in about four years.Phase 2 provides an initial integrated tools capability,Table 3. DevelopmentstrategyPhaseCapabilityFunction1AT automatic conflict detection advisoriesATConcept feasibility demonstrationAC manual notification of potential conflicts to area through voice communicationAutomatic conflict detection Dynamic conflict displaySC/ST conflict detection and provisionalSTplanning through an auxiliary displayAuxiliary displayDescent advisory aidsManual and limited automatic conflictdetectionSpacing advisory aidsProvisional planning aids2AT automatic conflict detection advisoriesAll Phase 1 functionsInitial operating capabilityAT cost-effective to ACresolution advisories displayedATProvisional planning aidsAC/AT provisional planningCost-effectiveresolutionsAT advisories passed to ST (approved by AC)STSC/ST automatic conflict detection, provisionalMature display interfaceplanning, and spacing aids display to an auxiliary controller through a fully developed interfaceFull automatic conflict detection3AT conflict detection and resolution advisoriesAll Phase 2 functionsFull operating capabilityAT advisories passed to ST (monitored by AC) SC/ST automatic conflict detection, provisionalATCost-effectiveresolutions enhanced toplanning, and spacing aids display at sector viainclude AC response to AT resolutionan advanced display interfacerejectionFully developed logic for ST probing of ATSTresolutions and display to sector controllerAT resolutions probing logicFully developed logic for AT response toDisplay configurationlogicresolution rejectionDeployabledisplay interface (such asDSR) + + + +The Federal Aviation Administration is trying to make its air traffic management system more responsive to the needs of the aviation community by exploring the concept of "free flight" for aircraft flying under instrument flight rules.A logical first step toward free flight could be made without significantly altering current air traffic control (ATe) procedures or requiring new airborne equipment by designing a ground-based system to be highly responsive to "user preference" in en route airspace while providing for an orderly transition to the terminal area.To facilitate user preference in all en route environments, a system based on an extension of the Center/TRACON Automation System (CTAS) is proposed in this document.The new system would consist of two integrated components.An airspace tool (AT)focuses on unconstrained en route aircraft (e.g., not transitioning to the terminal airspace), taking advantage of the relatively unconstrained nature of their flights and using long-range trajectory prediction to provide cost-effective conflict resolution advisories [o sector controllers.A sector tool (ST) generates efficient advisories tor all aircraft, with a focus on supporting controllers in analyzing and resolving complex, highly constrained traffic situations.When combined, the integrated AT/ST system supports user preference in any air route traffic control center sector.The system should also be useful in evaluating advanced free-flight concepts by serving as a test bed for future research.This document provides an overview of the design concept, explains its anticipated benefits, and recommends a development strategy that leads to a deployable system. +SUBJECT TERMSAir traffic control, Free flight, User-preferred routing, UPR + + + + + + + Introduction to RTCA/DO-254 + 10.1201/b17833-1 + + + Airborne Electronic Hardware Design Assurance + Washington D.C. + + CRC Press + Jan. 1995 + + + + Report of the RTCA Board of Directors' Select Committee on Free Flight. RTCA Inc., Washington D.C., Jan. 1995. + + + + + + DCHart + + Free Flight: An Evolutionary Revolution + + + Hart, D. C.: Free Flight: An Evolutionary Revolution. + + + + + + + Avionics + + + Apr. 1995 + + + + Avionics, Apr. 1995, pp. 50-51. + + + + + Interactions of airline operations control centers with the Air Traffic Management System + + PhilipSmith + + + ElaineMccoy + + + JudithOrasanu + + + RebeccaDenning + + + CharlesBillings + + + AmyVa + + 10.2514/6.1996-3854 + + + Guidance, Navigation, and Control Conference + San Diego, Calif + + American Institute of Aeronautics and Astronautics + July 1996 + + + Smith, P. J., et al.: Interactions of Airline Operations Control Centers with the Air Traffic Manage- ment System. AIAA Paper 96-3854, presented at the Guidance, Navigation, and Control Conference, San Diego, Calif., July 1996. + + + + + Design of an Automated System for Management of Arrival Traffic. NASA TM-102201 + + HErzberger + + + WNedell + + + Also + + + LEngle + + + June 1989. Oct. 1989 + + + Conflict Detection Tool. Addendum to TM-102201 + Erzberger, H.; and Nedell, W.: Design of an Automated System for Management of Arrival Traffic. NASA TM-102201, June 1989. Also, Engle, L.: Conflict Detection Tool. Addendum to TM-102201, Oct. 1989. + + + + + Exobiology in earth orbit : the results of science workshops held at NASA Ames Research Center / edited by D. DeFrees ... [et al.] ; prepared at Ames Research Center. + + DDefrees + + 10.5962/bhl.title.4099 + + + NASA Ames Research Center + + May 2, 1996 + National Aeronautics and Space Administration, Office of Management, Scientific and Technical Information Division, + + + Presented to FAA/AUA-500 management + Erzberger, H.: Conflict Probing and Resolution Algorithms. Presented to FAA/AUA-500 management, NASA Ames Research Center, May 2, 1996. + + + + + Field evaluation of Descent Advisor trajectory prediction accuracy + + StevenGreen + + + RobertVivona + + 10.2514/6.1996-3764 + + + Guidance, Navigation, and Control Conference + San Diego, Calif + + American Institute of Aeronautics and Astronautics + July 1996 + + + Green, S.: and Vivona, R.: Field Evaluation of Descent Advisor Trajectory Prediction Accuracy. A1AA Paper 96-3764, presented at the Guidance, Navigation, and Control Conference, San Diego, Calif., July 1996. + + + + + Environmental Benefits Assessment of Air Traffic Control Automation Tools + + SGreen + + + WDen Braven + + + DWilliams + + 10.2514/6.2020-2876.vid + + Apr. 1993 + American Institute of Aeronautics and Astronautics (AIAA) + + + Green, S.; den Braven, W.; and Williams, D.: Development and Evaluation of a Profile Negotiation Process for Integrating Aircraft and Air Traffic Control Automation. NASA TM-4360, Apr. 1993. + + + + + Mitigating Impacts of Natural Hazards on Fishery Ecosystems + + SGBenjamin + + + KJBrundage + + + LLMorone + + 10.47886/9781934874011.ch17 + + + + Mitigating Impacts of Natural Hazards on Fishery Ecosystems + East-West Highway, Silver Spring, MD + + American Fisheries Society + 1994. 1325. 20910 + + + Technical Procedures Bulletin No. 416 + NOAA/NWS + Benjamin, S. G.: Brundage, K. J.; and Morone, L. L.: The Rapid Update Cycle. Part I: Analysis Model Description. Technical Procedures Bulletin No. 416, NOAA/NWS, 1994. [National Weather Service, Office of Meteorology, 1325 East-West Highway, Silver Spring, MD 20910. Also available on the World Wide Web at http://www.fsl.noaa.gov/frd-bin/tpbruc.cgi.] + + + + + Interlace Requirements Specification for the Sensor Data Acquisition and Processing System (SDAPS) + + RSlattery + + + SMGreen + + Contract No. DCA 100-94-D-0015 + + Jan. 1994. Mar. 1996 + NASA Ames Research Center by BDM Federal, Inc + 80301 + Boulder + + + Conflict Free Trajectory Planning for Air Traffic Control Automation + Slattery, R.; and Green, S. M.: Conflict Free Trajectory Planning for Air Traffic Control Automation. NASA TM-108790, Jan. 1994. Interlace Requirements Specification for the Sensor Data Acquisition and Processing System (SDAPS). Contract No. DCA 100-94-D-0015, prepared for NASA Ames Research Center by BDM Federal, Inc., Boulder, CO 80301, Mar. 1996. + + + + + + diff --git a/file771.txt b/file771.txt new file mode 100644 index 0000000000000000000000000000000000000000..1fbc319fa2541c759924576e647f9e3943c270f0 --- /dev/null +++ b/file771.txt @@ -0,0 +1,267 @@ + + + + +Nomenclature +I. IntroductionO enable advanced air traffic management concepts, such as trajectory-based operations, the NextGen concept in the U.S. is increasing the amount of data shared between automation in the National Airspace System. 1 Trajectory-Based Operations utilize predicted aircraft four-dimensional trajectories to "assess the effects of proposed trajectories and resource allocation plans, allowing both service providers and operators to understand the implications of demand and identify where constraints need further mitigation."Understanding the functional and performance differences between disparate trajectory predictors (TPs) is critical for enabling the interoperability between these systems.If trajectory and aircraft intent data are to be effectively shared, the variation in requirements must be properly understood.This understanding is necessary to determine the minimum data that must be shared to achieve interoperability without degrading performance.Over the past decade, several US and European efforts to understand the differences between disparate trajectory predictors met with little success.These efforts attempted to survey and document the current and future DST requirements for trajectory prediction capabilities to identify commonalities and differences.The content varied widely from one organization to another, the material was inconsistent in what was documented (little comparable overlap), and the scope of information was significantly incomplete.The wide range of approaches taken to implement TPs mad it difficult to understand the similarities (and true differences) the requirements.A team of companies with expertise in TP technology (Lockheed-Martin, L-3 Communications, and GE-Aviation) documented the capabilities of five TPs from different domains, environments and functional support 2,3,4,5 .The five TPs were chosen to cover a range of current day TP applications, including air-and ground-based support of control, traffic flow management and flight planning systems in both research lab and operational deployment environments.These documents all followed a common document structure, referred to as the Requirements Framework, which was developed by NASA, Lockheed Martin and L-3 Communications to enable a clear, consistent, complete and cross-comparable documentation of TP requirements and capabilities.This common format allowed a formal comparison of the TPs to be conducted.The paper begins by describing the five TPs that were the focus of the TP comparison and by the approaches taken to perform the comparison.Next, detailed descriptions of the results of the TP comparison are presented.Finally, an interoperability scenario is described where airborne FMS data is sent via datalink to ground-based systems performing conflict detection and resolution functions.Using the comparison results as a foundation, this scenario illustrates the challenges disparate TP capabilities can provide in trying to achieve system interoperability. +II. Trajectory Prediction ComparisonIn support of the FAA/Eurocontrol Research & Development Committee (Action Plan 16) on Common Trajectory Prediction, NASA has been studying state-of-the-art trajectory predictor (TP) technologies in an attempt to better understand commonalities and differences in TP capabilities and requirements.Under this project, the capabilities of four state of the art TP technologies were captured, documented and compared:1. NASA's Center/TRACON Automation System (CTAS) En Route TP 2 2. NASA's CTAS Terminal Area TP 3 3. NASA's Future Air Traffic Management Concepts Evaluation Tool (FACET) TP 4 4. GE Aviation Flight Management System (FMS) TP 5 These four documents all follow a common capabilities framework that leveraged the latest version of the AP16 Common TP structure to enable effective cross-comparison of the documented capabilities.During the development of these documents, a minimal level of cross-comparison was also performed to identify documentation techniques T that could further enhance effective capability comparisons.The primary results of these early comparisons were modifications to the documentation approach (e.g., the use of abstraction and the use of standardized tables in capturing certain capabilities) to further enhance the cross-comparability of the documents.Under a separate Space Act Agreement with NASA, the capabilities of a fifth TP technology were captured by Lockheed Martin and added to the comparison: 5. Release 1 of the FAA's En Route Automation Modernization (ERAM) TP 6 This document was not a part of the early comparison effort, but did follow a similar document framework (based on an earlier revision of the AP16 Common TP structure) to the other four documents.In developing an approach to compare the five TP technologies, it was desired to achieve two main goals:1. Compare specific capabilities that are currently known to illustrate similarities or differences of interest between the TPs 2. Compare all capabilities in a systematic way in an attempt to identify new, currently unknown similarities or differences of interest between the TPsThe first goal was relatively easy to meet after the capabilities of known interest were identified.The second goal was much more difficult to achieve.To meet both goals, two separate comparison approaches were developed and performed.The first goal was met by identifying a series of capabilities of interest to the organizations involved in this project (termed "orthogonal looks" by NASA) and directly comparing them.These comparisons were performed verbally during a face-to-face meeting between representative experts for each TP at the Engility facility in Billerica, MA and through follow-up emails.The second goal was met by generating a comparison matrix of major capabilities from the TP capability documents.This comparison matrix focused on "top-level" capabilities, i.e., those that define the range of situations handled by the TP rather than details on how those situations are handled.Both approaches are described in more detail next. +A. Orthogonal Looks Comparison ApproachThe orthogonal look approach was designed to compare TP capabilities of specific interest to the organizations involved in the project.The organizations were NASA (CTAS and FACET TPs), Lockheed Martin (ERAM TP) and GE Aviation (FMS TP).§ All three organizations were requested to provide specific questions (orthogonal looks) they would like answered by each of TP capabilities.Both NASA and Lockheed Martin submitted a list.GE Aviation had a single question that was equivalent to a Lockheed Martin question.Both lists of orthogonal looks were answered and discussed in a face-to-face meeting at the Engility facility in Billerica, MA on March 25, 2009.TP experts from each organization were present and the majority of questions were discussed directly during the meeting.Questions that were either not covered during the meeting (due to time limitations) or which required additional investigation by one or more TP experts to answer properly were covered through follow-up emails. +B. Top-Level Capability Comparison ApproachDefining an approach to identify new, currently unknown similarities and differences between the five TP technologies presented some challenges.The simplest approach would have been to compare the five capabilities documents, section by section.To perform such a comparison was the original objective of using a common framework based on the AP16 Common TP structure.Unfortunately, using this type of comparison approach at this stage of the framework's development posed some potentially significant issues: It relied on every document using the exact same framework sections: This was not the case since the ERAM document used an earlier version of the framework than the version used by the other documents.Even in the case of the four TP documents that followed the same framework version, the effectiveness of the framework to support such a comparison was still unknown.It was desired to have an approach that was robust to potential issues with the framework, while providing potential insight into how to improve the framework to support such a section-by-section comparison.§ Engility Corporation is not the owner of a particular TP product and was responsible for the ultimate comparison results and, as such, did not provide specific comparisons of interest.It relied on every TP expert filling out the framework sections in the same way: The AP16 common TP structure (the foundation of the framework) is an effective tool in describing a wide range of TPs, but it takes skill to map the capabilities of a specific TP to the AP16 Common TP structure.Since the AP16 Common TP structure is still being developed and is not yet widely familiar to the general TP community, it is not surprising that different TP experts interpret it differently and that this could lead to comparable capabilities being captured in different sections of the capabilities documents.Identified during the early comparisons of four of the TP technologies, the boundary between the TP and its client application chosen by the documenting TP expert impacts where in the framework capabilities were captured.Every TP expert had to decide what capabilities to include within the TP and which to consider part of the client application.** This decision not only impacted where, but whether certain capabilities would be captured in the document.Different decisions by the TP experts could impact a section-by-section analysis.It did not provide any approach to scoping the comparisons: The framework was designed to capture the capabilities of a TP to a very detailed level.Decisions about how much detail to provide were left to the TP experts, but all five documents captured very detailed information about their TP's processing.It was not clear to what level an initial search for new interesting comparisons should be performed and a section-by-section comparison did not help to answer this question.Actually, it was feared that covering all sections would tend to lead to many comparisons at the lowest level (where differences are expected), potentially leading to a failure to identify the most significant similarities/differences at a higher level (i.e., hard to see the forest through the trees).An alternative approach was developed to:1. try and identify the most significant capability similarities/differences while giving insight into potential additional comparisons that could be performed later 2. be robust to potential limitations in the current framework with the idea that the comparison results would feed back into framework improvementsThe approach compared the "top-level" capabilities for each TP, focusing on available types of input data, output data, math models, and major behaviors that could be predicted without getting into details on how the TP actually developed a trajectory.Seven sets of data were captured for each TP in an excel spreadsheet (called the comparison matrix):• Input state data • Constraints handled • Behavior models used • Math models used • Input atmospheric models • Update processing supported • Output trajectory dataEach TP capability document was read from beginning to end and any reference to a capability falling under one of the seven data sets was captured.This eliminated the chance that comparable capabilities captured in different framework sections would be missed.Four out of the five documents were all read by the same TP expert and the data entered into the comparison matrix (the FACET document being the only exception; this data was entered by a NASA TP expert).The final comparison matrix was reviewed by all of the TP experts to check for correctness.The entire comparison matrix was also reviewed at the face-to-face meeting at the Engility facility in Billerica, MA on March 24-25, 2009.TP experts from each organization were present and the matrix was updated with corrections, as necessary.** This decision would appear to be a simple one, but in practice is quite complex.See reference 7 for details. +III. TP Comparison ResultsThe complete set of comparison results, for both the orthogonal look and the top-level capability comparisons, are provided in reference 8. Below is a subset of some of the results for the top-level capability comparisons.This data will be pared down in the final paper to just the most relevant results. +A. Input State DataInput state data supports several capabilities in the TP, including:• Defining the initial condition of the trajectory • Identifying the initial lateral and vertical maneuvers • Identifying missing intent data (e.g., when input state is not on the aircraft's flight plan route)Differences in the input state data available to different TPs reflects differences in the environments in which these TPs operate.It also illustrates differences in data available for determining initial maneuvers and intent modeling.Table 1 provides the lateral input state data for the five TPs.All five TPs receive position and true course data.The two CTAS TPs receive position in the local X/Y coordinate frame used by the Host computer, where the rest receive position in latitude and longitude.The use of a local Cartesian coordinate frame by CTAS reflects the limited airspace supported by each instance of this system (i.e., one Air Route Traffic Control Center (ARTCC)).True course data is available to all of the ground-based TPs, but since it is generated by processing previous radar tracks, it is susceptible to certain types of errors.The GE FMS TP has access to both true course and magnetic heading from the onboard avionics.Both the GE FMS and ERAM TPs receive indicators of whether the initial state is in a turn.The GE FMS TP receives the current roll angle from the onboard avionics, where the ERAM TP receives an "In Turn" indication from ERAM.† † Knowledge of the actual roll angle provides the most information about an aircraft that is currently in a turn.• CTAS TA • • FACET • • GE FMS • • • • ERAM r1 • • •Table 3 shows the input time, altitude and speed data for the five TPs.All five TPs receive both time and corrected altitude (corrected for altimeter setting).The ERAM TP also receives an established altitude which is the last Mode C or pilot-reported altitude from the aircraft.All five TPs also receive ground speed, which is the main form of speed available from ground sensors (e.g., radar tracks).The CTAS TPs can also accept a true airspeed (TAS) input, which is used primarily for non-active aircraft (i.e., using flight plan information, including the filed speed, which is a true airspeed, when no track data has been received).The GE FMS TP receives the full suite of speeds, including Mach, calibrated airspeed (CAS), TAS, and ground speed, all of which are available from the onboard avionics.All but the FACET TP also receive the current vertical speed of the aircraft.Table 5 shows a range of other input data related to the state of the aircraft.All TPs are aware of the aircraft type (e.g., Boeing 757-200).Weight is received by the CTAS En Route and the GE FMS TP, though the CTAS value is typically down-linked from the aircraft where the GE FMS TP receives weight from an onboard weight determination system.The GE FMS TP receives extensive information on the engines and aircraft configuration.The ERAM TP receives the equipage onboard the aircraft as well as whether the state data was generated by "coasting" previous track information.The GE FMS TP also knows the aircraft's navigation equipage.† † ERAM evaluates past radar tracks to determine whether the current track should be considered in a turn.• • • • • CTAS TA • • • • • FACET • • • GE FMS • • • • • • • ERAM r1 • • • • •• • CTAS TA • FACET • GE FMS • • •* • ERAM r1 • • • * Engine Bleed ( +B. Constraints HandledA TP generates a trajectory to achieve, if possible, the input constraints on the aircraft.These constraints may be explicitly defined as input intent data (e.g., a clearance to maintain a given heading), may be derived directly from input intent data (e.g., an input flight plan route is decomposed into a series of turn waypoints though route parsing), or may be added by the TP through intent modeling to create a sufficient set for integration (e.g., adding a waypoint capture constraint to connect the aircraft initial condition to an input route).Comparing the set of constraints that a TP is able to achieve provides insight into the scope of trajectory problems that the TP can predict. +Lateral ConstraintsTable 7 shows the waypoint-defined lateral constraints that each of the five TP's can accommodate.Each TP is able to end the trajectory at a specified waypoint.Though trajectory end conditions will be discussed in more detail later, this does indicate that all of the TPs have a similar approach to ending the trajectory (e.g., none of them require ending the trajectory at a fixed look-ahead time).Each TP can also handle waypoint-defined turns, such as those found in typical flight plan routes (e.g., inside turns).This was expected since all of the TPs are using some form of FAA flight plan route definitions.• • * CTAS TA • • FACET • • GE FMS • • • • • • ERAM r1• • * * CTAS and ERAM handle Hold clearances, but do not build the race track.The GE FMS does handle some additional waypoint-defined constraints that are specific to its environment.‡ ‡ For example, the "DME Arc From Fix" constraint is used in very specific (and infrequent) terminal procedures.‡ ‡ The GE FMS is able to handle all of the waypoint-based constraints defined in the ARINC 424 standard.The different types of 424 leg constraints are specific to the way the onboard guidance and navigation systems will interact to laterally control the aircraft.The ARINC 424 constraints were evaluated and generalized for comparison Similarly, the "Procedure Turn" is used by the FMS to perform a very specific direction reversal maneuver.Handling of these additional constraints is not required by the environments of the other four TPs.The handling of a "Hold" constraint is a very interesting example of the difference in requirements placed on a TP by different client applications.Three of the client applications (CTAS, ERAM, and GE FMS) handle Hold constraints.Only one (the GE FMS) uses the TP to build the actual "race track" path of the aircraft within the Hold and the transition to the racetrack from the previous route leg.Since the GE FMS TP is supporting the aircraft guidance function during a Hold maneuver, it makes sense that it would have this type of requirement.Neither the CTAS ER nor the ERAM TP builds the race track path, but both do handle the effects of Hold constraints within the client application.For example, CTAS protects the airspace around the Hold constraint during conflict detection and ERAM does not conflict probe downstream of a Hold constraint.Table 9 shows the route/radial-defined lateral constraints that each of the five TP's can accommodate.Each TP can handle routes defined by great circle path route segments.Similar to the waypoint-defined turn constraints, this was expected since these are the typical route segment definitions found in FAA flight plans.Three of the TPs (all except the CTAS Terminal Area and FACET TPs) can handle waypoint capture constraints to transition to a route.A waypoint capture constraint indicates that the aircraft should immediately turn and proceed directly to the identified waypoint and then follow the route after that waypoint.Again, this is a typical FAA routing approach, as well as a common intent modeling method for connecting an off-route initial condition to a route, so it is not surprising that most of the TPs support this constraint.The CTAS Terminal Area TP does not have a flight plan route in the terminal area to connect to, so this capability isn't applicable.The FACET TP does not include this capability because its TP is limited to predicting aircraft following their routes.§ § The CTAS En Route TP can accept an additional capture delay time constraint with a waypoint capture constraint.This constraint defines how long (in time) the aircraft should travel before turning to capture the identified capture waypoint.CTAS uses this to develop lateral delay absorption advisories.• • • CTAS TA • FACET • GE FMS • • • • • • ERAM r1 • • •Both the GE FMS and ERAM TP can accommodate an intercept route leg constraint.This constraint indicates that the aircraft should transition to the input route when the current path of the aircraft intersects the input route.The GE FMS also has the ability to handle two other intercepts: intercept DME distance and intercept radial.These constraints appear to be only required for the GE FMS TP's environment.The GE FMS TP can also accept a lateral offset constraint.This constraint defines the start, end, and required dimensions (e.g., displacement of the offset from the input route) to create an offset path that is laterally displaced from the input route.This type of offset path could also have been generated outside of the TP and sent in as a new route using other constraints previously discussed.The use of a lateral offset constraint by the GE FMS TP appears to support the specific operational environment of the FMS, where it is desired to maintain knowledge of the original route while predicting the offset path. +Vertical ConstraintsTable 6 shows the speed constraints that each of the TPs can accommodate.All five TPs can accommodate some type of cruise speed constraint, whether the constraint be in Mach, CAS, or TAS (the specific speed types accepted are based on the environment of the TP).Since most aircraft spend the majority of their flight in level cruise, it is to be expected that all of the TPs would handle speed constraints in cruise.All but the CTAS Terminal Area TP also handle speed constraints in Climb and Descent.The CTAS En Route, FACET, and GE FMS TPs all directly handle with the other TP constraint handling.Only a few (Constant Radius Arc, DME Arc From Fix, etc.) were considered different enough from the other constraints to warrant explicit identification in these tables.§ § The capabilities of the entire FACET system are not limited to just these cases, but it only uses its TP for cases where the aircraft is on its route.For more information, see reference 4.a climb or descent speed profile (Mach and CAS) constraint.ERAM uses an empirically derived performance model, which defines TAS as a function of altitude for each aircraft type.This TAS schedule is based on observations of actual aircraft performance, which means it is indirectly modeling a range of climb and descent speed profiles.The CTAS Terminal Area TP and GE FMS TP both accommodate an approach speed since they develop trajectories for the terminal area.The GE FMS TP also handles speed constraints for Hold, Takeoff, and Engine Out scenarios.Mach, TAS Empirical Empirical * In terminal area, the speed constraint impacts both level flight and descending path segments.The differences in handling climb and descent speeds, especially between the CTAS, GE FMS and ERAM TPs, illustrate the differences in TP requirements from their client systems.The CTAS En Route TP supports a range of client applications (e.g., CTAS decision support tools).One of them is the En Route Descent Advisor, which develops descent speed profile (Mach and CAS) clearance advisories for conflict-free schedule-conformant aircraft trajectories.In this environment, the TP is used to predict the impact of a specific descent Mach and CAS profile, hence the TP's ability to directly use descent Mach and CAS as constraints.When the CTAS En Route TP is used in environments where the Descent Speed profile is not known (e.g., in support of the Traffic Management Advisor), it uses default speed values based on company procedures for the aircraft type.This enables the TP to support both environments.On the other hand, the ERAM TP does not need to support the advisory generation environment of the CTAS En Route TP.Its approach of using an empirically-based model of aircraft speed (and vertical rate) has been developed to be effective when the actual speed profile is not known.For the GE FMS TP, the aircraft is always aware of its own Mach and CAS, so it obviously uses this information to its advantage.Table 7 and Table 8 show the non-waypoint-defined and waypoint-defined altitude constraints, respectively, that each of the TPs can accommodate.The FACET TP can only handle a cruise altitude constraint.This is consistent with its lower fidelity approach to trajectory modeling.For non-waypoint defined altitude constraints, the CTAS En Route, GE FMS and ERAM TPs all handle cruise altitude, interim altitude and departure/arrival speed limit altitude constraints (e.g., 250kt speed limit at 10,000 ft), which is not unexpected since these are common altitude constraints in today's operations.Both the CTAS TPs and GE FMS TP handle a defined altitude at which to apply an externally entered altimeter setting, though ERAM does not.The GE FMS TP also handles an Optimal Step Altitude (defines a target change in cruise altitude) and two altitude constraints related to takeoff (Acceleration Altitude and Thrust Reduction Altitude).The CTAS Terminal Area TP deals with an altitude "constraint" that defines the top of the TRACON airspace, which it uses to separate en route pilot behavior from TRACON behavior.*** *** The CTAS terminal Area TP receives initial radar tracks when the aircraft is still in en route airspace.The Max TRACON Altitude constraint defines at what altitude the aircraft enters the terminal area.CTAS ER • • • • CTAS TA • • FACET • GE FMS • • • • • • • ERAM r1 • • •For waypoint-defined altitude constraints, the GE FMS and ERAM TPs handle the most types, which is consistent with their deployment-level operational focus.All TPs (except the FACET TP) can handle AT constraints.The two CTAS TPs don't handle AT or Below constraints, which is consistent with their historical emphasis on arrival traffic handling.The GE FMS TP does not handle Remain AT constraints, which is consistent with the fact that these are very, very rare (added for a very specific operational situation).Both the GE FMS and ERAM can handle window constraints, which are just a combination of an AT or ABOVE and an AT or BELOW constraint at the same waypoint.It should be noted that window constraints are not allowed within RNAV procedures because not all FMS can handle them.CTAS ER • CTAS TA • • FACET GE FMS • • • • • ERAM r1 • • • • • •Table 9 shows a range of other waypoint-defined constraints that the TPs can accommodate.All of the TPs (except the FACET TP) can handle both Start AT and End AT speed constraints, though only ERAM handles Mach versions of these constraints.Both the CTAS En Route and GE FMS TPs can handle time constraints, which is not surprising since meet-time (required time-of-arrival) functionality has been a part of these systems for decades.The ERAM TP can handle a delay constraint added at a waypoint.ERAM uses this delay constraint to handle delays filed in a route (e.g., a military aircraft could be off performing a refueling operation or practicing a bombing run).The GE FMS TP also handles a waypoint-define (start and end) temporary change to a new Mach speed and a flight path angle constraint; both constraints are unique to the FMS environment.• • • CTAS TA • • FACET GE FMS • • • • • ERAM r1 • • • • +C. Behavior Models UsedBehavior models define the maneuvers predicted by a TP.They are typically defined in layers, with more general maneuvers (e.g., climb) being broken up into more specific maneuver components (e.g., constant CAS climb at max climb thrust followed by a constant Mach climb at max climb thrust).Comparing the different behavior models used by the different TPs highlights similarities and differences in which maneuvers are predicted and to what level of detail are they predicted. +Lateral Behavior ModelsTable 10 shows the different behavior models used by the TPs to define the aircraft's maneuvers when following the straight (i.e., non-turning) portion of a route.All five of the TPs model some form of following a great circle path (newer GE FMS TPs use WGS84 paths).The FACET TP is an interesting case in that it is unable to follow the great circle path defined by two waypoints (the typical flight plan route case).It is only able to model following a great circle path from the current position to a waypoint, which is only equivalent to the two waypoint case when the aircraft is initially on the path defined by two waypoints.The GE FMS TP also has the ability to model following a constant course or heading to a waypoint or some other termination point.• CTAS TA • FACET • GE FMS • • • • • • ERAM r1• * The GE FMS can follow a great circle path, constant course or constant heading to a variety of termination points, including a specified distance, DME distance, an altitude, a radial or manual termination point.Table 11 shows the different behavior models used by the TPs for performing turns associated with following a route.All of the TPs are able to model immediate turns to capture the next waypoint in the route (e.g., used when the initial condition is in a turn or the current track is not along the route direction) and model the aircraft performing an "inside" (i.e., non-fly over) turn.These are both typical behavior model maneuvers associated with flying a flight plan route.The CTAS TPs and the GE FMS TP also have the ability to model a turn overshoot maneuver, used by aircraft for extreme (typically greater than 135 degree) inside turns.ERAM doesn't require such a behavior model since its math model uses instantaneous turn modeling (see Math Models below), which would not show any difference between an overshoot and a non-overshoot turn.The CTAS TPs and the GE FMS TP also model two types of fly-over waypoint turns: one that Begins AT the turn waypoint and the other that Ends AT the turn waypoint.Again, the ERAM TP's choice for turn modeling would make such a distinction irrelevant.Also, though possible in en route airspace, these are more typical in modeling terminal airspace maneuvers.This may be the main reason that FACET does not model these types of turns.The GE FMS TP also performs two additional, specialized turn types: constant radius and DME arc turns.ER • • • • • CTAS TA • • • • • FACET • • GE FMS • • • • • • • ERAM r1 • •Table 12 shows the behavior models used for capturing a route.The CTAS Terminal Area and FACET TPs do not support capturing a route because their initial conditions are never off of their routes (see earlier discussion).The CTAS En Route, GE FMS and ERAM TPs all support waypoint capture behaviors, including a modeled delay before initiating the turn to the capture waypoint, the turn, and a great circle path to the capture waypoint.Only the GE FMS TP supports route intercept maneuvers, including both constant heading and constant course intercepts, to accommodate its input intercept constraints (see Table 9).It should be noted that the ERAM TP does handle intercept route leg constraints, but the intercept point is geometrically calculated and converted into a turn waypoint constraint, so no intercept behavior model is required.Only the CTAS En Route TP supports the Path Stretch behavior model to accommodate its Capture Delay Time input constraint (see Table 9).• • • • • CTAS TA FACET GE FMS • • • • • • ERAM r1 • •• * This delay models lag time in pilot action, not an externally entered delay constraint (that is handled by the Path Stretch behavior model). +Vertical Behavior ModelsTable 13 shows the behavior models used to model an aircraft during a climb maneuver.The CTAS Terminal Area TP does not support climb trajectories, since it is focused on arrival traffic.All of the other TPs model the climbing capture of a desired climb speed.For the CTAS En Route and GE FMS TPs, the speed captured can be either a target Mach or CAS.For the FACET and ERAM TPs, the speed constraints in climb are a performance model defined TAS schedule (see earlier for a discussion of ERAM's empirically-based TAS schedule), so the speed captured is always a TAS.For the majority of the climb, the CTAS En Route and GE FMS TPs can use either a constant speed (Mach or CAS) and known throttle setting (e.g., max climb thrust) or a constant speed and a constant vertical rate.The FACET and ERAM TPs do not define the behavior model in any more detail than as a climb.Both rely on their math model representation for accuracy.Additionally, the GE FMS and ERAM TPs have the ability to model level off behavior in the climb maneuver, both in capturing a desired speed and maintaining that speed at the constant altitude.The GE FMS TP also has the ability to model takeoff behavior, using two behavior models (ground roll acceleration and constant CAS takeoff) for this purpose.Table 14 shows the behavior models supported for aircraft in their cruise phase of flight.All of the TPs support constant speed cruise segments, as well as level flight transitions to a new cruise speed.The remaining two behavior models are to support changes in the cruise altitude.Neither the CTAS Terminal Area nor the FACET TP support changes in cruise altitude.For step descents to a new cruise altitude, both the CTAS En Route and GE FMS TPs can do either constant speed (Mach or CAS) and known thrust setting (e.g., idle thrust) or constant speed and constant vertical speed maneuvers.The ERAM TP does not define the maneuver in any more detail than as a descent (again relying on its empirical aircraft performance model for accuracy).The GE FMS and ERAM TPs have the added ability to model a speed change during the step descent.The situation is similar for step climbs, except that the CTAS En Route and GE FMS TPs only use a constant speed, known throttle setting (e.g., max climb thrust) model for these maneuvers.Table 15 shows the behavior models supported for aircraft in their descent phase of flight.For the descending maneuvers, each TP has its own behavioral modeling approach.The GE FMS TP has the most extensive approach, using idle thrust, constant vertical speed, and constant flight path angle versions of capturing a speed and holding a speed (Mach or CAS) constant.The CTAS En Route TP uses the same idle thrust and constant flight path angle versions of holding a speed (Mach or CAS) constant, but performs a non-specific (relies on the mathematic model) descending capture of a speed.Due to its low altitude trajectories, the CTAS Terminal Area only does constant flight path angle versions of capturing and maintaining a speed (CAS only at these altitudes).Both the FACET and ERAM TPs use the same type of model as described above for their climb behavior models.All but the FACET TP can also handle a level off segment during a descent.The CTAS En Route and GE FMS TPs can handle both capturing and maintaining a target Mach or CAS.The CTAS Terminal Area TP can capture and maintain a CAS.The ERAM TP can capture and maintain the TAS from its empirical model. +D. Math Models UsedBehavioral models are mapped to mathematical models within the TP to define either a set of equations of motion to integrate or a geometric algorithm to predict the aircraft trajectory.Comparing TP math modeling approaches gives insight into the accuracy achievable by the TP.First, the general integration approaches between the TPs are compared.Next, the specific lateral and vertical math models used by the TPs are compared. +General Integration ApproachTable 16 summarizes the general integration approaches used by each of the five TPs.Only the FACET TP integrates equations of motion for the lateral path of the trajectory.Each of the other TPs geometrically approximates the lateral path of the aircraft prior to integrating the vertical profile.In all cases but ERAM, these TPs create an approximation of the vertical profile to calculate required turn model parameters (e.g., turn radius and center).† † † The FACET TP performs a "fast time simulation" to create its trajectory and its integration of both lateral and vertical equations of motion only in a forward direction is consistent with this design.For the vertical approach, the CTAS En Route and GE FMS TPs perform both forward and backward integration (depending on the specific math model) of the vertical (both altitude and longitudinal) equations of motion.The CTAS Terminal Area TP only integrates longitudinal equations of motion; the altitude path is geometrically calculated from the fixed flight path angles defined in the behavior model (see Table 15).The ERAM TP does estimate certain vertical trajectory parameters (e.g., top of descent) using backward integration, but the final trajectory is generated using forward integration only.• * ERAM uses reverse altitude integration to estimate the TOD location, but then uses forward integration assuming the TOD is at the estimated value to build the actual trajectory (refining the TOD location, if necessary).As for integration type, the FACET and ERAM TPs use only Eulerian integration, the CTAS Terminal Area TP uses a combination of Eulerian and 2 nd Order Runga Kutta (depending on the specific math model), the CTAS En Route TP uses 2 nd Order Runga Kutta for all math models, and the GE FMS uses a proprietary approach. +Lateral Math ModelsTable 17 shows the lateral math models used by the different TPs.The FACET and ERAM TPs use great circle equations for all of their straight (non-turning) behavior models.This is consistent with the fact that both of these TPs use only great circle behavior models (see Table 10 andTable 12).The GE FMS TP uses three types of straight math models: great circle (newer versions are WGS84), constant heading and constant course.This is also consistent with the lateral behavioral models used by this TP (see Table 10, and Table 12).It is interesting that the CTAS TPs only use constant course math models, even though they only have great circle and constant heading ‡ ‡ ‡ behavioral models (see Table 10 andTable 12).The CTAS TPs take advantage of the fact that the distances they need to predict over are limited to a single ARTCC airspace and make the mathematical approximation of constant course being equivalent to great circle paths for short distances.For the constant heading segments, the approximation of constant course enables the lateral path to be geometrically approximated rather than integrated (see earlier discussion).An interesting apparent inconsistency is that the GE FMS TP supports constant heading math models, but does not integrate these lateral paths (the lateral path is geometrically approximated, as described earlier).This is possible because the GE FMS TP only uses constant heading segments when it is predicting paths off of the route and its atmospheric model does not use a varying wind field in these cases.This means that the track associated with the constant heading value does not vary over the segment, which enables the GE FMS TP to geometrically determine this path.• • CTAS TA • • FACET • • GE FMS • •* • • ERAM r1• • * The GE FMS TP does not have a varying wind field off the aircraft route.Hence, there is no difference between their approach to Constant Heading and Constant Track except for the selection of the initial track value.For turns, the CTAS TPs and the GE FMS TP use a constant radius turn model for all turns.The turn radius and turn center are approximated using the ground speed from an estimated vertical profile, as discussed previously.The FACET TP uses a variable bank angle model, which includes a model for the "roll in, roll out" behavior to start and end the turn.This is possible because the FACET TP does not require the predicted trajectory to accurately leave one route segment and connect to the next (i.e., it's lateral behavior models for straight segments do not include a great circle path between two waypoints).During forward integration of the lateral path, the FACET TP continually checks (using a simplified algorithm) to see if it needs to initiate a turn and when it does, it turns until it can follow a great circle path to the next waypoint, regardless of the original path between route waypoints (the FACET TP is only used for cases where the aircraft is following its flight plan route).The ERAM TP uses instantaneous turn models, so no integration or geometric algorithms are required. +Vertical Math ModelsTable 18 shows variations of two types of vertical math models used by the TPs: those using equations of motion requiring knowledge of the aircraft thrust and those using equations requiring altitude to be constant.The known thrust models are only used by TPs that use kinetic equations of motion, which rules out the CTAS Terminal Area, FACET, and ERAM TPs.Both the CTAS En Route and GE FMS TPs use known thrust models, including max cruise thrust and idle thrust versions.The GE FMS TP has several additional models beyond those required by the CTAS En Route TP.These are required for the additional vertical behavior models supported by the GE FMS TP (see Table 13, Table 14, and Table 15).The constant altitude models are used by TPs that use kinematic equations of motion.All of the TPs use some form of a constant altitude, constant speed model (e.g., used to model the constant speed, level flight cruise behavior models in Table 14).The CTAS Terminal Area, FACET and ERAM TPs do not use kinetic math models (e.g., known thrust and constant altitude) for level flight acceleration/deceleration behavior models, but instead use a defined acceleration/ deceleration rate at constant altitude model.§ § § The GE FMS TP has the option to use either kinetic models (known thrust, constant altitude) or kinematic models (constant altitude, defined accel/decel) for these behaviors.Table 19 shows variations of three additional types of vertical math models used by the TPs: those using equations of motion requiring vertical speed to be constant, flight path angle to be constant, or requiring a defined (not necessarily constant) vertical speed and TAS change rate.The constant vertical speed models are used by the CTAS En Route and GE FMS TPs to cover behavior models explicitly requiring a constant vertical speed (see Table , Table , andTable 1).The GE FMS TP has the option to use a defined accel/decel rate to perform climbing/ descending speed changes using a kinematic model.The constant flight path angle models are similarly used by the CTAS TPs and the GE FMS TP for behavior models (specifically, descent models) explicitly requiring a constant flight path angle (see Table 1).Again, the GE FMS TP has the option to use a defined accel/decel rate to perform climbing/descending speed changes using a kinematic model.The CTAS Terminal Area TP does all of its descending accel/decel segments using this type of model.The defined vertical speed, defined TAS change rate model is used by both the FACET TP and the ERAM TP for all of their climb and descent behavior models (see Table andTable 1).This model uses a vertical speed value and a target TAS value, both retrieved from the performance model, and both of which vary with altitude.• • • • CTAS TA • • FACET • GE FMS • • • • • • ERAM r1•* * TAS change rate is a combination of constant accel/decel rate and the rate required to follow ERAM's empirical model's TAS variation with altitude.Observed values for speed and altitude rate can modify the empirical values.Table 20 shows two polynomial-equation-based math models.These two specialized models are used by the GE FMS TP to model its ground roll acceleration and constant CAS takeoff behavior models (see Table ).21 compares some of the TP use of aircraft performance model (APM) data.The CTAS En Route, FACET and ERAM TPs support ground-based systems that deal with a large number of aircraft types, so they contain a range of APM model types, including jet, turboprop, and prop models.The FACET TP uses BADA, a standard § § § The CTAS TA TP uses a fixed value for the acceleration/deceleration. The ERAM TP uses a value from its performance model.The FACET TP uses a value that will accelerate the aircraft in one integration step.APM that provides data for 295 aircraft types (99.2% of European air traffic), 103 of the aircraft types are directly modeled and the rest are cross-referenced. 9The CTAS En Route and ERAM TPs use their own APM databases and both use cross-referencing between aircraft of similar types to reduce the number of native models required to cover the full range of aircraft types.The CTAS Terminal Area TP does not use aircraft type specific models, so it does not need to have a similar range of aircraft types.The GE FMS TP is installed on a single aircraft, so it only needs modeling data for a single jet aircraft.The CTAS En Route and GE FMS TPs use math models with kinetic equations of motion (e.g., the known thrust models in Table 18).These math models use thrust and drag data from their APM to support these equations of motion.The FACET and ERAM TPs use vertical rate and TAS schedules from their APM; both sets of data vary with altitude and reflect estimates of complicated pilot procedures.The CTAS Terminal Area, FACET **** , ERAM and (optionally) the GE FMS TP all use acceleration/deceleration rates from the APM to model speed changes (see Table 18 and Table 19).The CTAS En Route, FACET, and GE FMS TP all use fuel flow data from their APM to determine fuel burned over the length of the trajectory. +E. Output Trajectory DataThe output trajectory is the final result of the TP.The data included in the final trajectory, as well as the format are typically, at least to some degree, dictated by the client application that uses the trajectory.In all of the tables below, the computed (internal) trajectory was used for the FACET TP since it does not actually output a trajectory to FACET.Table 22 shows the lateral trajectory data returned by the TPs.The GE FMS TP does not return lateral data at each predicted point of the trajectory.Instead, it returns a series of straight lines and arcs.This data is used by the FMS to perform lateral guidance when in the Lateral Navigation (LNAV) flight mode.This data can be correlated to the vertical data to provide along path distance and true course, but it is not done explicitly in the output trajectory.The CTAS TPs, FACET and ERAM TP all provide position data (either in X/Y or Lat/Lon) and some combination of course and/or heading data.All but the FACET TP also provide along path distance at each predicted point.The CTAS TPs and the GE FMS TP also provide explicit turn geometry data for each turn waypoint, including the turn radius and turn center position.**** FACET uses a maximum acceleration/deceleration rate from its performance model to limit its speed changes, which it attempts to perform in one integration step.• • • Magnetic • • • CTAS TA • • • Magnetic • • • FACET • True GE FMS • • o o • • • ERAM r1 •* • •* ERAM converts output to geodetic latitude and longitudeTable 23 shows the time, altitude and speed data provided by each TP at each predicted point.All of the TPs provide time, predicted altitude, and true airspeed data.The GE FMS TP also returns a reference altitude at each point.This altitude relates to a stored trajectory that the vertical guidance is currently using and gives an indication of predicted vertical deviation from this reference path.The CTAS TPs and the GE FMS TP also return Mach (not for CTAS Terminal Area TP) and CAS at each point and the CTAS TPs, FACET and ERAM also return ground speed.All of the TPs except the GE FMS return vertical speed and the CTAS En Route, and GE FMS TP return flight path angle data.• • • • • • • • CTAS TA • • • • • • FACET • • • • • GE FMS • • •* • • • • ERAM r1 • • • • • * GE FMSmaintains a reference theoretical descent profile.This data is returned with each trajectory for guidance purposes. +IV. Interoperability ImpactsDifferences in TP capabilities of disparate TPs create requirements for successful integration when sharing trajectory data.The following example illustrates the issues.This text was taken from a final deliverable document.It will be rewritten in a form more appropriate for a conference paper.For this scenario, it is assumed that the suppliers of trajectory data are aircraft equipped with a conventional Flight Management System (FMS) being used for Navigation/Guidance.The GE FMS is assumed to be the FMS onboard all aircraft.For LNAV/VNAV guidance modes, the GE FMS is assumed to be the provider of behavior model data.For all other guidance modes, the behavior models are based on extensions of the GE FMS behavior modeling.For this scenario, the receiver of trajectory data is a ground-based separation management system.This system is supporting a controller with automated conflict detection (CD) and conflict resolution (CR) capabilities.There are two options for the ground-based system performing separation management: En Route Descent Advisor (EDA) and En Route Automation Modernization (ERAM).In Figure 1, aircraft A is an arrival aircraft descending in LNAV/VNAV guidance to an arrival fix at the TRACON boundary (the figure is a vertical view of the two aircraft).The aircraft's descent profile is created within the aircraft's FMS to meet an altitude and speed restriction at this fix.The arrival fix and its altitude and speed constraints are also known by the ground systems as part of the aircraft's flight plan.Aircraft B (crossing traffic) is flying level in LNAV/VNAV.Laterally, both aircraft are flying their routes as defined in their flight plans.A loss of separation between aircraft A and B will occur along aircraft A's descent trajectory.If no data are shared between the ground and air systems, the conflict might not be detected (a missed alert) and/or an incorrect conflict might be detected (a false alarm).Data sharing should provide enhancement to the ground knowledge of the descent trajectory improving prediction accuracy and therefore reducing the chances of a false alarm or missed alert. +A. Intent Data SharingOne option to improve the predictability on the ground would be to downlink additional intent information from the aircraft.For this scenario, it would be sufficient for the aircraft to downlink the following information to represent its lateral and vertical intent: The information available to the ground systems that are not contained in the flight plan would then be the descent speed profile (Mach and IAS) and the TOD location.In the case of an aircraft supplier of trajectory data, the FMS predicted TOD location can be considered part of the intent information because it is an input to the aircraft's guidance system (it is used to define the initiation point of the descent).A benefit of sharing TOD location information would materialize only if the ground systems are able to use this information.Currently, neither ERAM nor EDA can introduce this information into their TP process.The use of downlinked TOD data is not required for this approach to achieve benefits.For EDA, the knowledge of the aircraft's current descent speed profile will improve the accuracy of the EDA TP to better determine whether the aircraft needs to be delayed to meet its scheduled time of arrival and to better detect conflicts prior to issuing a meet-time or conflict resolution advisory.If EDA is creating resolution or meet-time advisories that do not include changing the aircraft's descent speed profile, knowing the aircraft's current descent speed profile would also improve the effectiveness of the generated advisories.If EDA generates a descent speed from the same basic math modeling.For conflict probing, since EDA's behavior modeling in this case matches the GE FMS behavior modeling, there is actually no improvement over intent data sharing.In cases where EDA's modeling was unable to identify accurately the behavior model just from the intent information, this approach would provide better accuracy for the EDA conflict probe.For conflict resolution, the intent data sharing approach is expected to be more desirable because the descent speed profile is clearly identified as a constraint and can be reused, as necessary in candidate resolution trajectory generation.‡ ‡ ‡ ‡ For ERAM, the cruise maneuver before TOD is the same as that for the GE FMS, so no conversion is required.During the descent, ERAM uses a much simpler behavior model (Descent: Descending: Descent) than either the GE FMS or EDA.This reflects ERAM's design decision to add its trajectory generation complexity to its empirical math modeling as opposed to its behavior modeling.§ § § § The final ERAM level maneuver is similar to that for the GE FMS except that for ERAM the speed to capture is defined in TAS instead of CAS.The arrival fix speed constraint sent with the GE FMS behavior model would have to be converted by ERAM.As discussed in the intent data sharing approach, ERAM does not accept descent speed profile constraints as inputs.This causes the same negative impact to the behavior model sharing approach, namely, without rewriting the way ERAM handles descent speeds, it does not appear possible to use behavior model sharing to synchronize ERAM trajectories for this scenario. +C. 4D Trajectory SharingFor this scenario, the 4D trajectory predicted by the GE FMS would be downlinked to the ground systems.This would be represented by a series of predicted future 4D states that include, at a minimum, position (latitude and longitude), altitude and time.For EDA, downlinking a 4D trajectory may improve the performance of its conflict probe since the FMS predicted TOD is a part of the downlinked trajectory.That being said, it is not expected that this approach will provide a significant improvement over the other two data approaches in terms of conflict detection since both intent data and behavior model sharing work very well in improving EDA's prediction accuracy for this scenario.This approach would only be significantly more effective than the other two if the aircraft's math modeling were significantly better than EDA's.Due to the aircraft's atmospheric modeling deficiencies, the aircraft trajectory could actually be less accurate than one generated by EDA.For conflict resolution, downlinking a 4D trajectory doesn't provide any support for generating resolutions and is therefore the weakest of the three approaches.For ERAM, downlinking a 4D trajectory is the first data sharing approach that enables an improvement to the ERAM conflict probe.This is because the impact of the descent speed profile is included in the downlinked 4D trajectory, which is an intent error source for ERAM generated trajectories.For conflict resolution, downlinking a 4D trajectory still doesn't provide any support for generating "what-if" solutions on the ground.This approach to data exchange is therefore no worse for ERAM than the other three approaches for conflict resolution purposes. +V. ConclusionThe aircraft trajectory prediction capabilities of five state-of-the-art trajectory predictors were captured and compared to illustrate the commonality and differences between current-day trajectory prediction technologies.The approach taken and detailed results for the comparisons of the input state data, constraints handled, behavior and math models used, input atmospheric models, and output trajectory data are presented.The comparisons illustrate how design approaches chosen by each TP to meet the objectives of its client application create significant differences between the capabilities of the different TPs.A scenario was presented to illustrate the issues that arise when attempting to achieve interoperability between air traffic management automation through trajectory information sharing.An air-ground scenario was chosen where airborne flight management systems downlink trajectory information to support two options (ERAM and EDA) for a ground-based system providing controller conflict detection and resolution support.Three forms of trajectory data sharing (intent, behavior model, 4D trajectory) were analyzed to show the strengths and weaknesses of each approach in achieving interoperability.Due to the similarities in capabilities between the GE FMS TP and the EDA TP, interoperability is more achievable between these systems than between the GE FMS and ERAM systems.The scenario illustrates how different design approaches chosen by different TPs to meet the needs of their specific ‡ ‡ ‡ ‡ Though the descent speed profile is available from the behavior model, it is most likely not going to be stated explicitly as a constraint.This adds at least a minimal chance for incorrectly inferring these speed constraints from the behavior model data (though arguably a negligible chance in this specific scenario).§ § § § For more discussion on this topic, see reference 8.environments, specifically the ERAM use of simplified behavior models and empirical math models to deal with the lack of known aircraft speeds, can cause significant impediments to achieving system interoperability.AP16=FAA/Eurocontrol Action Plan 16 on common trajectory prediction APM = aircraft performance model BADA = Base of Aircraft Data (Standard Aircraft Performance Model) CAS = calibrated airspeed CTAS = Center/TRACON Automation System CTAS ER = CTAS en route TP CTAS TA = CTAS terminal area TP DME = Distance Measuring Equipment ERAM = En Route Automation Modernization ERAM r1 = ERAM release 1 FACET = Future Air Traffic Management Concepts Evaluation Tool FMS = flight management system FPA = flight path angle GE = General Electric IAS = indicated airspeed ITWS = Integrated Terminal Weather System LNAV = lateral navigation guidance TAS = true airspeed TBO = trajectory-based operations TOD = top of descent TP = trajectory predictor VNAV = vertical navigation guidance +air-conditioning and anti-ice), Engine State (Engines Off, All Engines On, Engine Out), Configuration information (slats position and flap position), and Weight On Wheels indication. +has the option of using either a constant accel/decel or target thrust.** Other target thrust values: Derated Max Climb, Takeoff, Derated Takeoff and Max Continuous (Engine Out).*** Other target thrust values: Derated Max Climb and Max Continuous (Engine Out). +Figure 1 :1Figure 1: Descending aircraft scenario. +•current FMS route lateral waypoints including any associated altitude/speed constraints • cruise altitude • current commanded speed profile • cruise Mach • descent speed profile (Mach and IAS) • TOD location +Table 1 : State data: lateral Position Attitude TP X/Y Lat/Lon True Course Magnetic Heading Roll Angle In Turn? CTAS ER •1 +Table 3 : State data: time, altitude and speed3AltitudeSpeedVerticalTPTimeCorrectedEstablishedMachCASTASGroundSpeedCTAS ER +Table 5 : State data: other TP5AircraftAdditional Engine &NavigationCoastTypeWeightConfiguration DataEquipageIndicatorCTAS ER +Table 7 : Lateral constraints: waypoint-defined TP7TrajectoryConstantDME Arc FromProcedureEndTurnRadius ArcFixHoldTurnCTAS ER +Table 9 : Lateral constraints: route/radial-defined TP Great Circle Path Waypoint Capture9CaptureInterceptDelayDMEInterceptInterceptLateralTimeDistanceRoute LegRadialOffset +Table 6 : Vertical constraints: speeds6EngineDescentClimbOutTPCruise SpeedSpeedSpeedApproachHoldTakeoffSpeedCTAS ERMach, CAS, TASMach, CASMach, CASCTAS TACAS*•FACETMach, CAS, TASMach, CASMach, CASGE FMSMach, CASMach, CASMach, CAS••••ERAM r1 +Table 7 : Vertical constraints: altitudes TP Cruise Alt7Departure/TransitionAccelThrustArrivalMaxAlt(V2 toReductionInterimSpeedTRACOOptimal(altimeterClimb)AltLimit AltN AltStep Altsetting) +Table 8 : Vertical constraints: waypoint-defined altitude TP AT AT or ABOVE AT or BELOW Window Remain AT Change AT8 +Table 9 : Vertical constraints: waypoint-defined speed, time & other Speed (AT or BELOW)9Initiate/EndTimeDelayFlightStart ATMachPathTPMachIAS/CAS& End ATAngleCTAS ER +Table 10 : Route-following behavior models: straight10Great CircleConstant CourseConstant HeadingTPWP to WPTo a WPTo Other*To a WPTo Other*To Other*CTAS ER +Table 11 : Route-following behavior models: turns Inside Turn TP Turn to Capture WP Normal Overshoot Begin AT WP End AT WP11ConstantRadius ArcDME Arc +Table 12 : Route capturing behavioral models12Waypoint CaptureRoute InterceptPath StretchDelayTurn toGreatDelayConstantConstantTurn toDelayBeforeCaptureCircleBeforeCourseHeadingCaptureBeforeTPCapture*WPto WPInterceptInterceptInterceptHeadingCaptureCTAS ER +Table 13 : Climb behavior models Climbing Level TP Ground Roll Accel Constant CAS Takeoff Capture Climb Speed* Climb Constant Speed, Thrust Setting Constant Speed, Vertical Speed Capture Speed Constant Speed13CTAS ERCAS, MachCAS, MachCAS, MachCTASTAFACETTAS**•GE FMS••CAS, MachCAS, MachCAS, MachCAS, MachCAS, MachERAM r1TAS***•TASTAS* Includes both End AT and Begin AT speed captures. FACET is only TP that can only do Begin AT.** FACET uses a performance model derived TAS schedule.*** ERAM uses an empirically derived TAS schedule, modified (in the case of its Aircraft Trajectory only) by afactor based on observed speed. +Table 14 : Cruise behavioral models14* Both CTAS and the FMS could do Idle Thrust step descents, but currently do constant vertical speed.LevelAltitude Step: DescentAltitude Step: ClimbConstantConstantConstantSpeed,Speed,Speed,ConstantCaptureCaptureThrust**VerticalCaptureThrustTPSpeedSpeed*SpeedDescentSettingSpeedSpeedClimbSettingCTAS ER••CAS, Mach CAS, MachCAS, MachCTAS TA••FACET••GE FMS•••CAS, Mach CAS, Mach•CAS, MachERAM r1••••••* Includes Begin AT and End AT as well as acceleration and deceleration.* +Table 15 : Descent behavioral models Descending Capture Speed Constant Speed Level TP Descent Descent Idle Thrust Vertical Speed Flight Path Angle Idle Thrust Vertical Speed Flight Path Angle Capture Speed Constant Speed15CTAS ERCAS, MachCAS, MachCAS, MachCAS, Mach*CAS, MachCTAS TACASCASCASCASFACET•TASGE FMSCAS, MachCAS, MachCAS, MachCAS, MachCAS, MachCAS, MachCAS, MachCAS, MachERAM r1•TAS***TASTAS* CTAS does not decelerate to Mach, though it has the capability.** FACET uses a performance model derived TAS schedule.*** ERAM uses an empirically derived TAS schedule, modified (in the case of its Aircraft Trajectory only) by afactor based on observed speed. +Table 16 : Math models: integration approach16LateralIntegrationVertical IntegrationIntegration TypeAltitudeLongitudinalFor-Back-For-Back-For-Back-2nd OrderTPwardwardwardwardwardwardEulerianRunga Kutta ProprietaryCTAS ERNoNoYesYesYesYes•CTAS TANoNoNoNoYesYes••FACETYesNoYesNoYesNo•GE FMSNoNoYesYesYesYes•ERAM r1NoNoYesNo*YesNo +Table 17 : Lateral Math Models17StraightTurnsConstantConstantGreatConstantVariableTPCourseHeadingCircleRadiusBank AngleInstantaneousCTAS ER +Table 18 : Vertical math models: target thrust & constant altitude models Known Thrust Constant Altitude TP Constant Altitude Constant CAS Constant Mach Constant FPA18ConstantDefinedDefinedVerticalAccel/Accel/ConstantSpeedDecelDecelSpeed +Table 19 : Vertical math models: vertical speed & FPA models Constant Vertical Speed Constant Flight Path Angle TP19DefinedVerticalSpeed,DefinedDefinedTASDefinedAccel/ConstantConstantChangeAccel/ConstantConstantDecelMachCASRateDecelMachCASCTAS ER +Table 20 : Vertical math models: polynomial equations TP Constant Fuel, Distance, Time 2 nd Order, Fuel,20Distance, TimeCTAS ERCTAS TAFACETGE FMS•*•**ERAM r1* Used for Ground Roll Accel behavior.** Used for Constant CAS Takeoff behaviorTable +Table 21 : Math models: aircraft performance model TP Model Types21Cross-Accel/ReferenceDragThrustTASDecelTypesDataDataAltitude RateScheduleRateFuel FlowCTAS ERJ, T, P••*•*•*CTAS TA•FACETJ, T, P•Climb, Descent•Max Only•**GE FMSJ•***•***•• ***ERAM r1J, T, P•Climb, Descent****••J = Jet, T = Turboprop, P = Prop* CTAS drag data includes clean, flaps, gear, and speed brakes; Thrust data includes Idle, Max Climb, Max Cruise,and Corrected; Fuel Flow includes Idle, Max Climb, and Corrected.** FACET Fuel Flow includes Corrected.*** GE FMS has extensive Drag data for the aircraft in which it is installed; Thrust and Fuel Flow data includesIdle, Max Climb, Derated Max Climb, Takeoff, Derated Takeoff, Max Continuous (Engine Out), and Corrected**** ERAM empirical altitude rates (and speeds) can be modified by observed data (Aircraft Trajectory only). +Table 22 : Trajectory data: lateral22Separated DataData at Each Predicted PointTurn DataAlongTurnStraightLat/PathTrueWay-TurnTurnTPLinesArcs X/YLonDistanceCourseHeadingpointRadiusCenterCTAS ER +Table 23 : Trajectory data: time, altitude, and speed23AltitudeSpeedsVerticalTPTimePredictedReferenceMachCASTASGroundSpeedFPACTAS ER + † † † ERAM's use of instantaneous turn modeling does not require estimate turn parameters.‡ ‡ ‡ The Delay Before Capture behavior model is a constant heading straight model. + † † † † This does not address the potential architectural benefits of having the aircraft broadcast the data, removing the need for EDA to track this data internally. + + + + +VI. AcknowledgmentsThe authors would like to thank the many people who worked to capture the capabilities of the five aircraft trajectory predictors and supported the TP comparison efforts.They include Ed McKay and Sergio Torres from Lockheed Martin, Joel Klooster and Ana Del Amo from GE Aviation, and Vincent Kuo and David Karr from Engility Corporation. +VII. References + + + +profile advisory, however, EDA would typically retain knowledge of these clearances, so from then on the downlinked descent speed profile data would be less beneficial.† † † † For ERAM, there does not appear to be any benefit to sharing this intent data.The reason is that ERAM does not accept descent speed profile constraints as inputs.ERAM uses an empirically defined model of true airspeed (TAS) variation with altitude for its descent trajectory generation.This is consistent with the environment for which ERAM was developed (where no data sharing is available), but it does not allow ERAM to take advantage of the additional shared intent data.Without rewriting the way ERAM handles descent speeds, it does not appear possible to use Intent Data sharing to synchronize ERAM trajectories for this scenario. +B. Behavior Model Data SharingIn this scenario, the aircraft are responsible for generating behavior models and downlinking them to the groundbased systems.For a description of behavior model sharing as a form of achieving interoperability, see reference 7.For the GE FMS, the behavior model for aircraft A descending to the arrival fix in LNAV/VNAV are presented in Table 11 and Table 13.For aircraft B, all three systems use the same behavior model, so no conversion is required.The rest of the discussion focuses on aircraft A.For EDA, the entire descent behavior model matches the GE FMS behavior model (Table 11), so no conversion is required.This is expected since EDA's trajectory generation was built to be "FMS-quality" and has its origins + + + + + + + NextGen next generation air transportation system: NextGen policy issues + 10.1109/icnsurv.2011.5935406 + + + + 2011 Integrated Communications, Navigation, and Surveillance Conference Proceedings + + IEEE + June 2007 + + + Joint Planning and Development Office: Concept of Operations for the Next Generation Air Transportation System, Version 2.0, June 2007. Available for public download from http://www.jpdo.gov/library/NextGen_v2.0.pdf. + + + + + Center-TRACON Automation System (CTAS) En Route Trajectory Predictor Requirements and Capabilities + + L-3Communications + + + + Version 1.10 + L-3 Communications, "Center-TRACON Automation System (CTAS) En Route Trajectory Predictor Requirements and Capabilities", Version 1.10. + + + + + Center-TRACON Automation System (CTAS) Terminal Area Trajectory Predictor Requirements and Capabilities + + L-3Communications + + + + Version 1.0. + L-3 Communications, "Center-TRACON Automation System (CTAS) Terminal Area Trajectory Predictor Requirements and Capabilities", Version 1.0. + + + + + Future Air Traffic Management Concepts Evaluation Tool (FACET) Trajectory Predictor Requirements and Capabilities + + L-3Communications + + + + Version 1.0. + L-3 Communications, "Future Air Traffic Management Concepts Evaluation Tool (FACET) Trajectory Predictor Requirements and Capabilities", Version 1.0. + + + + + Aircraft performance requirements and flight safety + + JoopWagenmakers + + 10.1201/9780429070372-27 + + + Aviation Safety + + CRC Press + + + + + Version 1.20 + GE Aviation, "GE Aviation Flight Management System Trajectory Predictor Requirements & Capabilities", Version 1.20. + + + + + En Route Automation Modernization (ERAM) Trajectory Predictor Algorithm Description and Requirements Framework + + LockheedMartin + + + + Draft, Version 1.0. + Lockheed Martin, "En Route Automation Modernization (ERAM) Trajectory Predictor Algorithm Description and Requirements Framework", Draft, Version 1.0. + + + + + Abstraction Techniques for Capturing and Comparing Trajectory Predictor Capabilities and Requirements + + RobertVivona + + + StevenGreen + + + KarenCate + + 10.2514/6.2008-7408 + + + AIAA Guidance, Navigation and Control Conference and Exhibit + Honolulu, HI + + American Institute of Aeronautics and Astronautics + 2008 + + + Vivona, R. A., Cate, K. T., and Green, S.M. "Abstraction Techniques for Capturing and Comparing Trajectory Predictor Capabilities and Requirements," AIAA Guidance, Navigation, and Control Conference, Honolulu, HI, 2008. + + + + + Cross-Comparison of Aircraft Trajectory Prediction Technologies: CTAS, ERAM, FACET, FMS + + RAVivona + + + April 30, 2009 + + + NASA NRA Contract # NNA07BB30C deliverable + Vivona, R. A., "Cross-Comparison of Aircraft Trajectory Prediction Technologies: CTAS, ERAM, FACET, FMS", NASA NRA Contract # NNA07BB30C deliverable, April 30, 2009. + + + + + European Organization for the Safety of Air Navigation (EUROCONTROL) + + AndriesStéphanie + + 10.1093/law:epil/9780199231690/e637 + + + + Eurocontrol website on BADA + + + Oxford University Press + + + 9 Eurocontrol website on BADA: http://www.eurocontrol.int/eec/public/standard_page/proj_BADA.html 10 + + + + + Introduction to Flight — Third edition. J. D. Anderson. McGraw-Hill Book Company (UK) Limited, Maidenhead. 1989. 616 pp. Illustrated. £33.95. + + JAAnderson + + 10.1017/s0001924000023022 + + + The Aeronautical Journal + Aeronaut. j. + 0001-9240 + 2059-6464 + + 94 + 937 + + 1989 + Cambridge University Press (CUP) + + + 3rd edition + Anderson, J. A., "Introduction to Flight", 3rd edition, McGraw Hill, 1989. + + + + + Review of theNational Oceanic and Atmospheric Administration (NOAA) website(http://www.noaa.gov/) + + RosannaTLindquist + + 10.1080/10496505.2014.890004 + + + + Journal of Agricultural & Food Information + Journal of Agricultural & Food Information + 1049-6505 + 1540-4722 + + 15 + 2 + + + Informa UK Limited + + + Raytheon Corporation website on ITWS + U. S. Department of Commerce, National Oceanic & Atmospheric Administration website on the Rapid Update Cycle: http://ruc.noaa.gov/ 12 Raytheon Corporation website on ITWS: http://www.raytheon.com/capabilities/products/itws/ + + + + + + diff --git a/file772.txt b/file772.txt new file mode 100644 index 0000000000000000000000000000000000000000..2065c042ea37149ac0c4f367d4ea921d2a78dbe5 --- /dev/null +++ b/file772.txt @@ -0,0 +1,300 @@ + + + + +I. Introductionecently, the global Air Traffic Management (ATM) research community has increased its focus on the conceptual design, development and use of aircraft trajectory prediction capabilities, both on the ground and in the air, to support advanced ATM concepts.In the United States, the Next Generation Air Transportation System (NextGen) identifies "Aircraft Trajectory-Based Operations" and "Performance-Based Operations and Services" as two of the eight capabilities needed to achieve its goals of increasing the safety, security, and capacity of air transportation operations. 1 Trajectory-Based Operations utilize predicted aircraft four-dimensional trajectories to "assess the effects of proposed trajectories and resource allocation plans, allowing both service providers and operators to understand the implications of demand and identify where constraints need further mitigation."Performance-Based Operations and Services provides a foundational change in the National Airspace System, defining regulations and procedural requirements in performance terms rather than in terms of specific technology or equipment, "thus potentially maximizing the value of service providers' and users' investment" in existing equipment.Similarly in Europe, the Single European Sky Air Traffic Management Research (SESAR) identifies the four-dimensional trajectory as the core of their future ATM system, with the main driving principle of "each single flight [being] executed as close as possible to the intention of its owner." 2 In both the NextGen and SESAR concepts, the sharing of trajectory information between many ATM automation systems will be necessary for successful operations. +RUnderstanding the functional and performance differences between disparate trajectory predictors (TPs) is critical for enabling the interoperability between these systems.If trajectory and aircraft intent data are to be Operational Air Traffic Control No +Table 1. Current TP Capability Documents +II. BackgroundTo date, many research and development (R&D) organizations have developed and validated their own individual trajectory prediction technologies to meet the specific needs of their automation concepts and systems of interest.With few exceptions, these efforts have been conducted with little to no leveraging of previous work from other organizations.Over the past decade, several US and European efforts to understand the differences between disparate trajectory predictors met with little success.These efforts attempted to survey and document the current and future DST requirements for trajectory prediction capabilities to identify commonalities and differences.The primary goal of the surveys was to obtain and compile this information in a way that would facilitate cross comparisons, the results of which would point to opportunities to develop "common" TP capabilities to be shared with and leveraged by the R&D community.The first effort was conducted in 1999 in the form of a 2-day workshop involving senior technical leads from MITRE Corporation's Center for Advanced Aviation System Development and NASA.The scope was limited to a small set of ATM automation applications, primarily en route, including the Center TRACON Automation System (CTAS), 8 the User Request and Evaluation Tool (URET), 9 and envisioned enhancements to both.The results of the workshop, captured in an annotated briefing, included a side-by-side comparison of the major TP capabilities of each application highlighting similarities and differences.The comparison also included a summary of the major drivers behind the development of the capabilities.This exercise did provide an initial understanding of both the high-level similarities and differences in TP capabilities for these applications, as well as some insights regarding how to compare TP requirements and capabilities.However, the details were at too high of a level to point to significant opportunities for common capabilities.A major impediment was the lack of documentation of the salient details related to each TP.The second major effort was a joint activity conducted under the auspices of the US-Europe ATM R&D Action Plan 16 (AP16) for Common Trajectory Prediction Capabilities.AP16 is a team of senior TP experts representing the Federal Aviation Administration (FAA), NASA, Eurocontrol research labs, and major industry R&D organizations developing air traffic control and airborne (e.g., avionics and airframe) automation systems.The team conducted a broader survey requesting the details of their TP requirements and capabilities (in any form) from approximately twenty research and operational organizations.The request introduced the survey and described the specific aspects of TP for which information was needed.Key TP-related technical leads were contacted directly through a formal letter of request, email, and follow-up phone calls.After a year of effort, many indicated they had nothing to offer.Of the few that did respond with relevant details, the content varied widely from one organization to another, the material was inconsistent in what was documented (little comparable overlap), and the scope of information was significantly incomplete.Several key conclusions were generated based on feedback from these past efforts and analysis of the limited material obtained.First, documentation of this type is a systemic challenge for the community, particularly the research labs where the changes in requirements and capabilities can occur quite frequently.Second, the focus of most developers of automation concepts/systems seems to be limited to higher-level requirements for their automation system.Many TP clients have difficulty in specifically determining what their automation concept/system requires from its supporting TP capabilities.Third, the wide range of approaches taken to implement TPs, for research and operational systems, makes it difficult to understand the similarities (and true differences) of two or more TPs.For significant progress in the development and reuse of TP capabilities across the R&D community, a simple and concise methodology is needed to "standardize" the documentation of clear, consistent, complete, and cross-comparable TP requirements and capabilities.A key element of this methodology is the development of a document structure (i.e., the Requirements Framework) that enables effective capture of these TP requirements and capabilities.The documenting and comparison of the five TP technologies described in Table 1 is an initial attempt to develop such a document structure. +III. Lessons LearnedThough a formal comparison of the five TP capabilities documents was not yet scheduled, an opportunity to perform an initial comparison of four of the documents existed because these documents were all being developed by a single team of contractors (L-3 Communications and GE -Aviation).After the completion of early drafts of these four documents, this basic comparison of the documented capabilities was performed.Though the current structure of the Requirements Framework did support this basic comparison, differences in techniques for describing capability details impeded the ability to compare the capabilities.The main result of this initial comparison was the identification of an overarching requirement needed to perform a successful comparison: the TP capabilities documents needed to be written in a more generalized way and less from a TP implementation perspective (i.e., describing how the specific TP capabilities were implemented).To successfully compare TP capabilities using the original, non-generalized documents, the person performing the comparison was required to abstract or generalize the capability descriptions for each TP to try to identify commonalities and/or differences.The problem was that no single person could be expected to have enough specific knowledge of all of the TPs to effectively perform this abstraction.Therefore, the only realistic approach to performing the comparisons was to have the TP capability document developers themselves rewrite their capability descriptions using abstracted concepts and language.Since these documenters are experts for their specific TP, they would have the required knowledge to successfully perform such an abstraction.An additional benefit of performing this abstraction in the original documents is that the capability documents would now be more easily understood by a wider audience.To maximize the chances that the abstraction of the various TP capabilities by the different TP experts would result in easily cross-comparable capability descriptions, a common set of abstraction techniques to be used by all of the TP experts was developed.These techniques are described next. +IV. Abstraction TechniquesIn the following sections, each of the three main abstraction techniques developed is described.The description starts with a discussion of the issue that the specific technique is designed to resolve and then provides a description of the technique itself.For each technique, the resultant benefits to the development of the Requirements Framework and to efforts beyond the documentation and comparison of TP capability documents are also described. +A. Separating Behavioral from Mathematical ModelingFigure 1 illustrates a conceptualized model for the typical minimum set of inputs and outputs of a trajectory predictor (TP).In the model, the TP accepts aircraft initial condition data, represented by the state input.The state data is typically generated by sensors within the TP client application's operational environment (e.g., radar track data for ground-based systems or GPS data for airborne systems).The TP also accepts a wide range of data that may impact the future path of the aircraft, represented by the intent input.Intent data can include any type of information available to the TP client application and varies greatly based on the client application's operational environment.Examples include: flight plan data, controller issued altitude and time constraints, and aircraft guidance mode settings.Intent data is often equated with constraint data, but any data that can impact the future path of the aircraft is acceptable intent data, including pilot and/or controller preferences and objectives.The final input data is atmospheric data, which is a model of the airspace's winds, temperature and pressure aloft.The output of the TP is a 4-dimensional (4D) trajectory representing the predicted future states of the aircraft.The 4D trajectory output starts at the input state location and predicts an aircraft's future path that is consistent with the input intent data.It is the responsibility of the TP to convert the state and intent input data into a set of mathematical representations that can be numerically integrated to generate the output 4D trajectory.Each TP implements this process differently, attempting to maximize efficiency and effectiveness given the specific constraints of its application environment (e.g., available input data, implemented mathematical equations of motion, required computational performance).To identify common capabilities between TPs whose implementations vary significantly, a generalized approach for describing the modeling capabilities of a TP was created and used as the basis for abstracting each TP's capabilities.Figure 2 shows this generalized approach, which separates the details of the predicted aircraft maneuvers from the numerical integration approaches used to predict these maneuvers (i.e., generalizing the description of the TP process for turning state and intent data into mathematical representations). +Trajectory Predictor +State +Figure 2. Generalized Approach for Generating the Mathematical Representation.In Figure 2, the state and intent inputs are first converted into a behavioral model for the aircraft.As described above, the intent represents the known constraints, objectives, etc. of the aircraft.The behavioral model is the TP's representation of how the aircraft (pilot, guidance system, etc.) responds to those constraints and objectives.In other The mathematical model defines the mathematical representation of each maneuver defined in the behavioral model; the mathematical model describes how each maneuver will be integrated, including the required equations of motions, integration time step, integration algorithm, etc. Effective behavioral modeling is done in layers, describing aircraft maneuvers in ever-increasing levels of detail.Figure 3 illustrates an example of vertical behavior modeling for an aircraft whose initial state is not at its cruise altitude.The input data to the TP includes the aircraft's initial altitude (part of the state data) and the cleared cruise altitude (part of the intent data).From this information, the TP models two maneuvers representing the top level (layer) of the aircraft's behavioral model: the aircraft will start in a CLIMB maneuver and transition to a CRUISE maneuver when reaching the cruise altitude at the top of climb (TOC).The TP may require defining the CLIMB and CRUISE maneuver models in more detail (e.g., to achieve its functional and performance requirements).In Figure 4, the CLIMB maneuver has been expanded into a series of sub-maneuvers.The aircraft is now modeled as starting in an IAS CLIMB sub-maneuver, where the aircraft is maintaining maximum climb thrust and an indicated airspeed (IAS).This sub-maneuver transitions into a MACH CLIMB sub-maneuver at the Mach/IAS transition point, where now maximum climb thrust and a Mach number are maintained until reaching the TOC.These two sub-maneuvers could be further broken down into smaller maneuvers if further modeling detail were required.For example, if an acceleration segment from the initial IAS to the target climb IAS was required for the IAS CLIMB sub-maneuver or if a "flare" maneuver near the TOC was required for the IAS Mach maneuver.§ The behavior is the result of attempting to achieve the constraints & objectives in the intent because the aircraft may not be able to meet all of these constraints and objectives.Each maneuver in the top-level behavior model should be broken down into smaller sub-maneuvers until the desired level of maneuver detail is reached.Once this is completed, then a mathematical model can be defined for each of the lowest level sub-maneuvers.Table 2 illustrates a possible final behavioral and mathematical modeling for the CLIMB maneuver.For each mathematical model, a complete set of information must be provided to enable the required numerical integration.Table 2 provides a subset of the required information for a mathematical representation that uses point-mass equations of motion with a kinetic aircraft performance model.¶ +Top of +Behavioral Model Mathematical Model +Maneuver +Impact on the Requirements FrameworkSeveral major benefits are achieved by separating behavioral and mathematical modeling in the documentation of capabilities within the Requirements Framework.One benefit is that describing the behavioral modeling of a TP is a direct form of abstraction that supports more effective comparisons.Aircraft behavior is, by definition, independent of any specific TP.The aircraft will perform some series of maneuvers based on its objectives and constraints, independent of whether there are one or more TPs trying to predict its behavior.A TP's behavioral modeling description is therefore directly comparable to any other TP's behavioral modeling, since they are all trying to model the TP independent aircraft behavior.Each TP may model this behavior to a different level of detail, or some TPs may not be able to predict all maneuvers due to a lack of input intent information, but these are exactly the types of differences the comparison is attempting to identify.Differences in mathematical representations are ¶ An aircraft performance model (APM) is a subset of the mathematical model, providing a mathematical representation of just the aircraft's dynamics.Kinetic APMs typically provide either altitude rate (climb or descent rate) and acceleration data or thrust and drag values from which altitude rates and accelerations can be calculated.also important, but behavioral modeling capabilities are much more likely to represent potential functional requirements for future TPs.For example, the type and detail of aircraft maneuvers required by the TP are much more likely to be defined as a functional requirement than whether the TP needs to use point-mass equations of motion with a kinetic aircraft performance model (which is more likely to be a design approach to meet a performance requirement).Another benefit of documenting behavioral modeling separate from mathematical modeling, more specifically through the use of layered modeling descriptions, is the ability to evaluate very complex, detailed behavior modeling logic at different levels of detail.For example, the behavior model can be initially evaluated at only the highest level, as in Figure 3, describing just major maneuvers (climb, cruise, descent, etc.) without the complication of evaluating detailed modeling for each of these maneuvers.This enables TP comparisons to identify which major maneuvers the TPs have in common and which maneuvers are unique to one or the other TP.Then, for each common major maneuver, differences in modeling details can be evaluated.This method for comparing TP behavioral modeling, enabled by the layered description approach, provides additional enhancements to the comparisons:1. Two TPs with significantly different levels of behavioral modeling can be easily compared 2. Two TPs with complex behavioral and mathematical modeling can clearly identify if differences existed in the maneuvers predicted, the level of detail of the maneuver modeling, or in the mathematical representation of these maneuversIn the case of two TPs with significantly different levels of behavioral modeling detail, both TPs will typically use the same categories for their major behavior modeling.For all TPs, the major behaviors revolve around the conventional description of aircraft phases of flight: climb, cruise, and descent.If both TPs are modeling a climb maneuver, even if they differ greatly in the level of detail with which they model a climb (e.g., a single climb maneuver versus the two sub-maneuver behavioral model described in Table 2), they both typically identify the major maneuver as a climb.Therefore, the comparison can successfully identify whether (and when) the two TPs predict climb maneuvers.This comparison could be difficult without the layered approach to behavioral modeling since the detailed behavior models and underlying mathematical representations can be very different.Since the layered approach to behavioral modeling defines maneuvers at a high level and then drills down into lower levels of detail for each maneuver, two TPs with complex behavioral modeling can be compared to identify the level of detail at which the two approaches diverge.The difference may occur at the top-most behavioral modeling level (e.g., one TP does not predict climbs), at some intermediate level of behavioral modeling (e.g., one TP does not model constant IAS and constant Mach sub-maneuvers), or at the mathematical modeling level (e.g., one TP uses a kinetic aircraft performance modeling approach and the other uses a kinematic modeling approach).This ability to compare disparate TP capabilities is enabled by breaking up the complex behavior modeling into smaller, easier to evaluate component behaviors. +Impacts Beyond the Requirements FrameworkThe separation of behavioral modeling from mathematical modeling has additional benefits beyond those described above for enhancing the Requirements Framework.First, by describing the behavior modeling independent of the mathematical representation, the assumptions and approximations of the prediction model can be fully evaluated.This is a useful process when designing new TP capabilities (e.g., adding a new maneuver to an existing TP) or when evaluating pre-existing TP capabilities, as in the context of identifying potential modeling improvements to meet required performance requirements during validation.Second, as described above, behavioral model representations of aircraft maneuvers are independent of TP implementation.This makes this representation of predicted aircraft behavior an excellent candidate for transferring data between disparate TPs to enable successful interoperability.Each of these benefits is described in more detail.Modeling a New Maneuver Figure 5 and Table 3 illustrate an example of using behavioral and mathematical modeling during the design of a new maneuver to add to an existing TP.In this example, the maneuver being modeled is the aircraft's guidance system response to adding a STEP climb constraint to an FMS route, which causes the aircraft to climb to the new cruise altitude (target altitude) defined in the STEP climb constraint.The left column of Table 3 describes the (simplified) results of a theoretical discussion with a guidance expert on how the guidance system works when given a STEP climb constraint.Given the behavior description, a high-level STEP CLIMB maneuver can be divided into two sub-maneuvers: CAPTURE MACH and CONSTANT MACH CLIMB.Furthermore, evaluation of the actual aircraft behavior identifies that the finite time for initial throttle and elevator deflection has a negative impact on the predicted maneuver's altitude accuracy if not accounted for in the behavior model, so the CAPTURE MACH sub-maneuver should be divided into two lower-level sub-maneuvers: SPOOLUP and SPEED CAPTURE.Therefore, the definitions for the behavior model's three lowest-level submaneuvers are as follows:• SPOOLUP: Elevator is deflecting based on an undefined control law to capture the target Mach number.The autothrottle is moving to capture maximum climb thrust.• SPEED CAPTURE: Elevator is deflecting based on an undefined control law to capture the target Mach number.The autothrottle is maintaining maximum climb thrust.• CONSTANT MACH CLIMB: Speed is being maintained at the target Mach number.Throttle is being maintained at maximum climb thrust.If it is assumed that the TP uses point-mass equations of motion for all of its mathematical models, then attempting to define the aircraft behavior for the SPOOLUP sub-maneuver at a lower level of detail is not desirable since the ability to model elevator deflection does not exist in the mathematical model.A simplified mathematical model (a simple lag) could be used for this sub-maneuver if the main purpose for this model is to initiate the SPEED CAPTURE sub-maneuver at approximately the correct along-path location (to eliminate the altitude error bias over the entire STEP CLIMB maneuver that exists if the SPEED CAPTURE maneuver is initiated immediately at the STEP constraint location).For the SPEED CAPTURE sub-maneuver, maintaining the throttle at maximum climb thrust can be modeled directly by a kinetic aircraft performance model, but again, the elevator deflection can not be modeled directly.In this case, the mathematical model could use a constant vertical rate at the minimum 500 fpm if this is a close enough approximation of the aircraft's actual vertical rate.The CONSTANT MACH CLIMB submaneuver, which should cover the majority of the STEP CLIMB maneuver, can be mathematically modeled directly using point-mass equations of motion and a kinetic aircraft performance model.If during validation the STEP CLIMB maneuver does not meet all of the TP's performance requirements (e.g., altitude prediction accuracy), both the SPOOLUP and SPEED CAPTURE behavioral and mathematical models should first be re-evaluated for acceptable prediction accuracy since they contain the most significant approximations. +TP InteroperabilityA critical element of TP interoperability is to share data between systems, both on the ground and in the aircraft, in a way that can be correctly interpreted by the disparate TPs used by these systems.In Europe, the REACT consortium was created to elicit requirements for a standard language to exchange trajectory-related information between ATM systems 10 , but efforts similar to this in both Europe and the United States are still attempting to define the data content that can be unambiguously shared between a wide range of ATM systems.Behavioral modeling, as described above, presents one possible answer to defining this shared data.Because the behavioral model is TP independent, it can safely be shared among disparate TPs without losing its ability to be correctly interpreted.When received by a TP, the behavioral model can be converted into an associated mathematical model that is specific to the receiving TP.Each receiving TP will interpret the behavioral model at a level that is consistent with the mathematical modeling capabilities of that TP.For example, using the STEP CLIMB maneuver described in Figure 5 and Table 3, the behavior model can be described to as low a level of detail as is required by the conceptual objective for sharing the data.# Using the theoretical TP from Table 3, if the behavior is modeled to a level that includes elevator deflection and throttle position, this TP will only interpret the behavior model to a level consistent with its point-mass (mathematical) modeling assumptions, ignoring the additional behavioral modeling details.If a second TP of significantly less fidelity (e.g., a TP used for long-range traffic flow management applications) also receives the behavioral modeling data, it may only interpret the behavioral model at the highest STEP CLIMB level.In both cases, the same maneuver has been consistently interpreted by the disparate TPs, which is the required objective.If a third TP were to require data beyond that contained within the shared behavioral model, then this TP would need to model the unknown behavior details based on internal processing, similar to how current day systems perform "intent modeling" when missing critical intent data. 11 +B. Client Application -TP Boundary DefinitionIdentifying the set of TP capabilities within a given client application would appear to be a straight-forward exercise, but the definition of the boundary between client application and TP capabilities is often blurred through close integration of these capabilities in the application's implementation.For example, a ground-based DST may add an outer loop within the implementation of its "trajectory predictor" that iterates on a trajectory degree-offreedom (e.g., cruise speed) to meet a time constraint at a lateral waypoint.This implementation may take advantage of the meet-time algorithm's close integration with other TP capabilities to efficiently and effectively find a value for the degree-of-freedom that meets the time constraint, but does this make the meet-time capability a part of the client application's set of TP capabilities?This may at first appear to be an academic question, but an answer to this question is necessary to define the scope of TP capabilities captured within the Requirements Framework and hence, define the ultimate scope of the final comparisons.The first aspect of the Requirements Framework that defines a conceptual boundary between the TP and its client application is the use the AP16 Common Trajectory Predictor Structure (reproduced in Figure 6) as the Framework's outline. 12This structure has four process components:1. Preparation Process -responsible for processing the state, intent and other inputs from the client application into a mathematical representation.2. Trajectory Prediction Process -responsible for numerical integration of the mathematical representation to create an output 4D trajectory 3. Trajectory Update Process -responsible for internal TP iteration to meet all input constraints 4. TP Export Process -responsible for formatting and processing of the final 4D trajectory before sending to the client application # The required data content will be ATM concept specific, but the approach described should be valid for any level of data content required. +Figure 6. AP16 Common Trajectory Predictor StructureThis structure was developed by the AP16 core team to conceptually cover all potential descriptions of trajectory prediction capabilities, with the full knowledge that the conceptual boundary between the TP and the client application is an on-going subject for debate.The Trajectory Prediction Process, the inner most process in the Common structure, is the one process that is generally accepted as being required within the TP.The other three processes all include some processing that could conceptually be considered either within the TP or within the client application.For example, the meet-time algorithm described above, if considered a TP capability would exist within the Trajectory Update Process.Therefore, the Common TP structure was necessary, but not sufficient for clarifying the client application -TP boundary when documenting the TP capabilities in the Requirements Framework.An additional level of conceptual modeling was required.A general rule of thumb (referred to as the TP Boundary Rule) was applied to further distinguish, for the sake of documenting the TP capabilities within the Requirements Framework, the conceptual boundary between the client application and TP capabilities: if the capability directly supports a key function of the client application, then the capability is considered a client application capability and outside the scope of the TP.Therefore, for the groundbased meet-time algorithm discussed above, if the algorithm directly provides clearance advisories to the client application user (e.g., a sector controller), then this would be considered outside the scope of the TP since this capability directly supports a function of the client application.The TP Boundary Rule is useful for deciding whether a capability in one of the three "outside" processes of the Common TP Structure (Preparation Process, Trajectory Update Process, and TP Export Process) should be included in or excluded from the TP capabilities description.The handling of constraint relaxation, a common capability that falls within the grey areas of being considered a TP or client application capability, provides an example of this decision making process.Constraint relaxation is the process of identifying and changing or removing (i.e., relaxing) the requirement to achieve a particular constraint if all of the constraints on the aircraft can not be simultaneously achieved.A classic example of constraint relaxation is when a required descent to an altitude constraint can not be achieved within the current performance limits of the aircraft.Figure 7 illustrates this situation.The identification that the altitude constraint can not be achieved is a capability of the TP since it requires a prediction of the required top-of-descent (TOD) location.The question is whether the relaxation of the altitude constraint should also be considered a TP capability or whether it should be considered a client application capability. +Altitude ConstraintPredicted TOD +Relaxed Constraint Altitude ConstraintPredicted TOD Relaxed Constraint +Figure 7. Constraint Relaxation ExampleUsing the TP Boundary Rule, this capability should be considered a TP capability only if it does not directly support a function of the client application.The following are two possible client application scenarios, one where the relaxation should be considered a client application capability and the other where it should be considered a TP capability:1. Client Application Capability Scenario: After identification that all constraints cannot be met, an algorithm is performed to identify that the altitude constraint should be relaxed to meet a client application objective (e.g., orderly traffic flow).2. TP Capability Scenario: After identification that all constraints can not be met, an algorithm runs to modify the pilot procedure to achieve the altitude constraint as closely as possible.The client application displays the amount by which the altitude constraint was not met to the user.In the client application capability scenario, the identification of which constraint should be relaxed is tied directly to specific client application functionality.Because the selection of the altitude constraint for relaxation is the direct result of obvious client application functionality (specifically, the functionality that creates an orderly traffic flow), this capability is considered part of the client application and not the TP.In the TP capability scenario, no client application functionality requires control over which constraint is relaxed, because the client application only displays which constraint(s) will not be met and by how much based on an assumed pilot procedure.In this scenario, the client application is not using the relaxation of constraints to directly support its functionality, so the capability can be considered part of the TP. +Impact on the Requirements FrameworkImproving the definition of the conceptual boundary between the TP and the client application has significant impacts on the capabilities captured within the Requirements Framework.First, the definition of this boundary directly defines the state, intent and other input data to the TP.This boundary also directly impacts the definition of the capabilities included in the Preparation Process, since this is the process that handles input data.Identification of which capabilities should be included in both the Trajectory Update Process and TP Export Process are also impacted.Because the TP Boundary Rule is simple enough to be consistently applied by the various TP capability documenters, a level of consistency between the TP capability documents was achieved beyond that from just using the Common TP structure as the outline of the Requirements Framework. +Impacts Beyond the Requirements FrameworkThe TP Boundary Rule described above does not eliminate all ambiguity in defining what processing should be included in the Common TP Structure.Any capability that directly supports a key function of the client application can be safely considered outside the scope of the TP, but it is not correct that all other capabilities are within the scope of the TP.This can be best illustrated with an example.Assume that a ground-based DST is predicting the behavior of an aircraft that has just been issued a requiredtime-of-arrival (RTA) constraint.From the ground-based DST perspective, the aircraft guidance system's algorithm that iterates on FMS cost index to meet the RTA constraint does not have a direct impact on the DST's functionality.So the temptation might be to consider the capability to predict this algorithm's impact on aircraft behavior as a TP capability (i.e., this algorithm is considered part of the Trajectory Update Process in Figure 6).The problem with this conclusion is that including this type of capability inside the TP can, for more complex scenarios, require adding capabilities to the TP that are out of scope for TP processing.For example, take the same scenario, but now assume that the aircraft being predicted is a fully autonomous aircraft that has airborne self-separation assurance (ASAS) equipment on board.This new aircraft would include the ability to create a conflict-free path that meets the RTA constraint, possibly using both lateral and vertical degrees of freedom.The capability to predict this aircraft's behavior would require emulating a complex meet-time algorithm integrated with conflict detection and resolution capabilities.Including this level of capabilities, with the required addition of other aircraft trajectories as TP input data, is outside the scope of trajectory prediction capabilities.This issue illustrates that the conceptual boundary between the TP and the client application may benefit from an additional layer of processing.As illustrated in Figure 8, a new process (the Trajectory Management Process) is inserted between the client application and the TP process (which includes all of the processes in the Common TP Structure, Figure 6).This new process is responsible for capabilities outside of the client application that predict aircraft behavior, but that are also considered outside of the scope of TP capabilities.This conceptual structure solves the issues identified in the scenario of the ground-based DST attempting to predict the trajectory of an autonomous aircraft with ASAS equipment.In the new structure, all of the aircraft's algorithms associated with creating a conflict free trajectory that meets the RTA constraint, including conflict detection and resolution, can be added to the Trajectory Management Process.This removes the need to include these capabilities within the Common TP structure, while still enabling the client application to consider an Aircraft Prediction Process (i.e., the combination of the Trajectory Management Process and the TP) as separate from the client application. +C. Use of Abstracted Instead of Domain-Specific LanguageAn additional impediment to the effective comparison of TP capabilities is when capability descriptions use language that is too closely tied to the domain of the client application.This domain-specific language can cause difficulty in identifying common capabilities, since TPs from very different client application domains (e.g., an airborne TP versus a ground-based TP) can use different names for similar predicted behavior.The use of language specific to the implementation of a TP can cause a similar issue, since TPs can implement the same capability in different ways.The solution to these problems is to use abstracted terminology that generalizes the capability being described.Figure 9 and Table 4 illustrate this situation by describing a common aircraft maneuver in airborne-specific, ground-specific and abstracted language.The abstracted description describes the series of aircraft maneuvers in domain independent language.** The aircraft performs an immediate turn to a pre-defined heading and then follows this heading until capturing (lateral distance within 2 nmi.) a defined route.Once the route is captured, the aircraft ** All three descriptions are related to behavioral modeling, but the use of domain independent language is a separate issue from the behavioral modeling discussion above.turns until capturing a ground track that will provide a constant track path directly to the first turn waypoint in the route ahead of the capture location (note this is a different turn definition than the previous turn).The aircraft maintains this ground track until performing the turn at the first turn waypoint.After this turn, the aircraft follows the remainder of the route (as pictured).The airborne-and ground-specific descriptions exhibit both a lack of proper detail and design abstraction in describing the expected maneuvers.The airborne description defines the aircraft's lateral maneuver in terms of airborne guidance mode capabilities.Heading Select is an aircraft guidance mode that will perform a turn until capturing a target heading and then will maintain this heading until the guidance mode is changed.In this case, there is an allusion to the use of "LNAV armed", which is a state of the guidance system that will cause the aircraft to capture and follow the route (in an FMS, this route is called the active route) when the aircraft gets within 2 nmi. of this route.The Lateral Navigation (LNAV) mode follows the route when engaged.Without knowing how airborne guidance systems work, it would be difficult to identify that the abstracted description is equivalent to the airborne description.Similarly, the ground description uses ground-based language, namely the reference to a heading clearance and the following of a flight plan route.The abstracted description not only provides clear language for describing the development of the aircraft's lateral path, it also forces the documenter to identify and describe modeling details (such as the development of the transition from the constant heading segment onto the route via capturing the first waypoint ahead of the aircraft when the route is captured). +Impact on the Requirements FrameworkThe major impact on the Requirements Framework of using abstracted language is an increased ability to correctly identify similarities and differences in TP capabilities.An additional benefit is the documenter's enhanced ability to identify an appropriate level of detail to describe the TP capability.As future TP capabilities are documented and compared, an opportunity exists to identify common abstracted terms that can be reused by future capability documents to enhance the ease with which these new capability documents can be compared against existing documents. +Impacts Beyond the Requirements FrameworkThe definition of common terms was a major objective of early AP16 activities. 13The identification of common terms is a critical component of enabling TP experts from different client application domains to discuss and share information about trajectory prediction techniques and technology.The development of abstracted language to enable enhanced comparisons of TP capabilities will provide many abstracted terms that are perfect candidates to be added as common terms to the AP16 list of Common TP terminology. +V. Additional Benefits/Issues Uncovered by Abstraction +A. Library of Common Trajectory Prediction FunctionsOne of the key insights described above is the decomposition of intent into the basic elements of aircraft behavior and the mathematical representation of those behaviors.Such decomposition allows a broad range of mathematical models to be considered independently from the actual behavior.This approach would enable TP developers to identify and collect a broad variety of mathematical models (or develop new ones) with standardized interfaces.A common set of mathematical models would enable developers to compare and contrast the advantages and disadvantages of specific models for representing the aircraft behaviors that are key to the success of their application.Similarly, developers could identify and collect (or define new) aircraft behaviors that are key to meeting the prediction accuracy requirements for specific applications.The creation of these sets lends itself to the development of common libraries of aircraft behaviors and mathematical-modeling functions.TP developers can then leverage these libraries to improve their trajectory predictor or to create a new one.A compelling analogy may be drawn from the Object-Oriented (OO) methodology in software engineering where the concept of "abstraction" is fundamental to the methodology.Consider aircraft behavior as a type of "object" representing a set of intended maneuvers.OO development defines what an object is and does before deciding how it should be implemented.Internally, objects are comprised of specific functions that represent how the object responds (its behaviors) to changes in state.Externally, objects are defined by an external interface.Any object can then utilize and build upon other objects simply through the use of the external interface established for each object.This allows the detailed implementation of each object to remain "hidden" and avoids the complexity internal to an object's implementation.Objects (and their internal functions) can then be grouped and maintained within libraries that can be accessed by a broad range of applications.Aircraft behaviors (trajectory maneuvers and sub-maneuvers) can be viewed in a similar manner, e.g., as objects.Mathematical models correspond to the functional implementation of an object.Libraries of standard aircraft behaviors and mathematical modeling functions could be created and used by one or more TPs supporting a variety of different decision support automation systems and concepts.Standardization would require clear definitions of the inputs to each model function and corresponding outputs.By separating the behavioral model from the mathematical model, there is a strong potential for reuse of TP modeling functions developed in the future if they are developed with a common interface.Different mathematical models could then be interchanged within a particular behavior model.If a TP had originally been developed to use one particular mathematical modeling approach (e.g., using a kinematic aircraft performance model), the results of using an alternative approach (e.g., a kinetic aircraft performance model) could be analyzed by swapping those functions with others from the library.When selecting the appropriate mathematical models for a new TP application, a developer would consider such factors as the available input data and the required level of prediction accuracy.Lack of input data required to use the particular model needed to achieve required prediction accuracy would demand that additional information be made available.This would provide significant insight into what should be represented in a standard language for trajectory exchange.With respect to the specific TPs studied in this paper, it is interesting to note that several were developed using object-oriented languages and many of the behaviors were encapsulated within the source code.While an OO implementation of a TP can facilitate the abstraction of aircraft behaviors and corresponding mathematical models, it does not ensure extensibility to a library of common capabilities for multiple TPs.The logic and objects would not be compatible across (and interchangeable between) the separately developed TPs.As such, the functions and objects within the study TPs are not immediately transferable to a standard library.However, to be extensible, a full abstraction of the aircraft behaviors and mathematical models is needed to ensure that they are compatible across (and interchangeable between) multiple TPs.Future TP development activities could benefit significantly from the creation and leveraging of such common capabilities.Potential benefits include reductions in the efforts necessary for TP validation and overall life-cycle costs. +B. Trade-Off between Compatibility of Disparate TPs and Individual TP Prediction AccuracyThe lessons learned from abstraction also lend themselves to help answer an important question related to the definition of TP requirements.While trajectory prediction accuracy is a significant concern for many of the advanced NextGen and SESAR concepts, the system needs for interoperability of TP-based automation systems are equally important.The interoperability of automation systems (e.g., air and ground) is a function of the synchronization of the supporting TPs, among other things.In other words, two different automation systems may develop/advise conflicting decisions if their supporting TPs are not synchronized well enough to result in "similar" predictions (i.e., common situational awareness).The need for synchronization of predicted trajectories can conflict with the TP accuracy needs of any one system.In terms of TP requirements, this leads to the question of which is more important: the compatibility of functional requirements for two or more TPs, or the accuracy of any one TP?If the TP for one client application has access to (e.g., weight, thrust, flap schedules) or mathematical models that improve TP accuracy but cannot be shared with another client application/TP for which synchronization is critical, there may be significant differences in the trajectory predictions supporting each client application.This could pose an issue for an operational concept that depends on the sharing of trajectory information.This would be a case where synchronization (resulting in similar trajectory predictions across the two systems) may take precedence over a higher level of modeling accuracy in either system.Here accuracy refers to the degree of conformance between the predicted position and/or velocity of a vehicle (e.g., aircraft) at a given time and its true position and/or velocity. 13The earlier section discussing TP interoperability explains how the behavior models can be used to define the data that must be shared to achieve synchronization.If one TP does not have the capability to reconcile a particular mathematical model but is generating the trajectory prediction that is critical to the concept, it may require that its mathematical model be used for synchronization.If the higher fidelity mathematical model produces the critical trajectory, it may put a new requirement on the other TP to adopt its capabilities.The question "which client application's trajectory is more correct" is entirely dependant on the operational concept defining the roles of each client application and the context for how each client application's trajectory prediction will be used.This also illustrates how key functional requirements for a TP for one client application (that must be interoperable with another) can be identified through the cross comparison and analysis of the types of trajectories that must be predicted by both client applications.The level of sameness necessary for the trajectories will dictate the level of sameness in the abstraction of the behavior and mathematical models.This may drive the specification of requirements down to the mathematical model level where critical similarities and differences can be identified.The abstraction process can be used as an aid in this process.A comparison of the detailed modeling for a behavior, such as that done in Table 3, for the individual TPs could identify whether the differences in mathematical models may have a critical effect on compatibility.If this is examined during the early stages of development, it would save time and cost by uncovering potentially hidden requirements. +VI. Concluding RemarksThe abstraction techniques provided several significant benefits.Most importantly for the TP capabilities documentation effort, they improved the ability to perform effective comparisons between documented TP capabilities.This benefit is crucial for meeting the overall objective of defining a Requirements Framework that is clear, consistent, complete and cross-comparable.In addition, the abstraction techniques have led to the development of enhanced TP conceptual modeling and approaches for identifying TP common terms.Finally, additional benefits to the development of common trajectory prediction functions and defining the trade-offs between the compatibility and accuracy of disparate TPs were discovered.These wide-ranging results will support U.S./Europe activities in defining Common TP approaches, including the development of common validation methodologies and defining data sharing requirements for disparate TP interoperability.The next step in developing the Requirements Framework will be to perform a formal comparison of the five TP capabilities documents from Table 1.The objective of the first comparisons will focus on identifying commonalities and differences between the capabilities described in these documents.Differences will be evaluated to determine whether they represent real differences in TP capabilities or whether the difference is caused by a deficiency in the Requirements Framework.For example, it is possible that a lack of specificity in the Requirements Framework may cause a failure to identify that a particular capability needs to be captured.Another example is when a deficiency in the Requirements Framework causes two documenters to add the same capability to different sections of the Requirements Framework, resulting in an inaccurate comparison.These deficiencies in the Requirements Framework will be identified and resolved until the existing TP capabilities documents can be clearly and consistently cross-compared.After completing the comparisons of the existing TP capabilities documents, additional TP technologies from a wider range of operational domains will be targeted for documentation and comparison with the original set of documents.The application to a wider range of TP technologies should help to continue the development of the Requirements Framework until it is sufficiently broad in its scope to cover the desired range of future TP requirements and capabilities.Ultimately, the Requirements Framework will be applied to define requirements for a new, advanced TP being developed for one of the future ATM concepts being proposed in the United States and Europe.Through the successful analysis of existing TP capabilities, the result should be a clear, consistent, complete and cross-comparable documentation of these TPs requirements.Figure 1 .1Figure 1.Trajectory Predictor Inputs and Outputs. +Figure 3 .3Figure 3. Behavioral Modeling Example: CLIMB Transitioning to CRUISE Maneuver +Figure 4 .4Figure 4. Behavioral Modeling Example: Expanded CLIMB Model +Figure 8 .8Figure 8. Addition of Trajectory Management Process outside Common TP Structure +Figure 9 .Table 4 .94Figure 9. Example of a Lateral Aircraft Maneuver for Illustrating use of Abstracted Language +is a model of the aircraft's behavior while attempting to meet the constraints and objectives represented in the intent data.§ The second step is then to convert each element of the behavioral model into a mathematical model.words, itState State StateBehavioral Behavioral BehavioralMathematical Mathematical MathematicalIntent Intent IntentModel Model ModelModel Model Model +Table 2 . Behavioral and Mathematical Models for Example CLIMB Maneuver2Sub-ManeuverEquations of MotionSpeedAircraft Performance ModelIntegration DirectionIntegration AlgorithmTimestepCLIMBIAS CLIMB MACH CLIMBPoint-Mass: Known Speed & Thrust Point-Mass: Known Speed & ThrustConstant IAS Constant MachKinetic; Thrust from Max Climb Table Kinetic; Table Thrust from Max ClimbForward Forward2 nd Order Runga-Kutta 2 nd Order Runga-Kutta60 seconds 60 seconds +Example of Adding a New Maneuver using Behavioral & Mathematical Modeling.Target TargetSTEP STEPAltitude AltitudeConstraint ConstraintCONSTANT MACH CLIMB CONSTANT MACH CLIMBSub-Maneuver Sub-ManeuverSPOOL SPOOLSPEED CAPTURE SPEED CAPTURESub-Maneuver Sub-ManeuverSub-Maneuver Sub-ManeuverFigure 5. Behavior Model Behavior Description Maneuver Sub-ManeuversMathematical ModelAt the step constraint, the autothrottle drivesSPOOLUPConstant Altitude and Speedthe throttle to maintainmaximum climb thrust. The elevator is deflected to capture the target Mach number. Minimum climb rate is 500 fpm.STEP CLIMBCAPTURE MACHSPEED CAPTUREConstant Vertical Rate (500 fpm) and Max Climb ThrustAfter capturing thetarget Mach number,the Mach number is maintained whileCONSTANT MACH CLIMBConstant Mach and Max Climb Thrustclimbing at maximumclimb thrust. +Table 3 . Example of Adding a New Maneuver using Behavioral & Mathematical Modeling3 + American Institute of Aeronautics and Astronautics + + + + +VII. AcknowledgementsThe authors would like to express gratitude to the developers of the TP capabilities documents, David Karr (L-3), Vincent Kuo (L-3) and Joel Klooster (GE -Aviation), for all of their input during the development of the abstraction techniques.The authors would also like to thank Lockheed Martin, in particular Rich Smolen, Sergio Torres, and Ed McKay, for their continued efforts in supporting the development of the Requirements Framework.Finally, the authors would like to recognize their colleagues in the AP16 group for all of their efforts in the advancement of Trajectory Prediction technology, in support of which this effort was conducted. +VIII. References + + + + + + + + + The Next-Generation Air Transportation System's Joint Planning Environment: A Decision Support System + + EdgarWaggoner + + + ScottGoldsmith + + + JoshElliot + + 10.2514/6.2009-7011 + + + 9th AIAA Aviation Technology, Integration, and Operations Conference (ATIO) + + American Institute of Aeronautics and Astronautics + Version 2.0, 2007 + + + Joint Planning and Development Office + Joint Planning and Development Office, "Concept of Operations for the Next Generation Air Transportation System," Version 2.0, 2007. + + + + + The ATM Target of Operations + + Sesar Consortium + + No. DLT0612-001-02-00 + + 2007 + Toulouse, France + + + Technical Report + SESAR Consortium, "The ATM Target of Operations," Technical Report No. DLT0612-001-02-00, Toulouse, France, 2007. + + + + + Center-TRACON Automation System (CTAS) En Route Trajectory Predictor Requirements and Capabilities + + L-3Communications + + + + Version 1.10 + L-3 Communications, "Center-TRACON Automation System (CTAS) En Route Trajectory Predictor Requirements and Capabilities", Version 1.10. + + + + + Center-TRACON Automation System (CTAS) Terminal Area Trajectory Predictor Requirements and Capabilities + + L-3Communications + + + + Version 1.0. + L-3 Communications, "Center-TRACON Automation System (CTAS) Terminal Area Trajectory Predictor Requirements and Capabilities", Version 1.0. + + + + + Future Air Traffic Management Concepts Evaluation Tool (FACET) Trajectory Predictor Requirements and Capabilities + + L-3Communications + + + + Version 1.0. + L-3 Communications, "Future Air Traffic Management Concepts Evaluation Tool (FACET) Trajectory Predictor Requirements and Capabilities", Version 1.0. + + + + + GE-Aviation Flight Management System Trajectory Predictor Requirements and Capabilities + + Ge-Aviation + + + + Version 1.20 + GE-Aviation, "GE-Aviation Flight Management System Trajectory Predictor Requirements and Capabilities", Version 1.20. + + + + + En Route Automation Modernization (ERAM) Trajectory Predictor Algorithm Description and Requirements Framework + + LockheedMartin + + + + Version 1.0. + Lockheed Martin, "En Route Automation Modernization (ERAM) Trajectory Predictor Algorithm Description and Requirements Framework", Version 1.0. + + + + + Design of Center-TRACON Automation System + + HErzberger + + + TJDavis + + + SMGreen + + + + AGARD Meeting on Machine Intelligence in Air Traffic Management + Berlin, Germany + + May 1993 + + + + Erzberger, H., Davis, T. J., and Green, S. M., "Design of Center-TRACON Automation System," AGARD Meeting on Machine Intelligence in Air Traffic Management, Berlin, Germany, 11-14 May 1993. + + + + + Assessment of field trials, algorithmic performance, and benefits of the User Request Evaluation Tool (URET) conflict probe + + DJBrudnicki + + + KSLindsay + + + ALMcfarland + + 10.1109/dasc.1997.637321 + + + 16th DASC. AIAA/IEEE Digital Avionics Systems Conference. Reflections to the Future. Proceedings + Saclay, France + + IEEE + 1997 + + + Brudnicki, D. J., and McFarland, A. L., "User Request Evaluation Tool (URET) Conflict Probe Performance and Benefits Assessment", Air Traffic Management R&D Seminar, Saclay, France, 1997. + + + + + The Aircraft Intent Description Language: A key enabler for air-ground synchronization in Trajectory-Based Operations + + JavierLopez-Leones + + + MiguelAVilaplana + + + EduardoGallo + + + FranciscoANavarro + + + CarlosQuerejeta + + 10.1109/dasc.2007.4391836 + + + 2007 IEEE/AIAA 26th Digital Avionics Systems Conference + Dallas, TX + + IEEE + 2007 + + + Lopez-Leones, J., Vilaplana, M., Gallo, E., Navarro, F., and Querejeta, C., "The Aircraft Intent Description Language: A Key Enabler for Air-Ground Synchronization in Trajectory-Based Operations," Digital Avionics Systems Conference, Dallas, TX, 2007. + + + + + Applications of User Intent in Cockpit- and Ground-Based Decision Support Tools + + RobertVivona + + 10.2514/6.2002-4556 + + + AIAA Guidance, Navigation, and Control Conference and Exhibit + Monterey, CA + + American Institute of Aeronautics and Astronautics + 2002 + + + Vivona, R., "Applications of User Intent in Cockpit-and Ground-Based Decision Support Tools," AIAA Guidance, Navigation and Control Conference, Monterey, CA, 2002. + + + + + Trajectory Clustering Based on Trajectory Structure and Longest Common Subsequence + + Guang-YaZhang + + + JingZhang + + 10.12783/dtcse/ceic2018/24525 + + + DEStech Transactions on Computer Science and Engineering + dtcse + 2475-8841 + + ceic + + DEStech Publications + + + Version 2.2, 2004 + FAA/Eurocontrol Cooperative R&D Action Plan 16, "Common Trajectory Prediction Capability, Common Trajectory Predictor Structure," Version 2.2, 2004. + + + + + Attack and Defense on Aircraft Trajectory Prediction Algorithms + 10.2514/6.2022-4027.vid + + + FAA/Eurocontrol Cooperative R&D Action Plan 16 + + American Institute of Aeronautics and Astronautics (AIAA) + + + + Version 2.0, 2004 + FAA/Eurocontrol Cooperative R&D Action Plan 16, "Common Trajectory Prediction Capability, Common Trajectory Prediction-Related Terminology," Version 2.0, 2004. + + + + + + diff --git a/file773.txt b/file773.txt new file mode 100644 index 0000000000000000000000000000000000000000..acac66ea8cf20d6cad2e3f3ad65a359db27ab815 --- /dev/null +++ b/file773.txt @@ -0,0 +1,377 @@ + + + + +I. IntroductionHE prediction of four-dimensional (4D) aircraft trajectories is a foundational requirement for many advanced air traffic control (ATC) and air traffic management (ATM) concepts.In particular, it is a trajectory predictor (TP), generating 4D (position, altitude, time) trajectories for all aircraft of interest that is the engine that drives airborne and ground-based decision support tool (DST) technology.The effectiveness of DST capabilities, including conflict detection, conflict resolution, and clearance or maneuver advisory generation, is directly dependent on the capabilities and performance of the DST's TP.T With an ever increasing number of research and development efforts using DST technology to meet the demands of the U.S.'s NextGen 1 and Eurocontrol's Single European Sky ATM Research (SESAR) 2 concepts, validation of DST technology is a significant concern.Since most of these DSTs rely on aircraft trajectory prediction, validation of TP technology is similarly a significant issue.In the past, a DST's TP has typically not been validated independent from the validation of the DST itself.In the future, it is desired to develop common TP capabilities so that new DSTs do not have to provide their own customized TP.The use of common TP capabilities would require the ability to validate existing TP capabilities against new requirements imposed by each new DST application, most likely requiring or being greatly simplified by a separate TP validation methodology.Even when validating a DST with its own dedicated TP, the process of independently validating its TP may provide valuable insight into why the DST was unable to meet a specific TP-dependent requirement.In the extreme, TP validation may enable the validation of the DST's TP even if the DST fails to meet all of its requirements, reducing the amount of additional effort required to achieve full DST validation.This paper presents a new methodology (Section II and III) for performing TP validation.This methodology significantly expands the concepts for TP validation originally developed by the Federal Aviation Administration (FAA)/Eurocontrol Action Plan 16 (AP16) Committee on Common Trajectory Prediction. 3 Recognizing the inherent complexities in validating a TP, the new methodology defines a set of techniques and a multi-staged procedural framework designed to reduce the effort in identifying and resolving validation failures.The techniques, most new to TP validation, were inspired by the lessons learned from a number of field trials, flight tests and simulation experiments for the development of trajectory-predictor-based automation.When applied within the new procedural framework that adds structure to the TP validation process, these techniques are designed to support TP validation in avoiding the potentially large costs associated with failures during a single-stage, pass/fail approach.To illustrate the real-world applicability of the methodology, the validation effort performed by the FAA for its En Route Automation Modernization (ERAM) system is analyzed as a case study (Section IV).This initial validation effort identified that the ERAM system failed to achieve six of its eight requirements associated with trajectory prediction and conflict probe. 4All requirements have since been achieved, but to illustrate how the methodology could have benefitted the FAA effort, the validation techniques used for ERAM are analyzed.Next, additional techniques from the new methodology that could have been performed to help identify and resolve validation issues prior to the completion of the initial FAA validation effort are proposed (Section V).Finally, a demonstration of some of the proposed techniques using data from the ERAM validation effort is described (Section VI).This demonstration illustrates how some of the TP issues that ultimately caused failures in the ERAM validation could potentially have been identified and rectified earlier, using less costly validation techniques, an objective of the new methodology. +II. Validation Methodology OverviewA methodology is the collection of techniques, practices, and procedures used for some discipline.This section starts by defining validation, verification and requirements in the context of TP validation.It ends by describing the core techniques of the new methodology.In Section III, these techniques are put into a new, multi-staged procedural framework that describes the use of different data sources to achieve full validation. +A. Verification & ValidationAccurate and efficient trajectory prediction is crucial for many ATM automation systems to achieve their potential for increasing the safety, security, and capacity of air transportation operations.In particular, the underlying TP produces the predicted 4D trajectories to enable air traffic service providers and aircraft operators to assess the impacts of predicted trajectories and provide advisories for meeting ATM objectives in a timely manner.In order to achieve the accuracy and performance requirements of such automation, a thorough process to validate the automation's TP is required.TP validation primarily focuses on two exercises:1. Functional verification 2. Performance validation Functional verification is the process that confirms the TP performs its required functions under the required conditions.These required functions are defined within the TP's functional requirements.Performance validation is the process where the TP's performance is confirmed to be within acceptable limits.Those performance limits are defined within the TP's performance requirements.The two primary forms of performance requirements are computational speed and prediction accuracy.Validation of a TP's prediction accuracy requires additional clarification.This type of validation is the process of determining the degree to which a model is an accurate representation of the real system being modeled from the perspective of the model's intended uses.In the case of TP validation, the model is the TP's generation of a predicted trajectory and the real system being modeled is an aircraft's actual flown trajectory in the operational environment.The validation process is performed through a comparison of predicted and actual system behavior.The validity of the model is determined by the acceptability of the differences between these predicted and actual system behaviors (i.e., are prediction errors within acceptable limits).Therefore, selection of comparison metrics and the sources for the predicted and actual system data becomes a crucial step for performing the validation process.The metrics selected should explicitly characterize those aspects of the TP which have defined limits on prediction accuracy.Selected data sources, including simulations, field tests and operational data, should include or reflect appropriate error sources that would be experienced in actual operations to effectively excite model inaccuracies. +B. Direct versus Indirect TP RequirementsOne of the greatest challenges for a TP validation methodology is dealing with the variety of TP requirements.There are two classes of TP requirements against which validation is performed: **• Direct TP requirements• Indirect TP requirements Direct TP requirements are those that define explicit requirements for TP functionality or performance.Examples include required accuracy for predicted altitude profiles and trajectory response time (i.e., time from receipt of TP inputs to return of a trajectory).It is logical to expect that most TP requirements would be of this type, but unfortunately that is typically not the case. 5ndirect TP requirements are those that define required overall system performance for functions (e.g., DST functions) that are dependent on TP performance.Examples include required false alarm/missed alert rate for a conflict probe and, in the extreme, user acceptability requirements for a conflict probe that rely on a series of empirical tests to validate.Indirect TP requirements require sufficient TP performance to support the dependent functionality in achieving its performance requirement.Because indirect TP requirements never explicitly specify a required level of TP performance, they rely on a different validation approach than direct TP requirements.Unfortunately, it is currently common for a TP to have more (sometimes far more) indirect TP requirements than direct TP requirements, though it is not uncommon for a TP to have a mixture of requirements from both classes.As such, the validation methodology must handle both types of requirements. +C. Core Validation TechniquesThe methodology uses a set of core validation techniques at different stages of the validation process.These core techniques are described below.The overall process as applied to the validation of a TP, referred to as the validation framework, is presented in Section III. +Use of Validation StagesTP validation can be an expensive exercise.The collection of validation data, including high-fidelity simulation, field-test, and operational data (especially if collecting operational validation data requires deployment of a system in the field) can be, in the extreme, cost prohibitive.This reality inspired the AP16 committee to explore the development of a validation database, where validation data can be pooled to share the cost over the international community.Unfortunately, issues with the applicability of using validation data collected by a different organization create difficulties in sharing such data even under the best of circumstances.In addition, it is probable that indirect TP requirements will continue to make up the majority of requirements for many TPs, at least in the near future, and that this will increase the chances that a TP might not meet all of its requirements the first time through the validation process.This could, theoretically, require the TP to undergo the same costs multiple times to ultimately achieve validation.To deal with these issues, the methodology was developed as a series of potentially iterative validation stages, starting with simpler, cheaper efforts that could identify significant impediments to validation before investing in more expensive efforts.If the costs for all tests were minimal, then the methodology could just recommend a single pass/fail stage (most likely using data from the actual operational environment or a high-fidelity simulation environment like at the FAA Technical Center) and the TP would iteratively attempt to pass the validation tests in this environment.† † Because the costs of testing in these environments are significant, this could easily be a costprohibitive approach.Alternatively, if an inexpensive effort could either invalidate or at least indicate a high risk that the TP would not meet one or more of its requirements, then the TP could be improved before attempting to validate via an expensive effort.By building the confidence in the TP's ability to achieve its requirements in stages, iteration can hopefully be performed early in the cycle before the most significant costs have been incurred. +White Box TestingIn the context of TP validation, white box tests are those that take advantage of knowledge regarding the internal processing of the TP being validated.These tests are chosen to excite specific internal processing in the TP to achieve some objective of the validation process.Black box tests, on the other hand, do not require such knowledge and are designed to be independent of any specific TP processing.In support of a validation process, the concept is to use a series of white box tests, focusing on different levels of processing within the TP, to "build up" the validation until the entire TP is validated.This approach is primarily envisioned to support validation of prediction accuracy performance requirements.Though the creation of specific white box tests depends on the specific TP being validated, a useful way to identify TP processing targets within white box tests is to analyze the common processing levels found in all TPs.The Common TP structure 6 as developed by the AP16 Committee is useful for this purpose.This Common TP structure, simplified and divided into three modeling levels, is presented in Figure 1.The lowest level of TP modeling is the Math Modeling level.This is where modeled aircraft behaviors are turned into mathematical equations for integration or geometric approximation.Error sources at the Math Modeling level include equation of motion approximations, aircraft performance model limitations, atmospheric modeling limitations, and selection of an inaccurate math model for the behavior (e.g., using a constant track math model when the behavior is following a great circle path).The next higher level of TP modeling is the Behavior Modeling level.This is the identification of required aircraft maneuvers (behaviors) to meet identified constraints.Error sources at the Behavior Modeling level include behavior model approximations when actual maneuver details are not known (e.g., simplified modeling of complex guidance behavior) and selection of an inaccurate behavior model for the constraint when the actual behavior is not known (e.g., using a constant vertical rate maneuver for a cruise altitude change when the pilot chooses a max climb thrust maneuver).The next higher level of TP modeling is the Constraint & Initial Condition Modeling level.This is the identification of the trajectory's initial condition and required constraints that must be achieved by processing the input state and intent data.Error sources at this level include incorrect identification of constraints from the input intent and inaccurate modeling of constraints when input intent is missing (e.g., no information on if/how to reconnect an aircraft that is currently off its input flight plan route).The general white-box testing approach is to isolate and test different sources of errors during the validation process so that if iteration to meet a requirement is necessary, the validation process itself supports the identification of where (e.g., at what modeling level) the TP needs to be improved.Since higher level modeling tends to add errors to those from lower level modeling, isolating errors typically focuses on separating out higher level modeling errors so that the lower level modeling of the TP can be validated first.If the lower level modeling is insufficient to meet a requirement, then no improvement to the higher level modeling will enable the requirement to be met.For example, lateral intent errors (constraint modeling) due to inaccurate route modeling (e.g., strategies for reconnecting the aircraft position to a flight plan route when the aircraft is not currently on its flight plan route) can cause significant along-path errors.These errors are in addition to along-path errors caused by other error sources that exist with or without the lateral intent error, such as errors in ground speed due to wind errors from an inaccurate atmospheric model (math modeling).If an along-path prediction accuracy requirement cannot be met solely because of wind modeling errors, then no improvements to TP's route modeling will enable the TP to meet this requirement.In this case, it is beneficial to isolate, analyze and resolve the wind modeling errors prior to dealing with the lateral intent errors.The ultimate benefit of this approach is that the relative contribution of errors at different modeling levels can be determined.This is valuable if the TP fails to meet its performance requirements, because it provides insight into where the most significant improvements can be made to the overall performance of the TP.One theoretical approach to isolating the different modeling errors would be to literally extract the TP functions for different modeling levels (or combinations of modeling levels) from the TP, create input data for these extracted functions, and then analyze their outputs.In practice, it would be very difficult (often impossible) to extract functions that map directly to the conceptual models of TP processing in Figure 1 since actual TP implementations rarely segregate their modeling levels into separate functions.Therefore, a more useful approach is to test the entire TP while controlling which modeling level errors are excited in the input data (see Figure 2).For example, the Math Modeling level can be tested using cases that are absent of any significant errors in initial condition, constraint and behavior modeling.This would require the initial condition and constraints, as well as the resultant aircraft maneuvers, to be well defined and known from the TP inputs.For example, when the input aircraft state is on its flight plan route and the aircraft is flying that route at a known altitude and speed, the behavior of following the great circle paths and turns defined by the input route is fully known.In this situation the math modeling would be the source of trajectory prediction errors.To test the Behavior Modeling level, test cases without initial condition and constraint errors would be used.An example would be a change in cruise altitude where the final altitude (constraint) is known but the maneuver to achieve this altitude is not known.Behavior model testing includes the effects of errors in the math models associated with those behavior models.A rigorous process for applying this approach would be to start testing at the lowest modeling level (math modeling) and work up each modeling level until either:1.The given modeling level fails to meet the requirements, or 2. The given modeling level is too close to failing the requirement to believe higher levels will succeed, or 3.All levels of modeling have been tested and have met the requirements.In practice, the use of white box testing doesn't need to rigorously follow the above approach.In other words, all lower level modeling does not need to be validated before moving to the next modeling level.It is expected that white box tests that cover a few major modeling-level issues will be designed to isolate specific error sources of concern.For example, the prediction accuracy of the TP without lateral intent errors (as described above) may be validated prior to validating the TP with these errors to isolate the impact of this particular error source on the TP's prediction accuracy. +Test Bench TestingTest bench testing is the process of testing a TP independent of its client system by feeding input data directly into the TP's interface and evaluating the TP's output.The major benefit of this approach is that specific, controlled inputs can be sent to the TP and the resultant outputs evaluated.Test bench testing can be used for both black box and white box testing, though only direct TP requirements can be validated since the entire automation system is not run in the test bench environment.Test benching testing is expected to provide different levels of benefit to the three main validation efforts:Functional Verification: Since functional verification is the process of verifying that the TP performs its required functions under specified input conditions, test bench testing is ideal for this effort.TP inputs for each specified condition can be created directly without the need to develop a client system-level (operational) scenario.Since no actual trajectory data ‡ ‡ (as needed for prediction accuracy analyses) is required, the TP output can be evaluated directly for each specified scenario to verify proper TP functionality.Prediction Accuracy Validation: Prediction accuracy validation requires both TP predicted and actual trajectories for a given aircraft, so it is unlikely that initial prediction accuracy efforts will use test bench testing as an approach.Where test bench testing is expected to support prediction accuracy validation is during iteration after a TP has failed to meet a prediction accuracy requirement.High-fidelity validation tests, including high-fidelity simulation, field-test, and operational data tests are typically expensive in terms of resources and cost.If a TP fails to meet its performance requirements during one of these tests, it would be expensive to rerun the test with an updated version of the TP to validate the new TP performance.With proper planning, it is possible to use test bench testing to replace the need to rerun the high-fidelity test.For example, if during the first running of the high-fidelity test, sufficient TP input data to rerun the TP in a test bench environment and the resultant actual trajectory data is stored, then this data can be used to run a test bench test with the same fidelity as the high-fidelity test for validation purposes.The new TP would be given the same input data as the previous TP was exposed to during the highfidelity test and the resultant predicted trajectory would be compared to the actual trajectory recorded during the high-fidelity test.In the majority of cases, the results should be equivalent to rerunning the high-fidelity test with the new TP.§ § Computational Performance Validation: Though computational performance validation requires full clientsystem level testing to ultimately validate (since TP computational performance is not typically independent of client system impacts), test bench testing can provide some benefit.Test bench testing can be used to measure the TP's computational speed under controlled conditions (i.e., for a specific set of inputs) without the need for expensive validation data since actual trajectory data is not required.Hence, these tests are often less expensive to run than full client system level tests.A TP whose performance under test bench conditions is either not acceptable or only marginally so has a high risk of not being acceptable under full client system level testing.Test bench testing can be used to run quick, cheap initial computational performance tests or to roughly check computational performance when rerunning high-fidelity validation results during TP capability iteration to meet a requirement.Though not all current TP's can be extracted from their client systems to be tested in a test bench environment, it is expected that if test bench testing proves to be a highly effective validation technique, future TPs will be designed with interfaces that enable this testing approach. +III. Validation Procedural FrameworkThe procedural framework for applying the new validation methodology is illustrated in Figure 3.The process begins with functional verification and then proceeds through a series of performance validation stages of increasing fidelity and, typically, cost.Not all stages are required to be performed.Depending on the operational concept of the TP's client system, some levels of testing may not even be feasible.For example, current-day operational data will often not provide suitable validation data for a TP that will require input data from a future operational environment (e.g., clearance information that exists only after the TP's client system has been deployed).At any verification/validation stage, if the TP either fails or is too close to failing (e.g., has a high risk that at a later stage will fail) to meet a performance requirement, then the TP needs to be revised before proceeding to later stages of testing.If the TP revisions invalidate any previously performed validation efforts, then these tests should be rerun and additional TP capability revisions performed, as required, until the current TP capabilities achieve acceptable results.The methodology assumes that, by default, the testing will either treat the TP or the entire client system (including the TP) as a black box.White box testing is performed only if:1.It is believed that the TP has a high risk of failing the current validation stage 2. The TP has failed a validation stage (i.e., during TP revision efforts)In the former case, the black box test cases can be performed in a sequence based on "white-box" knowledge, isolating specific modeling layers in order to isolate potential error sources.The only difference between white box and black box testing in this case may just be the order in which the test cases are performed.In the later case, more ‡ ‡ An "actual" trajectory consists of a series of sensed aircraft states as the aircraft flies consistently (or inconsistently) with the clearances associated with the intent inputs for a predicted trajectory.§ § In cases where the inputs to the TP are changed to improve the accuracy of the TP, these new inputs would need to have been available and recorded during the high-fidelity test so the new TP inputs can be created for the test bench test.detailed white box testing may be used to help identify the appropriate modeling layer changes to enable the revised TP to pass the failed validation stage.Details of the specific verification/validation stages are described next. +A. Functional VerificationFunctional verification testing is where the functional response of the TP to specified inputs is compared to the functional requirements to verify that the TP is responding in a required manner.Functional verification should be performed prior to any validation tests because TP capability changes required to achieve functional verification have a high potential to invalidate the results of previously performed validation tests.Test bench testing is the preferred approach for functional verification because in this testing approach, the functional response of the TP *** is a direct result of the test (TP) inputs.During test bench testing, inputs can be created that, based on the functional requirements, which may be directly defined for specific TP inputs, require a specific TP response.The TP can then be given those inputs and the output of the TP analyzed to determine whether the TP's functional response was acceptable.Verification testing does not require a specific operational fidelity of input data.On the contrary, verification testing may require inputs that are rare in actual operations (e.g., only possible in significantly degraded situations) to excite the desired functional response.Verification testing also does not require actual trajectory data for comparisons.Only the TP outputs are required, since they can be evaluated directly against the functional requirements. +American Institute of Aeronautics and Astronautics +B. Test Bench Computational Performance ValidationTest bench testing can be used to measure the TP's computational speed under controlled conditions, i.e., for a specific set of inputs.These tests are often less expensive to run than full client system tests and can be valuable in identifying potential computational speed issues.Since these tests do not need high-fidelity validation data (just representative input data is required), they can be run before the more costly prediction accuracy validation tests.There are two main issues in using test bench testing for computational performance validation.The first is that only direct TP requirements can be tested, since the TP is extracted from the rest of the client system during these tests.The second is that these tests are typically not conservative in their assessment of TP computational speed since other client system processing that would impact the TP computational performance in the operational environment (e.g., through shared processors) are not included.Even with these limitations, test bench testing can be an effective tool in estimating whether the TP will meet its computational performance requirements prior to performing higher fidelity, more costly testing methods.Certainly, if the TP performance during test bench testing does not meet the computational requirements, it is safe to conclude that the TP will fail to meet these requirements *** The "functional response of the TP" can be measured in many ways, and the specific response measurement used is dependent on the type of testing being performed (e.g., white box or black box) and the particular functional requirement being verified.For example, a black box test of a TP given a new cruise altitude input may verify that the trajectory contains a cruise segment at this new altitude in the output trajectory. in the operational environment.Also, if the impact of system processes on the TP performance can be estimated (e.g., will decrease the TP performance by approximately 10%), then the test bench testing can provide an estimate of whether the TP will meet its computational requirements.One additional benefit of test bench testing is the ability to perform extreme condition testing, where the TP is run through scenarios where computational loads are predicted to be at their highest.These conditions may be difficult to create during client system testing.Test bench testing, given the caveats described above, may provide the only approach to test these conditions. +C. Archived Operational Data ValidationData from several current operational systems (e.g., ARTCC Host systems, ETMS) are regularly recorded by the FAA to keep an historical record of the National Airspace System (NAS).This data is used by the FAA and other organizations to perform offline analyses, often to baseline current-day NAS performance.The benefit to TP validation is that this recorded data typically includes both the TP's client system inputs (e.g., flight plans, altitude clearances) as well as the actual trajectory data (e.g., radar tracks) for aircraft whose trajectories need to be predicted by the TP.Often the operational data that is recorded are also the input data sources for new automation technologies added to the NAS.If this data contains enough client-system-level inputs to run the system in standalone mode then it is possible that computational load performance validation can be performed for both direct and indirect TP requirements.If the data also includes actual aircraft trajectory data, then it may be possible to perform prediction accuracy performance validation for both direct and indirect TP requirements.Since the data is regularly recorded by the FAA, use of this data is an inexpensive alternative to collecting validation specific operational data from field tests or operational facilities.A possible limitation for performing prediction accuracy validation from archived operational data is the impact of controller actions on the actual trajectories.Since the data was recorded from real operations, this means that the actual aircraft were impacted by clearances issued by the controllers at that time.Many of these clearances are not directly reflected as changes to today's operational data, though they impact the aircraft's actual trajectory.For example, a verbal clearance for an aircraft to intercept its flight plan route at a downstream waypoint may not be reflected as a change to that aircraft's flight plan.If the TP being validated would have access to intent sources beyond those available or archived today, then using archived operational data would add an additional error source that would not exist in the TP's operational environment, making this data unacceptable for its validation.The existence of these controller-issued clearances also impacts the acceptability of using archived operational data for the validation of indirect TP requirements related to a conflict probe.For example, to effectively determine the false alarm rate of a conflict probe, the aircraft must be allowed to fly unimpeded through predicted conflicts to see if a loss of separation actually occurs.Any clearances issued to the aircraft prior to reaching the predicted loss of separation will eliminate the ability to determine whether predictions before the clearance accurately or inaccurately predicted a conflict.Since new conflict probe technology should not be less accurate than the existing technology, it is expected to be rare that a controller would not issue a clearance to resolve a conflict detected by the new conflict probe.Hence, it is unlikely that the archived operational data would be useful for validating an indirect TP requirement based on conflict probe false alarm rate.Some TP validation efforts (see Ref. 7) have gotten around these limitations by using time-shifted operational data to test indirect TP requirements for conflict probes.This approach takes raw operational data and then adjusts specific aircraft by shifting all of their data (e.g., actual trajectory, clearances) in time to artificially create conflicts that did not exist in the real operational environment.This is actually closer to a high-fidelity simulation approach than an archived operational data approach since the raw operational data is manipulated to create TP validation events (e.g., conflicts).This time-shifting approach, if applicable, could be used as a lower cost alternative to high-fidelity simulation.Archived FAA data is not the only source for archived data that could be used at this stage.If a validation database, such as the one proposed by the AP16 group, were available, then this data would be suitable for prediction accuracy validation at this stage.It is assumed that validation database data would only be useful for direct TP requirements since its purpose would be to support TP validation independent of any particular automation system. +D. High-Fidelity Simulation ValidationThe Test Bench Computational Performance and Archived Operational Data Validation efforts previously described represent opportunities to perform inexpensive validation efforts to limit the chance of validation failure during more expensive validation efforts.As discussed above, they will not be appropriate for all validation efforts, but can be effective under the right conditions.The use of simulation, though significantly more expensive than these previous two validation efforts, should be an appropriate approach for most, if not all TP validation efforts and can significantly reduce the risk of validation failure during field-test or field-operations validation.Simulation used for validation purposes requires a sufficiently high level of fidelity in modeling aircraft motion and/or the operational environment to be used as a surrogate for actual operational data.The required simulation fidelity for validating different performance requirements is discussed below.The major benefit to simulation is the added control over error sources.In a simulation environment, the number of error types and magnitudes can be controlled, enabling a wide range of error situations to be covered by a much smaller number of validation runs.In the actual operational environment, either some (during field-test validation) or all (during field-operations validation) of the error sources are uncontrolled.In operational environments, it may take a large number of runs to capture rare events (e.g., combinations of errors), which may make it impractical to try to validate the TP in this environment for those situations.This makes high-fidelity simulation a preferred environment for testing less-common and extreme conditions when those conditions occur infrequently in the actual operational environment.The TP may even be validated in high-fidelity simulation under greater than maximum error cases to provide a margin of error for unexpected, extreme situations in the operational environment.Aside from the costs, the major disadvantage to using high-fidelity simulation is that not all of the error sources in the operational environment may be well understood.This makes modeling these errors challenging.This is especially true for modeling highly-correlated error sources that tend to occur in tandem.The type of high-fidelity simulation used for validation is dependent on the class and type of performance requirement being validated.The requirements for high-fidelity simulation are presented in Table 1 for each combination of direct/indirect and computational/prediction accuracy requirement. +Computational Performance RequirementsComputational performance, whether for direct or indirect requirements, requires a simulation environment that closely emulates the actual operational environment, including a high-fidelity simulation of:• Client system data load -creates proper impact of client system processing on TP processing via shared processors, etc. • Client system TP triggers -different uses of the TP by the client system may have different speed requirements • Computationally intensive TP scenarios -e.g., exciting constraint relaxation or conditional constraint handling • Hardware configurationsThe simulation must accurately model the impact of interacting TP and client system processing on shared hardware resources (processors, memory, etc.).Since the aircraft actual trajectory data is not relevant for computational performance validation, the aircraft simulation can be of any fidelity level that suits the above simulation requirements.Typically, a high-fidelity aircraft simulation should not be required.High-fidelity simulation facilities that simulate ATC/ATM operations, such as those at the FAA Technical Center, are appropriate for this stage of TP validation.Lower-fidelity environments could also be used, if they can meet the above requirements.For example, a standalone version of the client system, running on deployment-level hardware and fed by proper data to emulate the conditions above, could meet these requirements.For indirect TP requirements, the simulation environment must be able to simulate enough of the operational environment so client system metrics of performance can be properly measured. +Prediction Accuracy RequirementsFor validation of direct TP prediction accuracy requirements, a high-fidelity aircraft simulator should be used to generate simulated actual trajectory data for comparison against the TP predicted trajectories.It is assumed that a high-fidelity operational environment simulation (see previous) would not be required, but some mechanism for creating the proper TP inputs would need to exist.Potentially, test-bench testing techniques could be used to validate the TP for these requirements.If high-fidelity TP inputs cannot be created in any other way, then a high-fidelity operational environment simulation which includes a high-fidelity aircraft simulation is required.The goal of prediction accuracy validation is to validate that the TP's prediction error (PE) when predicting an aircraft's actual trajectory in the real operational environment () is within required limits.If simulation data is being used for validation, then the best estimate of the TP's PE that can be measured is with respect to the simulated aircraft's actual trajectory ( ).The difference between and is the trajectory error (TE) between the simulated aircraft's actual trajectory and the actual trajectory of the aircraft in the real operational environment:(With respect to TP validation, a high-fidelity aircraft simulator is one in which the simulator's is low enough such that the ability of the TP to meet its requirements using simulated data ( ) is a good indicator that the TP will meet its requirements when using operational data ( ).If is too large, then TP validation testing using simulated data may not accurately reflect the performance of the TP in the actual operational environment.For validation of indirect prediction accuracy requirements using high-fidelity simulation, all of the requirements described for computational performance requirements also apply.An additional requirement is that the simulation environment must also provide a high-fidelity aircraft simulation to provide accurate actual trajectory data.This can be difficult to achieve, in practice, if the high-fidelity operational environment requires the simulation of many aircraft simultaneously.It is typically not practical to hook up a large number of high-fidelity aircraft simulators to such an environment.In this case, the fidelity of the available aircraft simulator will need to be assessed (i.e., assessing the magnitude of the TE term in equation 1) to determine whether prediction accuracy validation can be performed.If this validation stage is used primarily to reduce the risk of failing during later field validation efforts, then the use of lower fidelity aircraft simulators may be acceptable for prediction accuracy validation at this stage. +E. Field-Test ValidationThe goal of field-test validation is to test the TP's capabilities under near-operational, but highly observed and controlled conditions.Though validation efforts are performed within the actual operational facility in which the TP's client system will operate, the environment is characterized as only near-operational because the procedures used are specific to the test and do not completely represent current operational procedures.Therefore, field-test TP validation typically includes:• temporary implementation of the TP (or its client system) within an operational facility• collecting operational data (e.g., radar tracks, Host flight plan data) directly from the facility's operational systems • using actual commercial (with line pilots) or test (with test pilots) aircraft• using controlled operational test procedures that may be outside the scope of current-day operations Though not always necessary, this field-test validation can be a critical component to an overall validation approach.The advantage of this type of validation is that it enables testing to be performed under somewhat controlled operational conditions.While some variables such as wind magnitude and direction can only be observed, other critical conditions such as pilot procedures can be controlled.One disadvantage of field testing, due to the nature of controlling the operational environment, is the limited number of test scenarios that can be studied.Finally, though not necessarily as expensive as full field-operations validation, field-test validation is typically resource intensive, often including large test teams and significant facility coordination efforts.Therefore, field-test validation is primarily used when the desired operational scenarios can be properly defined and controlled in a field test and either the:• cost of performing the field tests is significantly less than the cost of potentially failing the operational tests, or • the operational scenario is difficult or impossible to create under current-day operations One main use of field-test validation is to isolate high risk scenarios in the validation process before committing to full operational validation.These can include:• extreme condition testing -scenarios which put maximum stress on TP capabilities, creating a high risk of unacceptable TP performance • safety critical testing -scenarios for which extra safety procedures may be required to remove potential risks to aircraft safety Extreme condition scenarios may occur rarely in actual operations.If a critical scenario can be created in a fieldtest environment, then field-test validation enables this case to be examined without needing to wait for the situation to occur in normal operations (a potential cost savings).Another use of field-test validation is to validate the TP under the sort of operational conditions that will exist only after its client system is deployed.If the TP's client system supports a new ATC/ATM operational concept that does not exist then the only way to validate the TP's performance in this new environment with operationallevel fidelity is to create, under controlled conditions, a limited version of this environment within an operational facility.This is exactly the environment created during field-test validation. +F. Field-Operations ValidationField-operations validation is performed in the actual operational environment in which the TP's client system will operate.Similar to field-test validation, these efforts are performed within an operational facility and data is collected directly from the facility's operational systems.The distinction between field-operations and field-test validation is that field-operations validation is based only on the use of actual aircraft (e.g., commercial traffic) and actual operational procedures (no test aircraft or test procedures).This may limit field-operations validation to only those TP client systems which do not alter or significantly alter current-day operations.The goal of field-operations validation is to validate the TP's performance in its actual deployed environment, the ultimate level of validation fidelity.The major limitation of field-operations validation, other than its reliance on current-day procedures, is the inability to control the environment variables.Since all of the operational data is based on actual operations, the test scenarios are limited to whatever conditions occur the day of the testing.To capture all of the desired test conditions, particularly extreme conditions, is unrealistic due to time and cost constraints.It can also be difficult to capture large amounts of data for statistically significant results if the validation efforts impact the facility operations in a significant way.For these reasons, it is just not practical to perform all validation at this level.It is possible that in the extreme, this level of validation will only be used to "confirm" the results of the previous validation tests, i.e., validate the validation process.This would occur if the numbers of validation scenarios far exceed what can be captured during field-operations validation.Therefore, the explicit role of this validation level is dependent on the TP's client system and its operational environment. +IV. Case Study: ERAM TP ValidationTo illustrate the real-world applicability of the methodology, a recent validation effort performed by the FAA is analyzed as a case study.The FAA is deploying a new ATC system to replace both the existing Host Computer System (HCS) and its decision support tool in the en route domain, called the User Request Evaluation Tool (URET).The replacement system is called ERAM (En Route Automation Modernization).The formal Factory Acceptance Test (FAT) to validate ERAM performance was conducted in September 2007. 4The focus of the FAT was on two key functions of ERAM: trajectory prediction and the prediction of future losses of separation between two aircraft or between an aircraft and a Special Use Airspace (SUA).The validation tests used metrics defined to measure the ability of ERAM to perform these functions and an extensive set of computer analysis tools, developed over several years, to quantify these metrics.These metrics and tools, documented in Ref. 4, had also been effectively used in past FAA validation efforts, particularly for URET.The performance requirements used to validate ERAM during the FAT were derived from the requirement that the ERAM Flight Data Processing (FDP) and Conflict Probe Tool (CPT) subsystems must perform at least as well as the legacy HCS and URET.As summarized in Table 2, the principal requirements were altitude prediction accuracy (requirements FDP9389 and FDP9390), strategic missed conflict alert rate and strategic false conflict alert rate (requirements ERD1879-C3 through C6), and warning time metrics (requirements ERD1879-C1 and C2).Unfortunately, ERAM passed only two of the eight requirements during the formal test.Detailed results of the FAT are documented in Ref. 4. Because it failed a significant number of its validation tests, the FAA initiated an effort to correct the ERAM deficiencies through iterative modifications to the ERAM software.Over the following year, ERAM was modified and re-evaluated until finally in November of 2008, ERAM passed all of its requirements.Lessons learned from the ERAM validation effort indicated that the approaches taken to achieve ERAM validation, though ultimately successful, could certainly be improved.This paper proposes specific techniques from the new validation methodology that had the potential to improve the ERAM validation effort.These techniques are presented in the next section.To provide some context, it is useful to first identify how the approaches actually used during the FAT relate to the techniques and framework of the new methodology.The first two requirements in Table 2 relate to the altitude accuracy of the ERAM TP, making them direct requirements.For these two requirements, the FAA used data recorded from the HCS and the primary radars to run ERAM and compare the ERAM TP predictions against recorded actual aircraft data.This is an example of Archived Operational Data Validation from Figure 3.For the indirect conflict probe requirements, the operational data were altered, by adjusting the message times (i.e.time shifting) for each individual flight by a constant, to create known conflict situations. 7Since the recorded operational data had been impacted by air traffic control precisely to maintain aircraft separation, using the unaltered operational data to validate the performance of the CPT would have been difficult.Using the time shifting process to generate conflicts enabled the validation team to evaluate the performance of the CPT under controlled conditions.This is an example of a High-Fidelity Simulation Validation (Figure 3) using operational data.Even though operational data is used, because the conflict events are "created" through the alteration of this data, the validation is actually using simulated conflict events.This simulation approach creates a high-fidelity simulation because the time shifted archived operational data retains many of the idiosyncrasies that occur in the field (e.g.surveillance radar inconsistencies, message delays, etc.).All testing performed for the FAT followed black box testing procedures.However, the iterations and follow-on testing that took place after the FAT did include some additional white box techniques to identify error sources.During ERAM development for the FAT and through the iterations leading to ultimate validation, the process of identifying and resolving trajectory and conflict prediction errors was a costly effort, taking years of iteration to reach acceptable levels.With limited existing methods for approaching TP validation, the methods employed by the ERAM team were developed ad hoc and were not always systematic in their approach to identifying errors.In hindsight, the new validation methodology offers several techniques for structuring validation tests to support systematic identification and resolution of prediction errors, which have the potential to improve the ERAM iteration process to meet its requirements.Section V describes several techniques from the methodology that could have been applied to the ERAM FAT and Section VI presents quantitative examples of their applicability, retrospectively, using actual data from the FAT. +V. Proposed Techniques for ERAM TP ValidationAt the time of the formal test in 2007, six out of the eight ERAM performance requirements were not achieved.As an outcome of these results, the test team, composed of the development contractor and FAA participants, continued to correct issues and verify the automation against the requirements until the system passed all of its requirements approximately one year later.This was a costly exercise, but necessary to ensure ERAM performed correctly before deployment.The new validation methodology uses techniques designed to identify and resolve issues that could potentially result in failed requirements as part of the validation process.The application of additional techniques from the new methodology may have, at minimum, enabled the identification and resolution of problems in ERAM sooner, with less cost, and probably would have resulted in a more positive FAT outcome.Evaluating the approaches used by the test team, it was determined that the following additional techniques from the methodology had the potential to improve the effectiveness of the ERAM validation effort:1. Perform the validation in stages, starting with validation of requirements whose success/failure reduces risk in validating later requirements 2. Apply "white box" testing techniques during validation testing, not just during iterative development 3. Use test bench testing techniques during iterative development to successfully meet the requirements Due to the generalized nature of the methodology, it should not be assumed that this is the only set of additional techniques that could have been applied.The selection of specific techniques is left to the judgment of the group responsible for validation; these additional techniques should be viewed as one possible application of the methodology.It should also not be assumed that the lack of these (or similar) techniques implies the original validation was performed incorrectly.The goal of the methodology is to apply techniques that reduce the overall effort in achieving successful TP validation.The end result of a successful validation, as was the case for ERAM, should be the same in all cases.Three detailed examples of applying these techniques are discussed next.The first example describes the benefits of ordering the requirements for testing.The next two examples discuss specific applications of white-box and test-bench testing techniques for validating the two direct requirements. +A. Order the Requirements for ValidationThis technique focuses on performing the validation of the requirements in a specific order, as opposed to performing the validation for all requirements simultaneously as was done in the original ERAM validation effort.Specifically, the validation effort should:• Validate the direct requirements (FDP9389 and FDP9390) before the indirect requirements• Validate the direct requirement for level flight (FDP9389) before the direct requirement for non-level flight (FDP9390)The approach is to validate the requirements in sequence, iterating on the TP capabilities at each stage (if necessary) until the current requirement is met before moving on to validate the next requirement.The main advantage of this approach is that by ordering the requirements based on potential error sources, problems identified and resolved in achieving earlier requirements can remove potential risks in achieving later requirements.For ERAM, since TP altitude prediction accuracy directly impacts acceptable conflict probe performance, the TP altitude prediction accuracy requirements (FDP9389 and FDP9390) should be validated before the conflict probe performance requirements.† † † It is not guaranteed that failure to meet the altitude prediction accuracy requirements will cause the conflict probe performance requirements to fail, but it certainly increases the risk of this failure.Similarly, any improvement to the altitude prediction accuracy of the TP should only improve the conflict probe performance.Since both direct requirements failed to be met and four of the six indirect requirements also failed to be met (see Table 2), iteration to meet the direct requirements has the potential to either achieve one or more of the failed indirect requirements or at least reduce one error source's (vertical prediction inaccuracy) contribution to these requirements not being met.An additional benefit of focusing on the direct requirements first is that it takes less effort to perform this validation, since no additional processing on the operational data (time-shifting) is necessary for testing or retesting the direct requirements.(as opposed to the full system test), this iteration process can be quickly run multiple times, as necessary.After the Furthermore, the requirement for level flight TP prediction accuracy (FDP9389) should be validated prior to the requirement for non-level flight (FDP9390).If the TP doesn't achieve the vertical accuracy requirement for level flight, then one of the following occurs more often than is acceptable:• The actual trajectory is level when the predicted trajectory is in transition (climb or descent), or• Both the actual and predicted trajectories are level, but at different altitudes In both cases, the error is most likely caused by incorrect intent modeling.In the first case, it is possible that level errors could be caused by climb or descent rate errors (e.g., inaccurate aircraft performance or wind models causing predicted top-of-climb or top-of-descent errors), but intent errors should dominate these cases.Identifying and resolving the level flight errors have the potential to also positively impact the TP accuracy for non-level flight (FDP9390).On the other hand, TP vertical prediction accuracy when non-level is significantly impacted both by intent errors (e.g., predicted level flight when the aircraft is actually climbing or descending) as well as other modeling errors (e.g.inaccurate modeling of climb or descent rate).Therefore, it would be preferable to validate the non-level flight requirement after the level flight requirement has identified and resolved as many intent modeling errors as possible.This isolation of the (primarily intent) errors in the level flight cases is a form of white box testing. +B. White Box and Test Bench Testing Techniques: Phase of Flight AnalysisThe application of white box techniques to isolate error sources and test bench techniques to efficiently iterate on TP capabilities to achieve required ERAM performance can be illustrated by the application of these techniques to achieving the first direct requirement: FDP9389.Requirement FDP9389 is focused on the altitude accuracy of ERAM's TP for flights in level segments.Since the metric focuses on those segments where the aircraft's actual trajectory is level, these segments can belong to either the aircraft's cruise, descent or climb phase of flight.In the climb and descent phases of flight, level segments are typically caused by interim altitude clearances that level the aircraft off before reaching its final altitude (cruise altitude for climbs, waypoint defined constraint altitudes for descents).Errors in handling these interim altitudes, including inaccurate models of when the aircraft will be released from an interim altitude constraint or missing/incorrect prediction of procedurally defined interim altitudes, can impact the TP's ability to meet the level flight accuracy requirement.In the cruise phase of flight, the errors in predicting actual level flight segments is expected to be less likely, since the altitude constraint is typically the flight-plan-defined cruise altitude.Since the error sources vary based on the different phases of flight, it is beneficial to identify the phase of flight to which each actual level flight segment belongs.By adding this additional piece of information to the validation analysis, this allows the ERAM level flight accuracy (LFA) performance metric to be divided into three sub sets of data, one for each phase of flight:( 2 ) Then, if ERAM performance fails to meet the total level flight accuracy requirement, the data can immediately identify which phase of flight contributed the largest accuracy errors.‡ ‡ ‡ The benefit of this white box technique is that it helps separate the sources of error in the analysis, which helps in identifying which error sources should be focused on in the TP capability iteration process.Moreover, inputs to the TP should also be recorded when running the initial full operational data test § § § in order to enable the TP to be run in a test bench fashion during any required TP capability iterations to meet the requirement.From the white box testing results, the phase of fight that contributed most to the requirement failure is identified; a small representative set of these flights can then be down-selected and run through the test bench TP for each TP capability modification until the altitude accuracy of these cases reaches a target level that improves the chances that the requirement will be met.**** Due to the smaller effort to run the limited number of test bench cases ‡ ‡ ‡ The FAA metric is actually (see Ref. 4) the ratio of failed level flight segments (i.e., those which exceed an acceptable altitude error limit) to total level flight segments, so "contributed the largest accuracy errors" in this context means contributed the most failed level flight segments.§ § § It is desired to initially run the full set of (not time-shifted) operational data through ERAM to test whether FDP9389 is met, since no additional effort would need to be performed if ERAM met this requirement.**** Choosing an appropriate target level requires engineering judgment based on the specific situation.The target level, when achieved, represents when iteration on the representative set should end and re-evaluation of the TP performance for all flights is required.test bench tests are completed, the full scenario with all the flights must then be run to determine whether the iteration has successfully met the level flight requirement.If not, the process is performed again using the latest validation results as a starting point.Using test bench iteration, the effort of iterating to meet the requirement should be reduced over running the full system test each time the TP is modified. +C. White Box and Test Bench Testing Techniques: Lateral Intent Error Analysisequirement, FDP9390.This req those flights with and without lateral intent error would isolate this additional source of error: (4) (5) t in these cases uld also exist in cases with lateral intent errors (but the reverse is not always true).If the accuracy for flights without lateral intent errors () is unacceptable, then a test bench analysis and iterative resolution approach should be used to improve the TP's capability in predicting non-level flight segments without lateral intent errors.Once the cases without lateral intent errors have reached a target improvement value, the full operational data test of ERAM needs to be run again to see if the new TP capabilities meet the full requirement ().If not, and assuming that the results for flights without lateral intent errors has met its target value, † † † † then representative cases with lateral intent errors may be selected, analyzed and iteration performed to improve the TP's capabilities in dealing with these errors.After the iteration is completed, the full operational data test with all the flights may then be run to validate that the iterations have indeed successfully met the full non-level flight accuracy requirement.During this full operational data test, the level flight requirement (FDP9389) must also be rechecked to ensure that any changes made to meet the non-level requirement have not inadvertently led to a new failure to meet the level flight requirement.Further iteration may need to be performed until both requirements have been met.The processing required to run the indirect requirement validation tests (i.e., time-shifting to create conflicts) and the validation of the indirect requirements is performed after the two direct requirements are met.The expectation is that any TP capability modifications made to meet the direct requirements will either enable the indirect requirements to be met without modification or will reduce the degree to which the indirect requirements are not met. +VI. Demonstration of New ERAM TP Validation TechniquesTo demonstrate the potential benefits of applying the new methodology to the ERAM validation, two examples are presented.The first example demonstrates the benefits of validating the requirements in a specific order, namely, validating the level flight vertical prediction accuracy requirement FDP9389 first.It also demonstrates the benefits of white box testing based on phase of flight.The second example demonstrates the benefit of using white box testing techniques to separate the lateral intent errors associated with aircraft non-adherence to their flight plan routes.The data used for both examples are the legacy operational data sets from the original ERAM FAT in 2007. +A. White Box Testing of Level Flight Altitude Prediction Accuracy (FDP9389) Before Other RequirementsThe basic element of the ERAM TP Level Flight Accuracy (LFA) metric is an actual trajectory segment window where the actual trajectory is level (see Ref. 4).Each flight in the operational data typically has many level and nonlevel windows along its actual trajectory.For each level window, if the maximum error between the actual and predicted trajectories exceeds 500 ft, then the entire window is considered to have unacceptable error.The LFA metric is defined as the ratio of unacceptable windows to total windows for all flights.For the test results in 2007, evaluation of FDP9389 indicated that about 0.9 percent of the level flight segments have unacceptable error, while the requirement was to be less than approximately 0.2 percent (0.0088 ERAM result and 0.0016 requirement from Table 2).Thus, about 5 times as many level flight segments were in error, indicating one or more systematic errors causing altitude prediction inaccuracy.Post-FAT analyses performed by the FAA determined that errors in ERAM's handling of procedural level-off constraints during descents were a significant factor contributing to ERAM's unmet requirements during the FAT.This result could have been identified quicker and resolved earlier as part of the FAT through the use of white box testing.By breaking the LFA metric into three components, one for windows in each of the climb, cruise and descent phases of flight, the test team should have been able to identify that a significant number of windows with unacceptable error were in the descent phase of flight.For illustrative purposes, a flight sample was selected from the original archived data set that exhibited significant trajectory prediction errors during level segments in the descent phase of flight.The sample flight is a Boeing 737-700 that departed from Owen Roberts International Airport, Cayman Islands with a destination of Newark Liberty International Airport in New Jersey.The traffic sample begins within Washington Air Route Traffic Control Center (ARTCC) in level cruise at an altitude of 39,000 feet.Figure 5 is a vertical view that plots the aircraft's reported transponder altitude (right-most trajectory above 21,000 ft, in red) versus its ERAM predicted trajectories, recalculated at various times along its cruise portion of flight.The sample flight travels in a north easterly path until it transitions control (well after top of descent) to the adjacent New York ARTCC and arrives at its final destination, Newark Airport.The predicted trajectories begin their descent quite early with respect to the aircraft's actual top of descent (TOD).This is the main reason for the large vertical errors during the level (and non-level) window segments in the descent.The cause of these early TOD predictions can be seen from the figure to result from the improperly modeled restriction at 21,000 ft.This improper restriction is not specific to just this example flight, but negatively impacted the descent predictions of many similar arrivals in Washington ARTCC airspace.The use of the phase of flight white box testing technique would have enabled the testing team to identify that there was a significant issue with aircraft in their descent phase of flight, leading to the identification and a resolution of this source of intent error.It should also be noted that this intent error, which is relatively easy to detect when analyzing level flight window segments, is also a source of error for some of the non-level flight windows in descent.From Figure 5, the majority of the aircraft's actual trajectory where it is descending and a predicted trajectory is level would result in an unacceptable non-level window when calculating the non-level flight accuracy (NLFA) metric used for requirement FDP9390.If the incorrectly modeled altitude restriction were identified and removed in the process of achieving the level flight requirement FDP9389, it is reasonable to expect a noticeable improvement in achieving the non-level flight requirement FDP9390 before any analysis for this requirement has even begun.This is one of the expected benefits of performing the requirement validation in stages.Another of the expected benefits of validating direct requirements before indirect requirements is that modifications to meet a direct requirement are likely to improve the system's ability to meet the indirect requirements.This beneficial impact can be illustrated for the ERAM validation effort by analyzing the 36 flights that generated unacceptable level flight errors, contributing to ERAM's failure to meet the direct requirement FDP9389 (one being the sample flight above).For these 36 flights, the number of level flight segments per flight with unacceptable errors ranged from one to 42 with an average of about 13.Twenty of these 36 flights also generated 36 conflict predictions that were identified as false alerts, contributing to ERAM's failure to meet the indirect false alarm rate requirement ERD1879-C4.Figure 6 illustrates a histogram of the reduction in the number of unacceptable level flight segments achieved after corrective actions were applied (post the FAT) to meet all the requirements in Table 2. Though one of the 36 flights actually had more unacceptable level flight segments (the bar with a negative value in Figure 6), the corrective actions significantly improved the remaining 35 flights by removing most of their unacceptable level flight segments.Of the 36 corrected flights, only 12 contributed to 22 false alert conflict predictions, compared to 20 contributing 36 false alert predictions before the corrections, a reduction in both the number of flights with, and the total number of, false alerts.Therefore, iterating to meet the direct level requirement FDP9389 likely would have resulted in an improvement in ERAM's false alert rate before attempting to achieve requirement ERD1879-C4. +B. White Box Testing to Isolate Lateral Intent ErrorsTo illustrate another example of beneficial white box testing, the ERAM validation data was analyzed to determine the benefit of isolating cases with lateral routing errors from those without these errors.In theory, identifying and fixing prediction errors for flights without lateral routing errors should also improve flights with lateral routing errors, while providing a simpler environment to identify and resolve common error sources (since lateral routing errors can obscure the impacts of other error sources).To separate out the flights with lateral routing errors, the lateral adherence of a flight to its flight plan route was used.An effective method for defining lateral adherence was defined in Ref. 8. The lateral-adherence algorithm used for this analysis calculates the lateral adherence state for each aircraft's surveillance position report.If the position is within 1.0 nautical mile and heading within 30 degrees of intercepting the known route of flight, the aircraft is considered in lateral adherence, otherwise out.‡ ‡ ‡ ‡ First, the lateral-adherence algorithm was evaluated for its effectiveness in isolating lateral routing errors.For the 2007 ERAM validation data set, the trajectory prediction accuracy was measured for each flight in the sample.Three metrics were analyzed including the mean horizontal error (straight line, time coincident distance between the predicted and actual positions), the mean unsigned cross-track error (lateral side-to-side, spatially coincident distance between the predicted and actual position), and mean unsigned along-track error (longitudinal along-path distance between predicted and actual position).§ § § § These metrics were correlated and partitioned by in-and out-ofadherence state.The results are presented in the Scatter Plot Matrices illustrated Figure 7 and Figure 8, respectively. 10or Figure 7 and Figure 8, the three trajectory metrics are correlated by producing a three by three matrix with a total of nine cells.The frequency histograms are provided on the diagonal cells for mean horizontal, cross-track, and along-track errors, respectively.The histogram's y-axis represents the frequency count of mean errors (x-axis), scaled equally for each error metric.Each non-diagonal cell plots each flight's metric as a function of the other metric (e.g.mean horizontal error versus cross-track error).Figure 7 shows that for the in-adherence state, a flight's mean horizontal error is highly correlated to the mean along-track error, while this is not the case for mean crosstrack error.The correlation metric ***** for horizontal to cross-track prediction errors is only 0.17, while the horizontal to along-track error is near one at 0.99.The linear relationship between the along-and cross-track errors To further investigate the impacts of separating out the in-adherence cases for analysis, it was desired to determine whether these flights could be shown to have had a significant effect on failing the false alert requirement, ERD1879-C4.Like the previous example for level altitude errors, more than half of the 158 laterally adhering flights with significant along-track errors generated false alerts.These conflict prediction errors occurred both before and after the corrective actions were implemented.The hypothesis being tested assumes that the number of selected flights with and without false alerts does not change between the FAT ERAM run and the corrective action run.If this hypothesis can be rejected and the number of flights with false alert events is indeed lower for the corrective action run, it indicates a significant reduction in the false alarms after the corrective actions were taken.To effectively test the hypothesis, a formal categorical statistical method, called a contingency table, is employed and a chi-squared statistical method applied. 13,14The results are presented in Table 3. From the formal test, of the 158 laterally adhering flights, 84 flights exhibited at least one false alert event and 74 didn't exhibit any false alerts.After the corrective actions, only 58 flights exhibited false alert events and 100 didn't exhibit any.Formally, the test determines if the number of flights with and without false alerts between the two runs can be considered statistically equivalent as expressed in (6) by calculating the squared difference between the two cells from Table 3 where the two runs disagree, divided by their sum.† † † † † The test statistic, , is defined as follows: ( 6 ) where, n 21 is the quantity of flights in the second row, first column of the table n 12 is the quantity of flights in the first row, second column of the table If the hypothesis is true, this ratio will follow a chi-squared distribution with one degree of freedom (see Ref. 13 and 14 for further details).The application of (6) produces a very significant effect in the form of a very small pvalue of 0.000.‡ ‡ ‡ ‡ ‡ As a result, the hypothesis that the number of flights with false alert events is equivalent before and after the corrective actions can be rejected at a significance level of at least 0.05.Therefore, it can be stated that there is a statistical correlation between the corrective actions on the in-adherence flights and a reduction in ERAM false alarm rate.This illustrates that separately analyzing the performance of these flights during validation testing of requirement ERD1879-C4 could have effectively supported the identification of errors and necessary ERAM capability iterations to ultimately meet this requirement. +VII. Concluding RemarksA new methodology has been proposed to support the validation of aircraft trajectory predictors for ATC/ATM applications.This methodology includes a collection of techniques and a multi-stage procedural framework for TP validation designed to reduce the effort in identifying and resolving validation failures, avoiding the potentially large costs associated with failures during a single-stage, pass/fail approach.As a case study, the FAA's validation of the ERAM TP, which initially failed to achieve six of its eight TP requirements, was analyzed and specific techniques from the methodology that could have improved the ERAM validation were applied.Two examples evaluated the ERAM TP direct and indirect performance requirements in stages, using white box techniques to isolate error sources.Using actual data from the ERAM validation effort, the results illustrate that the application of additional techniques from the new validation methodology could have, at minimum, identified the problems in ERAM sooner, potentially reducing iteration costs and quite possibly improving the validation outcome of the formal test.Though the development of the methodology was focused on supporting TP validation, the same techniques and processes could be useful in the research and development of a new trajectory predictor.To control validation costs, the methodology was designed to be done in stages, progressively building up confidence in the TP's ability to meet its requirements.Due to the significant risk of failing to meet a requirement at any stage, the methodology was designed to be iterative, supporting the identification of why a requirement was not met (through white box testing) and low effort retesting (through test bench testing) during iterative TP development to meet the requirement.Since many TP requirements are defined in terms of their client automation's functionality (indirect requirements), these techniques are designed to support validating such requirements.These techniques, originally designed to resolve validation issues, are equally valuable to TP developers during the development of a new trajectory predictor, especially during the research and development phase.In the early development of a TP, since TP performance requirements are rarely available, issues with TP performance are typically first identified as issues in the performance of its client automation's functionality (e.g., conflict probe).Identifying how to modify the TP to enable proper performance of a client application function is equivalent to iterating on the TP capabilities to meet a failed indirect performance requirement.The methodology's iteration and white box testing techniques should be directly applicable to identifying and resolving such TP modeling issues.† † † † † In Ref. 14, the test is referred to as the McNemar's test and is specifically designed for testing two data sets that are not independent.This is clearly the case in this study where the same flights are examined between two runs of ERAM.‡ ‡ ‡ ‡ ‡ Ref. 11 defines the p-value as the smallest level of significance at which the null hypothesis would be rejected.If the p-value is small and less than the required value, often set at 0.05 in common practice, the null hypothesis should be rejected.Figure 1 :1Figure 1: Common TP structure -modeling layers. +Figure 2 :2Figure 2: Test case set divided into modeling-level subsets. +Figure 3 :3Figure 3: Framework for applying validation methodology. +Figure 4 :4Figure 4: Impact of lateral intent modeling error on vertical accuracy.Therefore, analyzing non-level flight accuracy (NLFA) in each phase of flight, but now separating this analysis into +Figure 55Figure 5: Vertical View -Altitude Versus Time +Figure 6 :6Figure 6: Reduction of Vertical Deviations from Corrective Action +Figure 9 :9Figure 9: Histogram on Impact of Corrective Actions on Along-Track Error +Table 1 : Type of simulation environment by requirement class and type.1Computational PerformancePrediction Accuracy PerformanceRequirementsRequirementsDirect TP RequirementHigh-Fidelity Operational Environment SimulationHigh-Fidelity Aircraft SimulationIndirect TP RequirementHigh-Fidelity Operational Environment SimulationHigh-Fidelity Operational & Aircraft Environment Simulation +Table 2 : ERAM testing results from the original FAT 4 .2RequirementDescriptionRequirement ERAMMetRequirementNumberResultRequirementTypeFDP9389Vertical trajectory accuracy -0.00160.0088NoDirectaltitude for level flightFDP9390Vertical trajectory accuracy -0.14310.1785NoDirectaltitude for non-level flightERD1879-C1 aircraft-to-aircraft immediate854 seconds740YesIndirectconflict prediction warningsecondstime > 10 min.ERD1879-C2 aircraft-to-aircraft immediate104 seconds128YesIndirectconflict prediction warningSecondstime < 10 min.ERD1879-C3 aircraft-to-aircraft missed0.0250.067NoIndirectconflict alert rateERD1879-C4 aircraft-to-aircraft false0.16, 0.28,0.23, 0.063,NoIndirectconflict alert rates0.007, 0.0050.016,0.008ERD1879-C5 aircraft-to-airspace missed0.020.062NoIndirectconflict alert rateERD1879-C6 aircraft-to-airspace false0.14, 0.007,0.08, 0.01,NoIndirectconflict alert rates0.01, 0.003,0.002,0.0030.001, 0.03 +Table 3 : Contingency Table for False Alerts of Sample Flights3Flights AfterFlights from FATCorrective ActionWith False AlertWithout False AlertTotalWith False Alert471158Without False Alert3763100Total8474158χ 2 =14.083, df=1; p-value=0.000 + † † † In general, it is expected that direct requirements should be validated before indirect requirements since direct requirements are only impacted by TP performance and indirect requirements are dependent on TP performance and other factors. + † † † † If the target value for flights without lateral intent errors has not been completely met, then iteration on a new representative set of these cases should be performed until an acceptable result is achieved. + + + + +VIII. AcknowledgementsThe authors would like to express their appreciation to the FAA/Eurocontrol Action Plan 16 (AP16) Committee for all of its work in the development of Common Trajectory Prediction capabilities.This group's accomplishments in finding common ground across TPs supporting a disparate set of ATM and airborne automation applications in the US, Europe and Australia have directly supported the development of the approaches described in this paper. +IX. References + + + +American Institute of Aeronautics and Astronautics ‡ ‡ ‡ ‡ This algorithm uses tighter bounds than used by ERAM for lateral adherence.ERAM uses conformance boxes centered at the predicted trajectory extending 2.5 nautical miles side-to-side.§ § § § These metrics are defined in detail in Ref. 9. ***** This term is defined in numerous texts as the Pearson R linear correlation coefficient.It measures how well pair wise data fits a straight line (see Refs. 11 and 12).If near +1.0 or -1.0, data fits a line well and near zero does not.for the in-adherence state is very weak with a correlation coefficient of 0.03.Thus, for measurements within the lateral adherence, trajectory errors manifest themselves mainly in the along-track dimension, indicating an error in predicted ground speed.Figure 8 presents the scatter plot results for the out of adherence state.For out of adherence data, the horizontal error is highly correlated to both cross-and along-track errors, with correlation values of 0.72 and 0.95, respectively.The correlation between the metrics of mean cross-and along-track errors is 0.5, indicating they have a modest linear relationship.Therefore, Figure 7 and Figure 8 show that filtering on lateral adherence state can effectively filter out the lateral routing error.Thus, by focusing on the in-adherence flights first, errors in ground speed prediction manifested as along-track errors, many of which are common to both in-and out-of-adherence cases, can be investigated without the complicating impacts of lateral routing errors.Now that it is clear that the lateral adherence algorithm would have isolated cases with primarily along-track errors, the question is whether this set of cases was worth analyzing in isolation.Figure 9 is a histogram of the reduction in the flight's mean along-track error after the post-FAT corrective actions took place.For illustrative purposes, it includes only the top ten percent of the largest along-track errors of flights that were in lateral adherence.This amounted to 158 flights from the complete set of approximately 2200 flights from the original test.Of these 158 flights, eight had larger errors after corrective action (bars with negative values in Figure 9) and the remaining 150 flights were indeed improved.One of the eight flights with larger errors indicated a difference of 26 nautical miles, while the others are degraded less than five nautical miles.The 150 improved flights exhibited a reduction in mean along-track error ranging from slightly above zero to 32 nautical miles.Since the vast majority of these flights were eventually improved to pass the requirements, the results illustrate, albeit only retrospectively, that segregating out the in-adherence flights that exhibited large trajectory errors provides a good initial focus group to expend resources on corrective actions.Though not analyzed due to lack of data, it is expected that the TP modifications to achieve the benefits in Figure 9 would have improved the TP performance for out-of-adherence flights as well. + + + + + + + The Next-Generation Air Transportation System's Joint Planning Environment: A Decision Support System + + EdgarWaggoner + + + ScottGoldsmith + + + JoshElliot + + 10.2514/6.2009-7011 + + + 9th AIAA Aviation Technology, Integration, and Operations Conference (ATIO) + + American Institute of Aeronautics and Astronautics + Version 2.0, 2007 + + + Joint Planning and Development Office + Joint Planning and Development Office, "Concept of Operations for the Next Generation Air Transportation System," Version 2.0, 2007. + + + + + The ATM Target of Operations + + Sesar Consortium + + No. DLT0612-001-02-00 + + 2007 + Toulouse, France + + + Technical Report + SESAR Consortium, "The ATM Target of Operations," Technical Report No. DLT0612-001-02-00, Toulouse, France, 2007. + + + + + A structured approach for validation and verification of aircraft trajectory predictors + + SMondoloni + + + SSwierstra + + + CGarcia-Avello + + + SMGreen + + + MPaglione + + 10.1109/dasc.2004.1391292 + + + The 23rd Digital Avionics Systems Conference (IEEE Cat. No.04CH37576) + Salt Lake City, Utah + + IEEE + 2004 + + + Mondoloni, S., Paglione, M., et. al., "A Structured Approach for Validation and Verification of Aircraft Trajectory Predictors", 23rd Digital Avionics System Conference, Salt Lake City, Utah, 2004. + + + + + Evaluation of En Route Automation's Trajectory Generation and Strategic Alert Processing: Analysis of ERAM Performance + + HRyan + + + GChandler + + + + November, 2008 + + + FAA Technical Note + + + Ryan, H., Chandler, G., et. al., "Evaluation of En Route Automation's Trajectory Generation and Strategic Alert Processing: Analysis of ERAM Performance", FAA Technical Note, DOT/FAA/TC-TN08/10, November, 2008. + + + + + Characterization Method for Determination of Trajectory Prediction Requirements + + TamikaRentas + + + StevenGreen + + + KarenCate + + 10.2514/6.2009-6989 + + + 9th AIAA Aviation Technology, Integration, and Operations Conference (ATIO) + Hilton Head, South Carolina + + American Institute of Aeronautics and Astronautics + 2009 + + + Rentas, T., Green, S., Cate, K., "Characterization Method for Determination of Trajectory Prediction Requirements", 9th AIAA Aviation Technology, Integration, and Operations Conference (ATIO), Hilton Head, South Carolina, 2009. + + + + + Action Plan 16 + + /Eurocontrol + + + Faa + + + + Common TP Structure and Terminology in support of SESAR & NextGen + + January 29, 2010 + + + Version 1.0 + Eurocontrol/FAA Action Plan 16, "Common TP Structure and Terminology in support of SESAR & NextGen", Version 1.0, January 29, 2010. + + + + + Methodology for Generating Conflict Scenarios by Time Shifting Recorded Traffic Data + + MikePaglione + + + RobertOaks + + + KarlBilimoria + + 10.2514/6.2003-6724 + + + AIAA's 3rd Annual Aviation Technology, Integration, and Operations (ATIO) Forum + + American Institute of Aeronautics and Astronautics + November 2003 + + + Paglione, M., R. D. Oaks, K. D. Bilimoria, "Methodology for Generating Conflict Scenarios by Time Shifting Recorded Traffic Data," Proceedings of the American Institute of Aeronautics and Astronautics (AIAA) Technology, Integration, and Operations (ATIO) Technical Forum, November 2003. + + + + + Lateral Intent Error’s Impact on Aircraft Prediction + + MikePaglione + + + IbrahimBayraktutar + + + GregMcdonald + + + JesperBronsvoort + + 10.2514/atcq.18.1.29 + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 18 + 1 + + 2010 + American Institute of Aeronautics and Astronautics (AIAA) + + + Paglione, M., I. Bayraktutar, G. McDonald, J. Bronsvoort , "Lateral Intent Error's Impact on Aircraft Prediction," Air Traffic Control Quarterly, Vol. 18 (1), 29-62, 2010. + + + + + Implementation and Metrics for a Trajectory Prediction Validation Methodology + + MikePaglione + + + RobertOaks + + 10.2514/6.2007-6517 + + + AIAA Guidance, Navigation and Control Conference and Exhibit + + American Institute of Aeronautics and Astronautics + August 20-23, 2007 + + + 9 Paglione, M., R. D. Oaks, "Implementation and Metrics for a Trajectory Prediction Validation Methodology," Proceedings of the American Institute of Aeronautics and Astronautics (AIAA) Guidance, Navigation, and Control Conference, August 20-23, 2007. + + + + + Learning SAS by example: a programmer's guide. Ron Cody, SAS Institute Inc: Cary, NC, 2007. No. of pages: xxxii + 626 (CD included). Price: $69.95. ISBN: 978-1-59994-165-3 + + TimothyDobbins + + 10.1002/sim.3542 + + + Statistics in Medicine + Statist. Med. + 0277-6715 + 1097-0258 + + 28 + 9 + + 2007 + Wiley + Cary, NC + + + SAS Institute + + + SAS Institute, 2007, JMP® Software: ANOVA and Regression Course Notes, SAS Institute Inc. Cary, NC. + + + + + + JayLDevore + + Probability and Statistics for Engineering and the Sciences, 5th Edition + Belmont, CA + + Duxbury Press + 2000 + + + Devore, Jay L. 2000, Probability and Statistics for Engineering and the Sciences, 5th Edition, ISBN 0-534- 37281-3, Duxbury Press, Belmont, CA. + + + + + + + GeorgeE PBox + + + JHunter + + + Stuart + + + WilliamGHunter + + + + Statistics for Experimenters, Design, Innovation, and Discovery, Second Edition + + 2005 + John Wiley & Sons + Hoboken, NJ + + + Box, George E. P., Hunter, J. Stuart, Hunter, William G., Statistics for Experimenters, Design, Innovation, and Discovery, Second Edition, Hoboken, NJ, John Wiley & Sons, 2005. + + + + + Statistical Analysis, An Interdisciplinary Introduction to Univariate and Multivariate Methods + + Kachigan + + + 1986 + Radius Press + + + Kachigan, Statistical Analysis, An Interdisciplinary Introduction to Univariate and Multivariate Methods, Radius Press, 1986. + + + + + Categorical Data Analysis + + AlanAgresti + + 10.1002/0471249688 + + + Categorical Data Analysis + + Wiley + 2002 + + + Agresti, A., Categorical Data Analysis, 2nd ed., John Wiley and Sons, 2002. + + + + + + diff --git a/file774.txt b/file774.txt new file mode 100644 index 0000000000000000000000000000000000000000..3b4069a2cea5a880d38f7dfaf19e53c53020ff74 --- /dev/null +++ b/file774.txt @@ -0,0 +1,130 @@ + + + + +The D-BRITE provides controllers with the flight identification (ID) of aircraft in the terminal airspace.The ASDE, which presently operates at many large airports, provides a map display of the airport surface that shows the locations of aircraft and other vehicles.The map display provides aircraft location information in an intuitive display that is similar to the controllers' out-the-window view.ASDE does not identify the aircraft flight number or provide any other flight-specific information because it is a primary (i.e., skin paint) surface surveillance radar.New surface surveillance systems, such as ASDE-X and a prototype that is being developed under the FAA's SafeFlight 21 program, will provide real-time information about the location and identity of aircraft.The ASDE map display, flight strips, and D-BRITE provide a good picture of the current state of the airport.However, data regarding future departure demand on airport resources is not currently available.NASA Ames Research Center, in cooperation with the Federal Aviation Administration (FAA) is developing a decision support tool known as the Surface Management System (SMS).The project is supported by NASA's Advanced Air Transportation Technologies (AATT) Project.SMS uses information provided by the new surface surveillance systems and departure plans provided by the air carriers in order to provide the Tower, TRACON, Center and air carriers with better information about current and future demand, thereby creating shared awareness of the departure situation and improving the capacity, efficiency, and flexibility of the airport 1 .SMS aids controllers with a variety of tasks including runway balancing and departure scenario optimization.Runway balancing is the task of ensuring that all active departure runways are equally busy in terms of imposed delay and usage.A departure scenario is defined as the mapping of departure fixes or gates § to a departure runway.The purpose of these runway assignment rules is to ensure that the airborne trajectories of aircraft that American Institute of Aeronautics and Astronautics takeoff from different runways do not cross.Departure scenario optimization is the task of ensuring that the current departure scenario provides the most efficient runway usage and leads to the least possible number of delays on the airport surface.Occasionally, it is possible to further enhance the efficiency of the airport surface by identifying aircraft that should be exceptions to the departure scenario.In this case, SMS provides runway advisories via map displays and timelines, advising the controller to taxi the aircraft to an alternate departure runway.SMS also provides additional advisories to help manage surface movements and departure operations.For example, SMS aids controllers with the task of sequencing departure aircraft by taking into account inter-departure gaps required by wake-vortex considerations and downstream departure flow constraints.SMS currently employs three types of user interfaces: map displays, timelines, and load graphs.Map displays of the airport surface provide a two-dimensional representation of the airport and include flight-specific information on data tags.Timelines provide flightspecific information and predictive time information, and load graphs provide aggregate data.Two simulations were conducted in order to solicit user feedback about the SMS concept, the preliminary user interfaces, and the algorithm performance.These realtime controller-in-the-loop simulations of SMS were conducted in the Future Flight Central (FFC) air traffic control Tower simulation facility at NASA Ames Research Center in September, 2001 and January, 2002.FFC is a 360-degree, high fidelity control Tower simulator designed to provide the look and feel of a Level V airport Tower cab.Developed as a joint effort between NASA and the FAA, FFC uses twelve large rear projection screens and computer-generated imagery to provide a 360-degree out-the-window view.Controllers use standard headsets to talk to the pseudopilots who control the individual aircraft movements.The initial simulation, held in September, 2001, consisted of three 45-minute runs.During each run, a different set of displays was presented to each Tower controller.Data were recorded during each run and the controllers completed questionnaires and participated in group debrief interviews after each run.Four active Dallas-Fort Worth (DFW) Tower controllers staffed the Local and Ground positions, while controllers from other airports observed and provided feedback.Observers were also present from Delta Air Lines and United Airlines.A large amount of expert user feedback was acquired through multiple discussions and informal debriefs with the controllers as well as through questionnaires and recordings of the simulation proceedings.The results of Simulation 1 indicated that map displays were well-liked by the Local and Ground controllers and that timelines had potential uses for them as well, but that both timelines and load graphs might be better suited for a Traffic Management Coordinator (TMC) or Supervisor position.The experimental design and results of Simulation 1 are described in detail in Reference 2. The feedback and human factors observations recorded during this simulation were incorporated into a refined version of SMS that was evaluated in a second simulation in January, 2002.The second simulation consisted of nine hour-long runs based on actual DFW traffic.Data were recorded during each run and the controllers completed questionnaires after specific runs.Additionally, group debrief interviews were held after each run.More than 30 participants were involved in the simulation, including five active DFW Tower controllers, a controller from the Memphis, TN airport (MEM) Tower, a supervisor from the Norfolk, VA Tower, and airline observers from FedEx, Northwest Airlines, UPS, American Airlines, and United Airlines.The five active DFW controllers staffed the Tower positions, while the other controllers observed and provided feedback.A large amount of expert user feedback was acquired through the informal debrief sessions as well as through questionnaires and data collected during the simulation.This paper summarizes the methodologies employed, the SMS displays tested, and the human factors lessons learned from the simulation.SMS will be further refined based on the feedback from the simulation and will be evaluated next in the FedEx ramp tower at MEM in the summer of 2002.An evaluation in the ATC Tower will follow in 2003. +SIMULATION DESCRIPTION Simulation EnvironmentIn nominal operating conditions, DFW operates two air traffic control Towers, one controlling the west half of the airport, the other controlling the east half.Since the FFC facility can simulate only one Tower, only one side of the airport could be modeled.Therefore, since the majority of the gates and runways are located on the east half of the airport, a modified version of the East Tower of DFW was modeled.Operations were only conducted in South Flow under Visual Flight Rules (VFR) conditions.Figure 1 is a diagram of DFW airport.The box encloses the area that was modeled for the SMS simulation.American Institute of Aeronautics and Astronautics Figure 1: Layout of DFW When DFW operates in South Flow, jet aircraft take off from runways 17R and 18L and land on runways 17C, 17L, and 18R.Runway 13L is used for prop departures and 13R is used for both prop and jet arrivals.As one of the goals of the simulation was to probe the efficacy of SMS in aiding the task of runway balancing, the procedures used on the east side of DFW needed to be altered to include two jet departure runways.Therefore, runway 17C was used for both arrivals and departures, thereby creating a second departure runway.Runway 13L was not used in the simulation. +ParticipantsNormal staffing of the DFW East Tower during busy traffic periods consists of eight positions: Supervisor, TMC, Clearance Delivery/Flight Data, Ground East 1 (Departure Ground or GE-1), Ground East 2 (Arrival Ground or GE-2), two Local controllers and a Cab Coordinator (CCE-1).Four of these eight positions, as well as an additional position created specifically for this simulation, were staffed by five active DFW controllers.The position that was created for the simulation was the "TMC Assistant" whose job it was to work with the TMC and to help evaluate the TMC displays.One of the active controllers, who is both a traffic manager and a controller at DFW, was assigned exclusively to the TMC position.The other four controllers rotated through the positions of Ground East 1 (GE-1), Ground East 2 (GE-2), Local East 1(LE-1), and TMC Assistant.A retired controller staffed the Cab Coordinator (CCE-1) position throughout the simulation.As a result of the altered DFW configuration, the roles of the Tower controllers changed slightly from their actual duties at DFW. • During the simulation, as in current DFW operations, GE-1 was responsible for taxiing departing flights from their spots ¶ to their departure runway and sequencing the aircraft for departure.The difference between the role of GE-1 in day-today operations and his role during the simulation was that during the simulation this controller queued aircraft to depart off of 17C in addition to 17R and did not send any aircraft to the west side of the airport.• At DFW, the GE-2 is responsible for taxiing arrivals that have cleared the active runways to their spot, mixing and merging flights coming to or leaving spot areas, and taxiing departures leaving from spots 50-53 to their departure runways.During the simulation, the GE-2 position incurred much less workload than the actual DFW position due to the fact that there were no aircraft arriving from the west side of the airport, as there are in current DFW operations.• The LE-1 is responsible for all activity on runways 17R and 17C including providing aircraft with arrival and departure clearance and taxiing aircraft across active runways.• CCE-1 is responsible for assisting LE-1 by scanning runways and taxiways for potential conflicts.• The TMC monitors the departure scenario and runway balance to ensure that the airport is operating efficiently.The TMC is also responsible for the coordination among the controllers in the Tower.For the purpose of the simulation, the TMC was asked to determine the optimal time to switch operations on 17C from departures to arrivals and relay this information to the controllers. +Simulation ScheduleThe simulation was comprised of nine hour-long runs.Three experimental conditions were tested and run three times each during the simulation.The traffic scenarios that were used approximated the 0800, 1130, and 1300 'rush' periods at DFW.However, in order to minimize learning effects, the scenarios were presented as if they were all occurring in the 1130 time frame.Each rush was characterized by a departure push followed by an arrival rush, but the specific overlap times differed in each scenario.The simulation began with an hour and a half of classroom training followed by one hour and 45 minutes of practice time in the FFC facility.When the ¶ A "spot" is the location on an airport surface at which an aircraft is transferred from ramp control to Tower control or vice versa.American Institute of Aeronautics and Astronautics training was complete, the nine simulation runs were conducted.Experiment Design SMS utilizes three types of displays to convey information and advisories: map displays, timelines, and load graphs.A map display is a two-dimensional representation of an airport outlining the airport surface (i.e., providing the outlines of the taxiways, runways, ramps) and showing the location and direction of travel for each aircraft.Timelines, which are referenced to a physical location (e.g., a runway threshold or taxiway intersection), show the predicted times when aircraft will occupy that location but do not explicitly show the current location of each aircraft.Load graphs display the amount of present and forecasted demand on a specified airport resource (e.g., a runway, departure fix, or taxiway intersection).Load graphs display aggregate demand information rather than flight-specific information.Both timelines and load graphs have been used in the Center-TRACON Automation System (CTAS) Traffic Management Advisor (TMA) tool 3 .The SMS display types are described in greater detail in Reference 2.Experimental Conditions Three experimental conditions were tested during the simulation.Using feedback that was obtained during the initial simulation, specific displays were designed for each experimental condition and controller position.These conditions and the corresponding displays are described below.Condition 1: Local and Ground displays Condition 1, the Baseline Condition, was intended to replicate the workings of a control Tower without SMS.During this Condition, the Local and Ground controllers' displays consisted of an airport map display, similar to what they will have after ASDE-X is deployed in the Tower.The aircraft symbols on the map display were labeled with their flight IDs. Figure 2 is a portion of a sample map display.The controllers had the ability to toggle the flight IDs on or off independently for arrivals and departures to de-clutter the map.Controllers were also able to highlight the datablock for any aircraft by moving the mouse pointer over the aircraft symbol for that aircraft.Figure 3 is an example of a highlighted datablock.The aircraft symbols on the map displays were labeled with expanded datablocks (i.e., flight specific American Institute of Aeronautics and Astronautics information such as aircraft type, departure gate and spot number was provided in datablocks).The specific information that was provided and the area covered by the map display depended on the controller position.Controllers had the ability to toggle through arrival and departure datablocks independently on the map display through three separate settings: no datablock, flight ID only, and expanded datablock.Figure 5 shows datablocks with flight-specific information.This datablock provides the flight ID and aircraft type on the first line and the initial departure fix and departure runway on the second line.The timeline displays that were shown to the Ground and Local Controllers differed according to controller position.The GE-1 had six timelines, four that were referenced to spot areas and two to the departure runways.The data tags on the timelines were colorcoded according to departure gate.The GE-2 had four timelines that were referenced to the spot areas that were color-coded such that the arrival data tags were white and the departure data tags were green.The LE-1 had four timelines, one for each departure queue (EF, EG, EH, and the queue for runway 17C).Again, the data tags were color-coded by departure gate.All of the timelines for the Ground and Local Controllers had a ten minute look-ahead time.In SMS, demand can be displayed as either constrained or unconstrained.Unconstrained demand means that the times that are being shown are calculated as if there were no other aircraft operating in the system.In other words, SMS does not run prediction algorithms to determine how the other aircraft will affect this flight's arrival or departure time.Constrained demand includes SMS predictions such as taxi delay and wake-vo rtex separations.The load graph in Figure 7 displays unconstrained demand.The TMC's timelines were referenced to the departure runways and displayed the expected arrival and departure traffic for those runways.Figure 9 is an example of the timelines that were presented to the TMC during Condition 2. The Configuration Change Advisory Tool, shown in Figure 10, provides the TMC with a suggested time to switch runway 17C from a departure runway to an arrival runway.The tool shows the arrival, departure, and total delays as a function of the time at which the configuration is changed and selects the point at which the predicted total delay is lowest.In this example the tool has suggested 18:12 as the appropriate change time.Human factors data were also collected via observation during each of the runs.Questionnaires were administered to controllers after the first and third run in each set of three runs.These questionnaires focused on the usability, suitability, and acceptability of the user interfaces.Usability refers to the ability of the controllers to readily obtain and use the information presented, suitability refers to the appropriateness of the user interfaces to the task requirements and information needs, and acceptability is the controller's trust in the information presented and his willingness to incorporate SMS into his/her task performance strategies.Most of the questions were based on a 7point Likert scale.However, some multiple choice and open-ended questions were included.The questionnaires were designed to be specific to the experimental condition and controller position.Therefore, not all controllers were asked the same questions.The usability, suitability, and acceptability questionnaires were administered to the Ground and Local controllers after each Baseline Condition and then again after the next two runs were completed (the next two runs consisted of Conditions 2 and 3, however not necessarily in that order).The TMC Assist position filled out questionnaires after each run, and the TMC filled out questionnaires after the first and last run in each experimental condition.Controllers also provided feedback via informal 30minute debrief sessions that were conducted after each run.During these debrief sessions, controllers were free to comment on any aspect of the simulation that had just occurred, including aspects of the SMS user interface, traffic flow, etc.These discussions included requests for additional information, rationale for decisions made during each run, and requests for display changes.Finally, the TMC participated in structured interviews that focused on the decision process used to determine when to change the airport configuration and how SMS supports other TMC tasks. +HUMAN FACTORS RESULTSThe human factors work that was conducted during the simulation sought to better understand how controllers make decisions and how they interact with our automation and with each other.As a result, the lessons that were learned fall into two categories: controllers' opinions about the different display types and information about how controllers do their work, either in terms of thought processes or interactions.For the purposes of this paper, only a subset of the data is described in detail.Although some numerical results are presented here, it is important to recognize that due to the small sample size these results represent trend information only. +Controller Opinions about Display TypesThe controllers' responses to the displays were heavily dependent on the position that they were staffing.The tasks performed by the TMC are strategic in nature.The TMC is responsible for ensuring the smooth flow of traffic on the airport surface and for determining the appropriate configuration for the airport.Local and Ground controllers, on the other hand, perform more tactical tasks in order to direct the individual aircraft around the airport surface.As a result of the differences in their tasks, the displays that the TMCs preferred were different than the displays preferred by the Local and Ground controllers. +Ground and Local Responses: Map DisplaysThe feedback that was received about the map displays, both from the questionnaires and the informal debrief sessions, was positive.All of the Ground and Local controllers were in favor of having a map display that provided both aircraft location and flight-specific information (via datablocks).Additionally, the map display was the preferred information source for all tasks performed by GE-1, GE-2, LE-1 when it was presented with expanded datablocks.Controllers indicated that they trusted the information on the map display.When they were asked, "How much did you trust the information provided to you on the map display?," on a scale of 1="Trusted Completely" to 7="Did Not Trust at All," they responded with mean scores of: American Institute of Aeronautics and Astronautics GE-1: x = 1.7, = 1.2 GE-2: x = 2.0, = 1.3 LE-1: x = 2.8, = 2.1.There are several hypotheses as to why controllers preferred using the map display to their other displays.First, the controllers who participated in the simulation were already familiar with map displays due to the fact that there is a map display currently in use the Tower at DFW.Also, map displays present an overhead view of the airport surface, much like the view the controllers see when they look out the windows of the Tower.Map displays present location information in a straightforward, intuitive manner.In addition, flight strips, which normally provide flight-specific information, were not available.The datablocks on the map display presented the easiest method of finding flight-specific information.Several issues were identified with the map display, the most important of which was clutter.On many areas of the airport, such as departure queues, aircraft line up close to each other.When this occurs, the datablocks from the various aircraft overlap and become unreadable.Figure 12 is a sample of a map display presenting the departure queues to runway 17R.Several controllers offered suggestions for how to deal with the clutter issue.For example, the GE-2 suggested moving aircraft type to the second line of arrival datablocks (but leaving it on the first line of departure datablocks) to reduce clutter.One feature that was implemented prior to the simulation in order to combat the clutter issue was the ability to highlight an aircraft so that its datablock was more readable.However, the controllers mentioned the highlighting feature was difficult to use.Instead they proposed a design in which relevant aircraft (i.e., the aircraft at the front of the departure queues or the first aircraft waiting to cross an active runway) were always highlighted.An issue that was mentioned by the GE-2 was that when the datablocks are set to display the flight ID only there is no cue as to which aircraft are arrivals and departures. +Ground and Local Responses: TimelinesThe Ground and Local controllers' responses to the timeline displays were less favorable than those to the map displays.The overwhelming response was that they would like the information from the timelines to be presented on the map display.However this may be due to a need for a single display rather than due to the nature of timelines themselves.During the simulation, timelines were provided on a separate monitor from the map display, which provided the majority of the flightspecific information.The controllers were concerned about spending too much heads-down time.As a result, the timelines were used less frequently than the map display.Local controllers found the timelines to be more useful than the Ground controllers did, as is evidenced by the results of the following question: "How useful was the information provided by the timelines?"on a scale of 1="Extremely Useful" to 7="Not at All Useful".The Ground controllers responded with mean responses of: GE-1: x = 5.7, = 0.6 GE-2: x = 5.0, = 2.0 LE-1: x = 3.3, = 1.5.This difference in ratings is directly related to the different tasks that are performed by the two different types of controllers.The Ground controllers used the timelines to inform them when an aircraft would be transitioning into or out of the active movement area.The Local controller used the bottom of the timelines to indicate which aircraft were available to be chosen next for departure.This information was also available on the local controller's map display, however due to the clutter issue described in the previous section, many datablocks were unreadable on the map display.Therefore, the timelines were valuable to the local controller who used them as a source for flight-specific information.For the Ground controllers, the usefulness of timelines is dependent upon the amount of location information that they provide.The arrival timelines were less useful than departure timelines due to the lack of location information implicit in an aircraft's presence on the timeline.Whereas the departure timelines clearly indicated that the aircraft was moving in the ramp area and would soon be approaching the spot in order to American Institute of Aeronautics and Astronautics transition into the active movement area, an aircraft's presence on the arrival timelines only indicated that the aircraft would be approaching the spot sometime soon.The controller then had to search for the aircraft, which could have been located anywhere on the airport surface.The GE-2 suggested that timelines should be referenced to the crossing points as opposed to the spot areas in order to have implicit location information, as aircraft approaching the runway crossing points can only be taxiing on specific taxiways.The GE-2 liked using the timelines for reading the flight-specific information because of map clutter due to datablocks.All of the Ground and Local controllers commented that the time duration of the timelines was unnecessarily large.Instead of having access to the predicted traffic over the next ten minutes, the Local controllers want to know the next one or two aircraft to depart or cross runways.As mentioned in the previous section, they suggested highlighting those aircraft on the map display instead of using timelines.Another issue that was brought up by the controllers was the fact that aircraft that are located at gates far away from the spot appear earlier on the timelines because they push back earlier than aircraft that are parked close-by.Although the information provided on the timelines is accurate, and the aircraft is actually moving in the ramp area, it is not beneficial to the controller to be made aware of it when it is still several minutes away from crossing the spot.The controllers also noted that the timelines were not sufficiently accurate.Timelines do not take into consideration aircraft that are not ready for pushback, gate arrival, etc, for various reasons.The GE-2 controller commented on some confusion because SMS provided inaccurate information when the aircraft taxied to a different parking gate than its assigned gate.Ground and Local Responses: Runway Advisories The concept of SMS providing runway assignment and departure sequence advisories was well received by the Ground controllers despite the fact that improvements to the current algorithms are necessary.The SMS algorithms that were used during the simulation did not consider the time needed for arrivals or departures to cross a runway, which led to advisories that did not reflect the way controllers would manage the aircraft.Despite these inaccuracies, the controllers stated that runway advisories freed up time for other tasks.Future research will focus on runway crossing delays.The GE-1 controller responses indicated that runway advisories decreased their workload slightly and increased the efficiency of the airport surface slightly.In response to the question "How much did using SMS runway advisories impact your workload?", the mean GE-1 response was x = 2.7, = 1.5 on a scale of 1="Decreased Workload" to 7="Increased Workload".Similarly, in response to the question "How much did the SMS runway advisories impact the efficiency of airport operations?" the mean GE-1 response was x = 2.3, = 1.2 on a scale of 1="Improved Airport Efficiency Greatly" to 7="Decreased Airport Efficiency Greatly".Additionally, in debrief sessions, the controllers said that SMS gave them a runway other than the one they would have preferred 50% of the time, nonetheless they claim to have followed the advisories 92% of the time.In debriefing, the controllers stated that they would like the runway advisories to display a clear strategy.For example, during the simulation, the TMC made a decision to change runway 17C from a departure runway to an arrival runway at a specific time.The TMC advised the controllers that the switch time was approaching, and the controllers began to depart all subsequent departures off of 17R.However, the algorithm that was determining whether or not to present a runway advisory saw that there were still several minutes before 17C became an arrival runway and advised the controllers to take several more departures off of 17C, contrary to the instructions of the TMC.Therefore, according to the controllers, when the system makes the call to stop sending departures to runway 17C, it should be a distinct change.The controllers are under no obligation to follow the advisories, and after the flight enters the queue for another runway, the information in the datablock is updated to reflect the correct departure runway.This update only occurs after the flight has joined the other queue, and it may take several minutes for the information in the datablock to reflect the correct departure runway.The Ground controllers also expressed that they liked the idea of having advisories, as long as they could override them whenever they wanted.Also, data show that controllers accepted more advisories during each successive run, which indicates increasing acceptance as their familiarity with advisories grew. +TMC ResponsesThe TMCs ranked timelines, Arrival/Departure Delay Load Graph, and Configuration Change Advisory Tool as the top three tools for performing their tasks.The map display was rarely used by the TMC.The TMCs ranked the timelines as the most useful tool for determining the time at which to change the configuration of the airport.The timelines presented a American Institute of Aeronautics and Astronautics clear visual picture of the demand on each of the runways.The TMC used the timelines to see the departure demand and determine the appropriate time to switch the runway from a departures to arrivals.The TMCs also reported using the timelines tactically in order to find B757s or heavies to use to create gaps in the traffic flow to allow waiting aircraft to cross the runways.In this case, the TMC would identify an upcoming 757 and then inform the controller that the 757 should be used to cross multiple aircraft over the active runway.The delay load graph was ranked as the second most important display because it supplements the timelines by providing information about how much delay is included into the predicted departure times.When a controller looks at a timeline of predicted traffic, it is impossible to determine how much delay each of the flights is absorbing.However, by correlating the demand shown on the timelines with the delay load graph, it is possible to determine if aircraft are being delayed or if they are scheduled to leave in the predicted traffic pattern.The delay load graph and timelines were also used together to provide a "what-if" functionality that allowed the TMC to choose a switch time, evaluate the ramifications, and pick a new switch time to examine if necessary.The TMC commented that the delay load graph was the most helpful of all the load graphs and that the queue load graph was never used.The third tool used by the TMC was the Configuration Change Advisory Tool, which informed the traffic manager of the theoretically optimal time to make the configuration switch.The TMC reported that the time was close, but not exactly the same as, the time that they would have chosen.The tool usually selected an earlier time than the TMC, which is logical due to the fact that the algorithms behind the configuration tool were not taking runway crossing times into account.It was, therefore, predicting lower taxi times overall and selecting an earlier configuration change time.The TMC noted that if the algorithms had taken runway crossing times into account, the tool would have been more accurate and, therefore, more useful. +Lessons about Controller Decision-MakingIn addition to administering questionnaires and conducting interviews with the controllers, the human factors work conducted during the simulation involved observation (i.e., watching how the controllers worked and with whom they interacted).Although each controller has specific tasks that must be done individually, each controller pulls information from the others and from the various displays in the Tower.Controllers collaborate with each other almost continuously.For instance, the Ground controllers, GE-1 and GE-2, spent much of their time observing the traffic on the LE-1 displays.Their goal was to monitor the LE-1's traffic load and the demand on each of the departure fixes.They attempted to make decisions that would minimize workload on the next person downstream.The TMC also spent a large portion of time standing between the GE-1 and GE-2 positions, advising the controllers about tactical decisions.However, the LE-1 only seemed to interact with TMC on occasion and with GE-1 via receiving flight strips.The main implication that this has for SMS is that each of the displays needs to be clearly visible to each of the other controllers, and the designs need to be standardized in terms of color-coding and symbology.The controllers told us that they do not follow any pattern of which aircraft go into which queue.However, in practice, they assign aircraft going to one fix to one queue and then put all "splitters # " on the other queues so that they can split up traffic as efficiently as possible.They view it as mixing up the traffic to operate as efficiently as possible.Different Ground controllers use different strategies to assign flights to departure queues.The goal is to set up the sequence so that no two flights in the final runway sequence have the same departure fix while also taking weight class into account.Another observation that was made was that the controllers do not pay close attention to how long aircraft are waiting at spots or to cross runways.If an aircraft has been waiting a long time then the TMC usually points it out to them.Only if the aircraft queue began to back up did they seem to take any measures to cross a large number of aircraft. +SUMMARYA major goal of the SMS Simulations was to learn about the roles of controllers, the methods they use to make decisions, and the types of information that they find useful.To this end, human factors studies investigated users' preferences via questionnaires, interviews, and observational studies.Information was acquired about what displays may be appropriate for each user in a Tower environment.The roles of each controller were better defined, and the potential uses of SMS were explored.It is clear that the Ground and Local controllers prefer using the map display for all of their tasks.The # A splitter is an aircraft placed in a departure sequence to "split up" two other aircraft headed to the same departure fix.American Institute of Aeronautics and Astronautics controllers stated that timelines are not useful for them and should be removed from their displays, especially if a new function can be integrated such that relevant aircraft are highlighted on the map display.This would address several issues (including clutter) and would consolidate all pertinent information onto one monitor.However, much of the negative response to the timelines was influenced by the controllers' unfamiliarity with the displays and the timelines' inaccuracy due to the prototype SMS algorithms.It is important to note that until the SMS algorithms are refined and include estimates of runway crossing delays, timelines should not be dismissed as a potential display for Tower controllers.Future research will be conducted with accurate timelines in order to determine their potential uses.Runway advisories were well-received by the Ground controllers.They appear to increase efficiency and decrease controller workload.However, the controllers stated that they must always be able to override the advisory and the advisories must display a clear strategy that is apparent to all users It was determined that timelines are well suited to the strategic tasks of the TMC as are the Delay Load Graph and the Configuration Change Advisory Tool.However, demand load graphs and load graphs displaying departure queues are not useful to the TMC.It was also found that because the tasks conducted by the TMC are much more strategic in nature than those conducted by the Ground and Local controllers, a longer look-ahead time was appropriate.Although each controller works individually, each of them also monitors the scenario downstream and takes the implications of their clearances on other controllers into consideration.Therefore, SMS displays must be consistent in terms of color-coding and symbology such that a controller glancing at another user's display is not confused by conflicting information.Additionally, the use of flight strips and SMS together creates additional workload.More research will be conducted, either to determine how to design automation to support the use of strips without increasing workload or to create a system that can fully replace flight strips.The human factors lessons learned from the second SMS simulation will be used to refine SMS.In addition, feedback from other user groups, such as ramp tower controllers, Airline Operations Center users, and TRACON and Center TMCs, will contribute to further development and refinement.SMS will first be demonstrated in the FedEx ramp tower at Memphis airport in the summer of 2002.ATC Tower demonstrations will follow in 2003.Figure 2 :2Figure 2: Map Display +Figure 4 :4Figure 4: Sample ETMS-like load graph Condition 2: Local and Ground displays During Condition 2, the Local and Ground controllers were presented with airport map displays and timelines.The aircraft symbols on the map displays were labeled with expanded datablocks (i.e., flight specific +Figure 5 :5Figure 5: Datablock with Flight-Specific Information +Condition 2 :2TMC displays During Condition 2, the TMC's display included timelines referenced to the runway thresholds, a map display covering the entire east side of the airport, and three load graphs: unconstrained arrival and departure demand, arrival and departure delay, and queue length to the departure runways. +Figure 66Figure 6 is a sample of the arrival and departure delay load graph.On the actual display, a white line (shown in the figure as a dark line) represents the delay on the arrival aircraft, a green line (shown in the figure as a light line) represents the delay on the departure aircraft.A red horizontal line (not shown in the figure) represents a theoretical delay limit and can be placed by the controller. +Figure 6 :6Figure 6: Delay Load Graph provided to TMC Figure 7 shows a sample arrival and departure demand load graph used by the TMC during Condition 2. On the actual display, a white line (shown in the figure as a dark line) represents the predicted arrival demand, and a green line (shown in the figure as a light line) represents the predicted departure demand. +Figure 7 :7Figure 7: Unconstrained Arrival and Departure Demand Load Graph provided to TMC The TMC also used a load graph depicting the queue lengths at each of the runways.Figure 8 is a example of that departure queue load graph.The queue for 17R is depicted in yellow (shown in the figure as a dark line), and the queue for 17C is shown in blue (shown in the figure as a light line). +Figure 8 :8Figure 8: Departure Queue Load Graph +Figure 9 :9Figure 9: TMC TimelinesThe TMC's timeline display provides future demand information for the three runways: 17R, 17C, and 17L.Departure information is displayed on the left timeline, and arrival information is displayed on the right.The left-hand side of the left timeline displays the departures from 17R, and the right-hand side of that same timeline displays the departures from 17C.On the right displays, the left-hand side displays the arrivals to 17C and the right-hand side displays the arrivals to 17L.The demand information that is provided on the timelines is constrained demand (i.e., it includes scheduling information provided by SMS).Condition 3: Ground, Local, and TMC Displays Condition 3 differed only slightly from Condition 2. All of the displays presented during Condition 2 were also used during Condition 3. The TMC received one additional display, the Configuration Change Advisory Tool. +Figure 10 :10Figure 10: Sample Configuration Change Advisory Tool Flight Strip Usage During the week prior to the simulation, several of the DFW controllers participated in dry runs to assess the readiness of the simulation.During these dry runs, the DFW controllers expressed concern that working with both flight progress strips (FPS) and SMS created too much workload, as there were multiple sources for the same information.They were wary of spending too much time "heads-down" and asked if it would be possible to attempt running the simulation without the flight strips.After several trial runs without flight strips, the controllers requested that flight strips be removed when flight-specific information was presented on the map display, during Conditions 2 and 3.It is important to note that it is not the intent of SMS to replace flight strips, and SMS does not replace several of the functions of flight strips.Air traffic controllers like flight strips for several reasons.The interface is familiar, easy-to-use, helps them instantly understand the current state of the traffic and lets them communicate without interrupting each other4 .The removal of flight strips during Conditions 2 and 3 introduced several issues.Controllers currently record clearances and departure queue assignments through various uses of flight strips, such as writing on them (see Fig.11), "cocking" them, or sliding the strip out, each signifying something different.In this way, physical flight strips serve as a memory aid.Without the flight strips, the controllers were forced to find new +Figure 11 :11Figure 11: Flight Strip from SMS Simulation Data Collection Several different types of data were recorded during the simulation.Visual and audio recordings were made of each run and SMS log files were recorded.SMS log files contain data such as the aircraft target positions, user keyboard entries, runway assignments and advisories, and runways used by each aircraft. +Figure 12 :12Figure 12: Clutter on the Map Display + + + + + + + + + Functionalities, displays, and concept of use for the surface management system [ATC] + + SAtkins + + + CBrinton + + + DWalton + + 10.1109/dasc.2002.1067903 + + + Proceedings. The 21st Digital Avionics Systems Conference + The 21st Digital Avionics Systems ConferenceIrvine, CA + + IEEE + October 27-31, 2002 + + + S. Atkins, C. Brinton, and D. Walton, "Functionalities, Displays, and Concept of Use for the Surface Management System," 21'st Digital Avionics Systems Conference, Irvine, CA, October 27-31, 2002. + + + + + Human Factors Lessons Learned from a Surface Management System Simulation + + DeborahWalton + + + CherylQuinn + + + StephenAtkins + + 10.2514/6.2002-5810 + + + AIAA's Aircraft Technology, Integration, and Operations (ATIO) 2002 Technical Forum + + American Institute of Aeronautics and Astronautics + 2002 + + + Planned to be published as a NASA TM + D. Walton and S. Atkins, "A Simulation Evaluation of the Surface Management System User Interfaces," Planned to be published as a NASA TM, 2002. + + + + + Denver TMA Assessment + + KHarwood + + + BSanford + + + + NASA Contractor Report + + 4554 + October 1993 + + + K. Harwood and B. Sanford, "Denver TMA Assessment," NASA Contractor Report 4554, October 1993. + + + + + Is paper safer? The role of paper flight strips in air traffic control + + WendyEMackay + + 10.1145/331490.331491 + + + ACM Transactions on Computer-Human Interaction + ACM Trans. Comput.-Hum. Interact. + 1073-0516 + 1557-7325 + + 6 + 4 + + December 1999 + Association for Computing Machinery (ACM) + + + W. MacKay, "Is Paper Safer? The Role of Paper Flight Strips in Air Traffic Control," ACM Transactions on Computer-Human Interaction, Vol. 6, No. 4, December 1999, pp 311-340. + + + + + + diff --git a/file775.txt b/file775.txt new file mode 100644 index 0000000000000000000000000000000000000000..fad0995032fb6f51cec689459acce38fb466bb61 --- /dev/null +++ b/file775.txt @@ -0,0 +1,437 @@ + + + + +IntroductionSuccessful integration of Unmanned Aircraft System (UAS) operations into airspaces populated with manned aircraft relies on an effective Detect and Avoid (DAA) System.A DAA system provides surveillance, alerts, and maneuver guidance (referred to as guidance in this report) to keep a UAS "well clear" of other aircraft [1][2][3][4].With research contributions from National Aeronautics and Space Administration (NASA), industry, and the Federal Aviation Administration (FAA), the RTCA Special Committee 228 (SC-228) published the Minimum Operational Performance Standards (MOPS) for DAA systems [5] and air-to-air radar [6] 2 Detect and AvoidThe DAA system aims to keep the UAS "well clear" of other manned aircraft.The DAA system consists of surveillance components, alerting and guidance algorithms, ground control station for UAS operator and pilot, and command and control systems.The DAA MOPS defines quantitatively a DWC around other aircraft the UAS should avoid.Alerts and guidance aim at assisting the UAS operator or pilot in maintaining separation defined by the DWC.The DAA MOPS also defines display requirements for alerts and guidance [7].The DAA MOPS assumes the UAS is flying by instrument flight rules (IFR), and the UAS pilot or operator is expected to coordinate with air traffic control (ATC) for a conflict avoidance maneuver.During the scripted encounters, a cylindrical DWC definition of 450 ft vertical separation and 2200 ft horizontal separation was utilized [8].A loss of DWC (LoDWC) occurs when the horizontal and vertical separations are simultaneously violated [9].The DAA alerting structure is comprised of three alert types: preventive, corrective, and warning.A preventive alert is a caution level alert that requires a maneuver to avoid a predicted DWC violation, but advises the pilot to maintain the UAS's current altitude to avoid conflicts.A corrective alert is a caution level alert that advises the pilot to coordinate with ATC before maneuvering.A warning alert requires immediate action from the pilot to start maneuvering to maintain DWC [7].The scripted encounters in FT6 only trigger corrective and warning alerts.Figure 1 shows the alerting timeline and the corresponding guidance.The guidance includes ranges of heading and altitude, shown as bands on a display system (see Section 3.6), predicted by the DAA system to have a high likelihood of leading to a LoDWC given a look-ahead time of typically 2 to 3 minutes.There is a corresponding guidance type for each alert type.Aircraft performance parameters such as turn, climb, and descent rates can be used for computing the ranges of heading and altitude.To execute a DAA maneuver to maintain DWC, a UAS pilot is expected to select and maneuver the UAS to a heading or altitude that is outside the bands.If the ownship gets too close to the intruder, a LoDWC may become inevitable even with maneuvers.In this situation, the guidance bands fill up all ranges of heading and altitude, but at the same time computes "regain well clear" bands to assist the ownship in maneuvering in order to regain well clear effectively.Regain-well-clear is also referred to as well clear recovery (WCR).For DAA systems assuming finite turn rates and climb/descent rates for the unmanned aircraft (UA), the WCR usually occurs earlier than a LoDWC during an encounter [7].A DAA system's surveillance components include ADS-B, active surveillance, and an onboard radar.ADS-B and active surveillance detect cooperative aircraft, i.e., aircraft that can broadcast ADS-B out messages and/or respond to interrogations by active surveillance.Both ADS-B and active surveillance have decent detection ranges beyond 10 nmi that provides more than enough alerting times for DAA.The onboard radar detects all aircraft, including the non-cooperative aircraft that are without a functioning transponder and cannot be detected by ADS-B or active surveillance.Previous research results show that, for low SWaP operations, a susitable surveillance range for the onboard radar lies between 2 and 3.5 nmi [7,10], considering both safety and operational suitability metrics such as alerting timelines. +Flight Test OperationsFT6 was conducted at NASA Armstrong Flight Research Center, CA in restricted airspace R-2515.FT6 included radar characterization, scripted encounters, and full mission.Nine flight test days were dedicated to the scripted encounters from August 2019 to December 2019.The data analysis in this report focuses on the scripted encounters.The following subsections are organized in this way: Section 3.1 presents the test aircraft.Section 3.2 discusses the encounter design.Section 3.3 describes the specific DAA algorithm for this flight test.Section 3.4 discusses the ground control station and its display.Section 3.5 describes the pilot procedure.Section 3.6 gives details of data collection and the post-processing steps. +Test AircraftThe flight test elements consist of one ownship (the UA) and one intruder manned aircraft per encounter.A detailed introduction on the ownship and intruder is provided in this section.The NASC TigerShark Block 3 XP (N1750X), equipped with the Honeywell DAPA-Lite radar for non-cooperative aircraft active surveillance, served as the ownship during the scripted encounters.It is a medium-altitude and long-endurance singleengine pusher aircraft.The specifications of the Tigershark are shown in Table 1 [9].Table 1: Specifications of Tigershark [9,11] The T-34C Mentor (NASA865), TG-14 Super Ximango (NASA856), and Beechcraft B200 King Air (NASA801) served as the intruders.They were flown against the ownship to evaluate the performance characteristics of the Honeywell DAPA-Lite radar and the DAA system.The T-34C Mentor, TG-14 Super Ximango, and Beechcraft B200 King Air have a maximum speed of 214 kts, 132 kts, and 292 kts, respectively [9].The TigerShark was equipped with ADS-B and Honeywell's DAPA Lite radar.ADS-B detects manned aircraft that are equipped with ADS-B out.Three panels of DAPA-Lite Radar were installed at the nose of TigerShark and were expected to detect and track traffic within its theoretical field of regard (±15 • elevation and ±110 • azimuth) regardless of whether that aircraft has other electronic means of identification [9].The TigerShark's Sagetech XP transponder offers ADS-B out capability, which provides position, altitude, velocity, and other information to the test team.The MXS transponder on the TigerShark receives NMEA GPS data from the Piccolo II autopilot through a serial interface.These data provided situational awareness of the Tigershark's altitude and position to the test team during all of the FT6 flights.The intruder aircraft were also equipped with ADS-B out [9].DAPA-Lite turned out to be unable to provide consistent tracks.Therefore, the test team decided to emulate a low SWaP sensor by filtering ADS-B surveillance data with a simulated surveillance volume.The simulated surveillance has a ±15 • elevation range and ±110 • azimuth range [9].Its range (inter-aircraft distance) varies between 2.0, 2.5, and 3.5 nmi across encounters.The UAS pilots controlled Tigershark through an auto-pilot system called Piccolo II.Piccolo II is a self-contained flight management computer that consists of built-in 3-axis accelerometers, 3-axis gyros, GPS, command and control radio, and various inputs/outputs to interface with external components (e.g.servos, research payloads, transponders).The pilots would type in a heading value in the Piccolo, then press "Send" to uplink the command.For FT6, Piccolo II provided NMEA GPS messages to the Sagetech MXS ADS-B unit through a RS-232 serial interface.A second serial port on the autopilot was interfaced to the Unmanned Aircraft Processor (UAP) to provide telemetry from the air data system that was not available from the VN-210 EGI [9].Towards the second half of the scripted encounters, pilots switched to Vigilant Spirit Control Station (VSCS) for issuing heading commands to be more representative of an integrated UAS where DAA information and vehicle commands would be executed from a common display. +Encounter Scenario Design +Encounter GeometryThere were several types of encounter geometries: Head-on, crossing, maneuvering intruder, ascending and descending intruder.The encounters included five different headings of the intruder relative to the ownship, 0 • , 40 • , 50 • , 80 • , 120 • .At the closest point of approach (CPA), a combination of vertical and/or lateral offsets was required of each encounter for safety.The lateral offset was 0 nmi when the vertical offset was 500 ft, and 0.3 nmi when the vertical offset was less than 500 ft.The ascent rate and descent rate were either 500 fpm or 1000 fpm.For encounters with a maneuver intruder, the intruder turned into the ownship at a heading of 40 • or 80 • relative to the ownship. +Detect-and-Avoid AlgorithmThe open-source Detect and AvoID Alerting Logic for Unmanned Systems (DAIDALUS) is a reference DAA algorithm for the Phase 1 MOPS [7].The DAIDALUS reference implementation computations of alerts are based on alerters.An alerter consists of three sets of DWC thresholds labeled "FAR," "MID," and "NEAR."The FAR, MID, and NEAR threshold values are intended to correspond respectively to the preventive, corrective, and warning alerting volumes and times.DAIDALUS's guidance computation for this work is based on a constant turn rate, constant acceleration of the ownship, and constant-velocity projections of traffic aircraft.These bands are computed for the MID and NEAR threshold values provided in the alerter [6].In DAIDALUS, a corrective alert was configured to indicate a predicted LoDWC within 60 seconds or less.A warning alert was configured to indicate a predicted LoDWC within 30 seconds or less.The vertical alerting threshold for both corrective and warning alerts was increased from 450 ft, the DWC's vertical threshold, to a large value of 4000 ft.The purpose of expanding this threshold from 450 ft was to keep the alerts stable and remove the effect of vertical offsets in the flight cards on the alerts.The horizontal distance threshold was also increased from 2200 ft to 3342 ft in most encounters to reduce the impact of sensor uncertainties on the stability of alerts.The maneuver guidance provided by DAIDALUS is presented in the form of conflict bands, i.e., ranges of headings and altitudes that lead to a well-clear violation, or recovery bands, i.e., ranges of ownship maneuvers that recover from a present or unavoidable well-clear violation.WCR will show up when the guidance band is saturated to maximize separation and avoid Near Mid-Air Collision (NMAC) [6].During the flight test, DAIDALUS was invoked as a plug-in to the Java Architecture for DAA Extendibility and Model (JADEM) tool suite.JADEM saves aircraft states, alerts, and guidance data to files.An additional wrapper layer around JA-DEM, called SaaProc, provides real-time message handling, and filtering aircraft states by the intruder's relative position to the ownship [9,12]. +Ground Control StationThe Research Ground Control Station (RGCS) consists of the VSCS with integrated JADEM & DAIDALUS DAA functionality, an interface to the Live Virtual Constructive (LVC) gateway, and a Plexsys voice communications link to the pseudopilot and ATC stations.DAA maneuvers were sent to Piccolo, which relayed them to the UA.Command and Control datalinks, comprised of an uplink and a downlink, to the UA were provided by a Piccolo-based Ground Control Station (GCS) and a Silvus Technologies radio system [9]. +Pilot's Background & ProcedureThe recruited Subject Pilots Under Test (SPUTs) for FT6 were active military with a fixed wing license and have had recent flying experience including flying medium to large fixed wing remotely piloted aircraft within the last year [13].During the scripted encounters, the UAS pilots were expected to select at their discretion a heading outside the conflict bands, i.e., a heading that was predicted to be conflict-free.WCR guidance was computed and displayed to pilots when the conflict bands saturate all the headings.In this situation, pilots were expected to select a heading within the range of the positive guidance provided by the WCR.The mitigated test matrix instructs the UAS pilot to execute a maneuver 8 seconds after a corrective alert is triggered or 3 seconds after a warning alert is triggered.Some flight cards required that the UAS pilot execute maneuver upon initiation of a warning alert or an alert that appears first.Results show that, due to various reasons, UAS pilots did not adhere strictly to the instructions regarding timing and, in some encounters, executed maneuvers very late. +Data Collection and Post-ProcessingData were collected during the flight test and post-processed.Figure 4 shows the data collection processes and post-processes.The solid arrows represent data flow in real time; the dotted arrows represent data flow in post-processing.The dashed boxes stand for processes or devices, and the solid boxes stand for data storage.The Sense-and-Avoid Processor (SAAP) recorded and stored the ownship's surveillance data from individual sensors.The surveillance data were sent to an onboard fusion tracker from Honeywell that performed track coordinate transformation, fil-tering, association, and fusion.The LVC Gateway received flight state messages, time-stamped them, and forwarded them to the SaaProc [14].The SaaProc read the flight states, computed alerts and guidance, and sent the guidance to the LVC Gateway.The LVC Gateway not only logged the flight state messages in LVC message files but also received guidance messages from SaaProc and forwarded them to VSCS.VSCS also sent pilot's maneuvers to the LVC Gateway.SaaProc recorded the flight states, alerts, and maneuvers as JADEM log files [12].It was necessary to post-process some flight test data to see if the alert was active when the pilot-selected avoidance maneuver heading was captured.FT6 LVC messages from the scripted encounters were post-processed into encounter files that included aircraft state using an LVC translator.Afterwards, the encounter files were utilized to re-compute alert and guidance using SaaProc configuration files that were configured to remove the simulated radar field of view to produce a new set of JADEM log files.4 Analysis and Results +Flight Cards SummarySystem checkout (SCO) flights were flown in August of 2019 prior to the data collection days for radar characterization and scripted encounters (SE).Flight data from System Checkout flight days and scripted encounters showed good quality and were included for data analysis.Figure 6 depicts the overall maneuver outcome categorization.Data analysis focused on the 90 mitigated encounters.These encounters were classified based on the outcome of alert, guidance, and effectiveness of pilots' maneuver in resolving conflicts.Twenty-four encounters were excluded from additional analysis due to data issues such as segmentation fault on the UAP computer.The remaining encounters were 66.A maneuver was ineffective if, at the completion of the maneuver, the alert was still active.Among the 66 encounters analyzed for maneuver effectiveness, 36 of these had maneuvers effectively resolving conflicts.Six encounters with maneuvers based on WCR were also excluded.The remaining encounters with ineffective maneuvers were further analyzed for potential causes.Among those 30 encounters with ineffective maneuvers, fourteen were attributed to pilots' decisions, one attributed to a true heading/true course mismatch, and the other 9 were attributed to trajectory prediction errors.Three sources of trajectory prediction errors were analyzed: change of intruder's velocity, wind error, and turn rate error.An encounter could exhibit multiple sources of trajectory error.Six of the encounters in the pilots' decision category also had contributions from change of intruders' velocities.The lead contributing factors were pilots' decision and change of intruders' velocities.During the scripted encounters flight test phase, a true heading and true course mismatch issue in the maneuver guidance execution was found.Pilots selected a horizontal maneuver based on the true heading conflict bands shown on the DAA display.However, when the maneuver was inputted to Piccolo, its heading value was interpreted by Piccolo as a true course.This mismatch caused the actual maneuver to deviate slightly from the intended one, the degree of which varies with the wind condition.It is necessary to investigate whether the corresponding target heading is in the conflict band at the maneuver.Target heading θ h & target course θ c difference is given by| θ h -θ c |= arccos V tas 2 + V 2 g -V 2 w 2V tas V g(1)Target heading θ h is given byθ h = θ c + arccos V tas 2 + V 2 g -V 2 w 2V tas V g(2)orθ h = θ c -arccos V tas 2 + V 2 g -V 2 w 2V tas V g(3)Target heading θ h can be calculated using Eq. 2 or Eq. 3 depending on the trigonometric relationship between the wind speed vector -→ V w , true airspeed vector --→ V tas , and ground speed vector -→ V g .Figure 7 shows whether the target heading or/and target course is/are conflict-free at the time of maneuver.The time of a maneuver is estimated from the ownship's state data by observing the first time the ownship started turning.Note this time may be a few seconds after the time the pilot submitted the heading command on the ground due to the system's latency.The circle represents "conflict-free," and the cross represents "in conflict."When both the target heading and the target course are in the conflict band, it will be called "Pilots' Decision."If the target heading is in conflict, but the target course is conflict-free, it will be called "True Course & True Heading Mismatch."The sources of error, such as change of intruder's velocity, wind error, and turn rate error, will be analyzed when the target heading is conflict-free. +Explanations of Categories with Examples +I. ExclusionThe exclusion category indicated that alerts were not properly triggered, or there were technical difficulties getting the data.When the timing was off, alerts were unstable, or Sensor Uncertainty Mitigation (SUM) led to very early WCRs, the flight cards were considered having improper alerts.The encounters were excluded if the segmentation faults occurred on the UAP computer that managed the payload, or data were not obtained. +II. Maneuver upon Well Clear RecoveryThe pilots selected a target heading based on the WCR guidance at the maneuver.Maneuvering upon WCR was undesirable because the UAS pilots maneuvered late. +III. Effective ManeuverEffective maneuver indicated that the ownship was conflict-free at the end of the maneuver.Namely, all the alerts disappeared when the turn was completed.There were 34 effective maneuvers.Figure 9 presented an example of an effective maneuver. +IV. Pilots' DecisionPilots' decision indicated that the ineffectiveness of the maneuvers was caused by both the target course that the pilots selected and its calculated target heading in the conflict bands.Figure 10 presented an example of pilots' decision. +VI. Change of Intruder's VelocityThe criteria is given by| --→ V int |=| --→ V (t) int - - → V start,M,OS |≥ 10kts(4)According to Eq. 4, the change of intruder's velocity is calculated by the magnitude of the vector difference between the intruder's velocity as a function of time and intruder's velocity at the start of ownship maneuver.There were five encounters in this category.Three out of the total five encounters had a maneuvering intruder.Figure 12 shows an intruder that kept turning, which caused difficulty to the DAA algorithm.A right turn appeared to be feasible for resolving the conflict at one time.However, the intruder maneuver "closed the gate" afterwards.According for intruder's accelerations may improve predictions. +VII. Wind ErrorThe criteria is given by| -→ w f --→ w i | | -→ w i | ≥ 40%(5)If the magnitude of the vector difference between the wind velocity before the maneuver -→ w i and wind velocity after the turn is completed -→ w f is over 40% of the magnitude of wind velocity | -→ w i | before the maneuver, the encounter will be classified as wind error.There were three encounters that belonged to this category.Figure 13 shows an encounter where both the magnitude of the wind velocity and the wind direction change were large. +VIII. Turn Rate ErrorThe criteria is given by| t act -t pred | t pred ≥ 50%(6)If the difference between the actual turn time t act and the predicted turn time t pred is over 50% of t pred , the encounter will be considered having turn rate error.t pred is calculated using a standard turn rate of 7 deg/s, and t act is measured from the start of the turn to the end of the turn with a cutoff turn rate of 0.75 deg/s.When the turn is small, the turn error tends to be larger because of its lower average turn rate.There were seven encounters in turn rate error category.Figure 14 presents an encounter where the turn rate error was 87.36 % and had an average turn rate of 3.48 deg/s because of its small turn. +Breakdown by Surveillance RangeFor the mitigated encounters, a simulated surveillance range was chosen to be 3.5 nmi, 2.5 nmi, and 2.0 nmi, which were potential candidates for required surveillance ranges for low SWaP sensors.The number of encounters that were flown with 3.5 nmi, 2.5 nmi, and 2.0 nmi surveillance ranges is 35, 36, and 19, respectively.After removing the flight cards from the exclusion and WCR category, the effectiveness and ineffectiveness rates for each surveillance range were tabulated in Table 3. Results suggest that encounters with 3.5 nmi had a higher probability of achieving effective DAA maneuvers. +Trajectory Error AnalysisThe DAIDALUS modeled the turn using an infinite roll rate at the start of the maneuver and at the end of the turn and a fixed turn rate during the turn.A simulation was run to emulate the turn using the initial and target conditions from the ownship state data collected during FT6.Initial conditions for the trajectory error simulation were wind speed, wind direction, true course, altitude, aircraft position, and aircraft ground speed with a standard turn rate of 7 deg/s.Target conditions were altitude, true course, and true heading.Wind speed, wind direction, aircraft ground speed, and altitude were held constant in the simulation.Figure 15 shows the discrepancy between the simulated and actual trajectories during the turn and after the turn.The simulated trajectory turned faster compared to the actual flight test trajectory.Figure 16 shows that the turn rate for the simulated turn was at 7 deg/s during the turn; however, the actual turn rate was increased to 7.4 deg/s within the first few seconds, then fluctuated between 5 to 7 deg/s, and eventually went down.Figure 17 indicates that the majority of the maximum turn rates for mitigated encounters was between 6 to 8 deg/s.The FT6 data showed that the range of turn rate was reasonable. +Buffered Heading and Encounter Effectiveness AnalysisAdding a buffer to the DAA's suggested minimum turn angle creates a "buffered heading." Figure 18 depicts how the buffered heading affected the effectiveness of the encounters.The analysis likely included the effects of the mismatch between true course and true heading, but the effect on the overall results is expected to be minor.The angle was calculated using the difference between the target course & the edge of the conflict band.When the angle was negative, the target course was in conflict.If the angle was positive, the target course was conflict-free.If the angle was less than -75 deg, band saturation might have occurred.The majority of the ineffective encounters had pilots' maneuvers inside the conflict band.Whereas, maneuvers were more effective when buffers were larger.Therefore, adding more buffer to the target heading may increase the probability of effectiveness. +ConclusionsThe analysis of FT6 data was focused on classifying mitigated encounters based on the effectiveness of the DAA maneuver in resolving conflicts.For maneuvers executed in a timely fashion before WCR, more than half of these maneuvers effectively resolved conflicts.When categorized by surveillance range, a 3.5 nmi surveillance range achieved a higher success rate (about 70%) than 2.5 and 2.0 nmi (about 50%).Enlarging the surveillance range from 2.0 nmi and 2.5 nmi to 3.5 nmi increased the effectiveness of the maneuver.When the maneuver turned out to be ineffective, additional analysis was done to determine the cause(s) of the ineffectiveness, such as pilots' decision, the true course and true heading mismatch, and trajectory prediction errors.Three main sources of trajectory prediction error were analyzed: change of intruder velocity, wind error, and turn rate error.Among all the ineffective encounters, the major contributing causes are pilots' decision and change of intruders' velocities.When analyzing the category of change of intruder's velocity, it was observed that maneuvering intruders presented a challenging case for the DAA algorithm.The data show that accounting for potential accelerations in the intruder's trajectory should be strongly considered for future flight tests.Trajectory error analysis successfully showed the discrepancies between the actual and simulated turn rates.Besides, the actual and simulated trajectories were plotted graphically to show the deviation between the two.Buffered heading & encounter effectiveness analysis was conducted.It was ascertained from the results that it may be beneficial for pilots to add more buffer to the target heading to increase the maneuver effectiveness.Figure 1 :1Figure 1: Alerting Timeline[7] +. The flight matrix shows initial point (IP)'s & CPA's locations, angle of intruder's heading relative to the ownship, ownship's altitude & ground speed, intruder's altitude & ground speed, vertical offset, CPA lateral offset, climb or descent rate, and maneuver point's (MP) location for each encounter.The flight time between an IP and a CPA is 2 minutes.A flight card for an encounter is comprised of 2 parts: ownship and intruder.Figure 2 and Figure 3 are dedicated to those ownship and intruder components, respectively.The flight card shows the top view graphics, horizontal view graphics, IP/CPA names and coordinates, altitudes, headings, distances, ground speeds, sensor selection, abort procedures, etc [9]. +Figure 2 :2Figure 2: Flight card for the Ownship[9] +Figure 3 : 9 ]39Figure 3: Flight card for the Intruder[9] +Figure 4 :4Figure 4: Data Flow Diagram +Figure 5 :5Figure 5: Vigilant Spirit Display +4. 22Categorization by Maneuver Outcomes 4.2.1 Flight Card Categorization Scheme & Statistics +Figure 6 :6Figure 6: Categorization by Maneuver Outcomes (Number of Encounters in Associated Classification) +Figure 7 :7Figure 7: Target Course & Target Heading Decision Chart +Figure 88shows an example of WCR.The upper left panel shows the true course and true heading of the ownship v.s.time.The red color and green color indicate the warning band and the WCR guidance, respectively.The upper mid panel shows the trajectories for the ownship and the intruder.The upper right panel shows the distance between the ownship and the intruder as the time elapses.The bottom left panel shows the wind direction relative to the true north measured from the ownship.The bottom middle panel shows the ground speed for the ownship and the intruder as well as the wind speed.The bottom right panel shows ownship's turn rate as the time elapses. +Figure 8 :8Figure 8: Example of WCR +Figure 9 :9Figure 9: Example of an Effective Maneuver +Figure 10 :10Figure 10: Example of Pilots' Decision +Figure 11 :11Figure 11: Example of True Course & True Heading Mismatch +Figure 12 :12Figure 12: Example of Change of Intruder's Velocity +Figure 13 :13Figure 13: Example of Wind Error +Figure 14 :14Figure 14: Example of Turn Rate Error +Figure 15 :15Figure 15: Trajectory Comparison +Figure 17 :17Figure 17: Turn Rate Histogram +Figure 18 :18Figure 18: Buffered Heading & Encounter Effectiveness Histogram +Flight Card Categorization Scheme & Statistics . . . . . . . .14 4.2.2Explanations of Categories with Examples . . . . . . . . . . .16 4.3 Breakdown by Surveillance Range . . . . . . . . . . . . . . . . . . .20 4.4 Trajectory Error Analysis . . . . . . . . . . . . . . . . . . . . . . . .21 4.5 Buffered Heading and Encounter Effectiveness Analysis . . . . . . .22NomenclatureADS-B= Automatic Dependent Surveillance-BroadcastATC= Air Traffic ControlCPA= Closest Point of ApproachDAA= Detect and AvoidDAIDALUS = Detect and AvoID Alerting Logic for Unmanned SystemsDWC= DAA Well ClearFAA= Federal Aviation AdministrationFT6= Flight Test 6GCS= Ground Control StationHITL= Human-in-the-LoopIP= Initial PointJADEM= Java Architecture for DAA Extendibility and ModelingLoDWC= Loss of DWCLVC= Live Virtual ConstructiveMOPS= Minimum Operational Performance StandardsNAS= National Airspace System1 Introduction NASA = National Aeronautics and Space Administration4NMAC= Near Mid-Air Collision2 Detect and Avoid RGCS = Research Ground Control Station SCO = System Checkout53 Flight Test Operations SPUT = Subject Pilots Under Test SWaP = Low Size, Weight, and Power6t act= Actual Turn Timeθ c= Target Courseθ h= Target Headingt pred= Predicted Turn TimeUA= Unmanned AircraftUAP= Unmanned Aircraft ProcessorUAS= Unmanned Aircraft SystemV g -→ V g -→ V start,M,OS = Intruder Velocity at the Start of Ownship Maneuver = Ground Speed = Ground Speed Vector --→ V int = Change of Intruder SpeedV tas --→ V tas --→ V (t) int= True Airspeed = True Airspeed Vector = Intruder VelocityVSCS= Vigilant Spirit Control StationV w -→ V w= Wind Speed = Wind Speed VectorWCR -→ w i -→ w f= Well Clear Recovery = Initial Wind Velocity = Final Wind Velocity5 Conclusions233.1 Test Aircraft . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .6 3.2 Encounter Scenario Design . . . . . . . . . . . . . . . . . . . . . . .8 3.2.1 Encounter Geometry . . . . . . . . . . . . . . . . . . . . . . .8 3.3 Detect-and-Avoid Algorithm . . . . . . . . . . . . . . . . . . . . . .11 3.4 Ground Control Station . . . . . . . . . . . . . . . . . . . . . . . . .11 3.5 Pilot's Background & Procedure . . . . . . . . . . . . . . . . . . . .12 3.6 Data Collection and Post-Processing . . . . . . . . . . . . . . . . . .12 4 Analysis and Results 13 4.1 Flight Cards Summary . . . . . . . . . . . . . . . . . . . . . . . . . .13 4.2 Categorization by Maneuver Outcomes . . . . . . . . . . . . . . . . .14 4.2.1 +in 2017.The corresponding Technical Standard Orders (TSO), TSO-C211 and TSO-C212, were published by the FAA in October 2017.A DAA system, according to the published DAA MOPS, includes surveillance components of Automatic Dependent Surveillance-Broadcast (ADS-B) In, airborne active surveillance, and air-to-air radar that can detect aircraft with or without transponders.Additional development, named Phase 2, for extending the MOPS to additional UAS categories and operations is underway.The rest of the paper is organized as follows: Section 2 provides background information on DAA, and Section 3 describes flight operations.Section 4 discusses data analysis results, such as effectiveness of DAA maneuvers, flight card categorization, alerting performace, and trajectory errors.Flight Test 6 (FT6) was conducted at NASA Armstrong Flight Research Centerfrom August to November 2019. The overall objective of FT6 was to collect data toinform the development of the Phase 2 DAA and low SWaP sensors' requirements.FT6 was conducted in 3 configurations, each having its own sub-objective: RadarCharacterization assessed the performance of Honeywell's DAPA-Lite. Scripted En-counters validated the performance of a DAA system under a limited surveillanceOne objective of the Phase 2 development seeks to define requirements for operations of UAS equipped with low size, weight, and power (SWaP) sensors, or low SWaP UAS.While ADS-B and active surveillance can fit in the payload of many mediumsized UAS, the large, high-power radar required by the Phase 1 MOPS is physically infeasible and/or economically impractical for many UAS operations.Low SWaP sensors have favorable payloads but provide smaller surveillance volumes.For safety and operational suitability, UAS pilots need sufficient alerting times to evaluate and execute DAA maneuvers in order to maintain separation defined by the DAA Well Clear (DWC).Thus, UAS equipped with Low SWaP sensors have speed restrictions to help ensure pilots have sufficient alerting time.The UAS Integration in the National Airspace System (NAS) Project at NASA established partnership with Honeywell International in 2017 to conduct a shared resource project for further development of a prototype airborne low SWaP surveillance system.Honeywell provided a prototype airborne radar called DAPA-Lite as a candidate for validating and verifying proposed performance requirements for low SWaP surveillance systems within a DAA system.NASA provided UAS integration support as well as flight test planning and execution.volume.Full Mission collected subject pilot's performance data performing DAA tasks in a real-world test environment to validate previous human-in-the-loop simulations.This report focuses on analysis of the scripted encounters data. +Table 2 :2Scripted Encounters Flight Matrix +Table 22lists ownship and intruder flight information +The flight days from which data are analyzed are SCO#5(08/13/19), SCO#6(08/22/19), SCO#7(08/28/19), SCO#8(08/29/19), SCO#9(09/24/19), SE1(10/01/19), SE2(10/03/2019), SE3(10/08/2019), and SCO#10(10/16/19).A total of 96 encounters were attempted.Among these 96 encounters, 72 led to a collection of alerting and guidance data successfully.The other 24 encounter data was not collected due to the triggering of improper alerts.The total flight hours were 29.8. +Table 3 :3Effectiveness & Ineffectiveness Breakdown by Surveillance Range + + + + + + + + + Well Clear Trade Study for Unmanned Aircraft System Detect And Avoid with Non-Cooperative Aircraft + + MinghongGWu + + + AndrewCCone + + + SeungmanLee + + + ChristineChen + + + MatthewWEdwards + + + DevinPJack + + 10.2514/6.2018-2876 + AIAA-2018-2876 + + + + 2018 Aviation Technology, Integration, and Operations Conference + + American Institute of Aeronautics and Astronautics + June 2018 + + + Wu, M. G.; Cone, A. C.; Lee, S.; Chen, C.; Edwards, E. W. M.; and Jack, D. P.: Well Clear Trade Study for Unmanned Aircraft System Detect And Avoid with Non-Cooperative Aircraft. 18th AIAA Aviation Technology, Integration, and Operations Conference, AIAA-2018-2876, June 2018. URL https://arc.aiaa.org/doi/pdf/10.2514/6.2018-2876. + + + + + A Recommended DAA Well-Clear Definition for the Terminal Environment + + MichaelJVincent + + + AnnaTrujillo + + + DevinPJack + + + KeithDHoffler + + + DimitriosTsakpinis + + 10.2514/6.2018-2873 + AIAA-2018-2873 + + + + 2018 Aviation Technology, Integration, and Operations Conference + + American Institute of Aeronautics and Astronautics + 2018. 2018 + + + Vincent, M. J.; Trujillo, A.; Jack, D. P.; Hoffler, K. D.; and Tsakpinis, D.: A Recommended DAA Well-Clear Definition for the Terminal Environment. 2018 Aviation Technology, Integration, and Operations Conference, AIAA-2018-2873, 2018. URL https://doi.org/10.2514/6.2018-2873. + + + + + Defining Well Clear for Unmanned Aircraft Systems + + StephenPCook + + + DallasBrooks + + + RodneyCole + + + DavisHackenberg + + + VincentRaska + + 10.2514/6.2015-0481 + AIAA-2015-0481 + + + + AIAA Infotech @ Aerospace + + American Institute of Aeronautics and Astronautics + 2015 + + + Cook, S. P.; Brooks, D.; Cole, R.; Hackenberg, D.; and Raska, V.: Defining Well Clear for Unmanned Aircraft Systems. Proceed- ings of AIAA Infotech@ Aerospace, AIAA-2015-0481, AIAA, 2015. URL https://doi.org/10.2514/6.2015-0481. + + + + + Characteristics of a Well Clear Definition and Alerting Criteria for Encounters between + + MJohnson + + + ERMueller + + + CSantiago + + + + + Europe Air Traffic Management Research and Development Seminar + + + 2015 + + + UAS and Manned Aircraft in Class E Airspace. Eleventh UAS + Johnson, M.; Mueller, E. R.; and Santiago, C.: Characteristics of a Well Clear Definition and Alerting Criteria for Encounters between UAS and Manned Aircraft in Class E Airspace. Eleventh UAS/Europe Air Traffic Management Research and Development Seminar , 2015, pp. 23-26. URL https://ntrs.nasa.gov/search.jsp?R=20190027490. + + + + + Minimum Operational Performance Standard (MOPS) for Helicopter Hoist Systems + 10.4271/as6342 + + 2017 + SAE International + + + Minimum Operational Performance Standards (MOPS) for Detect and Avoid (DAA) Systems. DO-365, RTCA. Inc., 2017. + + + + + AeroMACS minimum operational performance standards (MOPS) compliance field trials for Hitachi prototype + + RafaelApaza + + 10.1109/icnsurv.2015.7121331 + + + 2015 Integrated Communication, Navigation and Surveillance Conference (ICNS) + + IEEE + 2017 + + + Minimum Operational Performance Standards (MOPS) for Air-to-Air Radar for Traffic Surveillance. DO-366, RTCA. Inc., 2017. + + + + + Correction: Detect and Avoid Alerting Performance with Limited Surveillance Volume for Non-Cooperative Aircraft + + MinghongGWu + + + SeungmanLee + + + AndrewCCone + + 10.2514/6.2019-2073.c1 + + + + AIAA Scitech 2019 Forum + + American Institute of Aeronautics and Astronautics + 2019 + + + Wu, M. G.; Cone, A. C.; and Lee, S.: Detect and Avoid Alerting Performance with Limited Surveillance Volume for Non-Cooperative Aircraft. 2019. URL https://doi.org/10.2514/6.2019-2073.c1. + + + + + Detect-and-Avoid Alerting Performance for High-Speed UAS and Non-Cooperative Aircraft + + AndrewCCone + + + MinghongGWu + + + SeungManLee + + 10.2514/6.2019-3313 + + + + AIAA Aviation 2019 Forum + + American Institute of Aeronautics and Astronautics + Jun 2019 + + + Cone, A. C.; Wu, M. G.; and Lee, S.: Detect-and-Avoid Alerting Perfor- mance for High-Speed UAS and Non-Cooperative Aircraft. Jun 2019. URL https://arc.aiaa.org/doi/pdf/10.2514/6.2019-3313. + + + + + + MVincent + + + GWu + + + CRorie + + + SKim + + + MMarston + + + AFlock + + + THoang + + + RNavarro + + + + Integrated Test and Evaluation (IT&E) + Flight Test Series 6 Flight Test Report + + + Vincent, M.; Wu, G.; Rorie, C.; Kim, S.; Marston, M.; Flock, A.; Hoang, T.; and Navarro, R.: Integrated Test and Evaluation (IT&E) Flight Test Series 6 Flight Test Report. + + + + + Defining Well Clear Separation for Unmanned Aircraft Systems Operating with Noncooperative Aircraft + + ChristineChen + + + MatthewWEdwards + + + BilalGill + + + SamanthaSmearcheck + + + TonyAdami + + + SeanCalhoun + + + MinghongGWu + + + AndrewCone + + + SeungManLee + + 10.2514/6.2019-3512 + + + + AIAA Aviation 2019 Forum + + American Institute of Aeronautics and Astronautics + 2019 + + + Chen, C. C.; Gill, B.; Edwards, M. W. M.; Smearcheck, S.; Adami, T.; Calhoun, S.; Wu, M. G.; Cone, A. C.; and Lee, S.: Defining Well Clear Separation for Unmanned Aircraft Systems Operating with Non-cooperative Aircraft. AIAA, 2019. URL https://arc.aiaa.org/doi/pdf/10.2514/6.2019-3512. + + + + + Flight-Test Plan Design and Evaluation in a Closed-Loop Framework for a General Aviation Aircraft + + MMarston + + + SKim + + + VLoera + + + MVincent + + + RNavarro + + 10.2514/6.2022-2170.vid + + + Flight Test Series 6 Flight Test Plan Document + + + American Institute of Aeronautics and Astronautics (AIAA) + + + Marston, M.; Kim, S.; Loera, V.; Vincent, M.; and Navarro, R.: Integrated Test and Evaluation (IT&E) Flight Test Series 6 Flight Test Plan Document. + + + + + UAS Integration in the NAS Project: Flight Test 3 Data Analysis of JADEM-Autoresolver Detect and Avoid System + + CGong + + + MGWu + + + CSantiago + + + Dec 2016 + + + Gong, C.; Wu, M. G.; and Santiago, C.: UAS Integration in the NAS Project: Flight Test 3 Data Analysis of JADEM-Autoresolver Detect and Avoid System. , Dec 2016. + + + + + + + WBridges + + + RZach + + + + UAS-NAS Flight Test Series + + 6 + 2020 + + + Bridges, W.; and Zach, R.: UAS-NAS Flight Test Series 6, Ames Research Center Test Report, 2020. + + + + + RUMS - Realtime Visualization and Evaluation of Live, Virtual, Constructive Simulation Data + + GeorgeSoler + + + SrbaJovic + + + JamesRMurphy + + 10.2514/6.2015-1648 + + + + AIAA Infotech @ Aerospace + + American Institute of Aeronautics and Astronautics + 2015 + + + Soler, G.; Jovic, S.; and Murphy, J.: RUMS-Realtime Visualization and Eval- uation of Live, Virtual, Constructive Simulation Data. AIAA, 2015. URL https://ntrs.nasa.gov/search.jsp?R=20160007498. + + + + + + diff --git a/file776.txt b/file776.txt new file mode 100644 index 0000000000000000000000000000000000000000..a7e1bafc359890b1a0bb25b260f98bce48d8ddcf --- /dev/null +++ b/file776.txt @@ -0,0 +1,852 @@ + + + + +I. IntroductionContrails are artificial clouds that are the visible trails of condensed water vapor made by the exhaust of aircraft engines.Depending on atmospheric conditions, contrails may be visible for only a few seconds or minutes, or may persist for many hours which may a↵ect climate.Persistent contrails reduce incoming solar radiation and outgoing thermal radiation in a way that accumulates heat. 1 The global mean contrail cover in 1992 was estimated to double by 2015, and quadruple by 2050 due to air tra c increase. 2 Studies suggest that the environmental impact from persistent contrail is estimated to be three or four times, 3 or even ten times 4 larger than aviation-induced emissions.Therefore, strategies and policies to reduce aircraft induced persistent contrail need to be studied to minimize the impact on the global environment.There have been some research works accomplished to identify and reduce persistent contrail formation.Gierens 5 and Noppel 6 reviewed various strategies for contrail avoidance including changing engine architecture, enhancing airframe and engine integration, using alternate fuels, and modifying tra c flow management procedures.Among the tra c flow management solutions, Mannstein 7 presented a strategy to reduce the environment impact of contrails significantly by small changes to each aircraft's flight altitude.Campbell 8 proposed a mixed integer programming method to optimally reroute aircraft to avoid the formation of persistent contrails.Both methods require the onboard contrail detection system.Fichter 9 showed that the global annual mean contrail coverage could be reduced by decreasing the aircraft cruise altitude.Williams 10,11 proposed strategies for contrail reduction by identifying fixed and varying maximum altitude restrictions.These restrictions generally require more fuel burn and add congestion to the already crowded airspace at lower altitudes.Sridhar et al. 12 provided a set of strategies to reduce contrail formation.A 3D grid model was constructed based on weather data to describe the contrail potential areas and the cruising level of aircraft is adjusted to avoid these areas with the consideration of fuel-e ciency.The goal of this paper is to refine the research work in 12 by introducing two operationally feasible constraints from air tra c control perspective with the consultation from experienced and expert controllers.Moreover, we formulate the problem as integer programming.Because of the total unimodularity of its constraint matrix, solving the relaxed linear programming by simplex method will yield the optimal and integral solution with much shorter computation time.The rest of the paper is organized as follows.Section II introduces the atmospheric data, contrail formation model, aircraft data and contrail frequency index used in this paper.Section III presents the problem formulation and our contrail reduction strategy.The numerical results are shown in Section IV.Section V summarizes and concludes the paper.Contrails can be observed from surface data 13 and detected by satellite data. 14Duda 15 has related the observations to numerical weather analysis output and showed that persistent contrail formation can be computed using atmospheric temperature and humidity data retrieved from the Rapid Updated Cycle (RUC) data provided by the National Oceanic and Atmospheric Administration (NOAA).Contrails can persist when ambient air is supersaturated with respect to ice, which means that relative humidity with respect to ice (RHi) is greater than one hundred percent. 16The RHi can be computed from relative humidity with respect to water (RHw) and temperature, which are available in the RUC data.The one-hour forecast and the 40-km RUC data are used in this paper.The data have a temporal resolution of one hour, a horizontal resolution of 40 kilometers, and isobaric pressure levels from 100 to 1000 hectopascals (hPa) with 25 hPa increments.The vertical range of interest in this study is from 150 hPa to 400 hPa, which is equivalent to pressure altitude of about 23,600 feet to 44,400 feet.The temperature and RHw contours at 8AM eastern daylight time (EDT) on August 1, 2007 at pressure altitude 250 hPa, or 34,057 feet, are shown as the left and right subfigures in Fig. 1. +II. Data and Model +A. Atmospheric Data +B. Contrail Formation ModelThe potential persistent contrail formation areas (contrail areas) are defined as areas with RHi greater than or equal to 100%.RHi can be computed from RHw and temperature using the saturation vapor pressure coe cients of Alduchov, 17 formulated as RHi = RHw ⇥ 6.0612e 18.102T /(249.52+T ) 6.1162e 22.577T /(237.78+T ) ,where T is the temperature in Celsius.The atmospheric profile shown in the left and right subfigures in Fig. 1 can be translated to a contour of RHi, as shown in Fig. 2. The 40-km RUC data have (113 ⇥ 151) data points.The altitude level index l is defined as l = 1, 2, ..., 11 corresponding to isobaric pressure level at 400, 375, ..., 150 hPa.The level index, isobaric pressure level, and approximate aircraft cruising altitude are listed in Table 1.The potential persistent contrail formation matrix (contrail matrix) at time t at level l is defined asR l t = 0 B B @ r 1,1 r 1,2 . . .1 C C A ,(2)where r i,j is 1 if RHi 100% at grid (i, j) and 0 if RHi < 100%. +C. Aircraft DataContrails form when aircraft fly through a potential contrail formation area.Thus aircraft locations are needed to determine the contrail formation frequency.The aircraft data used in this paper are obtained from the aircraft locations provided by the Federal Aviation Administration's (FAA's) Aircraft Situation Display to Industry (ASDI) data.The ASDI has a sampling rate of one minute.The same 3D grid used in the RUC data is used to generate the aircraft position matrix.The aircraft position matrix is defined asA l t = 0 B B @ a 1,1 C C A ,(3)where a i,j is the number of aircraft within grid (i, j) flying closest to level l at time t.The aircraft position matrix indicates the air tra c density in the grid scale at di↵erent altitudes. +D. Contrail Frequency IndexContrail frequency index is defined as the number of aircraft that would fly through potential contrail formation regions during a period of time.Center contrail frequency index is used to indicate the contrail severity in a given center. 18To specify the location of the twenty U.S. centers in the grid scale, the center grid matrix is defined asC center = 0 B B @ c 1,1 c 1,2 . . . c 1,151 . . . . . . . . . . . . c 113,1 c 113,2 . . . c 113,151 1 C C A ,(4)where c i,j is one if the grid point is within the center and zero if not.The center contrail frequency index is defined as the number of aircraft flying through contrail area at time t at level l, formulated asCF I center,l,t = 113 X i=1 151 X j=1 r i,j a i,j c i,j ,(5)where r i,j , a i,j , c i,j are defined in Eqn. ( 2), ( 3), ( 4).The contrail frequency index derived in the previous section indicates the actual contrail activities.For strategic planning, prediction of the contrail frequency at time t at level l of a certain center is calculated byCF I center,l,t = 113 X i=1 151 X j=1 r 0 i,j a 0 i,j c i,j ,(6)where r 0 i,j is from RUC forecast data and a 0 i,j is the predicted aircraft locations from historical air tra c data. +III. Problem Formulation and Contrail Reduction StrategyThe center contrail frequency index can be used to identify the flight level that would have formed the most contrails and find an alternate cruising altitude with less contrail formations.The contrail frequency index after the contrail reduction strategy has been applied is formulated asCF I center,l,t = 113 X i=1 151 X j=1 r i,j c i,j âi,j ,(7)where r i,j and c i,j are defined in Eqn. ( 2) and ( 4) and âi,j is defined in Eqn.(3) with the aircraft location after applying the contrail reduction strategy, which can be considered as variables.Thus the contrail reduction strategy is to solve the optimization problem, whose objective is to minimize CF I under several constraints.In this paper, the authors consider the fuel e ciency constraint and the operationally feasible constraints. +A. Fuel E ciency ConstraintAs discussed in, 12 the original cruising altitude filed in flight plan is usually the optimal cruising altitude in term of fuel e cienccy and aircraft performance.Therefore shifting aircraft up and down to other further levels is more fuel consuming than flying them at original altitude.In this paper, we allow aircraft to alternate the cruise level within the range (block altitude clearance 19 ) [l l fe , l + l fe ], where l is the original cruising level and l fe is the level changing constraint based on fuel e ciency.l fe is set to 1 and 2 respectively in our simulation.As an example, the contrail freqency index matrix at Atlanta Center at 8AM EDT on August 1, 2007 isCF I center,t = 0 B B B B B B B B B B B B B B B B B B B @ 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 2 0 0C C C C C C C C C C C C C C C C C C C A ,(8)where the bold diagonal item CF I l center,t,l is the contrail freqency index at level l before applying contrail reduction and the o↵ diagonal item CF I l 0 center,t,l at (l 0 , l) is the contrail freqency index when guiding aircraft from level l to level l 0 .There are 11 levels in total as illustrated in Table 1, therefore the matrix ( 8) is 11⇥11.If the aircraft are only allowed to move two levels up or down ( l fe = 2) the matrix becomes0 B B B B B B B B B B B B B B B B B B B @ 0 0 0 ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ 0 0 0 0 ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ 0 0 0 0 0 ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ 1 0 0 0 0 ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ 31 33 10 52 68 ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ 25 13 76 104 148 ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ 28 105 128 209 132 ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ 62 47 36 22 6 ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ 45 35 19 6 0 ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ 36 19 6 0 ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ 19 0 0 1 C C C C C C C C C C C C C C C C C C C A ,(9)where ⇥ indicates an invalid move.More strictly, if the aircraft can only be shifted up or down one level ( l fe = 1) the matrix is0 B B B B B B B B B B B B B B B B B B B @ 0 0 ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ 0 0 0 ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ 0 0 0 ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ 0 0 0 ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ 33 10 52 ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ 13 76 104 ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ 105 128 209 ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ 47 36 22 ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ 35 19 6 ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ 19 6 0 ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ ⇥ 0 0 1 C C C C C C C C C C C C C C C C C C C A . (10)The following two constraints are called operationally feasible constraints for contrail reduction. +B. Center Level Capacity ConstraintIn, 12 the authors shifted all the aircraft at level l to the level with minimum CF I l 0 center,t,l .However, if there is one level always with the minimum CF I l 0 center,t,l , the previous approach may potentially form a congested center level by moving a large number of aircraft to the same level.In practice, the wake turbulence or wake vortex phenomenon 20,21 constrains the air tra c controller to maintain a minimum miles-in-trail.The waketurbulence separation criterion is currently a limiting factor in airspace capacities.The FAA is working with NASA to develop and demonstrate integrated systems technology for addressing separation criteria.Some work has been done on miles-in-trail separation and space metering. 22,23,24 I summary, it is potentially harmful to move too many aircraft to the same level because of smaller separations and associated higher risk.In Fig. 3, the miles-in-trail (MIT) changes from MIT1 to a smaller value MIT3 after the metering point because there are some aircraft merging from the lower level.Another reason to introduce this constraint is to keep the air tra c controller's workload below a reasonable threshold.In order to measure dynamic density and evaluate the controller's workload, several metrics were proposed. 25,26 mong all these dynamic density metrics, the aircraft count or the tra c density in one area is the first factor noticed by every metric.Therefore maintaining the aircraft count at each level below a threshold value will guarantee the workload of air tra c controller manageable.In this paper, we propose the center level capacity constraint to keep the aircraft amount at each level lower than a critical value.This idea is similar to the Monitor Alert Parameter of a sector, which is set to reflect controller's acceptable workload. 27We use our level capacity constraint as a straightforward and operationally intuitive method to keep the controller workload at a manageable level.Normally it is di cult for the controllers to visualize potential vertical conflicts with a 2-D radar scope. 19s shown in Fig. 4, controllers can observe and manage the horizontal aircraft (in black) at level l and level l + 1 respectively.However, if there are too many climbing aircraft (in blue) moving from level l to level l + 1, it will introduce extra workload to air tra c controllers. +C. Climbing/Descending Aircraft Count ConstraintAlmost all the metrics in 25,26 mentioned the number of climbing/descending aircraft or the number of aircraft with large altitude change as the dynamic density impact factor, which means a huge number of climbing/descending will increase the controllers' workload.In this study, suppose the solution of aircraft count at level l at time t is q ⇤ center,t,l , we restrict the di↵erences of current solution with the aircraft counts at time t 1 and t + 1 within a given threshold.i.e., |q ⇤ center,t,l q center,t 1,l |  q and |q ⇤ center,t,l q center,t+1,l |  q.By doing this, our contrail reduction strategy should not shift too many aircraft up and down compared to the original flight plans.Therefore less extra workload will be added to the controllers. +D. Problem FormulationInstead of choosing the aircraft count in each grid as the variables, we use the aggregate aircraft count at each level to take advantage of the special structure of contrail frequency index matrix in Eqn.(8).For ease of illustration, l fe is set to 1 at this time.Thus our variables are chosen based on the +/ 1 level changing scheme in Eqn.(10), and the unknown variables are arranged as a vector of length 31 [x 1,1 , x 2,1 , x 1,2 , x 2,2 , x 3,2 , ..., x 9,10 , x 10,10 , x 11,10 , x 10,11 , x 11,11 ] 0 , where x i,j is the number of aircraft which used to cruise at level j and now cruise at level i.Our goal is to guide aircraft to the level with minimum CF I value while satisfying all the constraints.According to Eqn. (10), the c vector in the objective function is: c = [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 33, 0, 10, 13, 52, 76, 105, 104, 128, 47, 209, 36, 35, 22, 19, 19, 6, 6, 0, 0, 0] 0 .Fuel E ciency Constraint tells us the aircraft used to cruise at level j can only be adjusted to level j 1, j and j + 1 (except the aircraft used to cruise at the lowest level 1 and the highest level 11).By flow conservation, we have matrix A 1 of 11 ⇥ 31 and vector b 1 of length 11: Center Level Capacity Constraint requires the total aircraft number at each level after contrail reduction is fewer than the capacity.Since center level capacity is not an implemented setting in practice, we estimated the values of the capacities for each level.For example, there can be three source streams of aircraft forming the aircraft cruising at level l, which are x l,l 1 aircraft from level l 1, x l,l aircraft staying at level l and x l,l+1 aircraft shifted from level l + 1.In summary, we obtain constraint matrix A 2 of 11 ⇥ 31 and vector b 2 of length 11:A 1 = 0 B B B B B B @A 2 = 0 B B B B B B @1 0 1 0 0 0 . . .0 0 0 0 0 0 0 1 0 1 0 1 . . .0 0 0 0 0 0 . . .Climbing/Descending Aircraft Count Constraint requires the absolute di↵erence between the adjusted aircraft at level l and the aircraft used to cruise at level l at time t 1 smaller than a given threshold.i.e., |q ⇤ center,t,l q center,t 1,l |  q.Similarly, the absolute di↵erence between q ⇤ center,t,l and q center,t+1,l should also be less or equal than q.It is easy to calculate the number of aircraft used to cruise at level l at time t 1 and t + 1 from ASDI aircraft data.However, q value is not necessarily the same under di↵erent altitudes, center or sector boundary shapes and geographic conditions.In the simulation part of this paper we assume all the q are the same.Ignoring the center subscript notion, we denote the number of aircraft used to cruise at t 1 and t + 1 at level l as q t 1,l and q t+1,l , which we can compute from the historical data.At each level, we restrict the maximal number of climbing/descending aircraft with q l .Then we have:q 1  x 1,1 + x 1,2 q t+1,1  q 1 , q 2  x 2,1 + x 2,2 + x 2,3 q t+1,2  q 2 , . . . . . . q 11  x 11,10 + x 11,11 q t+1,11  q 11 , q 1  x 1,1 + x 1,2 q t 1,1  q 1 , q 2  x 2,1 + x 2,2 + x 2,3 q t 1,2  q 2 , . . . . . . q 11  x 11,10 + x 11,11 q t 1,11  q 11 .Let's denote A 2 as A t+1,+ , (q t+1 + q) as b t+1,+ , A 2 as A t+1, , ( q t+1 + q) as b t+1, , A 2 as A t 1,+ , (q t 1 + q) as b t 1,+ , A 2 as A t 1, , ( q t+1 + q) as b t 1, , the Climbing/Descending Aircraft Count Constraint can be written as:0 B B B @ A t+1,+ A t+1, A t 1,+ A t 1, 1 C C C A x  0 B B B @ b t+1,+ b t+1, b t 1,+ b t 1, 1 C C C A .(13)By combining (11)(12)(13), we have the fule e ciency constraint (one equality) and two operationally feasible constraints (five inequalities in total) listed all as inequalities below:0 B B B B B B B B B B @ A 1 A 1 A 2 A t+1,+ A t+1, A t 1,+ A t 1, 1 C C C C C C C C C C A x  0 B B B B B B B B B B @ b 1 b 1 b 2 b t+1,+ b t+1, b t 1,+ b t 1, 1 C C C C C C C C C C A ,(14)which is simply denoted as:Ax  b.Finally, we formulate an integer programming problem:min c 0 x s.t. Ax  b,x 0, and x 2 I.1 C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C C A ,(17)where the part between the second and the third horizontal lines is the transformed submatrix A 2 .Since matrices A t+1,+ , A t+1, , A t 1,+ , A t 1, in (14) are either A 2 or A 2 , every line in the omitted part below the third horizontal line also only contains one single 1 or one single 1.Based on (43)(v) of, 29 if the transformed submatrices A 1 and A 1 above the second horizontal line in (17) is totally unimodular, then the original matrix A is also totally unimodular.Moreover, if the transformed A 1 is total unimodular, then the two parts of the transformed A 1 and A 1 above the second horizontal line are total unimodular.Now let's prove the following matrix is total unimodular:0 B B B B B B B B B B B @11 C C C C C C C C C C C A ,(18)where each column either contains only one number 1 or a 1/ 1 pair.By (43)(v) of, 29 the first two columns with one non-zero item 1 can be removed while preserving the total unimodularity of a matrix.Then according to (18) of, 29 the matrix whose each column contains exactly one 1 and exactly one 1 is totally unimodular.⌅Since matrix A is totally unimodular and vector b is integral, there must exist an optimal and integral solution for both the linear programming relaxation (16) and for the integer programming in (15), which is guaranteed by any simplex method.In this paper we are dealing with a medium size linear programming problem, so the simplex method in Matlab is capable to provide the e cient solution.The detailed simulation results are presented in next section. +IV. Simulation ResultsIn this section, we use the same RUC weather data and ASDI aircraft data as in ( 8), ( 9), (10).We first evaluate the performance of our refined strategy after introducing addtional operationally feasible constraints.Secondly, the aircraft count at each cruising level is examined to see if the result of our refined strategy can be governed by the estimated level capacity.Thirdly the numbers of aircraft shifted to each level are plotted which can also be understood as climbing/descending aircraft counts. +A. Contrail Reduction PerformanceIn Fig. 5, the indice 1, 2, 3 on x-axis represent the historical data (no contrail reduction), the original contrail reduction strategy in 12 and our refined strategy.The red curve is the result of contrail reduction by allowing +/ 1 level aircraft shifting.The blue curve is the contrail reduction by allowing +/ 2 levels aircraft shifting.The original strategy has better contrail reduction performance whenever it is the +/ 1 level shifting or the +/ 2 levels shifting.The +/ 2 levels shifting of the original strategy gives out almost 60% contrail reduction while the +/ 2 levels shifting of our strategy can provide a 20% contrail reduction. +B. Level Capacity ConstraintIn Fig. 6, the numbers 1, 2, ..., 11 on x-axis are the 11 cruising levels and the y-axis indicates the number of aircraft cruising at each level.The red curve describes the level capacity which helps the controllers keep their workload manageable.The blue curve is the adjusted aircraft count at each level provided by the original strategy.The purple curve is the contrail reduction result provided by our refined strategy.The left subfigure is the result of the +/ 1 level shifting while the right subfigure is the result of the +/ 2 levels shifting.The contrail reduction result from the original strategy will potentially guide the aircraft to the level with minimal contrail potential grids, which may form a very busy cruising level, i.e., level 9 in the right subfigure of Fig. 6.However, our strategy successfully adjust the aircraft to reduce the contrail frequency index and restrict the aircraft count below the level capacity at the same time.Fig. 7 shows us the shifted aircraft number at each level, which can also be understood as climbing/descending aircraft counts.The blue curve represent the result of the original strategy and the purple curve is provided by our refined strategy.The left subfigure is the result of the +/ 1 level shifting while the right subfigure is the result of the +/ 2 levels shifting. +C. Shifted Aircraft NumberNo matter it is the +/ 1 level shifting or the +/ 2 levels shifting, our strategy does great job on managing the aircraft count of each level at a relative low value.This feature can help the controllers process fewer climbing/descending aircraft at each level. +V. ConclusionThis study brings two additional operationally feasible constraints to the original contrail reduction problem.By introducing the center level capacity constraint and the climbing/descending aircraft count constraint, the paper formulates the 3D grid contrail reduction problem as an integer programming.With the total unimodular property, the relaxed linear programming provides the optimal and integral solution in a computational e cient manner, which is the most important advantage of our approach.Another benefit of the method is that instead of shifting all the aircraft from one level to another, our formulation supports the feature of splitting the total number of aircraft used to cruise at level l to multiple levels.i.e., part of the aircraft used to cruise at level l can stay at this level, part of them may be shifted to upper levels, and the remaining part will be guided to lower allowed cruising levels.Simulation results show that the refined contrail reduction strategy provides less contrail reduction than the original strategy presented in. 12owever, it strictly obeys the level capacity and the numbers of shifted aircraft at each level are restricted to maintain air tra c controllers workload.Future work will integrate operationally acceptable rerouting technique 30,31 into contrail avoidance trajectory optimization.Instead of the 3D grid aircraft shifting, the new avoidance trajectories and flight plans may provide more contrail reductions.Figure 1 .1Figure 1.Contours of temperature and RHw at 34,057 feet at 8AM EDT on August 1, 2007. +Figure 2 .2Figure 2. Contour of RHi at 34,057 feet at 8AM EDT on August 1, 2007. +Figure 3 .3Figure 3. Miles-in-trail is reduced after more aircraft merge from the lower level. +Figure 4 .4Figure 4. Aircraft are climbing from cruising level l to level l + 1. +b 1 =1[10, 10, 10, 15, 40, 200, 200, 100, 50, 20, 10] 0 , where b 1 (l) is the number of aircraft which used to cruise at level l in the original flight plans. +b 2 =2[15, 15, 20, 25, 50, 250, 250, 150, 70, 30, 15] 0 , where b 2 (l) is the center level capacity at level l. +Figure 5 .5Figure 5. Contrail reduction performance comparison. +Figure 6 .6Figure 6.Our refined strategy is strictly governed by the level capacities. +Figure 7 .7Figure 7.The climbing/descending aircraft count at each level. +Table 1 .1Level index, isobaric pressure level and approximate aircraft cruising altitude.Level index1234567891011Pressure level (hPa)400 375 350 325 300 275 250 225 200 175 150Cruising altitude (100 feet) 236 251 267 283 301 320 341 363 387 414 444 + American Institute of Aeronautics and Astronautics + 1 1 . . .0 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + + +AcknowledgementsThe authors would like to thank Dr. Craig Wanke from MITRE for the very helpful discussion on operational feasibility constraints and Mr. Dave Slosson, who worked as a FAA air tra c controller at ZID center and then an air tra c supervisor at Fort Wayne tower and TRACON, for his expert and experienced consultation. + + + +E. Solution -Refined Contrail Reduction StrategyIn this work the integer programming (15) is relaxed to a linear programming (16) for computational eciency.We study the total unimodularity of matrix A in (16) and prove that there exists an optimal and integral solution for the linear programming relaxation, which is also the optimal solution for the original integer programming (15).It is guaranteed to be optimal and integral when ( 16) is solved by the simplex method.Theorem 1 Matrix A in ( 16) is total unimodular.Proof 1 Here the dimension of matrix A is 66 ⇥ 31.Targeting to transform each row of submatrix A 2 (14) into the row with a single 1, through a series of elementary column operations, 28 we have transformed A as: + + + + + + + Radiative forcing by contrails + + RMeerkötter + + + USchumann + + + DRDoelling + + + PMinnis + + + TNakajima + + + YTsushima + + 10.1007/s00585-999-1080-7 + + + Annales Geophysicae + Ann. Geophys. + 1432-0576 + + 17 + 8 + + 1999 + Copernicus GmbH + + + Meerkotter, R., Schumann, U., Doelling, D., Minnis, P., Nakajima, T., and Tsushima, Y., "Radiative forcing by contrails," Annales Geophysicae, Vol. 17, 1999, pp. 1080-1094. + + + + + Future Development of Contrail Cover, Optical Depth, and Radiative Forcing: Impacts of Increasing Air Traffic and Climate Change + + SMarquart + + + MPonater + + + FMager + + + RSausen + + 10.1175/1520-0442(2003)016<2890:fdocco>2.0.co;2 + + + Journal of Climate + J. Climate + 0894-8755 + 1520-0442 + + 16 + 17 + + September 2003 + American Meteorological Society + + + Marquart, S., Ponater, M., Mager, F., and Sausen, R., "Future Development of Contrail Cover, Optical Depth and Radiative Forcing: Impacts of Increasing Air Tra c and Climate Change," Journal of Climate, Vol. 16, September 2003, pp. 2890-2904. + + + + + A Standing Royal Commission + + SusanOwens + + 10.1093/acprof:oso/9780198294658.003.0003 + + + Knowledge, Policy, and Expertise + London, UK + + Oxford University Press + 2002 + + + + "The Environmental E↵ects of Civil Aircraft in Flight," Tech. rep., Royal Commission on Environmental Pollution, London, UK, 2002. + + + + + Aircraft induced contrail cirrus over Europe + + HermannMannstein + + + UlrichSchumann + + 10.1127/0941-2948/2005/0058 + + + Meteorologische Zeitschrift + metz + 0941-2948 + + 14 + 4 + + 2005 + Schweizerbart + + + Mannstein, H. and Schumann, U., "Aircraft induced contrail cirrus over Europe," Meteorologische Zeitschrift, Vol. 14, No. 4, 2005, pp. 549-554. + + + + + A Review of Various Strategies for Contrail Avoidance + + KlausGierens + + + LingLim + + + KostasEleftheratos + + 10.2174/1874282300802010001 + + + The Open Atmospheric Science Journal + TOASCJ + 1874-2823 + + 2 + 1 + + 2008 + Bentham Science Publishers Ltd. + + + Gierens, K., Limb, L., and Eleftheratos, K., "A review of various strategies for contrail avoidance," The Open Atmospheric Science Journal, Vol. 2, 2008, pp. 1-7. + + + + + Overview on Contrail and Cirrus Cloud Avoidance Technology + + FNoppel + + + RSingh + + 10.2514/1.28655 + + + Journal of Aircraft + Journal of Aircraft + 0021-8669 + 1533-3868 + + 44 + 5 + + 2007 + American Institute of Aeronautics and Astronautics (AIAA) + + + Noppel, F. and Singh, R., "Overview on contrail and cirrus cloud avoidance technology," Journal of Aircraft, Vol. 44, No. 5, 2007, pp. 1721-1726. + + + + + A note on how to avoid contrail cirrus + + HermannMannstein + + + PeterSpichtinger + + + KlausGierens + + 10.1016/j.trd.2005.04.012 + + + Transportation Research Part D: Transport and Environment + Transportation Research Part D: Transport and Environment + 1361-9209 + + 10 + 5 + + September 2005 + Elsevier BV + + + Mannstein, H., Spichtinger, P., and Gierens, K., "A note on how to avoid contrail cirrus," Transportation Research. Part D, Transport and environment, Vol. 10, No. 5, September 2005, pp. 421-426. + + + + + An Optimal Strategy for Persistent Contrail Avoidance + + ScotCampbell + + + NatashaNeogi + + + MichaelBragg + + 10.2514/6.2008-6515 + + + AIAA Guidance, Navigation and Control Conference and Exhibit + Honolulu, HI + + American Institute of Aeronautics and Astronautics + August 2008 + + + Campbell, S. E., Neogi, N. A., and Bragg, M. B., "An optimal strategy for persistent contrail avoidance," AIAA Guidance, Navigation and Control Conference, AIAA, Honolulu, HI, August 2008. + + + + + The impact of cruise altitude on contrails and related radiative forcing + + ChristineFichter + + + SusanneMarquart + + + RobertSausen + + + DavidSLee + + 10.1127/0941-2948/2005/0048 + + + Meteorologische Zeitschrift + metz + 0941-2948 + + 14 + 4 + + August 2005 + Schweizerbart + + + Fichter, C., Marquart, S., Sausen, R., and Lee, D. S., "The impact of cruise altitude on contrails and related radiative forcing," Meteorologische Zeitschrift, Vol. 14, No. 4, August 2005, pp. 563-572. + + + + + Reducing the climate change impacts of aviation by restricting cruise altitudes + + VictoriaWilliams + + + RobertBNoland + + + RalfToumi + + 10.1016/s1361-9209(02)00013-5 + + + Transportation Research Part D: Transport and Environment + Transportation Research Part D: Transport and Environment + 1361-9209 + + 7 + 6 + + November 2002 + Elsevier BV + + + Williams, V., Noland, R. B., and Toumi, R., "Reducing the climate change impacts of aviation by restricting cruise altitudes," Transportation Research. Part D, Transport and environment, Vol. 7, No. 5, November 2002, pp. 451-464. + + + + + Variability of contrail formation conditions and the implications for policies to reduce the climate impacts of aviation + + VictoriaWilliams + + + RobertBNoland + + 10.1016/j.trd.2005.04.003 + + + Transportation Research Part D: Transport and Environment + Transportation Research Part D: Transport and Environment + 1361-9209 + + 10 + 4 + + July 2005 + Elsevier BV + + + Williams, V. and Noland, R. B., "Variability of contrail formation conditions and the implications for policies to reduce the climate impacts of aviation," Transportation Research. Part D, Transport and environment, Vol. 10, No. 4, July 2005, pp. 269-280. + + + + + Fuel efficient strategies for reducing contrail formations in United States airspace + + BanavarSridhar + + + NeilYChen + + 10.1109/dasc.2010.5655533 + + + 29th Digital Avionics Systems Conference + Salt Lake City, UT + + IEEE + Oct 2010 + + + Sridhar, B., Chen, N. Y., and Ng, H. K., "Fuel e cient strategies for reducing contrail formations in United States airspace," AIAA Digital Avionics Systems Conference, AIAA, Salt Lake City, UT, Oct 2010. + + + + + Contrail Frequency over the United States from Surface Observations + + PatrickMinnis + + + JKirkAyers + + + MicheleLNordeen + + + StevenPWeaver + + 10.1175/1520-0442(2003)016<3447:cfotus>2.0.co;2 + + + Journal of Climate + J. Climate + 0894-8755 + 1520-0442 + + 16 + 21 + + November 2003 + American Meteorological Society + + + Minnis, P., Ayers, J. K., Nordeen, M. L., and Weaver, S. P., "Contrail frequency over the united states from surface observations," Journal of Climate, Vol. 16, No. 21, November 2003. + + + + + Contrail coverage derived from 2001 AVHRR data over the continental United States of America and surrounding areas + + RabindraPalikonda + + + PatrickMinnis + + + DavidPDuda + + + HermannMannstein + + 10.1127/0941-2948/2005/0051 + + + Meteorologische Zeitschrift + metz + 0941-2948 + + 14 + 4 + + August 2005 + Schweizerbart + + + Palikonda, R., Minnis, P., Duda, D. P., and Mannstein, H., "Contrail coverage derived from 2001 AVHRR data over the continental United States of America and surrounding areas," Meteorologische Zeitschrift, Vol. 14, No. 4, August 2005, pp. 525-536. + + + + + Relating observations of contrail persistence to numerical weather analysis output + + DPDuda + + + RPalikonda + + + PMinnis + + 10.5194/acp-9-1357-2009 + + + Atmospheric Chemistry and Physics + Atmos. Chem. Phys. + 1680-7324 + + 9 + 4 + + February 2009 + Copernicus GmbH + + + Duda, D. P., Palikonda, R., and Minnis, P., "Relating observations of contrail persistence to numerical weather analysis output," Atmospheric Chemistry and Physics, Vol. 9, No. 4, February 2009, pp. 1357-1364. + + + + + Estimated contrail frequency and coverage over the contiguous United States from numerical weather prediction analyses and flight track data + + DavidPDuda + + + PatrickMinnis + + + RabindraPalikonda + + 10.1127/0941-2948/2005/0050 + + + Meteorologische Zeitschrift + metz + 0941-2948 + + 14 + 4 + + June-July 2003. 17 + Schweizerbart + Friedrichshafen at Lake Constance, Germany + + + Duda, D. P., Minnis, P., Costulis, P. K., and Palikonda, R., "Conus contrail frequency estimated from ruc and flight track data," European Conference on Aviation, Atmosphere, and Climate, Friedrichshafen at Lake Constance, Germany, June-July 2003. 17 + + + + + Improved Magnus Form Approximation of Saturation Vapor Pressure + + OlegAAlduchov + + + RobertEEskridge + + 10.1175/1520-0450(1996)035<0601:imfaos>2.0.co;2 + + + Journal of Applied Meteorology + J. Appl. Meteor. + 0894-8763 + 1520-0450 + + 35 + 4 + + April 1996 + American Meteorological Society + + + Alduchov, O. A. and Eskridge, R. E., "Improved magnus form approximation of saturation vapor pressure," Journal of Applied Meteorology, Vol. 35, No. 4, April 1996, pp. 601-609. + + + + + Prediction and Use of Contrail Frequency Index for Contrail Reduction Strategies + + NeilChen + + + BanavarSridhar + + + HokNg + + 10.2514/6.2010-7849 + + + AIAA Guidance, Navigation, and Control Conference + Toronto, Canada + + American Institute of Aeronautics and Astronautics + Aug 2010 + + + Chen, N. Y., Sridhar, B., and Ng, H. K., "Prediction and use of contrail frequency index for contrail reduction strategies," AIAA Guidance, Navigation and Control Conference, AIAA, Toronto, Canada, Aug 2010. + + + + + Gale Cengage Learning + + MSNolan + + 10.1163/1872-9037_afco_asc_2259 + + Feb 2010 + Brill + 5 + + + th ed. + Nolan, M. S., Fundamentals of Air Tra c Control, Cengage Learning, 5th ed., Feb 2010. + + + + + U.S. International Air Travel Statistics. Calendar Year 1994. Center for Transportation Information, Volpe National Transportation Systems Center, Cambridge, MA 02142. Various paging + + GCHay + + + RHPassman + + 10.1177/004728759503400286 + + + Journal of Travel Research + Journal of Travel Research + 0047-2875 + 1552-6763 + + 34 + 2 + + 1995 + SAGE Publications + Cambridge, MA + + + Hay, G. C. and Passman, R. H., "Wake Turbulence Training Aid," Tech. rep., John A. Volpe National Transportation Systems Center, Cambridge, MA, 1995. + + + + + Assessment of wake vortex separation distances using the WAVIR toolset + + LSpeijker + + + GVan Baren + + + LSherry + + + JShortle + + + FRico-Cusi + + 10.1109/dasc.2004.1391279 + + + The 23rd Digital Avionics Systems Conference (IEEE Cat. No.04CH37576) + San Diego, USA + + IEEE + 2004 + + + Speijker, L., van Baren, G., Sherry, L., Shortle, J., and Rico-Cusi, F., "Assessment of Wake Vortex Separation Distances Using the WAVIR Toolset," AIAA Digital Avionics Systems Conference, San Diego, USA, 2004. + + + + + Miles-in-Trail (MIT): A Perspective + + ParimalKopardekar + + + StevenGreen + + + JohnAston + + + TomRoherty + + 10.2514/6.2003-6700 + + + AIAA's 3rd Annual Aviation Technology, Integration, and Operations (ATIO) Forum + Denver, USA + + American Institute of Aeronautics and Astronautics + November 2003 + + + Miles-in-trail Operations: A Perspective + Kopardekar, P., Green, S., Roherty, T., and Aston, J., "Miles-in-trail Operations: A Perspective," AIAA's 3rd Aviation Technology, Integration, and Operations (ATIO) Tech, Denver, USA, November 2003. + + + + + Modeling Time and Space Metering of Flights in the National Airspace System + + PT RWang + + + CRWanke + + + FPWieland + + 10.1109/wsc.2004.1371463 + + + Proceedings of the 2004 Winter Simulation Conference, 2004. + the 2004 Winter Simulation Conference, 2004Washington, DC, USA + + IEEE + 2004 + + + + Wang, P., Wanke, C., and Wieland, F., "Modeling time and space metering of flights in the National Airspace System," Simulation Conference, Washington, DC, USA, 2004, pp. 1299-1304. + + + + + Joint Metering and Conflict Resolution in Air Traffic Control + + JeromeLeNy + + + GeorgeJPappas + + 10.2514/1.53253 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 34 + 5 + + 2010 + American Institute of Aeronautics and Astronautics (AIAA) + + + University of Pennsylvania + + + Tech. rep + Ny, J. L. and Pappas, G. J., "Joint Metering and Conflict Resolution in Air Tra c Control," Tech. rep., University of Pennsylvania, 2010. + + + + + Airspace Complexity and its Application in Air Tra c Management + + BSridhar + + + KSSheth + + + SGrabbe + + + + 2nd USA-Europe ATM Seminar + Orlando, USA + + December 1998 + + + Sridhar, B., Sheth, K. S., and Grabbe, S., "Airspace Complexity and its Application in Air Tra c Management," 2nd USA-Europe ATM Seminar , Orlando, USA, December 1998. + + + + + 5th ifac/ifors conference on dynamic modelling and control of national economies Budapest, Hungary, 17–20 June 1986 + + PKopardekar + + + SMagyarits + + 10.1016/0305-0548(86)90069-9 + + + Computers & Operations Research + Computers & Operations Research + 0305-0548 + + 13 + 1 + 103 + June 2003. 2008 + Elsevier BV + Budapest, Hungary + + + Measurement And Prediction Of Dynamic Density. Order JO 7210.3V + Kopardekar, P. and Magyarits, S., "Measurement And Prediction Of Dynamic Density," 5th USA-Europe ATM Seminar , Budapest, Hungary, June 2003. 27 Federal Aviation Administration, Facility Operation and Administration, Chapter 17, Section 7. Monitor Alert Param- eter , 2008, Order JO 7210.3V. + + + + + Total unimodularity and Degeneracy-Aware Dantzig-Wolfe Decomposition for Large-Capacity Cell Transmission Model + + PengWei + + + DengfengSun + + 10.1109/cdc.2011.6160210 + + + IEEE Conference on Decision and Control and European Control Conference + Orlando, USA + + IEEE + Dec 2011. 29 + + + Wei, P. and Sun, D., "Total Unimodularity and Degeneracy-Aware Dantzig-Wolfe Decomposition for Large Capacity Cell Transmission Model," 50th IEEE Conference on Decision and Control and European Control Conference, Orlando, USA, Dec 2011. 29 + + + + + Theory of Linear and Integer Programming + + ASchrijver + + + 1998 + Wiley + + + Schrijver, A., Theory of Linear and Integer Programming, Wiley, 1998. + + + + + Dynamic Generation of Operationally Acceptable Reroutes + + ChristineTaylor + + + CraigWanke + + 10.2514/6.2009-7091 + + + 9th AIAA Aviation Technology, Integration, and Operations Conference (ATIO) + Hilton Head, South Carolina + + American Institute of Aeronautics and Astronautics + Sep 2009 + + + Taylor, C. and Wanke, C., "Dynamic Generation of Operationally Acceptable Reroutes," 9th AIAA Aviation Technology, Integration, and Operations Conference, Hilton Head, South Carolina, Sep 2009. + + + + + Improved Dynamic Generation of Operationally-Acceptable Reroutes Using Network Optimization + + ChristineTaylor + + + CraigWanke + + 10.2514/6.2010-8077 + + + AIAA Guidance, Navigation, and Control Conference + + American Institute of Aeronautics and Astronautics + 2011 + + + Taylor, C. and Wanke, C., "Improved Dynamic Generation of Operationally Acceptable Reroutes Using Network Opti- mization," Journal of Guidance, Control, and Dynamics, 2011. + + + + + + diff --git a/file777.txt b/file777.txt new file mode 100644 index 0000000000000000000000000000000000000000..6b6442cfb506279f509511f5276f72d5f031769f --- /dev/null +++ b/file777.txt @@ -0,0 +1,685 @@ + + + + +Many strategies have been proposed to reduce contrail formation in the United States airspace.One approach is to build three-dimensional weather grids and have aircraft cruising at different flight levels to avoid the persistent contrail potential area with the consideration to constraints like fuel burn, flight level capacity, and total operations of climbing/descending aircraft count at each level.However, there is no air traffic control regulation defining the flight level capacity and the total number of climbing/descending aircraft operations for each flight level.This paper presents a contrail reduction scheme, which considers the defined Monitor Alert Parameter value as the sector capacity constraint.Instead of shifting (changing cruise altitude among vertical grids) for all aircraft in a center, the new scheme only shifts certain aircraft out of those grids that are within the persistent contrail potential area.Compared with the level shifting scheme, the grid shifting has a finer resolution and brings the benefits of more contrail reduction and less fuel burn.Furthermore, the one-hour planning interval is changed to one-minute, which provides higher temporal resolution in solution results.Numerical experiments are performed to compare the new vertical grid shifting scheme with the previous level shifting scheme. +NomenclatureT temperature in Celsius RHw relative humidity with respect to water RHi relative humidity with respect to ice t planning interval l cruising altitude level index R l t potential persistent contrail formation matrix (contrail matrix) at time t at level l A l t aircraft position matrix at time t at level l CF I center contrail frequency index S l m sector identification matrix of the mth sector at level l B l t sector congestion matrix at time t at level l +I. IntroductionThe visible trails of condensed water vapor made by the exhaust of aircraft engines are called contrails.Contrails may exist for only a few seconds or several minutes, or may persist for many hours which may affect climate.Persistent contrails reduce incoming solar radiation and outgoing thermal radiation in a way that accumulates heat. 1 The global mean contrail cover in 1992 was estimated to double by 2015, and quadruple by 2050 due to air traffic increase. 2Studies suggest that the environmental impact from persistent contrail is estimated to be three or four times, 3 to ten times 4 larger than aviation-induced emissions like CO 2 and NO x .Therefore, strategies and policies to reduce aircraft induced persistent contrails need to be studied to minimize the impact on the global environment.Several researchers have studied how to identify and reduce persistent contrail formation.Gierens, Limb and Eleftheratos in Ref. 5 and Noppel and Singh in Ref. 6 reviewed various strategies for contrail avoidance including changing engine architecture, enhancing airframe and engine integration, using alternate fuels, and modifying traffic flow management procedures.Among the traffic flow management solutions, Mannstein, Spichtinger and Gierens 7 presented a strategy to reduce the environmental impact of contrails significantly by small changes to each aircraft's flight altitude.Campbell, Neogi and Bragg 8 proposed a mixed integer programming method to optimally reroute aircraft to avoid the formation of persistent contrails.Both methods require an onboard contrail detection system.Fichter, Marquart, Sausen and Lee 9 showed that the global annual mean contrail coverage could be reduced by decreasing the aircraft cruise altitude.Williams, Noland and Toumi 10,11 proposed strategies for contrail reduction by identifying fixed and varying maximum altitude restrictions.These restrictions generally require more fuel burn and add congestion to the already crowded airspace at lower altitudes.Sridhar, Chen and Ng 12,13 defined the contrail frequency index to model and predict persistent contrail formation.They provided a set of strategies to reduce contrail formation.A three-dimensional (3D) weather grid model was constructed to describe the contrail potential areas and the flight level of aircraft is adjusted to avoid these areas with the consideration of fuel burn trade-off.Chen, Sridhar, Li and Ng evaluated the contrail reduction strategies based on aircraft travel distances. 14Sridhar, Ng and Chen integrated emission and climate models together with air traffic simulations to further understand the complex interaction between the physical climate system, emissions and air traffic. 15They analyzed the trade-offs between extra fuel burn and reduction of global surface temperature change.Wei, Sridhar, Chen and Sun 16 continued the research on flight level shifting (changing cruise altitude) strategy considering the constraints of flight level capacity and climbing/descending aircraft operations limit introduced by air traffic controllers' operational feasibility.However, in practice, there is no air traffic control regulation on capacity or operations limit at each flight level.Instead, the Monitor Alert Parameter (MAP) value is assigned to each sector to limit the aircraft count and maintain a manageable number of operations for air traffic controllers.Evans, Chen, Sridhar and Ng 17 discussed the contrail reduction strategies and emissions under MAP value constraint, in which the sector counts were monitored by the Future Air traffic management Concepts Evaluation Tool (FACET). 18n this paper, the sector capacity limits (MAP values) were integrated into the 3D grid model.In addition, considering that shifting all the aircraft from their current cruising level to another level results in more fuel burn, the finer grid shifting scheme was developed to change aircraft cruise altitude in each grid.In this work, the authors apply the MAP value to the 3D grid model in order to investigate the trade-off between contrail reduction and sector congestion.Shifting all the aircraft at one flight level 12,16 is replaced to shifting them to different vertical grids, which will reduce the fuel burn.In order to obtain better temporal resolution, the one hour planning interval proposed in 12,16 is changed to one minute.The new grid shifting scheme is developed with and without considering the sector capacity constraint.The simulation first compares the contrail reduction results of level shifting and free grid shifting without considering sector capacities.Then the performance of grid shifting satisfying sector capacity constraint is analyzed.The rest of the paper is organized as follows.Section II introduces the atmospheric data, contrail formation model, aircraft data and contrail frequency index used in this paper.Section III presents the problem formulation with the higher temporal resolution model.The solution methodology of grid shifting is presented in Section IV.The numerical results are shown in Section V. Section VI summarizes and concludes the paper. +II. Data and ModelA. Atmospheric Data Contrails can be observed from surface data 19 and detected by satellite data. 20Duda, Palikonda and Minnis 21 have related the observations to numerical weather analysis output and showed that persistent contrail formation can be computed using atmospheric temperature and humidity data retrieved from the Rapid Updated Cycle (RUC) data provided by the National Oceanic and Atmospheric Administration (NOAA).Contrails can persist when ambient air is supersaturated with respect to ice, which means that relative humidity with respect to ice (RHi) is greater than one hundred percent. 22The RHi can be computed from relative humidity with respect to water (RHw) and temperature, which are available in the RUC data.The one-hour forecast and the 13-km RUC data are used in this paper.The data have a temporal resolution of one hour, a horizontal resolution of 13 kilometers, and isobaric pressure levels from 100 to 1000 hectopascals (hPa) with 25 hPa increments.The vertical range of interest in this study is from 150 hPa to 400 hPa, which is equivalent to pressure altitude of about 23,600 feet to 44,400 feet.The temperature and RHw contours at 8AM eastern daylight time (EDT) on August 1, 2007 at pressure altitude 250 hPa, or 34,057 feet, are shown as in Fig. 1. +B. Contrail Formation ModelThe potential persistent contrail formation areas (contrail areas) are defined as areas with RHi greater than or equal to 100%.RHi can be computed from RHw and temperature using the saturation vapor pressure coefficients of Alduchov, 23 formulated as RHi = RHw × 6.0612e 18.102T /(249.52+T ) 6.1162e 22.577T /(237.78+T ) ,where T is the temperature in Celsius.The atmospheric profile shown in the left and right subfigures in Fig. 1 can be translated to a contour of RHi, as shown in Fig. 2. The 13-km RUC data have (337 × 451) data points that cover most U.S. airspace.The altitude level index l is defined as l = 1, 2, ..., 11 corresponding to isobaric pressure level at 400, 375, ..., 150 hPa.The level index, isobaric pressure level, and approximate aircraft cruising altitude are listed in Table 1.The potential persistent contrail formation matrix (contrail matrix) at time t at level l is defined asR l t =     r 1,1,l r 1,2,l . . .    ,(2)where r i,j is 1 if RHi ≥ 100% at grid (i, j) and 0 if RHi < 100%. +C. Aircraft DataContrail formation areas represent environmental conditions likely to produce contrails.Thus aircraft locations are needed to determine the contrail formation frequency.The aircraft data used in this paper are obtained from the aircraft locations provided by the Federal Aviation Administration's (FAA's) Aircraft Situation Display to Industry (ASDI) data.The ASDI has a sampling rate of one minute.The same 3D grid in the RUC data is used to generate the aircraft position matrix.The aircraft position matrix at level l is defined asA l t =     a 1,1,l a 1,    ,(3)where a i,j is the number of aircraft within grid (i, j) flying closest to level l at time t.The aircraft position matrix indicates the air traffic density in the grid scale at different altitudes. +D. Contrail Frequency IndexThe contrail frequency index (CFI) is defined as the number of aircraft that would fly through potential contrail formation regions during a period of time, and is used to indicate the contrail severity in a given region. 13he center contrail frequency index in center C n is defined as the number of aircraft flying through persistent contrail potential area at time t within center C n , formulated asCF I Cn,t = Cn i,j,l r i,j,l a i,j,l ,(4)where r i,j,l and a i,j,l are defined in Eqn. ( 2), (3).The contrail frequency index indicates the actual contrail activities in a given region defined on a 3D grid.For strategic planning, the contrail frequency at time t of a certain center C n is calculated byCF I Cn,t = Cn i,j,l r i,j,l âi,j,l ,(5)where r i,j,l is from RUC forecast weather data and âi,j,l is the predicted aircraft locations resulted from our optimal solutions.The goal of this work is to design an optimization method to decide âi,j,l 's such that the objective function CF I Cn,t is minimized under given constraints. +III. ProblemIn this work the 3D grid is used for not only modeling the RUC weather data but also for shifting aircraft position vertically by grid.Two constraints are considered.The aircraft performance constraint requires that aircraft position can only be shifted up or down by one grid.Sector capacity constraint is introduced to maintain the manageable aircraft amount inside each sector for air traffic controllers.In summary, the grid shifting scheme is formulated as a linear programming problem. +A. Aircraft Performance ConstraintAccording to Base of Aircraft Data (BADA) by EUROCONTROL, 24 most of the typical climb/descent rates of en route commercial aircraft are less than 3,000 feet per minute, which indicates that we can only shift aircraft up or down by one grid during one minute because based on Table 1, each level varies from 1,500 to 3,000 feet.For the a i,j,l aircraft used to cruise in grid (i, j, l), three variables x l,l i,j , x l,l-1 i,j and x l,l+1 i,j are adopted to denote how many aircraft will stay in current grid, descend to a lower altitude grid and climb to a higher altitude grid.Notice that for the grids at level 1 or level 11 (see in Table 1), there will be only x l,l i,j and x l,l+1 i,j or x l,l i,j and x l,l-1 i,j .a i,j,l =      x l,l i,j + x l,l+1 i,j l = 1, x l,l i,j + x l,l-1 i,j l = 11, x l,l i,j + x l,l-1 i,j + x l,l+1 i,j otherwise.(6)Similarly, the solution âi,j,l consists of x l,l i,j , x l-1,l i,j and x l+1,l i,jexcept the grids at level 1 or level 11 which only consists of x l,l i,j and x l+1,l i,j or x l,l i,j and x l-1,l i,j .âi,j,l =     x l,l i,j + x l+1,l i,j l = 1, x l,l i,j + x l-1,l i,j l = 11, x l,l i,j + x l-1,l i,j + x l+1,l i,jotherwise.(7) +B. Sector Capacity ConstraintThe Monitor Alert Parameter (MAP) value 25 is used as an indicator of a capacity limit for each sector of airspace.A sector's MAP value reflects the maximum aircraft count that can be safely handled by a sector controller.MAP values are set up as sector capacities in this work.In order to impose a MAP value to a sector, we need to identify the sector in the 3D grid model described in Section II.The 3D sector identification matrix S m is used to locate the grids inside the corresponding sector.S l m =     s 1,1,    ,(8)s i,j,l = 1 if grid (i, j, l) is inside sector S m and s i,j,l = 0 otherwise.Thus the sector capacity constraint for sector m can be formulated as: 337,451,11 i,j,l s i,j,l âi,j,l ≤ MAP m .In short, Eqn. ( 9) for sector S m can be written as:Sm i,j,l âi,j,l ≤ MAP m .(10)Similarly, the 3D center identify matrix is defined to locate all the grids in center C n , in which the element c i,j,l = 1 if grid (i, j, l) is inside center C n and s i,j,l = 0 otherwise.In this work, the contrail reduction scheme is applied within each center.Thus the objective function is to minimize the total contrail frequency index inside a certain center: min 337,451,11 i,j,l c i,j,l r i,j,l âi,j,lIn short, Eqn. ( 11) can be written as: min Cn i,j,l r i,j,l âi,j,l = min Cn i,j,lCF I i,j,l(12) +C. Problem FormulationWe consider the optimization problem inside one center within time interval t. r i,j,l 's are obtained from the weather forecast.The problem formulation is min Cn i,j,l CFI i,j,l s.t. ( 6), (7), Sm i,j,l âi,j,l ≤ MAP m , ∀S m ∈ C n .(13) i, j, l are the index for each 3D grid.CFI i,j,l 's are the CFI number generated by the aircraft in grid (i, j, l) in t.C n is the nth center.a i,j,l is the number of aircraft which used to cruise inside grid (i, j, l) in t. S m is the mth sector in center C n .MAP m is the MAP value assigned to S m . +IV. MethodThe grid shifting scheme is developed in this section, which provides better contrail reduction performance and less fuel burn than the level shifting scheme. +A. Level Shifting vs. Grid ShiftingThe level shifting scheme and grid shifting scheme are demonstrated in the left and right subfigures of Fig. 3.The blue aircraft are the positions before shifting and the black aircraft are the positions after shifting.In the left subfigure, the level shifting moves all the aircraft at level l to level l + 1 when it finds that cruising at level l + 1 generates fewer CFI in total.In the right subfigure, the grid shifting only moves those aircraft that produce fewer CFI at level l + 1 than at l. Therefore the grid shifting is more accurate, and is expected to provide better contrail reduction results and less fuel burn.Furthermore, sector capacity constraints can be integrated into grid shifting. +B. Grid Shifting Without Sector Capacity ConstraintAlthough the optimization problem in ( 13) is solvable by linear programming techniques, we can find solutions faster based on the special structure of the problem.We first introduce the free grid shifting scheme without considering sector capacity constraint.Then the grid shifting satisfying sector capacity constraint is proposed.To develop the free grid shifting, the aircraft performance constraint still needs to be satisfied, which means that the aircraft can be only moved up or down by one flight level.We first observed that without the sector capacity constraint, all shifting moves for aircraft inside a grid are decoupled.By ignoring the sector capacity constraint, all the local optimal solutions for each grid together provide the global optimal solution.The summation of all the local optimal solutions CFI i,j,l is the global optimal solution.Because Cn i,j,l min CFI i,j,l = min Cn i,j,l CFI i,j,l when there is no coupling sector capacity constraint among different grids.The local optimal solution is easy to find by the greedy method.Consider the coefficient in objective function ( 12) is r i,j,l which is either 1 or 0. Thus the idea is to check every grid and its upper and lower neighboring grids (grids at level 1 and level 11 only need to check one neighboring grid).If r i,j,l = 0 in this grid, there is no contrail formed and none of the aircraft in this grid need to be shifted.If r i,j,l = 1 in this grid, all the aircraft a i,j,l will produce persistent contrail.Now the potential persistent contrail formation values in its upper grid r i,j,l+1 and in its lower grid r i,j,l-1 are checked.If there is one of them equal to 0, we can avoid producing contrail in grid (i, j, l) by shifting a i,j,l to the corresponding neighboring grid.If both potential persistent contrail formation values of the neighboring grids are 0, shift all a i,j,l aircraft to the lower grid (or the upper grid).If r i,j,l = 1 in grid (i, j, l) and both potential persistent contrail formation values of its neighboring grids are 1, there is no contrail reduction shifting that can be performed.The greedy algorithm is summarized in Algorithm 1:Algorithm 1 Greedy Grid Shifting Algorithm 1: for i, j, l do 2:if r i,j,l == 1 then3: if r i,j,l-1 == 0 then 4:shift to lower grid: âi,j,l = 0, âi,j,l-1 = a i,j,l-1 + a i,j,lif r i,j,l+1 == 0 then 8:shift to upper grid: âi,j,l = 0, âi,j,l+1 = a i,j,l+1 + a i,j,lend if 11: end for +C. Grid Shifting With Sector Capacity ConstraintAn En Route Center is divided into smaller regions, called sectors, which can be further defined as low altitude sectors, high altitude sectors and super high altitude sectors.Each sector is monitored and controlled by at least one air traffic controller.In this work, high and super high sectors are considered by grid shifting because they can cover all 11 levels from 23,600 to 44,400 feet. +Identify the Congested SectorsIn order not to introduce extra congestion to a busy sector, we need to first identify the busy sector.A sector is called busy if the total aircraft count in this sector is larger than the assigned MAP value.337,451,11 i,j,l s i,j,l a i,j,l > MAP m .The sector congestion matrix at time t at level l is defined as:B l t =     b 1,1,l b 1,2,l . . . b 1,451,l . . . . . . . . . . . . b 337,1,l b 337,2,l . . . b 337,451,l     ,(14)where b i,j,l = 1 if grid (i, j, l) belongs to a busy sector and b i,j,l = 0 otherwise. +Grid Shifting With Sector Capacity ConstraintOur target is to balance the contrail reduction and sector congestion.Therefore we want to design a scheme that will not introduce extra congestion to those already busy sectors.Therefore, for grid (i, j, l) the first step is to check the value of the contrail formation coefficient r i,j,l .If r i,j,l = 0, no persistent contrail will be formed in current grid so contrail reduction shifting is not needed.If the contrail formation coefficient r i,j,l = 1, its neighboring vertical grids are checked.When one of the upper and lower grids has the contrail formation coefficient equal to 0, which means moving a number of a i,j,l aircraft to this neighboring grid can reduce the total CFI by a i,j,l , check the corresponding sector congestion value b i,j,l .If b i,j,l = 0, the neighboring grid is in a non-congested sector.The aircraft will be shifted.Otherwise, if b i,j,l = 1, the neighboring grid is in a busy sector.The shifting move will not be performed.If r i,j,l = 1 and both of its neighboring grids have 0 contrail formation coefficient, the aircraft in grid (i, j, l) will be shifted to the grid with b i,j,l = 0.If r i,j,l = 1 and both of its neighboring grids have contrail formation coefficient equal to 1, no shifting needs to be performed.The grid shifting scheme is fast and the contrail reduction computation for twenty U.S. centers in a day can be done in minutes.The corresponding flow chart is illustrated in Fig. 4. The complete algorithm is listed in Algorithm 2:Algorithm 2 Grid Shifting With Sector Capacity Constraint 1: for i, j, l do 2:if r i,j,l == 1 then 3:if r i,j,l+1 == 0 then 4:if b i,j,l+1 == 0 then 5:shift to upper grid: âi,j,l = 0, âi,j,l+1 = a i,j,l+1 + a i,j,l 6: if r i,j,l-1 == 0 then 10:if b i,j,l-1 == 0 then 11:shift to lower grid: âi,j,l = 0, âi,j,l-1 = a i,j,l-1 + a i,j,l 12:end if13:end if14:end if 15: end for +V. ResultsThe RUC weather data and the ASDI air traffic data on April 23, 2010 are used to build the 3D grid model.The high and super high sectors of twenty U.S. centers between flight level 1 and flight level 11 are considered for sector capacity constraint.24 hours contrail reduction results are shown. +A. Contrail Reduction by Level Shifting and Free Grid ShiftingIn Fig. 5 contrail frequency index results are listed for twenty U.S. centers.The blue bars are the contrail frequency indices without any contrail reduction strategy calculated from historical weather and air traffic data.Due to the different weather conditions and air traffic, 20 centers have various CFIs.We can see that on that particular day, Minneapolis center (ZMP) has the highest contrail frequency index, which is over 20,000.Also, the contrail frequency indices in Chicago Center (ZAU), Denver Center (ZDV) and Kansas City Center (ZKC) are higher than 5,000.New York Center (ZNY) has 0 CFI.Purple bars are the contrail frequency indices after the level shifting reduction method presented in [12].To make a fair comparison, the +/-1 level shifting method is implemented to compare with the +/-1 grid shifting method.The green bars are the contrail frequency indices after the grid level shifting reduction method presented in Section IV.The results show that the level shifting method significantly reduces the CFI in ZMP by more than 50% and it provides more CFI reduction in ZAU, ZDV, ZKC and other centers.Moreover, the grid shifting method reduces the CFI in ZMP to about 33%, which is a larger reduction than level shifting method.Similarly, the grid shifting method achieves better CFI reduction than level shifting method in all U.S. 20 B. Contrail Reduction by Free Grid Shifting and Grid Shifting With Sector Capacity Constraint Fig. 6 shows that the contrail frequency index reduction comparison between the grid shifting method without sector capacity constraint and the grid shifting method with sector capacity constraint.The CFI reduction performances of the two methods are about the same.In other words, the grid shifting method satisfies the sector capacity constraint while achieving about the same amount of CFI reduction.In this way, no extra air traffic congestion will be introduced to each sector.In detail, Table 2 lists the resulting CFIs of twenty centers after two types of grid shifting schemes.In most of the centers the grid shifting with sector capacity constraint has the close performance to the free grid shifting.However, in ZMP, the free shifting reduces CFI by about 100.The reason is that on the day of April 23, 2010, the potential contrail formation area overlaps with a busy sector in ZMP.To prevent introduce extra congestion into the busy sector, the grid shifting is not performed in the corresponding sector. +VI. ConclusionIn this paper the grid shifting contrail reduction strategy was proposed.The one minute planning interval is adopted to have a higher temporal resolution than our previous work.The finer grid shifting algorithm can reduce more contrails than the level shifting method in all twenty U.S. centers in a day.Moreover, the grid shifting can identify the busy sectors and then balances the contrail reduction and sector congestion.The more accurate grid shifting brings the potential benefit of less fuel burn than level shifting scheme.Figure 1 .1Figure 1.Contours of temperature and RHw at 34,057 feet at 8AM EDT on August 1, 2007. +Figure 2 .2Figure 2. Contour of RHi at 34,057 feet at 8AM EDT on August 1, 2007. +Figure 3 .3Figure 3.Comparison between level shifting scheme and grid shifting scheme. +Figure 4 .4Figure 4. Flow chart of the grid shifting with sector capacity constraint. +Figure 5 .5Figure 5.Comparison between level shifting scheme and grid shifting scheme. +Figure 6 .6Figure 6.Comparison between grid shifting with sector capacity constraint and grid shifting without sector capacity constraint. +Table 1 .1Level index, isobaric pressure level and approximate aircraft cruising altitude.Level index1234567891011Pressure level (hPa)400 375 350 325 300 275 250 225 200 175 150Cruising altitude (100 feet) 236 251 267 283 301 320 341 363 387 414 444 +centers.x 10 42before reductionlevel shifting reductiongrid shifting reductionContrail Frequency Index0.5 1 1.50ZSEZOAZLAZLCZDVZABZMPZKCZFWZHUZAUZIDZMEZOBZDCZTLZJXZMAZBWZNY +Table 2 .2Contrail Frequency Indices of twenty U.S. centers after grid shifting with sector capacity constraint and without sector capacity constraint.Centers Shifting With Constraint Free ShiftingZSE139139ZOA60ZLA66ZLC14131413ZDV23342303ZAB1816ZMP79507859ZKC51815179ZFW165165ZHU312312ZAU34423440ZID15311531ZME34743463ZOB239239ZDC2221ZTL14761469ZJX414414ZMA569569ZBW9595ZNY00 + of 12 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on August 23, 2013 | http://arc.aiaa.org| DOI: 10.2514/6.2013-5177Copyright © 2013 by Peng Wei, Banavar Sridhar, Neil Chen, Dengfeng Sun.Published by the American Institute of Aeronautics and Astronautics, Inc., with permission. + of 12 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on August 23, 2013 | http://arc.aiaa.org| DOI: 10.2514/6.2013-5177Copyright © 2013 by Peng Wei, Banavar Sridhar, Neil Chen, Dengfeng Sun.Published by the American Institute of Aeronautics and Astronautics, Inc., with permission. + of 12 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on August 23, 2013 | http://arc.aiaa.org| DOI: 10.2514/6.2013-5177Copyright © 2013 by Peng Wei, Banavar Sridhar, Neil Chen, Dengfeng Sun.Published by the American Institute of Aeronautics and Astronautics, Inc., with permission. + of 12 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on August 23, 2013 | http://arc.aiaa.org| DOI: 10.2514/6.2013-5177Copyright © 2013 by Peng Wei, Banavar Sridhar, Neil Chen, Dengfeng Sun.Published by the American Institute of Aeronautics and Astronautics, Inc., with permission. + of 12 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on August 23, 2013 | http://arc.aiaa.org| DOI: 10.2514/6.2013-5177Copyright © 2013 by Peng Wei, Banavar Sridhar, Neil Chen, Dengfeng Sun.Published by the American Institute of Aeronautics and Astronautics, Inc., with permission. + + + + + + + + + Radiative forcing by contrails + + RMeerkötter + + + USchumann + + + DRDoelling + + + PMinnis + + + TNakajima + + + YTsushima + + 10.1007/s00585-999-1080-7 + + + Annales Geophysicae + Ann. Geophys. + 1432-0576 + + 17 + 8 + + 1999 + Copernicus GmbH + + + Meerkotter, R., Schumann, U., Doelling, D., Minnis, P., Nakajima, T., and Tsushima, Y., "Radiative forcing by contrails," Annales Geophysicae, Vol. 17, 1999, pp. 1080-1094. + + + + + Future Development of Contrail Cover, Optical Depth, and Radiative Forcing: Impacts of Increasing Air Traffic and Climate Change + + SMarquart + + + MPonater + + + FMager + + + RSausen + + 10.1175/1520-0442(2003)016<2890:fdocco>2.0.co;2 + + + Journal of Climate + J. Climate + 0894-8755 + 1520-0442 + + 16 + 17 + + September 2003 + American Meteorological Society + + + Marquart, S., Ponater, M., Mager, F., and Sausen, R., "Future Development of Contrail Cover, Optical Depth and Radiative Forcing: Impacts of Increasing Air Traffic and Climate Change," Journal of Climate, Vol. 16, September 2003, pp. 2890-2904. + + + + + A Standing Royal Commission + + SusanOwens + + 10.1093/acprof:oso/9780198294658.003.0003 + + + Knowledge, Policy, and Expertise + London, UK + + Oxford University Press + 2002 + + + + "The Environmental Effects of Civil Aircraft in Flight," Tech. rep., Royal Commission on Environmental Pollution, London, UK, 2002. + + + + + Aircraft induced contrail cirrus over Europe + + HermannMannstein + + + UlrichSchumann + + 10.1127/0941-2948/2005/0058 + + + Meteorologische Zeitschrift + metz + 0941-2948 + + 14 + 4 + + 2005 + Schweizerbart + + + Mannstein, H. and Schumann, U., "Aircraft induced contrail cirrus over Europe," Meteorologische Zeitschrift, Vol. 14, No. 4, 2005, pp. 549-554. + + + + + A Review of Various Strategies for Contrail Avoidance + + KlausGierens + + + LingLim + + + KostasEleftheratos + + 10.2174/1874282300802010001 + + + The Open Atmospheric Science Journal + TOASCJ + 1874-2823 + + 2 + 1 + + 2008 + Bentham Science Publishers Ltd. + + + Gierens, K., Limb, L., and Eleftheratos, K., "A review of various strategies for contrail avoidance," The Open Atmospheric Science Journal, Vol. 2, 2008, pp. 1-7. + + + + + Overview on Contrail and Cirrus Cloud Avoidance Technology + + FNoppel + + + RSingh + + 10.2514/1.28655 + + + Journal of Aircraft + Journal of Aircraft + 0021-8669 + 1533-3868 + + 44 + 5 + + 2007 + American Institute of Aeronautics and Astronautics (AIAA) + + + Noppel, F. and Singh, R., "Overview on contrail and cirrus cloud avoidance technology," Journal of Aircraft, Vol. 44, No. 5, 2007, pp. 1721-1726. + + + + + A note on how to avoid contrail cirrus + + HermannMannstein + + + PeterSpichtinger + + + KlausGierens + + 10.1016/j.trd.2005.04.012 + + + Transportation Research Part D: Transport and Environment + Transportation Research Part D: Transport and Environment + 1361-9209 + + 10 + 5 + + September 2005 + Elsevier BV + + + Mannstein, H., Spichtinger, P., and Gierens, K., "A note on how to avoid contrail cirrus," Transportation Research. Part D, Transport and environment, Vol. 10, No. 5, September 2005, pp. 421-426. + + + + + An Optimal Strategy for Persistent Contrail Avoidance + + ScotCampbell + + + NatashaNeogi + + + MichaelBragg + + 10.2514/6.2008-6515 + + + AIAA Guidance, Navigation and Control Conference and Exhibit + Honolulu, HI + + American Institute of Aeronautics and Astronautics + August 2008 + + + Campbell, S. E., Neogi, N. A., and Bragg, M. B., "An optimal strategy for persistent contrail avoidance," AIAA Guidance, Navigation and Control Conference, AIAA, Honolulu, HI, August 2008. + + + + + The impact of cruise altitude on contrails and related radiative forcing + + ChristineFichter + + + SusanneMarquart + + + RobertSausen + + + DavidSLee + + 10.1127/0941-2948/2005/0048 + + + Meteorologische Zeitschrift + metz + 0941-2948 + + 14 + 4 + + August 2005 + Schweizerbart + + + Fichter, C., Marquart, S., Sausen, R., and Lee, D. S., "The impact of cruise altitude on contrails and related radiative forcing," Meteorologische Zeitschrift, Vol. 14, No. 4, August 2005, pp. 563-572. + + + + + Reducing the climate change impacts of aviation by restricting cruise altitudes + + VictoriaWilliams + + + RobertBNoland + + + RalfToumi + + 10.1016/s1361-9209(02)00013-5 + + + Transportation Research Part D: Transport and Environment + Transportation Research Part D: Transport and Environment + 1361-9209 + + 7 + 6 + + November 2002 + Elsevier BV + + + Williams, V., Noland, R. B., and Toumi, R., "Reducing the climate change impacts of aviation by restricting cruise altitudes," Transportation Research. Part D, Transport and environment, Vol. 7, No. 5, November 2002, pp. 451-464. + + + + + Variability of contrail formation conditions and the implications for policies to reduce the climate impacts of aviation + + VictoriaWilliams + + + RobertBNoland + + 10.1016/j.trd.2005.04.003 + + + Transportation Research Part D: Transport and Environment + Transportation Research Part D: Transport and Environment + 1361-9209 + + 10 + 4 + + July 2005 + Elsevier BV + + + Williams, V. and Noland, R. B., "Variability of contrail formation conditions and the implications for policies to reduce the climate impacts of aviation," Transportation Research. Part D, Transport and environment, Vol. 10, No. 4, July 2005, pp. 269-280. + + + + + Fuel efficient strategies for reducing contrail formations in United States airspace + + BanavarSridhar + + + NeilYChen + + 10.1109/dasc.2010.5655533 + + + 29th Digital Avionics Systems Conference + Salt Lake City, UT + + IEEE + Oct 2010 + + + Sridhar, B., Chen, N. Y., and Ng, H. K., "Fuel efficient strategies for reducing contrail formations in United States airspace," AIAA Digital Avionics Systems Conference, AIAA, Salt Lake City, UT, Oct 2010. + + + + + Prediction and Use of Contrail Frequency Index for Contrail Reduction Strategies + + NeilChen + + + BanavarSridhar + + + HokNg + + 10.2514/6.2010-7849 + + + AIAA Guidance, Navigation, and Control Conference + Toronto, Canada + + American Institute of Aeronautics and Astronautics + Aug 2010 + + + Chen, N. Y., Sridhar, B., and Ng, H. K., "Prediction and use of contrail frequency index for contrail reduction strategies," AIAA Guidance, Navigation and Control Conference, AIAA, Toronto, Canada, Aug 2010. + + + + + Evaluation of Contrail Reduction Strategies Based on Aircraft Flight Distances + + NeilChen + + + BanavarSridhar + + + JinhuaLi + + + HokNg + + 10.2514/6.2012-4816 + + + AIAA Guidance, Navigation, and Control Conference + Minneapolis, MN + + American Institute of Aeronautics and Astronautics + Aug 2012 + + + Chen, N. Y., Sridhar, B., Li, J., and Ng, H. K., "Evaluation of Contrail Reduction Strategies Based on Aircraft Flight Distances," AIAA Guidance, Navigation and Control Conference, Minneapolis, MN, Aug 2012. + + + + + Integration of Linear Dynamic Emission and Climate Models with Air Traffic Simulations + + BanavarSridhar + + + HokNg + + + NeilChen + + 10.2514/6.2012-4756 + + + AIAA Guidance, Navigation, and Control Conference + Minneapolis, MN + + American Institute of Aeronautics and Astronautics + Aug 2012 + + + Sridhar, B., Ng, H. K., and Chen, N. Y., "Integration of Linear Dynamic Emission and Climate Models with Air Traffic Simulations," AIAA Guidance, Navigation and Control Conference, Minneapolis, MN, Aug 2012. + + + + + A Linear Programming Approach to the Development of Contrail Reduction Strategies Satisfying Operationally Feasible Constraints + + PengWei + + + BanavarSridhar + + + NeilChen + + + DengfengSun + + 10.2514/6.2012-4754 + + + AIAA Guidance, Navigation, and Control Conference + Minneapolis, MN + + American Institute of Aeronautics and Astronautics + Aug 2012 + + + Wei, P., Sridhar, B., Chen, N. Y., and Sun, D., "A Linear Programming Approach to the Development of Contrail Reduction Strategies Satisfying Operationally Feasible Constraints," AIAA Guidance, Navigation and Control Conference, Minneapolis, MN, Aug 2012. + + + + + Trade-off between Contrail Reduction and Emissions under Future US Air Traffic Scenarios + + AntonyEvans + + + NeilChen + + + BanavarSridhar + + 10.2514/6.2012-5508 + + + 12th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference and 14th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference + Indianapolis, Indiana; Napoli, Italy + + American Institute of Aeronautics and Astronautics + Sep 2012. 18 Bilimoria,. June 2000 + + + FACET: Future ATM Concepts Evaluation Tool + Evans, A., Chen, N. Y., Sridhar, B., and Ng, H. K., "Trade-off between Contrail Reduction and Emissions under Future US Air Traffic Scenarios," AIAA Aviation Technology, Integration, and Operations Conference, Indianapolis, Indiana, Sep 2012. 18 Bilimoria, K., Sridhar, B., Chatterji, G., Sheth, K., and Grabbe, S., "FACET: Future ATM Concepts Evaluation Tool," 3rd USA/Europe Air Traffic Management Seminar , Napoli, Italy, June 2000. + + + + + Contrail Frequency over the United States from Surface Observations + + PatrickMinnis + + + JKirkAyers + + + MicheleLNordeen + + + StevenPWeaver + + 10.1175/1520-0442(2003)016<3447:cfotus>2.0.co;2 + + + Journal of Climate + J. Climate + 0894-8755 + 1520-0442 + + 16 + 21 + + November 2003 + American Meteorological Society + + + Minnis, P., Ayers, J. K., Nordeen, M. L., and Weaver, S. P., "Contrail frequency over the united states from surface observations," Journal of Climate, Vol. 16, No. 21, November 2003. + + + + + Contrail coverage derived from 2001 AVHRR data over the continental United States of America and surrounding areas + + RabindraPalikonda + + + PatrickMinnis + + + DavidPDuda + + + HermannMannstein + + 10.1127/0941-2948/2005/0051 + + + Meteorologische Zeitschrift + metz + 0941-2948 + + 14 + 4 + + August 2005 + Schweizerbart + + + Palikonda, R., Minnis, P., Duda, D. P., and Mannstein, H., "Contrail coverage derived from 2001 AVHRR data over the continental United States of America and surrounding areas," Meteorologische Zeitschrift, Vol. 14, No. 4, August 2005, pp. 525-536. + + + + + Relating observations of contrail persistence to numerical weather analysis output + + DPDuda + + + RPalikonda + + + PMinnis + + 10.5194/acp-9-1357-2009 + + + Atmospheric Chemistry and Physics + Atmos. Chem. Phys. + 1680-7324 + + 9 + 4 + + February 2009 + Copernicus GmbH + + + Duda, D. P., Palikonda, R., and Minnis, P., "Relating observations of contrail persistence to numerical weather analysis output," Atmospheric Chemistry and Physics, Vol. 9, No. 4, February 2009, pp. 1357-1364. + + + + + Estimated contrail frequency and coverage over the contiguous United States from numerical weather prediction analyses and flight track data + + DavidPDuda + + + PatrickMinnis + + + RabindraPalikonda + + 10.1127/0941-2948/2005/0050 + + + Meteorologische Zeitschrift + metz + 0941-2948 + + 14 + 4 + + June-July 2003. 23 + Schweizerbart + Friedrichshafen at Lake Constance, Germany + + + Duda, D. P., Minnis, P., Costulis, P. K., and Palikonda, R., "Conus contrail frequency estimated from ruc and flight track data," European Conference on Aviation, Atmosphere, and Climate, Friedrichshafen at Lake Constance, Germany, June-July 2003. 23 + + + + + Improved Magnus Form Approximation of Saturation Vapor Pressure + + OlegAAlduchov + + + RobertEEskridge + + 10.1175/1520-0450(1996)035<0601:imfaos>2.0.co;2 + + + Journal of Applied Meteorology + J. Appl. Meteor. + 0894-8763 + 1520-0450 + + 35 + 4 + + April 1996 + American Meteorological Society + + + Alduchov, O. A. and Eskridge, R. E., "Improved magnus form approximation of saturation vapor pressure," Journal of Applied Meteorology, Vol. 35, No. 4, April 1996, pp. 601-609. + + + + + Health hazard evaluation report: HHE-81-042-832, Federal Aviation Administration, New York Air Route Traffic Control Center, Ronkonkoma, New York. + 10.26616/nioshhhe81042832 + No. 10/04. 25 + + + Traffic Management National, Center, and Terminal + + 17 + Sep 2004. 2012 + U.S. Department of Health and Human Services, Public Health Service, Centers for Disease Control, National Institute for Occupational Safety and Health + Note + + + User Manual for the Base of Aircraft Data (BADA). Section 8.Monitor Alert Parameter + Eurocontrol Experimental Center (EEC), User Manual for the Base of Aircraft Data (BADA), Revision 3.6 , Sep 2004, Note No. 10/04. 25 Federal Aviation Administration: Air Traffic Organization Policy, chap. 17. Traffic Management National, Center, and Terminal, Section 8.Monitor Alert Parameter, 2012. + + + + + + diff --git a/file778.txt b/file778.txt new file mode 100644 index 0000000000000000000000000000000000000000..33c50dec58b473f912b8e6353f58c518f4596bfa --- /dev/null +++ b/file778.txt @@ -0,0 +1,318 @@ + + + + +INTRODUCTIONThe cost of air traffic delay grows each year as traffic and fuel costs increase.As part of the Center-TRACON Automation System (CTAS), NASA is developing the Final Approach Spacing Tool (FAST).FAST will increase throughput and reduce delays by providing advisories for runway balancing, arrival sequencing, and final approach spacing.Analyses, simulations, and field trials indicate that FAST could help controllers increase arrival throughput on busy runways by several aircraft per hour.Studies by Seagull Technology, Inc. and Logistics Management Institute (LMI) have estimated the potential savings from FAST at 10 major US airports. 1,2,3,4,5,6Both studies reviewed published field trial results and runway capacity considerations and concluded that FAST has the potential to increase throughput by about 4 arrivals per hour per runway by helping controllers reduce the variance of inter-arrival timing. 7This more precise arrival timing allows controllers to reduce interarrival spacing without increasing the incidence of separation violations.The studies employed independent demand and capacity estimates and used separate queuing engines to calculate the reduction in queuing delay.Their capacity models, although different in detail, each provided Visual Meteorological Conditions (VMC) and Instrument Meteorological Conditions (IMC) capacity estimates for the principal runway configurations at each airport.The Seagull study used an event-driven queuing model, whereas LMI integrated the Kolmogorov queuing equations to convert capacity improvements into delay savings.They also developed independent cost models to convert delay savings to cost savings.In spite of these differences, both studies predicted that delay reductions from such throughput increases at 10 major airports would save over $300M annually in direct operating costs.However, their results disagree on predicted delay savings for some airports.They omit some important airports because high simulation setup costs make it too costly to extend the models.They did not validate their predictions with actual delay measurements for the airports studied.Their models focus on delays in IMC and VMC and do not reflect the fact that FAST cannot predict flight trajectories and must cease operating when hazardous weather significantly disrupts arrival routes.Consequently, they also do not account for the fact that FAST will experience unusually large residual queues after the routes have re-opened.Both studies adopted similar limitations of scope.The dollar savings estimates of both were intentionally American Institute of Aeronautics and Astronautics conservative by focusing only on direct operating cost savings from currently defined FAST functionality.They did not include delay multipliers for downstream savings, passenger delay cost savings, schedule uncertainty cost savings, or savings from possible future FAST improvements.Downstream delay propagation effects, which are large for thunderstorm and IMC days, were omitted for lack of a model applicable to all airports.Passenger delay was excluded from both studies because its dollar value is controversial.The studies omitted dollar benefits for increasing schedule predictability.Although delay reductions will result in less delay uncertainty, there is no model for the magnitude of the savings from reducing the back-up resources needed for dealing with schedule uncertainties.The delay and cost reduction potential of possible future FAST enhancements, such as controller aids for hazardous weather, arrival sequence optimization, or spacing reduction based on wake-vortex measurements were omitted for lack of supportable predictions on their effects on throughput. +OVERVIEWThis study provides guidelines for reconciling disagreements and extending the benefit predictions to additional airports.It provides a rule of thumb for ranking benefits from queuing considerations and for validating the rankings against actual delays.It relates delay to ceiling and visibility data and uses Integrated Terminal Weather System (ITWS) logs to determine the relative importance of hazardous weather delays and runway queuing delays.It examines the correlation of delay between airports and estimates the impact of downstream delay on FAST benefits by using a recently published delay propagation model.It begins with definitions of FAA delay and conventions regarding IMC and VMC.It then examines the statistics of delay in IMC and VMC to calibrate the relationship of between visibility-induced capacity changes and actual airport delay statistics.It also examines the correlation between delays at key airports for both flight delays and schedule delays.It concludes by examining occurrences of hazardous weather in DFW terminal airspace, where strong correlation exists between thunderstorm days and delays. +CODAS DELAY DATAThis paper makes extensive use of data from the FAA's Consolidated Operations and Delay Analysis System (CODAS). 8CODAS reports two main types of delay: airborne and arrival.Both types of delay are given as averages in units of minutes per arrival.In deriving these averages, CODAS counts early arrivals as zero delay rather than negative delay.CODAS airborne delay is measured relative to the flight duration predicted at the time of departure.It is the actual flight duration minus the predicted flight duration.Airborne delay does not include departure delays and is relatively independent of interactions between airports.The direct operating cost of airborne delay can be readily calculated.Although some airborne delay can be caused by en route weather and traffic flow problems, normally one of its largest components is the terminal queuing delay that runway capacity improvements from FAST are intended to reduce.CODAS arrival delay is measured relative to scheduled arrival time.It is the actual arrival time minus the most recent OAG scheduled arrival time.If the flight duration predicted on take-off is the same as the scheduled flight duration, the CODAS arrival delay is the sum of the departure delay and the CODAS airborne delay.Arrival delay averages are always several times larger than airborne delay averages.Because departure delay includes pushback delay, ground hold delay, and taxi-out delay, arrival delay is more difficult to cost objectively than is airborne delay.The CODAS delay database also includes the meteorological conditions at each airport.CODAS defines Visual Meteorological Conditions as the combination of ceiling and visibility for which visual approaches are allowed.To support visual approaches the ceiling must be 500 ft above the minimum vectoring altitude, which is determined by airport elevation, terrain clearance, and other local factors.Thus CODAS VMC corresponds to "high Visual Flight Rules (VFR)".At Boston (BOS), visual approaches are permitted when the ceiling exceeds 2500ft and the visibility exceeds 5mi.At Dallas Fort Worth Airport (DFW), visual approaches are permitted only for ceilings above 3500ft and visibility greater than 5mi.Lower ceiling and visibility conditions are considered IMC.That is, CODAS IMC corresponds to "low VFR" and below.CODAS weather data come in either 15-minute or hourly summaries.Any hour with one or more 15minute intervals of IMC is considered to be an IMC hour.In our analysis, any day with one or more IMC hours between 6AM and midnight is considered to be an IMC day.The results of the Seagull study displayed similar behavior.The delay savings estimated by Seagull for 5 of the airports exceeded their airborne delays, with LAX, LaGuardia (LGA) and ORD savings estimates exceeding CODAS delays by large factors. +MODEL RESULTS AND CODAS DATALarge queuing delay savings may be caused in part by modeling errors since queuing models, being very sensitive to small errors in demand/capacity assumptions, can easily misestimate delays in peak demand periods.However, annual delay improvements cannot logically exceed the actual annual delay.If the flight duration predicted at the time of each departure were an unbiased estimate of the minimum achievable flight duration, and if the CODAS airborne delay for each airport correctly accounted for all flights into that airport, then the measured delay would always exceed any delay savings that might result from a small incremental improvement in runway capacity.However, CODAS may systematically underestimate airborne delay.If the aircraft operator at takeoff bases the flight duration prediction on the mean historical flight time for that route, rather than the shortest feasible flight time, CODAS will underestimate airborne delay.Underestimation of arrival delay from intentional schedule padding occurs in the same way.Non-reporting aircraft can also cause CODAS to incorrectly estimate delay.Most CODAS airborne delay values are estimated from automatic reports obtained from the Airline Service Quality Performance (ASQP) System supported by the 10 largest domestic airlines.However, these reports do not account for all the flights into each airport.ASQP does not contain information on small air carrier, commuter, air taxi, general aviation, cargo, military, or international flights.CODAS estimates the delay for non-reporting aircraft from the Enhanced Traffic Management System (ETMS) and from the airlines' Computerized Reservation System (CRS) schedule database. 9though it underestimates delay, CODAS is nevertheless useful for assessing changes in delay, correlating delay between airports, and studying the effects of weather on delay. +A SIMPLE RANKING RULEDelay trends are important in validating model predictions with delay statistics and in extending model results to additional airports.Steady-state queuing theory tells us that the benefit of a given capacity increase at an airport is proportional to N 2 /R, where N is the total traffic count and R is the number of active runways at the airport.This proportionality holds under the assumption that individual runways are loaded to similar values of mean traffic intensity (ratio of demand to capacity) at all airports.This appears to be a reasonable assumption for most major airports.Figure 2 shows the FAST savings estimates from the LMI model plotted as a function of N 2 /R for the 10 airports studied.The LMI benefit predictions are seen to follow the N 2 /R trend reasonably closely, thus lending credence to the savings prediction for LAX, in spite of its inconsistency with CODAS airborne delay.4) BOS( 4) DTW( 5) DFW( 7) LAX( 4) SFO( 4)JFK(3) EWR(3) ORD(7)LGA (2) Airport(No. of Runways longer than 1 mile) +Figure 2. LMI FAST savings estimates vs. (N 2 /R).The trend agreement also allows extension of the LMI results to other airports.We used the rule to provide preliminary benefit estimates for Philadelphia, Charlotte, Denver, Miami, Minneapolis/St.Paul, and St. Louis.The benefits for these six airports ranged from American Institute of Aeronautics and Astronautics $11M per year for Denver to $35M per year for Miami.Overall, results indicate that the four largest airports (LAX, ATL, DFW, and ORD) would benefit most from FAST, but if controllers using FAST could indeed have increased runway arrival throughput in 1997 by 4 aircraft per hour at all 16 airports, airline operators would have recovered an estimated $460M in direct operating costs. +IMC DELAYFAST is intended to help improve airport capacity.Transitions from VMC to IMC cause measurable statistical changes in airport capacity.Therefore, quantifying the relationship between local meteorological conditions and measured delay (i.e., using the transition from IMC to VMC as an analytical surrogate for a capacity increase) can provide baseline comparisons for capacity modeling results.The FAA's CODAS delay database includes local ceiling, visibility, and wind as well as a meteorological condition indicator that switches from IMC to VMC when visual approaches are allowed at each airport.We used this database to examine the dependence of actual delay data on local meteorological conditions at key airports.On VMC days, the mean was 1.9 minutes of delay per aircraft, the standard deviation was 0.83 minutes per aircraft, and the delay on the worst VMC day averaged 5.3 minutes of delay per aircraft.On IMC days, the means, standard deviations, and peak delays were 2 to 3 times larger than on VMC days.The CODAS arrival delay at DFW in 1997 was 3 to 4 times larger than the airborne delay by all statistical measures, and showed a similar factor of 2-3 increase in IMC.The tendency for all CODAS arrival and airborne statistics to be 2 times larger in IMC than in VMC appears to be unusual for an airport like DFW.On average, DFW has excess capacity that is not strongly influenced by reduced ceiling and visibility.However, DFW experiences hubbing peaks each day that temporarily exceed even the available VMC runway capacity.During these rushes, a small decrease in either en route or terminal capacity can cause a large increase in delay.In VMC the queues that build up in these brief periods of excess demand are quickly cleared after the demand subsides.It takes longer to clear these queues in IMC.CODAS defines IMC as that combination of ceiling and visibility for which visual approaches are no longer permitted.In 1997 before the new DFW runway became operational, arrival capacity could often be reduced by loss of a diagonal runway, resulting in larger queues during transient arrival rushes and longer residual recovery periods after the rushes subside.Figure 4 separates the CODAS data into IMC and VMC components for 7 important airports (Atlanta (ATL), DFW, LGA, BOS, Philadelphia (PHL), Newark (EWR), and LAX) with varying operational characteristics.Results similar to DFW were found for all of these airports: on IMC days the means, standard deviations, and peak delays were significantly larger than their values on VMC days.The observation that BOS and DFW delays were equally sensitive to IMC is somewhat unexpected.The sensitivity of arrival runway capacity to meteorological conditions differs significantly between these two airports.In some reconfiguration situations, Boston's arrival runway capacity can drop by nearly 50% in +American Institute of Aeronautics and AstronauticsIMC, whereas the biggest IMC arrival runway capacity drop possible at DFW in 1977 was about 33%. +DELAY CORRELATIONThe fact that all the airports experience larger delay means and variances on IMC days than on VMC days seems to support the notion of local causality: that is, if we can increase IMC arrival capacity at EWR, we should also reduce delays at EWR.However, CODAS airborne delay is not as local as one might suppose.When we examine the correlation between delays at airport pairs, we find evidence of systematic effects correlating delays over the region for CODAS airborne delays as well as arrival delays.This occurs even on mixed days when one airport experiences some IMC and the other experiences solid VMC.We also see that correlation decreased as the geographical separation between airports increased.Figure 5 shows the correlation between CODAS arrival delays at EWR and LGA for all days in 1997 for all four combinations of meteorological conditions.The correlation was strong.The correlation between CODAS arrival delays on the days in 1997 in which IMC prevailed at both airports was 0.85.Correlation was equally strong on those few "mixed" days when it was IMC at one airport and VMC at the other.The weakest correlation was for the majority of days when the weather was clear at both airports.Even on these days the correlation was significant and positive at 0.49.Strong correlation between EWR and LGA delays is to be expected.Their traffic is managed by a common TRACON.The airports are close to each other geographically and share common arrival and departure fixes 9 .Because of this physical proximity, weather conditions were also correlated between the two airports: in 1997 there were only 34 days -split 15/19in which one airport experienced some IMC and the other experienced solid VMC; there were 121 days when both experienced IMC; and there were 208 days when both experienced solid VMC.Positive correlation also occurs between CODAS arrival delays at other airport pairs since delays relative to schedule are correlated by the connectivity of the air transport network.What is surprising is that strong positive correlation occurs between CODAS airborne delays at airport pairs, even though those delays are measured relative to planned flight duration rather than scheduled arrival time.Figure 6 shows the correlation between EWR and PHL for CODAS airborne delays.There was strong correlation even though these were non-schedulerelated delays, the distance between the airports is greater, and their air traffic is managed by different facilities.Figure 7 summarizes the correlation coefficients between selected airports for CODAS airborne and arrival delays for all days in 1997.The correlation generally decreased as the distance between the airport pairs increased.There was a small positive correlation between delays at PHL and BOS, although most of this correlation was chance on days when it was VMC at both airports and the delay was small.The annual airborne delays were not correlated for widely separated airports.However, the annual arrival delays showed small positive correlation between all of these airport pairs because of downstream schedule impacts on high delay days.For example, the schedule-based arrival delays at DFW and ORD were positively correlated, probably because both airports are major hubs for American Airlines. +American Institute of Aeronautics and Astronautics +DOWNSTREAM DELAYDownstream delay caused by schedule connectivity can multiply the cost of large delay events.Late arrivals propagate through airline schedules and result in additional downstream delays.This delay multiplication effect multiplies the dollar benefits from reductions in initial delay.We developed an analytical model that allows us to estimate the magnitude of the downstream arrival delay resulting from direct arrival delay at DFW in 1997.The model is based on a published analysis by Beatty et al of empirical downstream schedule delay trees resulting from 500 delayed flights into DFW. 10hat analysis showed that the number of minutes of downstream delay resulting from each initial delayed flight is roughly proportional to the number of minutes that the initial flight was delayed.The relationship of the delay multiplier to the time and duration of the initial delay was modeled in the form DM=1+S*DD, where DM is the delay multiplier, DD is the number of minutes the initial flight was delayed, and the dimensionless delay-time factor S is an empirically derived function, which is greatest when the initial delay occurs at 6 AM.We found from that S decreases approximately linearly as the time of the initial delay increases from 6 AM to 10 PM, as shown in Figure 8.This delay-time factor can be used to estimate delay multipliers for CODAS daily average arrival delay (which is equivalent to the mean value of DD).At DFW in 1997 these daily averages ranged from about 2 minutes per flight to about 58 minutes per flight and accumulated 3.72 million minutes of initial delay for flights into DFW over the year.We estimated the value of DM for each day resulting from the reported mean of the initial delay DD for that day.The linearity of the factor S and the nearly symmetric time distribution of arrivals at DFW between 6 AM and 10 PM allow us to estimate the downstream delay for each day from the mean value of S (which corresponds to an initial delay occurring at about 1:30 PM).This approximation tells us that, on average, the downstream delay was about 20% of the initial arrival delay at DFW for all days in 1997.This result is shown in Figure 9, which also includes the delay multiplication factors for calm days and storm days.The total cumulative CODAS arrival delay relative to schedule for flights into DFW in 1977 was 1.43 million minutes on thunderstorm days and 2.29 million minutes on days without storms.The result of the downstream delay calculation for days with and without storms is shown as initial plus downstream delay in this stacked column chart.Because storm days had larger initial delay they also had larger downstream delay.Thus, the effective multiplier was about 1.3 for storm days compared to 1.13 for calm days. +HAZARDOUS WEATHERFAST is currently unable to predict flight trajectories when storms disrupt arrival routes.Thus, thunderstorms reduce the amount of time that FAST can be used.However, such route disruptions are infrequent and the benefit of extra runway capacity American Institute of Aeronautics and Astronautics increases disproportionately when the storm has passed and controllers must clear out residual storm queues.To determine the net effect of thunderstorms on FAST benefits it is necessary to quantify the relationship between hazardous weather and delay.We examined hazardous weather delays at DFW in 1997 and at EWR in 1999. 11Weekly report logs from the Integrated Terminal Weather System (ITWS) at DFW indicate that there were 94 days with thunderstorms within 50 nautical miles of DFW. 12 The DFW TRACON logs show that on about 50 of these days the storms involved enough disruption to air traffic to cause delays.At EWR there were 36 days with thunderstorms within 100 NM of the airport that caused major delays.These numbers are higher than the number of days in which thunderstorms were officially reported at DFW and EWR.Tower personnel report thunderstorms at an airport when they detect lightning or thunder.On average that occurs 45 days a year at DFW and 26 days a year at EWR. Figure 10 is a plot of the CODAS airborne delay at DFW on the 50 worst delay days in 1997 sorted by delay magnitude.The 14 worst days all had thunderstorm activity.Thirty-four of the 40 worst airborne delay days were thunderstorm days.Large airborne delays are strongly associated with thunderstorms.Yet, in spite of the fact that days with very high delay often experienced thunderstorms, the total annual delay on storm-free days was about 42% larger.Figure 11 shows the cumulative 1997 CODAS airborne delay separately for thunderstorm days and all other days at DFW sorted in order of descending airborne delay.We multiplied the average delay on each day by that day's arrival count to obtain the cumulative aircraft delay minutes.The cumulative annual CODAS airborne delay on thunderstorm days was about 415,000 minutes.The cumulative delay on other days was 591,000 minutes.The direct operating cost to airlines at DFW in 1997 can be estimated by multiplying the airborne Delays by the $19/minute estimate obtained from the Seagull and LMI benefit analyses.The results total $11.2M for storm-free days and $7.9M for thunderstorm days.The large savings for calm days occurred partly because there were more calm days and, to a lesser extent, because there were more arrivals on calm days.The cumulative minutes and dollars for thunderstorm days would be larger if the calculation included nominal delay and dollar equivalents for each cancelled flight.A complete cost accounting for downstream delay increases storm-related costs relative to costs on stormfree days because larger delays cause larger downstream ripple effects.We examine the magnitude of this effect below.Thunderstorms and IMC are both important contributors to large CODAS airborne delays at DFW.But there are also other sources of delay.Although the predominantly North-South orientation of the DFW runways makes it potentially vulnerable to crosswinds, DFW had only one day in 1997 that was free of IMC and thunderstorm activity but that had CODAS airborne delays greater than the average for an IMC day.On December 9, the delay built up during five hours of 20to 25-kt crosswinds after 1PM, but a long period of delay in the morning when the winds were below 10 kt also contributed to the high daily delay.Unlike EWR in 1998, where winds alone caused numerous large delay events, DFW in 1997 did not experience significant delay contributions from high winds.Delays can also be caused by inefficient handling of arrival traffic or by contention for air space and runways in peak arrival periods.We found that days with high average delay at DFW have statistically lower daily arrival counts.(This was seen at EWR also, where We further analyzed the effect of weather on delay at DFW in 1997 by dividing the days into categories with and without thunderstorms and with and without periods of IMC.We found that solid VMC days with thunderstorms had mean delays almost as small as VMC days without thunderstorms, likely because the storms were far from the airport and good visibility at the airport helped clear any queues that occurred from flow disruptions.Consequently it is not necessary to distinguish between the two types of VMC days.Figure 12 summarizes the CODAS airborne delay for the whole year for the three main weather combinations.As shown in part a), VMC days had the smallest average CODAS airborne delay (1.9 minutes per arrival).Days with IMC and no thunderstorms averaged 2.9 minutes of delay per arrival.Days with thunderstorms plus IMC averaged 6.1 minutes of delay by per arrival, more than double that of storm-free IMC days.Part b) shows the number of days at DFW in 1997 that experienced each weather category.237 days were solid VMC.79 days had one or more hours of IMC, but no thunderstorm activity within 50 NMI of the airport.And 49 days had one or more hours of IMC plus thunderstorms within 50 NMI of the airport.Part C gives the resulting cumulative annual delay for each of the three weather conditions.Because the many small VMC delays occurred regularly during daily arrival rushes they contributed 46%, of the annual total.The 79 IMC days without thunderstorms contributed 24% of the annual total.The 49 days that had both thunderstorms and periods of IMC contributed the remaining 30%, which was the second largest cumulative annual delay.These 49 days also included +ConclusionsThe overall direct savings estimates of prior simulation studies appear to be both consistent with their underlying capacity predictions and conservative.However, their conclusion that there is potential for a full-time capacity increase of 4 aircraft per hour is critical to their results and should be validated by analysis of operational and radar data for individual airports.FAST benefits will accrue in all weather conditions.However, because en route and terminal airspace congestion causes queuing delay every day during arrival rushes, VMC days will contribute most of the annual delay benefit.Large queuing delays are also caused when thunderstorms and strong winds disrupt routes, and although thunderstorms occur relatively infrequently, they contribute significantly to annual delay.Therefore the ability to clear out storm queues may be an important benefit potential for FAST that was not considered in prior studies.Downstream delay propagation data indicates that accounting for downstream delay can increase FAST benefits by an additional 20%.Figure 1 FASTFigure 1 .11Figure 1 compares the 1997 CODAS average annual airborne delay at 10 airports with the LMI model estimates for annual delay that would have been saved in 1997 by FAST and TMA at those airports.Although the general trends of the model data and delay data are similar, the delay savings estimates for the airports are +Figure 3 Figure 3 .33Figure 3 compares CODAS airborne delay on IMC and VMC days at DFW in calendar year 1997.35% of the days had one or more IMC hours between 6 am and midnight.The top 33 delay days were all IMC days and 38 of the top 40 delay days were IMC days. +Figure 4 .4Figure 4. CODAS arrival delay for seven airports on VMC and IMC days -1997. +Figure 5 .Figure 6 .56Figure 5. EWR and LGA -daily CODAS arrival delay correlation -1997 all days. +Figure 7 .7Figure 7. Correlation coefficients for CODAS delay for nine airport pairs -1997 all days. +Figure 8 .Figure 9 .89Figure 8. Delay-time factor S as a function of the time of the initial delay. +Figure 10 .10Figure 10.CODAS airborne delay on 50 worst days -DFW 1997. +Figure 11 .11Figure 11.Cumulative CODAS airborne delay on days with and without storms-DFW 1997. +5 of the 6 ground hold days for flights into DFW in 1997.Part d) of the figure shows the distribution of delays with weather when the downstream arrival delay multipiers of Figure 8 are used to multiply airborne delays.The percentage contribution of storm days increases slightly, but VMC days still predominate. +Figure 12 .12Figure 12.CODAS airborne delay statistics for three weather conditions -DFW 1997. +American Institute of Aeronautics and Astronauticsthe average number of cancellations per thunderstormor IMC event was more than 26 flights.) An airlinedoes not cancel a flight because the demand it will generate might cause delays. Airlines cancel flights because they anticipate-or are already experiencing-costly disruptions from other causes. Although high peak demand usually increases peak delay, high daily demand is negatively correlated with high daily delay at DFW.Delay (minutes/arrival)1 2 3 4 5 7 6a) Mean Delayb) Number of Days in Year VMC 237 IMC + Storm 49 IMC, no storm 790VMCIMC, no storm IMC + Stormc) Cumulative Delay (%)d) Cumulative with Downstream Delay (%)IMC + Storm 30%VMC 46%IMC + Storm 33%VMC 44%IMC, nostormIMC, no24%storm23% + + + + +*This work was performed for the National Aeronautics and Space Administration under Air Force Contract No. F19628-00-C-0002. + + + + + + + + + + GHunter + + Inc. TR96150-02 + + + CTAS Error Sensitivity, Fuel Efficiency, and Throughput Benefits Analysis + + July 1996 + + + Hunter, G et al, "CTAS Error Sensitivity, Fuel Efficiency, and Throughput Benefits Analysis," Seagull Technology, Inc. TR96150-02 (July 1996). + + + + + Initial Air Traffic Management (ATM) Enhancement Potential Benefits Analysis + + GCouluris + + + TWeidner + + + JSorensen + + Inc. TR96151-01 + + + Seagull Technology + + September 1996 + + + Couluris, G., Weidner, T., Sorensen, J., "Initial Air Traffic Management (ATM) Enhancement Potential Benefits Analysis," Seagull Technology, Inc. TR96151-01 (September 1996). + + + + + CTAS Benefits Extrapolation Preliminary Analysis + + TWeidner + + + GCouluris + + + GHunter + + Inc. TR96156-01 + + + Seagull Technology + + February 1998 + + + Weidner, T., Couluris G., Hunter, G., "CTAS Benefits Extrapolation Preliminary Analysis," Seagull Technology, Inc. TR96156-01 (February 1998). + + + + + Capacity-Related Benefits of Proposed Communication, Navigation, Surveillance, and Air Traffic Management Technologies + + TWeidner + + 10.2514/5.9781600866630.0269.0287 + + + Air Transportation Systems Engineering + Orlando, FLA + + American Institute of Aeronautics and Astronautics + December 1998 + + + + 2nd USA/EUROPE Air Traffic Management R&D Seminar + Weidner, T., "Capacity Related Benefits of Proposed CNS/ATM Technologies," 2nd USA/EUROPE Air Traffic Management R&D Seminar, Orlando, FLA, (December 1998). + + + + + Licensee contractor and vendor inspection status report. Quarterly report, April 1997--June 1997 + + Lee + + 10.2172/577000 + CR-1997-201682 + + April 1997 + Office of Scientific and Technical Information (OSTI) + + + NASA Contractor Report + Lee et al, "Estimating the Effects of Terminal Area Productivity Program," NASA Contractor Report CR-1997-201682 (April 1997). + + + + + Benefit Estimates of Terminal Area Productivity Program Technologies + + RHemm + + CR-1999-208989 + + January 1999 + + + NASA Contractor Report + Hemm, R. et al, "Benefit Estimates of Terminal Area Productivity Program Technologies," NASA Contractor Report CR-1999-208989 (January 1999). + + + + + Operational Test Results of the Passive Final Approach Spacing Tool + + TJDavis + + + DRIsaacson + + + JERobinson + + + WDen Braven + + + KKLee + + + BSanford + + 10.1016/s1474-6670(17)43820-1 + + + IFAC Proceedings Volumes + IFAC Proceedings Volumes + 1474-6670 + + 30 + 8 + + June 1997 + Elsevier BV + Chania, Greece + + + Davis, et al, "Operational Test Results of the Final Approach Spacing Tool," Proc. IFAC 8 th Symposium on Transportation Systems '97, Chania, Greece, (June 1997) + + + + + U.S. Airport Infrastructure and Sustainable Aviation Fuel + + KristiMoriarty + + Aviation Capacity Enhancement Plan + + + + AllisonKvien + + Aviation Capacity Enhancement Plan + + + 10.2172/1768316 + + + + ) and CODAS DataBase + + Office of Scientific and Technical Information (OSTI) + 1998. December 1998 + + + Federal Aviation Administration Office of System Capacity, 1998 Aviation Capacity Enhancement Plan," www.asc.faa.gov (December 1998) and CODAS DataBase, www.apo.data.faa.gov. + + + + + Complex, Congested Airspace and Impacts on Airline Operations + + Dearmon + + + 2000 + McGraw Hill + 34 + 495 + New York + + + Handbook of Airline Operations + DeArmon et al, "Complex, Congested Airspace and Impacts on Airline Operations," Chapter 34, p. 495, "Handbook of Airline Operations," McGraw Hill, New York, 2000. + + + + + Preliminary Evaluation of Flight Delay Propagation through an Airline Schedule + + RogerBeatty + + + RoseHsu + + + LeeBerry + + + JamesRome + + 10.2514/atcq.7.4.259 + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 7 + 4 + + December 1998 + American Institute of Aeronautics and Astronautics (AIAA) + + + 2nd USA/EUROPE Air Traffic Management R&D Seminar, Orlando FLA + Beatty, R. et al, "Preliminary Evaluation of Flight Delay Propagation Through an Airline Schedule," 2nd USA/EUROPE Air Traffic Management R&D Seminar, Orlando FLA (December 1998). + + + + + Delay Causality and Reduction at Newark International Airport Using Terminal Weather Information Systems + + SAllan + + + SGaddy + + + JEvans + + ATC-291 + + July 2000 + + + MIT Lincoln Laboratory Project Report + Allan, S., Gaddy, S., Evans, J., "Delay Causality and Reduction at Newark International Airport Using Terminal Weather Information Systems," MIT Lincoln Laboratory Project Report ATC-291, (July 2000). + + + + + The Integrated Terminal Weather System (ITWS) + + JEvans + + + EDucot + + + + Lincoln Laboratory Journal + + 7 + 2 + 1994 + + + Evans, J., Ducot, E., "The Integrated Terminal Weather System (ITWS)," Lincoln Laboratory Journal, Vol. 7, No. 2, (1994). + + + + + + diff --git a/file779.txt b/file779.txt new file mode 100644 index 0000000000000000000000000000000000000000..50c5516613b0c699e7d488540f0365c8bd16f3ac --- /dev/null +++ b/file779.txt @@ -0,0 +1,578 @@ + + + + +I. IntroductionA landing times and predictions thereof are of interest to many stakeholders for many reasons.Airlines provide value to their passengers by arriving at the destination when scheduled, particularly if passengers or crew need to make connecting flights.Airports and flight operators use arrival predictions to schedule support services for inbound flights (parking, fueling, loading, etc.).These support services often require the relocation of equipment and personnel.The limited nature of these resources makes timing critical for efficient operations at the airport.Air traffic controllers and Traffic Management Coordinators (TMC) keep aircraft safely separated and moving efficiently in the air and on the airport surface.Predictions of events like landing times helps TMCs to safely and efficiently control and manage arrival traffic.Air traffic management systems such as Time-Based Flow Management (TBFM) and the Airspace Technology Demonstration 2 (ATD-2) Integrated Arrival, Departure and Surface (IADS) scheduler [1] leverage such predictions to project upcoming demand of arrivals at airports.Given the importance of landing time predictions, it is no surprise that a variety of physics-based and machine learning (ML)-based approaches have been developed to predict these times.Within the Federal Aviation Administration (FAA), several National Airspace System (NAS) systems process trajectory data to predict arrival times for flights.Using an internal trajectory model, these systems predict and monitor arrival times for flights within the scope of the systems' operation.The arrival time predictions resulting from these trajectory models may be published on several available data feeds.For this project, we looked at the arrival time calculations produced by two FAA flow management systems.Brief descriptions of their physics-based landing times are provided below.• Traffic Flow Management System (TFMS): The Estimated Time of Arrival published on the TFMData System Wide Information Management (SWIM) feed is a prediction of a landing (also known as "wheels-on") time for a flight.As a flight progresses, this time is updated by TFMS.An airline-provided runway arrival time is also sometimes available on this feed, depending on the flight, but this time is not produced by TFMS.TFMS also publishes the full trajectory of many flights, describing a flight's traversal through waypoints along its route.Traversal data in TFMData flight messages also shows elapsed time from departure for each waypoint along the flight's route, with the final waypoint in the route correlating to the Estimated Time of Arrival for the flight.• Time-Based Flow Management (TBFM): TBFM publishes calculated landing times through the TBFM Metering Information Service (TBFM-MIS) SWIM feed.TBFM facilities produce landing times, Scheduled Time of Arrival and Estimated Time of Arrival, with and without expected traffic flow constraints, respectively.In the FAA's Air Traffic Control manual, TBFM's Scheduled Time of Arrival "takes other traffic and airspace configuration into account.[The Scheduled Time of Arrival] shows the results of the TBFM scheduler that has calculated an arrival time according to parameters such as optimized spacing, aircraft performance, and weather [2]."Unlike the Scheduled Time of Arrival, the Estimated Time of Arrival does not include the effects of constraints.These arrival time predictions are produced for different purposes specific to the system they support, so the calculated arrival times are not identical across systems.Some incorporate expected traffic management control actions, some do not, depending on the scope of the system.Those systems that do incorporate adjustments do not necessarily share those expected adjustments with the other systems.Previous efforts on Air/Ground Trajectory Synchronization (AGTS) have produced business rules that attempt to select the most accurate scheduled time of arrival at a meter fix [3], but we are not aware of any similar approaches for selecting from among candidate landing time predictions from disparate, operational physics-based trajectory models.The physics of aircraft flight is well-understood, so physics-based models can be very accurate when configured appropriately and provided with accurate input data, and given an unimpeded path.However, other flights are also competing for the same shared and constrained resources (e.g., air traffic controllers, runways, gates).In addition, the source data used by these systems sometimes lacks coverage, or is subject to message timing irregularities which can cause inaccuracies in the prediction.Data prediction issues are particularly noticeable around the time of a flight's departure, and just before the flight lands at its destination.Machine Learning (ML) models can be an alternative method to produce accurate predictions of future event times, particularly when sufficient training data is available.Previous ML approaches for predicting landing times have leveraged input features like distance-to-go, scheduled time-to-go, departure delay, airline, surveillance data describing the current state of the aircraft, meteorological conditions at the arrival airport, airport configuration, expected traffic volumes near and at the arrival airport, time-of-day, time-of-week, and season [4][5][6][7][8][9].For instance, Ayhan, Costas, and Samet developed features describing the meteorological conditions and airspace congestion likely to be encountered in the airspace along the expected route of travel [10].Our approach mixes physics-based predictions with ML-based predictions, which has been shown to be effective in other situations [11].None of the previous ML research explicitly leveraged the physics-based landing predictions as a feature to their ML-based predictions.Most modeling approaches construct separate models per arrival airport, or even per origin-destination pair, but some train a single model that applies to flights bound for various arrival airports [5,6]; origin and destination airport identifiers are input features for such models.Boosting and bagging tree-based ML model types such as gradient boosting machines and random forests performed relatively well in previous aircraft landing time prediction research [4,6,8,9].Multiple authors developed sequences of half a dozen or more regularized linear models and tree-based models, with downstream models predicting the residual of upstream models, sometimes using separate sub-sets of input features [5,8].Some approaches predict not only a point estimate of the landing time, but a probability distribution for the landing time as well, which can be useful for downstream risk-based decision-making [4,7].Although there is a fundamentally time-series nature to the data and landing time prediction problem, we found only one case in which authors explicitly formulated and solved a time-series prediction problem.Ayhan, Costas, and Samet developed a Long Short-Term Memory (LSTM) recurrent neural network (RNN) model, but its performance did not exceed that of a boosting and bagging tree-based ML model that incorporated no explicit notion of a time-series [10].To the best of our knowledge, none of these ML models resulted in operational use, and none addressed how to ensure all required input data might be made readily available for real-time use.In this research, we developed a novel approach for predicting landing times of airborne flights bound for many airports in the continental United States.First, we developed mediation rules to select from among three operationallyavailable physics-based landing time predictions at any time and for any flight.Second, we developed ML models that predict landing times based on input features describing the flight, arrival airport, other traffic, and, unlike previous research, an operationally-available physics-based landing time prediction (the mediated prediction).The performance of these predictions is then compared.Third, we placed a special a emphasis on developing deployable models by using only publicly-and readily-available data and minimizing dependence on adaptation data describing airports or airspace.This report is structured as follows.Section II describes the mediation and ML steps to predict landing times.Section III describes the data that was used to develop and evaluate the results.Section IV summarizes initial results comparing the performance of predictions.Finally, Section V provides next steps for future efforts. +II. ApproachThe first part of the approach was to select a single estimated runway arrival time from available physics-based estimates.The selected time is the "mediated landing time."The mediated landing time is then used as an input to the ML model.The output of the model is a correction factor, which is then applied to the mediated landing time, to be used as an improved landing time prediction.We started with a rather narrow focus when setting up the mediation rules, then widened the scope when training the ML model.For the mediation rules, we picked a day for which we had a good set of data available, 8/19/2019, to consider options for mediation rules.For the model training and testing, we expanded the input data to include flights from April 2020 through August 2020, but narrowed the focus of the airports to those considered for near-term deployment of ATD-2 capabilities. +A. Mediation of Available Physics-based Predictions of Landing Times +Data SourcesWe considered three operational landing time estimates.These values are described in Table 1.Each of these values are available in SWIM feeds as listed.The field names of the data elements within the SWIM messages are listed in the Field column.In this report, the data elements are referred to as the name in the Alias column to disambiguate their source.The TFMS ETA may start to be published up to approximately one full day prior to the flight's expected departure, with infrequent updates until the flight departs from its origin.Once the flight is airborne, the values are typically updated at a rate of approximately once per minute.There are multiple TBFM systems running in facilities in the NAS.The TBFM ETA is not produced by every TBFM system that produces data for a particular flight.A flight filed route may cross several airspaces captured by various TBFM systems, however, only the TBFM system used to manage the arrival flows at the destination will publish a TBFM ETA for the flight.If the TBFM system at the destination facility has not published any messages for the flight yet, there will not be a TBFM ETA, even if the flight is airborne.The TBFM ETA may be updated as frequently as about once every 6 seconds, but there may be 10 minutes or more between published updates.Like the TBFM ETA, the TBFM STA is only provided by the TBFM system at the destination.Not all flights will receive a TBFM STA.The TBFM STA is expected to be the time the flight will arrive at the runway, provided the flight will be controlled according to the assumed constraint applied at the scheduling fix STA (sta_sfx), which is a scheduled time at a constrained point typically near the arrival fix.Because scheduling is based on constraints at the fix rather than at the runway, the TBFM STA is likely less accurate than the sta_sfx.When used for arrival metering, the TBFM STA may become "frozen."All STA values for a flight may be frozen, once the flight enters a freeze horizon that is associated with a metering fix.The freeze horizons are optional and can be turned on or off as needed by TMCs.Prior to freezing, the TBFM STA will typically update at a frequency about the same as the TBFM ETA.When frozen, the TBFM STA will stop updating.The TBFM STA may be rebroadcast periodically when frozen.It is possible for a frozen TBFM STA to become unfrozen, either because freeze horizons are turned off, because the list of arrivals is reshuffled to address a sequencing issue, or because of a manual unfreezing action by a TMC to re-sequence flights. +Mediation RulesBecause of the differences in scope and design of TFMS and TBFM systems, there are differences in the landing times calculated for any given flight.The ML model we developed takes only one predicted landing time as a feature for its prediction.In this section, we describe the development of mediation rules used to decide which of the available landing times should be used as an input for the ML model.While we sought to develop mediation rules that selected the physics-based prediction with the lowest errors, we also imposed constraints on these rules.We aimed for a simple set of rules that would be easy to explain to users and to develop and maintain.Furthermore, we decided to derive a single set of rules that applied to arrivals at all airports in the NAS.Prediction accuracy could be improved by relaxing these constraints, but complexity of the rule set would increase.Before departure, use TFMS ETA Prior to departure, the TFMS ETA appears to be more accurate than TBFM ETA and TBFM STA.Flights that are still on the surface should not use the TBFM ETA or TBFM STA for a predicted landing times.Figure 1 shows accuracy of the estimates relative to the actual departure time.After departure, TBFM ETA and TBFM STA accuracy improves and stabilizes.The figure shows mean error from a sample day of flights operated on 8/19/2019, destined for airports with Digital-Automatic Terminal Information Service (D-ATIS) availability (76 relatively major airports).These airports were selected to include the usefulness of checking the consistency between TBFM runway assignments and current D-ATIS runway configuration.Because the correlation of these values did not result in an improvement in the mediated landing time, it is not discussed further in this paper.After departure, prefer TBFM STA TBFM systems compare the amount of expected traffic to the configured capacity of Meter Reference Elements (fixes and arcs) within their area of responsibility.When a TBFM system expects the capacity of its resources to be exceeded by the demand, it calculates a TBFM STA that includes the amount of time it expects the flight needs to satisfy the constraint of the affected resources.If the TBFM system does not detect such a demand-capacity imbalance, the TBFM STA and TBFM ETA will be equal to one another.However, because the TBFM STA and TBFM ETA are not published in the same message, the latest available TBFM STA or TBFM ETA visible to SWIM consumers may not precisely reflect the current expectations of the publishing TBFM system.When all three estimates (TBFM STA, TBFM ETA, and TFMS ETA) are available, the TBFM STA appears to be most often closer to the actual landing time.This can be seen in Figure 2, which plots the mean accuracy for each estimate across all available airports for the last 30 minutes of the flights.The lines show the mean absolute difference between value available from the estimate and the actual landing time, as the flight approaches actual landing time.The dashed line shows a theoretical "best" ETA, using whichever of the three available times is closest to the actual arrival time at each instant and for each flight.Typically within minutes after a flight departs, the flight starts being "tracked" by the destination TBFM system.This is about the time the flight is picked up by the radar surveillance system, and not necessarily the time the flight took off.When a flight is not tracked by the TBFM facility, the TBFM ETA may not be updated consistently, which may cause stale estimates.Ideally, our mediation rules would key off this tracked indicator.Because different TBFM systems may begin tracking the same flight at different times, and because the flight may be dropped and tracked again, it can be challenging to extract this data element from the data feed.For this effort, we choose the simple approach of using the flight's departure status as a proxy for the tracked indicator. +Account for inconsistencies near the arrival timeIn the last few minutes of a flight, the flight passes the arrival fix, and the accuracy of the TBFM ETA and TBFM STA can deteriorate prior to landing.Around this time, the flight is being handed off between the Air Route Traffic Control Center (ARTCC) and a Terminal Radar Approach Control (TRACON) facilities.The en route controller is effectively finished with the flight, and thus no longer separating the flight at the fix to which the TBFM STA is referenced.This can be seen in the last few minutes of Figure 2, as the mean absolute error for TBFM ETA and TBFM STA begins to increase slightly in the last few minutes to arrival, while TFMS ETA appears to improve.At this same time, the TFMS ETA may fall to a time in the past until it is corrected after the flight lands.The closer to landing on the runway, the more flights had TFMS ETA values in the past.Figure 3 shows counts of arrivals as they approach their actual landing time.The flights are split into those whose latest published TFMS ETA had a value which was still a future time relative to the current time, and those whose latest published TFMS ETA has gotten stale to the point where it implied the flight should have arrived already (at the bottom right of the figure). +Fig. 3 State of TFMS ETA relative to current time.Given that the TBFM ETA and TBFM STA appear to lose their predictive value when the aircraft is very near its landing time, the TFMS ETA is presumed to be the more accurate estimate in these cases, regardless of the existence of the TBFM STA.When the TFMS ETA indicates a time in the past, that time will be used at the mediated landing time. +Implemented mediation rulesThe mediation rules that were used are summarized in Algorithm 1. Applying these rules improved the mediated arrival estimate (black line in Figure 4) compared to only using the TFMS ETA.However, the "theoretical best" (dashed line), below the mediated error line, indicates there are more improvements that could be made to this estimate. +B. Machine Learning for Predicting Landing Times +Data sourcesNot all the available data used for modeling was machine-readable after being parsed from available data feeds.Three sets of "adaptation" data were required for the development of the ML models.Adaptation data describes characteristics of airspace, airports, aircraft, or other relevant entities in a way that is machine-readable and can be leveraged by ML model-training algorithms or other computer programs for which such data are relevant.Although adaptation data can greatly empower such programs and indeed is sometimes required, it is also expensive to produce and maintain in the sense that it requires substantial domain expertise and tedious effort to generate and update.To reduce this labor and facilitate easy deployment at new airports, we have deliberately avoided adaptation data wherever possible.In many cases, the information provided by adaptation data can be extracted from training data and/or it is not required by ML models.The first set of adaptation data we utilized is a classification of aircraft types into classes like narrow body, wide body, and regional jet.Aircraft types are typically stored as an encoded value in flight data.The FAA lists hundreds of such encoded values in their Aircraft Type Designators documentation [12].The encoded values in the flight data are a level of abstraction from one or more individual aircraft models, and the classes are a further abstraction to group together many of these encoded values.The second is a list of known runways at the airport for which a model is being developed.This could be learned from the data, but an explicit specification of candidate runway names facilitates coordination between this ML model and an upstream arrival runway prediction model.The FAA regularly publishes an updated list of airports and their runways in its 28-day chart updates [13].The third is FAA National Flight Data Center (NFDC) airport runway polygons.These runway polygons represent the area of each runway defined for each modeled airport.These polygons can be compared to surveillance data to determine when a particular aircraft has moved onto a runway surface.Essentially all of the data used in the landing time prediction models come from publicly-available FAA SWIM data feeds.This National Airspace System (NAS)-wide data was processed, synthesized, and loaded into a database by the Fuser system developed by NASA and Mosaic ATM [14].The FAA systems that produce data that were Fused and then used or intended to be used for this effort are:• Time-Based Flow Metering (TBFM),• Traffic Flow Management System (TFMS), and • Airport Surface Detection Equipment, Model X (ASDE-X).More particularly, we queried the Fused data to generate the following raw data for each flight bound to an arrival airport:• TBFM messages,• TFMS surveillance data,• Flight characteristics such as aircraft type and carrier,• Actual times known when flights are airborne, such as gate departure and take off times, and • Airline-scheduled times such as landing times.Although the actual time a flight lands at its destination airport is available in FAA SWIM data, the runway on which the flight lands is not.For this effort, we derived both of these data elements from surveillance data (ASDE-X or TFMS) and NFDC runway adaptation data [13] to ensure consistency and reliability of these crucial data elements.When surveillance located a flight within a runway polygon and with physics consistent with aircraft landing, it was straightforward to determine that the flight arrived on that runway at a certain time.In some cases, surveillance data was not available to definitively assess the arrival runway and time.Although this situation was relatively rare at large airports, we were able to approximate the values when necessary.We used the airborne surveillance data from TFMS to infer which runway was used, and at what time the aircraft landed.We accomplished this with a simple physics-based approach developed by Robert Kille [15].The results presented here are based only on flights for which we were able to directly assess the landing runway and time.We omit those cases for which we had to rely on airborne surveillance and the Kille Method, which is less reliable than the more direct assessment.Even so, predictions of landing times can be produced for flights regardless of which approach is used after the landing event to infer the actual landing runway and time.The target for the ML models is the actual landing time derived from surveillance data as described here, not any landing time recorded explicitly in SWIM data. +Input to real-time ML model serviceThe ML models developed for this effort have been carefully created to ensure they can be used in real-time with readily-available data.Table 2 describes the 13 input data elements accepted by the initial ML model services.Only the first three of the listed data elements are required to make a prediction, which is important for real-time scenarios in which data is often missing.Imputation strategies are described in the table.For timestamp input data, missing values are imputed with related timestamps.Missing categorical input like air carrier or arrival runway are encoded to binary values ("one-hot encoded").Each known category becomes a feature for the model and is set to 1 instead of 0 if and only if the input takes on that category value.For example, a flight operating under the major air carrier American Airlines (AAL) would have the feature major_carrier_AAL set to 1, but would have the feature major_carrier_DAL (representing Delta Air Lines) set to 0. If a flight is missing the category information altogether, all of the associated encoded feature columns are set to 0.Other input data are planned for inclusion in the future.These include surveillance data and expected runway demand.The TFMS surveillance data describes the current physical state of the aircraft, rather than just knowing the flight has departed from its origin.Demand at the flight's intended runway is planned to be approximated using counts of flights predicted to land on the same runway at roughly the same time as the flight in question. +Features provided to ML modelThe ML model uses features to determine the appropriate predicted landing time for a flight.The features that have been engineered and provided to initial ML models are shown in Table 3 for two sample airports.The first in this list of features, seconds to arrival runway best time, represents the seconds to the mediated physics-based landing time prediction, which is itself a prediction of the landing time.The mediated physics-based landing time prediction refers to the arrival time estimates published by TFMS and TBFM through SWIM messages, described earlier.Other features quantify the extent to which the flight is ahead of or behind the airline schedule, which previous research suggests is predictive of actual landing times [7].Many of the features are categorical, which are encoded as separate binary features for each categorical option.These categorical options can be particular to the modeled airport.For example, one of these categorical input features is the TBFM arrival runway.The TBFM arrival runway is not available for every flight.When the TBFM system is active, the TBFM arrival runway is produced by the TBFM system serving the flight's destination.It is an expectation by that TBFM system of the runway the flight will be assigned.This expectation is based on TBFM's own adaptation data for the airport, and air traffic controller input regarding current and planned airport configuration.In the case of KDFW, there are 7 physical runway surfaces.Each of these runways may be approached from either direction, making 14 possible values for an expected arrival runway.KJFK has 4 physical runway surfaces, therefore the number of possible values is 8 for that input.Because there are different numbers of categories at different airports, the total number of features can vary considerably from airport to airport.Another input to the model is the predicted arrival runway.This feature, by contrast to the TBFM arrival runway, is modeled for each flight, even for those with no TBFM arrival runway provided in the SWIM data.This input is actually generated from a separate ML model that predicts the arrival runway from some input features [16].As mentioned earlier, there are multiple TBFM systems serving the NAS.Each system has its own adaptation for its area of responsibility.Within each TBFM system, areas are further divided into TRACONs, which are divided again into airports.Each arrival airport may have its own set of defined metering resources ("meter fix"), and associated flows of air traffic ("stream class").This TBFM adaptation data is available in the TBFM-MIS SWIM feed, but is particular to a given facility.TBFM also publishes the stream class associated with a particular flight.A flight arriving at KDFW will be assigned a particular stream class by its destination TBFM system (ZFW, in the case of a KDFW arrival).A flight arriving at KDFW from a different direction will typically be assigned a different stream class.A flight arriving at KJFK will be assigned a stream class by a different TBFM system altogether (ZNY, in the case of a KJFK arrival).These stream classes are specific to a flow of traffic, and allow the ML model to group flights by flow without the use of origin airport or origin ARTCC as a feature.We intend to add more features for the ML model in the future.Other features that we intend to calculate and provide to the model include the planned additional input mentioned in the previous sub-section (future arrival runway counts) as well as differences between the current physical state of the aircraft and the typical physical state of flights at the moment they land on the predicted runway. +Training and testing data set engineeringWe trained the ML model with a sample that consisted of the state of a flight bound to an airport at a particular moment when it was airborne, as quantified by the model features, and the corresponding duration of time until it actually landed (the target value to be predicted).To construct data sets of samples for model training and testing, we first sampled the state of each flight bound to an airport every 30 seconds while it was airborne.Whereas the TFMS ETA is typically updated every minute, TBFM STA and TBFM ETA may be updated every few seconds.Sampling the data reduces some of the noise in the mediated physics-based landing time prediction values, and avoids bias when certain periods have much more frequent updates than others for a given flight.We then further pared down the data using random sampling.We randomly kept 2% of these samples, which equates to one sample every 25 minutes, on average.Similar random sampling of airborne times was used in previous research on predicting landing times [6].By down-sampling, we were able to reduce the volume of highly redundant samples in our data sets.Finally, to break the data into train and test sets, we randomly selected 30% of dates in the data set to be used for testing.All samples from flights that landed on those dates became test samples.This train-test split approach was used to avoid redundancy across samples, where flights could be represented in both the training and testing data sets.Such testing samples could be more highly correlated with training data samples than any future operational data would be and therefore lead to test set performance that is not representative of the performance that would be achieved by a deployed model. +ML model typesThe initial ML models have not been formulated to predict the time until the flight lands, but rather the error or residual in the mediated physics-based landing time prediction for a flight.More precisely, let denote the actual landing time of flight and let ˆ MPP ( ) be the mediated physics-based prediction (MPP) of the landing time for flight at time (where < ).Then the error or residual for flight at time is MPP ( ) = -ˆ MPP ( ).This error MPP ( ) becomes the target for the ML models.The trained ML models produce a prediction ˆ MPP ( ) of this error for flight at time , which can be converted into a predicted landing time ˆ ( ) by simply adding it to ˆ MPP ( ):ˆ ( ) = ˆ MPP ( ) + ˆ MPP ( ).(1)We set up the ML models to predict the errors in the mediated physics-based predictions of landing times for two main reasons.First, these predictions are readily available and in many cases already very accurate.Second, learning when and how an adequate predictor of a future event time is wrong is generally an easier ML problem than learning to predict the future event time without initial predictions.In fact, although we found no previous research in which ML models were trained to predict the errors of a physics-based predictor of landing times, multiple researchers have developed a sequence of models in which one model predicts the residual of the previous model [5,8] and indeed this idea is fundamental to "boosting" ML models.ML models are trained to minimize a loss function quantifying their predictive performance on training data, potentially with modifications to encourage model simplicity and prevent over-fitting to the training data.We have investigated Random Forest with mean squared error (MSE) loss, and XGBoost with MSE loss and Huber loss functions on two airports.This analysis is discussed in Section IV.B From our testing, we have found that in almost all error metrics we track, the XGBoost model with Huber loss performs comparably or often much better than the Random Forest model and the XGBoost model with MSE loss.Consequently, we trained XGBoost ML models with Huber loss function for four more airports.More specifically, we trained XGBRegressor models as implemented in the scikit-learn and XGBoost Python packages using the pseudo-Huber loss [17].Each model's hyperparameters were optimized using scikit-learn's GridSearchCV which performs an exhaustive search over specified parameter values using k-fold cross validation. +III. DataFor an initial training and testing of landing time prediction models, we collected five months (April 2020 through August 2020) of arrival data for Dallas/FortWorth International Airport (KDFW), George Bush Intercontinental Airport (KIAH), Dallas Love Field Airport (KDAL), John F. Kennedy International Airport (KJFK), Newark Liberty International Airport (KEWR), and Charlotte Douglas International Airport (KCLT).The training and testing data set size for each airport is reported in Table 4. +IV. Initial ResultsInitial results show a general improvement by the XGBoost model over the mediated physics-based prediction (MPP).Table 5 quantifies the performance of the MPP and ML model prediction (ML: XGBoost) of landing times in the test data set for each airport.For the XGBoost results in this table, we have used the Pseudo-Huber loss, as it resulted in superior performance.The duration of time until landing varies considerably across the training samples, so prediction errors of the same duration can have substantially different operational implications.Therefore, we selected two metrics that quantify the error (the difference between the actual and predicted arrival values) as a percentage of the true duration of time (the difference between actual departure and actual arrival): mean absolute percentage error (MAPE) and median absolute percentage error (MdAPE).The third metric is more aligned with operational use of the ML model.This metric is the percent of predictions that are within 30 seconds of the true value (Pct predictions within 30s).The window may be of operational interest because of the physical behavior of landing aircraft.Flights take time to slow down and exit the runway after touchdown, typically around a minute in duration.During that time, additional flights are not permitted to land on the same runway (barring special circumstances like formation flying).A prediction within this one-minute (±30 second) window around the actual landing time can be thought of roughly as predicting the "correct" arrival runway landing "slot."For almost all airports and metrics we see that the ML model prediction offers a modest to large improvement in performance over the mediated physics-based prediction.The few exceptions are that for KJFK and KCLT, where the mean absolute percentage error (MAPE) of the ML model predictions is considerably higher than the mediated physics-based predictions (MPP) for those same airports.However, the ML model still achieves lower median absolute percentage errors for these airports.This pair of results suggests that some outlier predictions with very large absolute 9.8% 15.6% 18.1% 17.0% 24.2% bold face indicates the method with the better performing value percentage errors exist, and are causing the higher mean absolute percentage error for the ML model predictions.These large errors in the predictions can occur especially when the true time to landing is very short, and relatively small landing time errors can equate to large percentage errors with respect to look-ahead time.At KJFK, neither model manages to predict even 10% of landing times within ±30 seconds of the true value.The phenomenon of large outliers is likely also the root cause of the surprisingly high mean absolute percentage errors for both methods at KIAH.The much smaller median absolute percentage errors at KIAH indicate that most percentage errors are within 5% of zero.In general, the further a flight is from its destination airport, the more variability in the arrival time prediction.Figure 5 shows the error distribution for KDFW and KJFK for the MPP and XGBoost models, respectively, by minutes prior to actual arrival.The largest error distribution is seen at the left, at 5 hours prior to landing.As the flight approaches its actual landing time (as assessed using surveillance data), the predictions tend to improve.The median error at each lookahead time prior to arrival is plotted as the solid blue line on the charts.In the case of KDFW, we can see that the median error is closer to zero for the XGBoost model than the MPP model for all lookahead times, however the 25 th to 75 th and 5 th to 95 th percentile range appears similar for both methods.For KJFK, the median is closer to zero and the percentile ranges appear smaller in the XGBoost model throughout the lookahead time.However, MPP XGBoost KDFW KJFK +Fig. 5 Error Distribution by Lookahead Timeas the minutes prior to actual arrival approach zero, the error becomes larger in the KJFK predictions in one direction.This means that we are starting to predict increasingly longer arrival times than the actual from approximately ten minutes out.At approximately this time, the flight is transitioning from en route to terminal airspace.The NAS systems most directly involved with the flight's systems are changing, and the flight is merging with local traffic for landing.Outliers are visible in the data.Figure 6 shows one point for each predicted value, with the predicted time to landing on the vertical axis, and the actual time on the horizontal axis.Both the MPP and XGBoost models are shown, for KDFW and KJFK.The KDFW plots show a very reasonable pattern with no large outliers visible.The KJFK plots, however, show a small number of large outliers in the lower left corner in which the actual time to landing was very large (around 1400 minutes) and the predicted time to landing was very small (under a few minutes).This occurs in both the MPP and XGBoost models therefore the large errors in the XGBoost model is due to the MPP data, as we are learning from the residuals of this model.We assume there were some data quality issues for these outliers.Matching flight data to a particular flight is a nontrivial task, and it is particularly troublesome when the source of flight data transitions between different automated NAS systems, as is happening shortly before the flight arrives at its destination and during arrival.Flight matching is further complicated by the reuse of flight data on sequential days.For example, a flight with a particular callsign can fly between the same city pair on different days, which may account for the approximately 1440 minutes (24h) difference, if the actual time to landing was assessed for the next day's flight.It is well known that air traffic has been greatly affected due to the pandemic.Not only did traffic counts greatly decrease, but the way air traffic operations functioned was affected at every level.For example, air traffic control facilities suffered staffing issues requiring the temporary closure of airspace and facilities, airline passengers required increased screening which complicated the boarding process, and aircraft required longer times between operations to be cleaned.In this section, we offer a comparison of the accuracy of the ML XGBoost model trained on 2019 data (mid-August 2019 to early December 2019) compared to the accuracy of the model trained on 2020 data (April 2020 through August 2020).The effect from year to year appears to be relatively minor regarding model accuracy.Table 6 shows the results of this comparison for KDFW, with the error values presented for each dataset.We had hypothesized that for models in which the overall traffic patterns would affect the predicted outcome that the performance might be worse.We did not expect, however, that the ML XGBoost model accuracy to be affected significantly, as it is only concerned with the arrival time for a given flight.Some of the input to the model could be less accurate (e.g.predicted arrival runway) however it was unclear how much this would impact the results.The comparison of the results across years does not indicate a systemic degradation of the model performance.The results showed an improvement with the 2020 data for the Pct predictions within 30s and MAPE metrics, and relatively small change in the MdAPE.We determined that no further exploration of the pandemic effects on the model were needed, and no changes were made in our procedure to accommodate for pandemic effects. +B. ML Model Algorithm and Loss Function ComparisonA potential point of contention when developing ML models is how to deal with outliers.The choice of loss function can affect how the model learns from outliers.More specifically, mean squared error (MSE) applies much larger importance to outliers compared to weighting them linearly as when using mean absolute error (MAE) loss function.The Huber loss aims to offer the advantages of both the MSE and MAE loss in one piecewise function in which MAE is used for values above a threshold and MSE for values below that threshold [18].In the XGBoost implementation of the Huber loss, the Pseudo-Huber loss is actually used, which ensures continuous derivatives [19].To investigate the loss function, we trained several algorithms for airports KDFW and KJFK.Table 7 displays the overall accuracy results on the test set of the ML algorithms; Random Forest with MSE loss, XGBoost with MSE loss, and XGBoost with Huber loss.The results indicate that using XGBoost with Huber loss leads to similar or more accurate results for these metrics in almost all cases.Additionally, the MAPE is reduced for both airports with Huber loss.Figure 7 shows the error distribution for the three algorithms: Random Forest, XGBoost with MSE loss, and XGBoost with Huber loss on the test data.For both airports tested, KDFW and KJFK, the XGBoost with Huber loss model (orange) displays a tighter distribution than the other two models.The XGBoost with Huber loss model distribution appears more centered around zero minutes of error, particularly in the KDFW chart.This centering indicates the model's predictions are less biased. +C. Feature Importance: SHAPFeature importance metrics can be a valuable tool to explain the output of a predictive model.This is especially true in the context of aviation.It is essential to operations to understand-why, for example, a certain flight is predicted to land at a different time than expected.If a flight operator or an air traffic controller is to be able to improve their operation in the future, they need to understand what brought about the successes or failures in previous operations.Conventional feature importance metrics generally look at the global importance of features to the model.With SHAP (SHapley Additive exPlanations) [20] we can derive the local importance of features.We can also get better information on global interpretability of feature importance.That is, we can look at a single prediction resulting from our model, and the SHAP algorithm will help us understand how each feature in the model influenced that specific prediction.With the encoded categorical features, the influence is spread across many discrete binary features.If the features are considered individually, the single features are ranked lower in the list.If we separated the encoded features, the variables seconds arrival runway late and seconds departure runway late would be the top two most influential features in the KDFW model.Another way to look at the relative importance of different features is to view their impact on the model output, prediction by prediction.Figure 9 shows the impact of each feature with the positive and negative relationships of features to the model output.Each dot is a Shapley value for a feature and data point in the training model.The x-axis represents the impact on the model for a feature.Points that occur on the same x-axis value are jitterred so that we can see the distribution.If that impact is positive, the feature leads to a higher prediction (a later predicted landing time).If the impact is negative, the feature leads to a lower prediction (an earlier predicted landing time).The color of Figure 9 represents the high (red) or low (blue) feature values for that single data point.The seconds arrival runway late feature had a high level of importance.It appears that seconds arrival runway late and seconds to arrival runway best time had a tempering effect on the prediction.When these features were low (blue), the prediction they affected still had a positive value (positive on the horizontal axis), indicating that enough of the other features combined to cause a later prediction.With seconds departure runway late, however, in general it followed that the higher the feature value, the higher the prediction value.The top ranked categorical feature in this list, is the feature arrival meter fix tbfm HOWDY in third position.This feature is an indicator telling the model whether or not the given flight was assigned to the TBFM arrival meter fix named HOWDY.HOWDY refers to a geographical point approximately 40 miles to the southeast of KDFW.HOWDY is one of many airspace waypoints that is adapted in the TBFM system for use as a metering fix by TBFM.Local feature importance can be assessed for individual predictions as well.Figure 10 shows the flight path of an In this example, the model was provided the feature set for the flight.Figure 11 explains how the features were combined to produce this single prediction.We can see that the features affected this prediction with a different relative importance level than the aggregated set.The TBFM arrival runway for this flight was set to 35C, which increased the value of the prediction.As a separate feature, the flight did not indicate an arrival at runway 17L (a runway in the southerly direction).This feature contributed to further increase the value of the prediction.In this example, representing a single prediction for a single flight, the expected value was 32.025.This "base value" (in gray) shows the prediction if we just took the mean value of all predictions.The model produced a value of 85.542, after incorporating all effects of the features.This means the model predicted that an extra 85 seconds of flying time should be added to the MPP.If the MPP was '2020-04-12T21:55:02Z' for this flight, which was its TBFM STA at the moment the model was called, the prediction would have provided '2020-04-12T21:56:27Z' as its improved landing time.Based on position data available on TFMData SWIM feed, this flight landed at approximately '2020-04-12T22:00:02Z', which indicates the flight added even more to its flight time than the model predicted.In the waterfall chart, we can see the amount that each feature contributes to the prediction from the mean prediction.For example, the feature seconds departure runway late (660) contributes +19.17 to the prediction.The red coloring means positive influence on the prediction.By contrast, seconds to arrival runway best time (558) reduced the output of the model by 25.7.Despite seconds to arrival runway best time being the largest contributor to the shift, and shifting in the negative direction, the output was still higher than the expected value.In this way, we can use the SHAP method to obtain interpretable operational information on how our model's predictions are affected by the input features. +V. Next StepsAs mentioned previously, we plan to extend the preliminary work described here in various ways.We intend to incorporate additional model input data and features.We plan to develop and evaluate models for arrivals at additional airports and with much larger data sets.Finally, we will help others evaluate how to make use of our results by providing a much richer evaluation of the models, both in terms of the quality of their predictions and the importance of various features in determining those predictions.Other extensions to this line of work are also under consideration.We may implement a hybrid mediation-ML approach in which a classification ML model is trained to select from among the physics-based landing time predictions using a custom loss function based on the error resulting from the selected physics-based landing time prediction.Additionally, we might evaluate another operationally-available landing time prediction that is neither ML nor physicsbased.The airline-provided arrival runway time in TFMS data (available after flight departure) may be based on airline operational goals or expectations, which can provide additional context and information for the model.Under certain circumstances it may provide superior predictions, or at least predictions with errors that are largely statistically independent of those of physics-based models.The airline provided prediction values could also be used as input to ML models.Fig. 11Fig. 1 ETA accuracy relative to runway departure time. +Fig. 22Fig. 2 Mean ETA accuracy (final 30 minutes). +Algorithm 1 :1Fig. 4 Mean ETA accuracy (final 3 hours). +Fig. 66Fig. 6 Predicted vs. Actual Time to Landing Scatterplots +Fig. 77Fig. 7 Histogram of Errors per Algorithm/Loss Function. +Fig. 88Fig. 8 Aggregated Feature Importance with Collapsed Categorical Variables (KDFW XGBoost only). +Figure 88Figure 8 shows the global feature importance for the KDFW model.The features are ranked with the top as the most influential on the model prediction.The set of encoded arrival runway features is shown to be the most influential on the predictions for the KDFW model.In this chart, the encoded categorical variables are collapsed into a single feature, by group.Using the SHAP method, we can see that the arrival runway expectation provided by TBFM is more important to the model than the arrival fix provided by TBFM, for example.With the encoded categorical features, the influence is spread across many discrete binary features.If the features are considered individually, the single features are ranked lower in the list.If we separated the encoded features, the variables seconds arrival runway late and seconds departure runway late would be the top two most influential features in the KDFW model.Another way to look at the relative importance of different features is to view their impact on the model output, prediction by prediction.Figure9shows the impact of each feature with the positive and negative relationships of features to the model output.Each dot is a Shapley value for a feature and data point in the training model.The x-axis represents the impact on the model for a feature.Points that occur on the same x-axis value are jitterred so that we can see the distribution.If that impact is positive, the feature leads to a higher prediction (a later predicted landing time).If the impact is negative, the feature leads to a lower prediction (an earlier predicted landing time). +Fig. 99Fig. 9 Aggregated Feature Importance: SHAP Beeswarm (KDFW XGBoost only). +Fig. 10 AAL100610Fig. 10 AAL1006 at the time of model prediction (10 NM range rings).Map data: Google, Landsat/Copernicus +Fig. 11 Single11Fig. 11 Single Prediction SHAP Waterfall Plot (KDFW XGBoost only). +Table 1 Trajectory Based Runway Arrival Estimate Descriptions1SWIM Service SystemFieldAliasDescriptionTFMS ETA is based on the TFMS model of theTFMDataTFMS ETA@timeValue TFMS ETAflight's route and is updated as the flight progresses.It is typically limited to one update per minute.TBFM ETA assumes there are no restrictions posedTBFM-MISTBFMeta_rwyTBFM ETAon the aircraft or airspace, so there are no adjust-ments made to the time to apply spacing betweenflights.Unlike the TBFM ETA, the TBFM STA accounts forTBFM-MISTBFMsta_rwyTBFM STAother flights, constraints, and airspace configuration, and adjusts the arrival time for spacing and otherimpacting factors. +Table 2 Input to real-time ML model service2NameTypeNotetimestamptimestamp time at which prediction is generatedmediated landing time predictiontimestamp selected from TBFM STA, TBFM ETA, and TFMS ETAfirst airborne surveillance timestamp timestamp from TFMS or TBFMactual take-off timetimestamp impute with first airborne surveillance timestampactual departure gate push-back timetimestamp impute with actual take-off timeairline-scheduled gate push-back timetimestamp impute with actualairline-scheduled take-off timetimestamp impute with actualairline-scheduled landing timetimestamp impute with mediated landing time predictionTBFM arrival stream classstringimpute with one-hot encodingTBFM arrival meter fixstringimpute with one-hot encodingTBFM arrival runwaystringimpute with one-hot encodingmajor air carrierstringimpute with one-hot encodingaircraft typestringimpute with one-hot encodingbold face indicates required inputs, all other input may be null +Table 3 Features provided to ML models at KDFW and KJFK3NameEncoding Number at KDFW Number at KJFKseconds to arrival runway best timenone11seconds arrival runway latenone11seconds departure stand latenone11seconds since departure runway actual timenone11seconds since time first trackednone11seconds departure runway latenone11arrival stream class tbfmone-hot4116arrival meter fix tbfmone-hot129arrival runway tbfmone-hot148major carrierone-hot33aircraft typeone-hot44predicted arrival runwayone-hot148total number of features9454 +Table 4 Number of samples in data sets by airport4NameKDFWKJFKKCLT KEWRKIAH KDALTraining 111, 005 99, 237 77, 364 58, 843 56, 537 77, 387Test25, 508 18, 083 17, 238 13, 248 12, 874 14, 906 +Table 5 Summary of prediction quality in test data sets5MetricModelKDFWKJFK KCLT KEWR KIAH KDALMAPEMPP ML: XGBoost9.7% 20.1% 18.7% 9.3% 25.1% 25.0% 14.7% 32.3% 14.1% 17.6% 30.9% 15.7%MdAPEMPP ML: XGBoost3.5% 3.0%4.1% 2.9%5.2% 4.9%3.4% 2.5%4.9% 3.2%3.1% 3.0%Pct predictions within 30sMPP ML: XGBoost 22.3% 19.2%6.4% 16.5%14.7% 11.1% 22.4% +Table 6 Prediction quality change in KDFW test data sets: 2019 vs 20206Metric20192020ChangeMAPE15.6%9.3%-6.3MdAPE2.8%2.9%+0.1Pct predictions within 30s 21.3% 22.3%+1.0bold face indicates the year with the better performing value +Table 7 Summary of prediction quality in test data sets ML algorithms and loss functions7MAPEMdAPE% within 30sModelKDFWKJFK KDFW KJFK KDFW KJFKRandom Forest12.6% 35.7%3.1% 3.0%19.9% 9.6%XGBoost MSE12.7% 38.3%3.2% 2.9%19.4% 9.5%XGBoost Huber9.3% 25.1%2.9% 2.9% 22.3% 9.8%bold face indicates the model with the better performing value + + + + +AcknowledgmentsThe authors gratefully acknowledge funding for this work from the National Aeronautics and Space Administration (NASA) as part of the Airspace Technology Demonstration 2 (ATD-2) project. + + + + + + + + + National Aeronautics and Space Administration (NASA) + + AGing + + + SEngelland + + + ACapps + + + MEshow + + + YJung + + + SSharma + + + ETalebi + + + MDowns + + + CFreedman + + + TNgo + + 10.1007/978-1-349-94186-5_816 + NASA/TM-2018- 219767 + + + + The Grants Register 2018 + + Palgrave Macmillan UK + 2018 + 2 + + + + National Aeronautics and Space Administration + + + Tech. Rep. + Ging, A., Engelland, S., Capps, A., Eshow, M., Jung, Y., Sharma, S., Talebi, E., Downs, M., Freedman, C., Ngo, T., et al., "Airspace Technology Demonstration 2 (ATD-2) Technology Description Document (TDD)," Tech. Rep. NASA/TM-2018- 219767, National Aeronautics and Space Administration, 2018. URL https://ntrs.nasa.gov/citations/20180002424. + + + + + Practice for Application of Federal Aviation Administration (FAA) Federal Aviation Regulations Part 21 Requirements to Unmanned Aircraft Systems (UAS) + 10.1520/f2505-07 + + + + JO 7110.65Y -Air Traffic Control," online + + ASTM International + 2019 + + + Federal Aviation Administration + Federal Aviation Administration, "JO 7110.65Y -Air Traffic Control," online, 2019. URL https://www.faa.gov/ regulations_policies/orders_notices/index.cfm/go/document.information/documentID/1036234. + + + + + Using Synchronized Trajectory Data to Improve Airspace Demand Predictions + + AliciaFernandes + + + DanWesely + + + BillHoltzman + + + DougSweet + + + NoureddinGhazavi + + 10.1109/icns50378.2020.9222909 + + + + 2020 Integrated Communications Navigation and Surveillance Conference (ICNS) + Herndon, VA, USA + + IEEE + 2020 + + + + Fernandes, A., Wesely, D., Holtzman, B., Sweet, D., and Ghazavi, N., "Using Synchronized Trajectory Data to Improve Airspace Demand Predictions," Integrated Communications Navigation and Surveillance Conference (ICNS), Herndon, VA, USA, 2020, pp. 3C2-1-3C2-15. https://doi.org/10.1109/ICNS50378.2020.9222909. + + + + + A Tree-Based Ensemble Method for the Prediction and Uncertainty Quantification of Aircraft Landing Times + + YGlina + + + RKJordan + + + MAIshutkina + + + + Proceedings of American Meteorological Society 10th Conference on Artificial Intelligence Applications to Environmental Science + American Meteorological Society 10th Conference on Artificial Intelligence Applications to Environmental ScienceNew Orleans, LA, USA + + 2012 + + + Glina, Y., Jordan, R. K., and Ishutkina, M. A., "A Tree-Based Ensemble Method for the Prediction and Uncertainty Quantification of Aircraft Landing Times," Proceedings of American Meteorological Society 10th Conference on Artificial Intelligence Applications to Environmental Science, New Orleans, LA, USA, 2012. + + + + + Predicting flight arrival times with a multistage model + + GTakács + + 10.1109/bigdata.2014.7004435 + + + + 2014 IEEE International Conference on Big Data (Big Data) + Washington, DC, USA + + IEEE + 2014 + + + Takács, G., "Predicting flight arrival times with a multistage model," Proceedings of IEEE International Conference on Big Data (Big Data), Washington, DC, USA, 2014. https://doi.org/10.1109/BigData.2014.7004435. + + + + + Data-driven aircraft estimated time of arrival prediction + + CSKern + + + IPDe Medeiros + + + TYoneyama + + + + Proceedings of Annual IEEE Systems Conference (SysCon) + Annual IEEE Systems Conference (SysCon)Vancouver, BC, Canada + + 2015 + + + Kern, C. S., de Medeiros, I. P., and Yoneyama, T., "Data-driven aircraft estimated time of arrival prediction," Proceedings of Annual IEEE Systems Conference (SysCon), Vancouver, BC, Canada, 2015. + + + + + Analysis of short-term forecasting for flight arrival time + + MyungSKim + + 10.1016/j.jairtraman.2015.12.002 + + + + Journal of Air Transport Management + Journal of Air Transport Management + 0969-6997 + + 52 + + 2016 + Elsevier BV + + + Kim, M. S., "Analysis of short-term forecasting for flight arrival time," Journal of Air Transport Management, Vol. 52, 2016, pp. 35-41. https://doi.org/10.1016/j.jairtraman.2015.12.002. + + + + + Prescriptive analytics in airline operations: Arrival time prediction and cost index optimization for short-haul flights + + AnnaAchenbach + 0000-0003-0851-1641 + + + StefanSpinler + + 10.1016/j.orp.2018.08.004 + + + + Operations Research Perspectives + Operations Research Perspectives + 2214-7160 + + 5 + + 2018 + Elsevier BV + + + Achenbach, A., and Spinler, S., "Prescriptive analytics in airline operations: Arrival time prediction and cost index optimization for short-haul flights," Operations Research Perspective, Vol. 5, 2018, pp. 265-279. https://doi.org/10.1016/j.orp.2018.08.004. + + + + + Predicting Aircraft Landing Time in Extended-TMA Using Machine Learning Methods + + IDhief + + + ZWang + + + MLiang + + + SAlam + + + MSchultz + + + DDelahaye + + + + Proceedings of 9th International Conference for Research in Air Transportation (ICRAT) + 9th International Conference for Research in Air Transportation (ICRAT)Tampa, FL, USA + + 2020 + + + Dhief, I., Wang, Z., Liang, M., Alam, S., Schultz, M., and Delahaye, D., "Predicting Aircraft Landing Time in Extended-TMA Using Machine Learning Methods," Proceedings of 9th International Conference for Research in Air Transportation (ICRAT), Tampa, FL, USA, 2020. + + + + + Predicting Estimated Time of Arrival for Commercial Flights + + SametAyhan + + + PabloCostas + + + HananSamet + + 10.1145/3219819.3219874 + + + Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining + the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data MiningLondon, United Kingdom + + ACM + 2018 + + + Ayhan, S., Costas, P., and Samet, H., "Predicting estimated time of arrival for commercial flights," Proceedings of KDD, London, United Kingdom, 2018. + + + + + Lessons Learned in the Application of Machine Learning Techniques to Air Traffic Management + + BanavarSridhar + + + GanoBrotoChatterji + + + AntonyDEvans + + 10.2514/6.2020-2882 + + + AIAA AVIATION 2020 FORUM + + American Institute of Aeronautics and Astronautics + 2020 + + + Sridhar, B., Chatterji, G. B., and Evans, A. D., "Lessons Learned in the Application of Machine Learning Techniques to Air Traffic Management," Proceedings of the AIAA AVIATION Forum, 2020. + + + + + Practice for Application of Federal Aviation Administration (FAA) Federal Aviation Regulations Part 21 Requirements to Unmanned Aircraft Systems (UAS) + 10.1520/f2505 + + + + JO 7360.1E -Aircraft Type Designators," online + + ASTM International + 2019 + + + Federal Aviation Administration + Federal Aviation Administration, "JO 7360.1E -Aircraft Type Designators," online, 2019. URL https://www.faa.gov/regulations_ policies/orders_notices/index.cfm/go/document.information/documentID/1036757. + + + + + Structure of situational tasks in the air traffic controllers� professional training + + KaterynaSurkova + 0000-0002-1388-7611 + + + MarynaLomakina + + 10.33251/2522-1477-2020-8-124-129 + + + + Scientific Bulletin of Flight Academy. Section: Pedagogical Sciences + SBFA + 2522-1477 + + 8 + + 2020 + Flight Academy of National Aviation University + + + Federal Aviation Administration, "National Flight Data Center 28 Day NASR Subscription," online, 2020. URL https: //www.faa.gov/air_traffic/flight_info/aeronav/aero_data/NASR_Subscription/. + + + + + National Aeronautics and Space Administration (NASA) + + SMGorman + + 10.1007/978-1-349-95810-8_847 + + + + The Grants Register 2019 + + Palgrave Macmillan UK + 2019 + + + + Gorman, S. M., "Fuser and Fuser in the Cloud," Tech. Rep. ARC-E-DAA-TN72700, National Aeronautics and Space Administration, 2019. URL https://ntrs.nasa.gov/citations/20190030496. + + + + + + RKille + + Summary of the July 1999 Internal Departure Delay Investigation for ZDV Along with New Departure Delay Data From ZOB, ZID, ZFW and ZDV + + 2004 + + + unpublished internal presentation + Kille, R., "Summary of the July 1999 Internal Departure Delay Investigation for ZDV Along with New Departure Delay Data From ZOB, ZID, ZFW and ZDV," unpublished internal presentation, 2004. + + + + + Predicting Arrival and Departure Runway Assignments with Machine Learning + + AndrewChurchill + + + WilliamJCoupe + + + YoonCJung + + 10.2514/6.2021-2400 + + + AIAA AVIATION 2021 FORUM + Washington, DC, USA + + American Institute of Aeronautics and Astronautics + 2021 + + + Churchill, A. M., and Coupe, W. J., "Predicting Arrival and Departure Assignments with Machine Learning," Submitted to AIAA AVIATION Forum, Washington, DC, USA, 2021. + + + + + Scikit-learn: Machine Learning in Python + + FPedregosa + + + GVaroquaux + + + AGramfort + + + VMichel + + + BThirion + + + OGrisel + + + MBlondel + + + PPrettenhofer + + + RWeiss + + + VDubourg + + + JVanderplas + + + APassos + + + DCournapeau + + + MBrucher + + + MPerrot + + + EDuchesnay + + + + Journal of Machine Learning Research + + 12 + + 2011 + + + Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V., Thirion, B., Grisel, O., Blondel, M., Prettenhofer, P., Weiss, R., Dubourg, V., Vanderplas, J., Passos, A., Cournapeau, D., Brucher, M., Perrot, M., and Duchesnay, E., "Scikit-learn: Machine Learning in Python," Journal of Machine Learning Research, Vol. 12, 2011, pp. 2825-2830. + + + + + The Elements of Statistical Learning + + TrevorHastie + + + RobertTibshirani + + + JeromeFriedman + + 10.1007/978-0-387-84858-7 + + 2009 + Springer New York + New York + + + Hastie, T., Tibshirani, R., and Friedman, J., The Elements of Statistical Learning, International series of monographs on physics, Springer-Verlag New York, 2009. + + + + + Table 9: XGBoost parameters optimised by CFAEE-SCA. + + XgboostDevelopers + + 10.7717/peerj-cs.956/table-9 + + + 2021 + PeerJ + + + XGBoost Developers, "XGBoost Parameters," https://xgboost.readthedocs.io/en/latest/parameter.html#learning-task- parameters, 2021. + + + + + A Unified Approach to Interpreting Model Predictions + + SLundberg + + + SLee + + + + 31st Conference on Neural Information Processing Systems + + 2017 + + + Lundberg, S., and Lee, S., "A Unified Approach to Interpreting Model Predictions," 31st Conference on Neural Information Processing Systems, 2017. + + + + + + diff --git a/file780.txt b/file780.txt new file mode 100644 index 0000000000000000000000000000000000000000..d36a3dd1f8c442678c51d72cd2b58f07e6814c2d --- /dev/null +++ b/file780.txt @@ -0,0 +1,3650 @@ + + + + +The NASA STI program operates under the auspices of the Agency Chief Information Officer.It collects, organizes, provides for archiving, and disseminates NASA's STI.The NASA STI program provides access to the NASA Aeronautics and Space Database and its public interface, the NASA Technical Reports Server, thus providing one of the largest collections of aeronautical and space science STI in the world.Results are published in both non-NASA channels and by NASA in the NASA STI Report Series, which includes the following report types: TECHNICAL PUBLICATION.Reports of completed research or a major significant phase of research that present the results of NASA Programs and include extensive data or theoretical analysis.Includes compilations of significant scientific and technical data and information deemed to be of continuing reference value.NASA counterpart of peer-reviewed formal professional papers but has less stringent limitations on manuscript length and extent of graphic presentations. +TECHNICALMEMORANDUM.Scientific and technical findings that are preliminary or of specialized interest, e.g., quick release reports, working papers, and bibliographies that contain minimal annotation.Does +Project SummaryThe purpose of this project is to create performance data for twelve Unmanned Aerial Systems (UAS) aircraft that can be used by many aviation models.The performance data are presented in two formats: the Base of Aircraft Data (BADA) format specified by EUROCONTROL, and the Multi Aircraft Control System (MACS) format specified by NASA.During the execution of the project, simulations were conducted using the Kinematic Trajectory Generator (KTG) for the BADA files, and the MACS software for the MACS files.Simulation output from KTG and MACS were examined and validated by the UAS manufacturers.Nine of the twelve UAS aircraft were validated using this process, although some discrepancies were found in the trajectory generators and are documented in this report.Three of the twelve UAS aircraft-two rotorcraft and one hybrid UAS-require different trajectory generators and will need to be validated at some future point.In addition to the twelve BADA and MACS formatted performance files, the project also conducted simulations using the communication, navigation and surveillance (CNS) capabilities of the UAS aircraft.CNS equipage files provided by the UAS manufacturers were used to configure and conduct the experiments using the Airspace Concept Evaluation System (ACES) with KTG.Finally, operational requirements and limitations of all twelve UAS aircraft are documented by the project.As UAS aircraft have some unique operating requirements-for example, some aircraft can be launched by a catapult while others cannot fly when the wind speed exceeds thirty knots-documentation of these limitations allows researchers to determine whether the weather conditions and availability of infrastructure limit or prohibit the conduct of UAS missions.The value to the aviation community of the work generated by this project is enormous.UAS aircraft perform very differently than piloted aircraft.UAS aircraft have vastly different cruise speeds, operating range, altitude ceilings, and departure and approach speeds than equivalent piloted aircraft such that finding a match between piloted aircraft performance and a UAS aircraft is impractical.Because the BADA and MACS files created by the project are specific to UAS aircraft, aviation researchers can use these UAS performance files to correctly experiment with UAS aircraft in the National Airspace System using virtually any standard aviation simulation tool. +List of Figures +List of Tables +IntroductionThe purpose of this project was to create performance data for twelve Unmanned Aerial Systems (UAS) aircraft in two formats usable by standard aviation models: the Base of Aircraft Data (BADA) that has been specified by EUROCONTROL [1], and the Multi Aircraft Control System (MACS) that has been specified by NASA [2].In addition, simulations were conducted to evaluate the communication, navigation and surveillance (CNS) capabilities of the UAS aircraft using the Airspace Concept Evaluation System (ACES).This report presents the industry data acquired for twelve UAS aircraft, the BADA and MACS files that were produced for these aircraft and tests to verify the data files.The twelve UAS aircraft are Shadow B, Global Hawk, Orbiter, Aerosonde, Predator A, Predator B, Gray Eagle, Predator C, Hunter, Cargo UAS, Fire Scout and NEO S-300 Mk II VTOL.The tests were able to identify and correct errors in the BADA data.Data for the twelve aircraft analyzed in this project were provided by AAI and General Atomics (GA).A summary of modeling, production and verification of the BADA and MACS files for these aircraft is shown in Table 1.Results from ACES simulations to evaluate CNS capabilities of the aircraft are also presented in this report.Manufacturer data for eight aircraft were provided by AAI: Shadow B (RQ7B), Aerosonde, Orbiter, Cargo UAS, NEO S-300 Mk II VTOL, Hunter UAS (MQ-9B), Global Hawk (RQ4A) and Fire Scout.Important specifications and basic attributes of these aircraft are shown in Table 2, Table 3, Table 4, Table 5, Table 6, Table 7, Table 8 andTable 9. +Industry Data PresentationIndustry data for only Shadow B are presented here as a sample and for brevity (Table 14).The data for all twelve UAS aircraft are presented in Appendix A. +BADA File PresentationResearch and development activities in the Air Traffic Management (ATM) and the Air Traffic Control (ATC) systems require accurate information on aircraft performance, expressed via an Aircraft Performance Model (APM).While the primary role of APMs is to provide aircraft performance data to ATM/ATC simulation tools, APMs should also be capable of computing the geometric, kinematic and kinetic aspects of an aircraft in flight.Furthermore, these performance models should also be applicable in all phases of flight and be available for a wide set of aircraft.1 Currently, APMs do not exist for UAS, and the task of developing them is complicated due to the significant heterogeneity in UAS configuration and operation.In this project, APMs were developed for 12 UAS aircraft and expressed in two formats: the Base of Aircraft Data (BADA) and the Multi-Aircraft Control System (MACS).The resulting UAS APMs, the assumptions used in their generation, and the limitations identified along the way are described in the remainder of this report.BADA is an APM developed and maintained by EUROCONTROL [1].BADA provides a set of ASCII files containing performance and operating procedure coefficients for approximately 300 different aircraft in all phases of flight.The coefficients include those used to calculate thrust, drag and fuel flow and those used to specify nominal cruise, climb and descent speeds.BADA is based on a kinetic approach to aircraft performance modeling, which models aircraft forces.The intended use of BADA is trajectory simulation and prediction in ATM research and development and strategic planning in ground ATM operations.Currently, several air traffic modeling and simulation tools such as ACES, FACET etc., use BADA for trajectory simulation.Four Base of Aircraft Data (BADA) files were generated for each UAS aircraft, consisting of stall speeds during different phases of flight, ascent and descent rates, fuel flow rate, empty and fuel masses, and aircraft speeds at different altitudes during the flight.These four files are:  Operational Performance File (.OPF): contains performance parameters for a specific aircraft type including drag and thrust coefficients  Airlines Procedures File (.APF): contains speed procedure parameters for a specific aircraft type  Performance Table File (.PTF): contains summary performance tables of true airspeed, climb/descent rate and fuel consumption at various flight levels for a specific aircraft type  Descent file (.DCT): contains descent rate and fuel consumption rate during descent.This file represents data in the .PTF file in a different format.For each aircraft, the .APF, .OPF and .PTF files were compiled by the Purdue team, whereas the .DCT file was compiled by IAI using the data in the .PTF file.The four BADA files for Shadow B are shown in Figure 1, Figure 2, Figure 3 and Figure 4.It should be noted that only those columns in the .DCT file that are relevant to simulating the flight using the Kinematic Trajectory Generator (KTG) were compiled.KTG is a flight trajectory simulation tool developed at IAI [3], which was used to simulate the UAS flights and validate the BADA files.Therefore, the .DCT file contains less data than the version provided by the EUROCONTROL.Only the BADA files for Shadow B (Figure 1, Figure 2, Figure 3 and Figure 4), and their corrected versions and the reasons for the corrections later on in this report, are presented here.The files for all the UAS aircraft (including Shadow B) were provided to NASA on a DVD, along with the option to download them from an ftp site: ftp://ftp.i-a-i.com.While it is safe to assume that the fuel flow equations and the climb/descent procedures provided in BADA can be used for large UAS, the lightweight aircraft may not perform as intended if modeled using the same formulas.Once the modeling is completed, simulation tools utilize the tables and parameters in the .PTF and .OPF files for each aircraft to describe its trimmed motion or transition at any specified altitudes. +Challenges with BADA File Format for UAS: Deficiencies and LimitationsBADA is primarily used for manned aircraft and its capability to model rotorcrafts, hybrids or electric aircraft is currently unknown.Current BADA format does not have provisions for simulating rotorcraft and electric engines (both frequently used in the UAS family).Performance characteristics and/or aircraft component types that are missing in BADA, but important for understanding the UAS-NAS integration, can be classified as deficiencies in BADA.These deficiencies are of the following types:  Aircraft type, class and size (e.g., rotorcraft are currently not considered in BADA)  Propulsion type (e.g., BADA currently handles only jets, turboprops and pistons; electric engines are not considered)Performance characteristics that are poorly modeled in BADA (fidelity too low to be used in existing simulations) can be classified as limitations of BADA:  Stall speed buffers that are too limiting  Climb/descent schedules that are often ill-suited for many UAS +DeficienciesAircraft type, class and size: BADA was primarily developed for manned, fixed-wing aircraft, and does not have provisions to include rotorcraft or hybrid aircraft.Additionally, BADA specifies wake categories based on aircraft weight: Small (up to 12,500 lb.), Medium (12,500 to 41,000 lb.), Large (41,000 to 255,000 lb.) and Heavy (more than 255,000 lb.).However, it does not include very-small/light aircraft such as Orbiter or Aerosonde.Consequently, BADA coefficients and procedures are not well-defined for such very-light aircraft.Considering these restrictions, some of the UAS aircraft could not be properly represented in BADA until modifications (revisions to the format) were in place: Rotorcraft: NEO S-300 Mk II VTOL and Fire Scout  Hybrid: Cargo UAS  Very Light Aircraft: Aerosonde and OrbiterThe following BADA fields, in particular, are difficult, or even impossible, to determine for the three aforementioned aircraft types: a) stall speeds, b) cruise, climb and descent speeds, c) rate of climb/descent coefficients, d) thrust coefficients, and e) ground movements. +Propulsion type:In its current format, BADA can accommodate three engine types: Jet, Turboprop or Piston.This prevents the representation of UAS that use electric motors, such as the Orbiter.Introduction of electric engine format into BADA requires changes to the .APF and .PTF files in BADA, particularly the fuel flow of the aircraft, in addition to the performance coefficients in the .OPF file. +LimitationsStall Speed Buffer: As described earlier, aircraft speeds in the .PTF file are currently set to accommodate transport aircraft, these buffer values need to be modified for realistic UAS representation.More specifically, current true airspeed values in the .PTF file have to be at least 1.3 times (1.2 in some cases) the stall speeds at different phases of flight.While this is justified in the case of transport aircraft for reasons of passenger comfort, implementing this in UASs alters their performance.The relationship between stall speeds and speeds in the .PTF file are shown in Table 30.Currently these rules are strictly followed while developing the BADA files for the aircraft in our list since most simulation software have a hard constraint on these conditions before flying an aircraft.Discrepancies resulting from this rule directly affect the performance of certain aircraft. +Ill-suited climb/descent schedules:In BADA, standard airline procedures are defined using speed profiles in different phases of flight.Procedures similar to that need to be defined in order to calculate rate of climb/descent, fuel flow etc., at different flight levels.In the case of commercial jet aircraft, BADA provides methods to calculate speed profiles at different flight levels, as exemplified in Figure 5, where, C Vmin refers to the stall speed buffer (Table 30) and Vd CL represents standard airline climb speed increments as shown in Figure 6.These procedures are also defined in BADA for manned aircraft with turboprops and piston engines (not shown here).Similarly, standard descent procedures are also defined for manned aircraft (not shown here).However, these definitions were not used in the development of BADA files for the UAS aircraft.Climb/descent speeds, rate and fuel flow are directly taken from the output of sizing tools (FLOPS, JSBSim, etc.) with the stall speed buffers being the only added constraint.Also, simulation software such as KTG and FACET do not hard-code these definitions.Considering the vast heterogeneity in design, such standard procedures may be hard to define for UAS aircraft. +MACS File PresentationThe Multi Aircraft Control System (MACS) is a comprehensive research platform used in the Airspace Operations Laboratory (AOL) at NASA Ames Research Center [2].It was developed to increase the overall realism and flexibility of controller-and pilot-in-the loop air traffic simulations [4].There are three functional classes of aerodynamic models in MACS with varying levels of fidelity, viz. the motion predictor class, the 4-DOF model and the 6-DOF model.These aero models use aircraft performance database files as parameters for the models.Currently, 434 aircraft files exist within the MACS database.Addition of new aircraft types for simulation in MACS requires adding database entries for those new aircraft.While MACS allows for simple mappings of aircraft and engines to those already in the database, an entirely new database entry was created for each UAS studied.This is due to the vast differences in size, weight, and flight envelope between UAS and aircraft already in the MACS database.The addition of a new aircraft in the MACS database is accomplished by essentially filling out the aircraft_specific_model_data.datfile.This master file (Figure 8) contains all top level information regarding an aircraft and has provisions to map the required drag model and engine model of the aircraft.Mach numbers for an aircraft.Further, where applicable, it also specifies changes to these coefficients for other flight parameters such as settings of flaps, landing gear and speed brakes. Flight parameters file: This file specifies the flight path in terms of origin and destination airports along with their location and altitudes, the waypoints, different operational speeds (climb, cruise, descent, approach and landing in knots of indicated air speed), cruise altitude, communication and navigational equipage, and flight-specific operational procedures (e.g., self-separation).The aircraft model data file for each UAS aircraft was produced by Purdue, whereas the flight parameters file was compiled by IAI by utilizing the data from the .OPF and .PTF BADA files.It should be noted that the 'AIRFRAME DRAG MODEL' file and 'ENGINE THRUST MODEL' file in Figure 8 are external files that are called to the motion class while executing a particular aircraft.If a particular UAS aircraft is similar to an existing aircraft in MACS, a simple mapping will accomplish this process (Table 15), but for other aircraft new drag model and thrust model have to be created.The three MACS files for Predator B are shown in Figure 8, Figure 9 and Figure 10, respectively.The MACS files for the twelve UAS aircraft were provided to NASA on a DVD, along with the option to download them from an ftp site: ftp://ftp.i-a-i.com.Since this project involves representing UAS data in two formats (BADA and MACS), there is a reasonable need for consistency between a MACS file and a BADA file for the same aircraft.Accordingly, a convention was developed such that a majority of the entries in a MACS file were mapped to specific entries in a BADA .OPF or .APF file as shown in Figure 7.The various attributes that distinguish UAS from traditional fixed-wing manned aircraft also imply difficulties in populating the aircraft_specific_model_data.datfile since some fields are either not applicable or are not available as a result of the UAS configuration. +Challenges with MACS File Format for UASThe various attributes that distinguish UAS from traditional fixed-wing manned aircraft also imply difficulties in populating the aircraft_specific_model_data.datfile since some fields are either not applicable or are not available as a result of the UAS configuration. +Lack of Airframe Drag ModelFor majority of the twelve UAS aircraft studied in this project, detailed airframe drag data was not available due to the propriety nature of the information.Efforts were made to substitute or map drag data from similarly sized aircraft to mitigate this problem.However, this was not possible for all twelve UAS aircraft due to vast differences in size between the smaller UAS and existing aircraft in the MACS database.Consequently, these UAS aircraft were not simulated in MACS: Orbiter, Cargo UAS, Fire Scout and NEO S-300 Mk II VTOL. +No Support for Electric Engines and RotorcraftIn the case of the Orbiter UAS, which is a battery powered fixed-wing aircraft; it was difficult to generate a MACS profile simply because the format only supports jets, turboprops or props.Furthermore, an attempt was made to match a similar engine based on output, but this was unsuccessful since the size of the Orbiter UAS (and its power plant) is much smaller than anything available in the database currently.Similarly, rotorcraft and hybrid engines are also not fully represented in MACS currently.More details regarding the MACS modeling of these aircraft are discussed in later sections. +MethodologyThe flowchart shown in Figure 11 shows the various steps involved in generating BADA and MACS aircraft performance models (APMs) for a UAS.The process includes validation via test in ATM/ATC simulation software, specifically trajectory generators used in ACES and MACS.The first step in the analysis involves collection of required UAS data to estimate its weights and performance.Data for UAS being studied in this project are collected from their respective manufacturers.Next, an aircraft sizing algorithm (FLOPS, DATCOM-JSBSim, etc.) uses the data to estimate weights, aircraft climb, cruise, descent performance, etc.A MATLAB-based tool was developed to generate BADA and MACS files using outputs from the sizing algorithms.In the last step, the complete APM files are examined via use in ACES/MACS for purposes of validation. +Flight Optimization System (FLOPS)The Flight Optimization System (FLOPS) is a multidisciplinary system of programs for conceptual and preliminary design and evaluation of advanced aircraft concepts.It consists of nine primary modules out of which the first five are used in this project: 1) weights, 2) aerodynamics, 3) engine cycle analysis, 4) propulsion data scaling and interpolation, 5) mission performance, 6) takeoff and landing, 7) noise footprint, 8) cost analysis, and 9) program control.The weights module uses statistical/empirical and analytical equations to predict the weight of each item in a group weight statement. +The aerodynamics module uses a modified version of the Empirical Drag EstimationTechnique (EDET) program to provide drag polars for performance calculations.Modifications include smoothing of the drag polars, more accurate Reynolds number calculations, and the inclusion of other techniques for skin friction calculations.Alternatively, drag polars can also be input, but so far we have been using the FLOPS calculated values until we get it in the same ballpark as the manufacturer provided values.FLOPS engine cycle analysis module provides the capability to internally generate an engine deck consisting of thrust and fuel flow data at a variety of Mach-altitude conditions.Engine cycle definition decks are provided for turbojets, turboprops, mixed flow turbofans, separate flow turbofans, and turbine bypass engines.Piston engine and propeller performance data can also be generated.Since very detailed engine decks were not available from manufactures due to security reasons, FLOPS' internal decks were used, while information such as baseline engine thrust, fuel flow, etc. were obtained from the manufacturer.The propulsion data scaling and interpolation module uses an engine deck that has been input or one that has been generated by the engine cycle analysis module, fills in any missing data, and uses linear or nonlinear scaling laws to scale the engine data to the desired thrust.It then provides any propulsion data requested by the mission performance module or the takeoff and landing module.The mission performance module uses the calculated weights, aerodynamics, and propulsion system data to calculate performance.Based on energy considerations, optimum climb profiles may be flown to start of cruise conditions.The cruise segments may be flown at the optimum altitude and/or Mach number for maximum range or endurance or to minimize NOx emissions, at the long range cruise Mach number, or at a constant lift coefficient.Descent may be flown at the optimum lift-drag ratio.FLOPS engine thrust output is validated by comparing the results to the manufacturer-provided thrust data.If the values differ by more than , the FLOPS engine cycle module is re-run by altering coefficients within the module (such as overall pressure ratio, bypass ratio for turbofans, and turbine entry temperature) until the difference is less than .In this project, the program is used in such a way that an optimal weight of the aircraft is estimated for a given range or endurance, thrust (engine parameters), geometric features etc. FLOPS results are then compared to manufacturer provided data.Cruise, climb and descent phases of flight where scheduled according to the following procedures after consulting with the manufacturers, a) Cruise: fixed Mach number at input maximum altitude or cruise ceiling, b) Climb: minimum fuel-to-distance profile, and c) Descent: descent at optimum lift-drag ratio.FLOPS can handle only fixed-wing aircraft of the following engine types: Jet, Turboprop and Piston.FLOPS is primarily designed for modeling manned aircraft and hence, it has limitations in modeling very light aircraft such as the Aerosonde and Shadow B. In this project, the following seven aircraft are modeled using FLOPS: Shadow B, Global Hawk, Predator A, Predator B, Gray Eagle, Avenger and Hunter UAS. +Constructing BADA models of UAS from Public Data/ Photos employing 3D Modeling, JSBSim, and DATCOMIn this analysis, publically available data and photographs of UAS are converted into detailed models.These models are used to measure the static performance of the UAS in order to create BADA models.This approach is appropriate for UAS when not enough data is available to characterize the aircraft performance for BADA.The following aircraft in the UAS set are modeled using the JSBSim/DATCOM integrated model: Aerosonde, Orbiter and Cargo UAS. +DATCOM Aerodynamics ModelIn 1976, the McDonnell Douglas Corporation was commissioned to convert the USAF Stability and Control DATCOM to an automated program.Implementation of the Digital DATCOM was completed in 1978.Since that time, it has undergone various updates and is still widely used in industry and academia today [5].The Digital DATCOM has several limitations.It assumes the fuselage is a body of revolution, so external fuel tanks and other large protrusions from the fuselage cannot be accounted for.There is also no method for a twin vertical tail, so this must be approximated as a single vertical tail.In addition, there is no method to compute the effect of rudder control so this must be estimated.The underlying methods of the DATCOM are based on charts and equations used in aircraft design.This technique of aerodynamics modeling is faster than a computational fluid dynamics based approach, but is also less accurate.Previously, our lab has conducted wind tunnel testing of a small UAS in order to validate the Digital DATCOM for application to this domain.The Digital DATCOM reads a data file describing the aircraft geometry.It then produces tables for the predicted aerodynamics.The lift, drag, and side force coefficients are available in the user manual.The DATCOM output is in the stability frame (rotated from the aircraft body frame by the angle of attack). +Propulsion ModelsUAS propulsion systems are modeled using existing methods within the JSBSim library [6].JSBSim provides models for piston, turbine, and turboprop engines and electric motors.The turbine engine produces its own thrust; however, the turboprop and electric motor must use a propeller to convert the engine power to thrust. +MethodologyThe main inputs required for analyzing each aircraft are the mass properties, propulsion characteristics, flight control, and aerodynamic properties.Several programs are used to provide inputs for JSBSim simulation.The aircraft visual model is generated by Blender [7], the aerodynamic properties are generated from DATCOM, and the engine and propeller files are generated from the Aeromatic website [8].Some other input data includes moments of inertia, which were calculated given the aircraft's configuration data and aerodynamic type, and stability-related characteristics, such as center of gravity and aerodynamic center, which were estimated from the blender model.The interactions between the different elements of this process are shown in +Gathering Publically Available Data/ PhotographsInformation on UAV performance specifications, dimensions, propulsion systems, aerodynamics, and mass properties can be found on the internet.Often this information is published as marketing information.Also, various photographs can be obtained on the internet.In addition to the general shape of the aircraft, these photographs provide information on the position of the control surfaces, landing gear, etc. that is typically not published. +7.2.2.2Constructing 3D Models in Blender Due to the sensitive nature of UAS dimension information; all of the dimensions of the aircraft required for input into the DATCOM aerodynamics program are not publicly available.To obtain reasonable estimates of this information, 3D models were constructed in the Blender 3D modeling program.If orthographic drawings are available, these drawings are employed to construct the 3D model as shown in Figure 13.The shape of the aircraft is modified until it agrees with all of the orthographic projection views provided. +Figure 13. Orthographic projection/picture based modelingWhen an orthographic projection drawing is not available, pictures can be utilized.The disadvantage to this method is that it is difficult to correctly account for the perspective distortions.If enough pictures are taken of the same aircraft, it is possible for some algorithms to recover the orthographic projection of the image; however, this approach was not utilized in this analysis.An example of employing a picture to aid in 3D modeling is shown in Figure 14. +7.2.2.3Measuring 3D Model to Create DATCOM Input File Once a 3D model has been created in blender it can be easily used to measure quantities required for the DATCOM aerodynamics input file.For instance, the wing section of the Cargo UAS is being measured (Figure 15). +Test Aircraft in Manual Flight SimulationThe FlightGear flight simulator is used to test the accuracy of each aircraft system [9].FlightGear takes the main JSBSim file for each aircraft as input (Figure 16).The JSBSim file includes file paths for the visual model of each aircraft from the AC extension file from Blender, the aerodynamic flight characteristics from DATCOM, engine and propeller information, flight control details, and ground reaction details.Each path contributes to the entire function of the model in the flight simulator and is then tested for each of the following:  The aircraft is observed on the runway to test accuracy of ground reactions. The simulation is initialized with the aircraft in free fall to test the aircraft glide characteristics.If necessary, stability augmentation systems are added at this stage to make manual flight easier. When applicable, the aircraft are tested for smooth and controlled takeoff. Control surfaces are checked for proper function. +JSBSim Trimming and Performance Table GenerationOnce the flight testing is completed, the model is trimmed at various conditions using the JSBSim trim program to generate the performance table.For each flight altitude, aircraft's weights are varied by three different fuel levels, low level, nominal level, and hi level.In the original BADA performance table, the corresponding aircraft's true airspeed for each flight level is based on the aircraft's flight procedure.However, such information is not available for most of the UAVs.True airspeed is instead chosen within the operational speed range provided by the manufacturers.Inputs of flight level and true airspeed are then fed into JSBSim as well as aircraft's weight.For cruise flight, flight path angle is set to zero and then JSBSim provides the fuel flow rate.However for climb and descent flights, simulation is conducted with increments of the flight path angle.The maximum flight path angle that ensures the aircraft's trim is then used in the following equation to calculate the rate of climb. +Modeling of Electric UAS Aircraft: OrbiterThe fundamental idea here is that fuel, fuel consumption, and fuel capacity of any sort can be decomposed into raw energy units (kW-h, BTU, etc.) as a middle ground.Using dimensional analysis, the energy content of an electrical battery is converted into kW-h and that capacity is then normalized by the energy content of a specified fossil fuel.The end result is a volume of fossil fuel (in liters) that contains the same amount of energy as the original electrical battery as shown in Eq. ( 1), where, B v is the published battery voltage (in volts), C v is the published battery capacity (in A-h), and E q is the energy content of the fossil fuel (in KW-h/L).1000 q v v eq E C B C  (1)However, this solution is not complete without a way to represent the rate of energy consumption.As with the energy capacity problem, the electric engine power consumption is converted to raw energy units (J/s, BTU/s, etc.), which is often specified by the manufacturer.This energy consumption rate is normalized by the specific energy content of a fossil fuel (J/kg) such that the flow rate is in terms of weight.The result of the conversion is a weight-based fuel flow value (in kg/min) that represents the same amount of energy flow as the electrical systems onboard the aircraft.This is shown in Eq. ( 2), where P v is rated electric engine power (in KW), E q is again the energy content of the fossil fuel (in (KW-h)/L), and ρ is the density of the fossil fuel.q v q v E P E P f 60 10 6 . 3 60000 6      (2)This dimensional analysis method, while convenient and simple, is not without its drawbacks.Weight is an important measure in aircraft mission performance analysis and this method does not account for the reality that a battery does not change in weight when it is being charged or drained.As a result, simulations that implement this solution can result in the aircraft losing more weight than possible due to "fuel" consumption. +Rotorcraft Modeling and Analysis: RPATRotorcraft performance was estimated using Rotorcraft Performance Analysis Tool, RPAT, developed at Cornell.This Microsoft Excel based performance analysis tool is capable of calculating hover performance, maximum gross weight, parasite and profile drag, and forward flight power consumption for given rotorcraft input parameters.At Purdue, the RPAT basic program went through serious modification to output the entire .PTF table for rotorcrafts, which includes the flight speed, fuel flow rates for different phases of flight, climb and descent rates for three different weight settings.The modified RPAT consists of several modules viz.Aircraft Specifications, Hover Performance, Parasite Drag Estimate, Profile Drag Estimate, Forward Flight Power Analysis, Forward Flight Summary and the BADA format .PTF table.As mentioned before, BADA equations are not suitable for rotorcrafts.The calculation follows preliminary design process and performance analysis based on rotorcraft energy equations [10].Results from the modified version of the RPAT were compared to the existing full scale helicopter performance data for verification.The flight profile assumes that the rotorcrafts climb at the best rate of climb and cruise at the best range speed.The descent profile is adjusted to match the performance characteristics given by the manufacturer.In the Aircraft Specifications module, the basic sizes of components and performance parameters are estimated using statistical/empirical equations based on 7 initial inputs: aircraft gross weight, range, maximum forward flight speed, number of blades in rotors, number of engines and specific fuel consumption [11].The estimated values are only used when specific data are not available.In the Hover Performance module, with complete aircraft specifications from aircraft specifications, 'out of ground effect' rotorcraft hover performance is calculated.In the modified RPAT, rotorcrafts are assumed to be 'out of ground'.The essence of hover performance calculation is to analyze distribution of required power to main rotor and tail rotor using iterations.For hover performance, power available at varying altitude is also calculated.From Parasite Drag Estimation and Profile Drag Estimation, power required correspond the drag components for varying altitude and varying forward flight speed are estimated.Parasite drag is estimated using Eq. 3, where D p is the parasite drag, f is equivalent flat plate area, V is forward velocity and is dynamic pressure.The flat plate area of the aircraft can be obtained by drag build up; however, since data was available, given flat plate area were used in both aircraft calculations.2 * V f q f D p    (3)Using parasite drag, atmospheric condition and flight velocity, parasite power can be calculated for the forward flight as given in Eq. 4, where hp p is the parasite power in Horse Power units.1100 550 * 3 V f V D hp p p    (4)Profile power caused by both main rotor and tail rotor is given by Eq. 5, where hp pro is the profile power in Horse Power units, C d is profile drag coefficient, Ω is angular velocity of rotor blades, A b is area of rotor blades, R is rotor radius and μ is rotor tip speed ratio.( ) ( )(5)Rotor disk angle of attack (α) is also calculated using the parasite drag as given in Eq. 6, where GW is gross weight of the aircraft.          p D GW 1 sin  (6)Rotor disk angle of attack calculation assumes that angle of attack is positive for forward flight.The estimated rotor disk angle of attack is then used in forward flight for induced velocity calculation.In the Forward Flight Power Analysis module, previously calculated power components are added to the induced power estimated.With the assumption that rotors are ideal, induced drag is calculated using the same equation used for a fixed wing aircraft (Eq.7), where T is thrust, A is disk area and ρ is the density of fossil fuel or air.(7) Using induced drag calculations, induced power is estimated using Eq. 8, where hp ind is induced power in Horse Power units.(8) By combining estimated power components, power required for forward flight is calculated using Eq. 9, where hp access is the access power.Access power was assumed to be zero for aircraft used in this project.(9) Power required is a function of forward flight velocity and thus can be represented in a graph known as the power curve, shown in Figure 17. +Figure 17. Sample power curveThe power required and available power data are produced for entire range of flight altitudes and for three different weight settings.Using the power required and power available data, cruise, climb and descent performance data are calculated for .PTF.When generating a .PTF table, the rotorcrafts are assumed to be flying at the most efficient flight profile: best rate of climb, maximum range speed at cruise and maximum glide range speed at descent.This results in a flight profile very similar to fixed wing aircrafts, where the rotorcraft does not perform any vertical flight, which is highly unlikely.First, the cruise performances are calculated using best range forward velocity setting.Best range forward velocity will maximize the UAS mission range.Speed is calculated assuming there are no head or tail wind and the engine models are turbine engines.The maximum range speed for cruise is determined at the speed where a line through origin is tangent to the power curve.For climb performance analysis, Eq. 10 is used to calculate the extra power required to climb.When the difference between the power available and power required from the power curve is maximum, the flight profile during climb corresponds to the best rate of climb.Unlike fixed wing aircraft, forward flight speed during best rate of climb is much different from that of cruise or descent for rotorcraft.Furthermore, the differences between the rate of climb for low, nominal and high mass configurations are large, because rotors are the source of both lift and thrust for rotorcrafts.Descent velocity is found at speeds for maximum glide range speed.This velocity is found by determining a point on the power curve where through the origin is tangent to the required power curve, similar to cruise speed.Fuel flow rates during descent are estimated by adjusting the throttle to match the manufacturer determined rate of descent.Using partial power of level flight setting, Eq. 10 is used to calculate the negative climb.In this project, the following aircraft are modeled using RPAT: Fire Scout and NEO S-300 Mk II VTOL.This section documents the sizing of the aircraft chosen for analyses, comparison of the sizing results with data provided by aircraft manufacturers, analysis of BADA and MACS files and the deficiencies or limitations associated with BADA and MACS in representing the aircraft.A summary of the manufacturer prescribed engines and the engine decks actually used in this project is provided in Table 16.High resolution data for the actual engines were not available due to security reasons and therefore, either an alternative deck was used to mirror the actual engine or an engine type within the modeling tool is used to duplicate the original.Mismatches between engines lead to several discrepancies, which are described in detail in the following subsections.If an internal engine cycle is used, FLOPS uses linear or non-linear scaling laws to scale the engine data to the desired thrust.If the maximum thrust at cruise for a particular vehicle is provided by its manufacturer, for example, this value is input to FLOPS before the execution of the program.The desired thrust values are sometimes not achieved due to conflicts in the FLOPS optimization regimes.Since priority is given to sizing the vehicle to the exact weights and configurations, the engine thrust values are sometimes compromised.An exact match between thrust values from data and FLOPS can lead to discrepancies in weights, configurations etc., and vice versa.Mismatches between engine thrust values for a number of aircraft are listed in the subsequent sections.In some cases the transport weight equation coefficients within FLOPS were altered by trial and error until the weights, configuration and engine thrust match the manufacturer data to provide a reasonable vehicle performance output.If a desired thrust value is not provided by the manufacturer, FLOPS chooses a default starting point for sizing, based on the type of the engine in use.Similar procedures were followed in the other sizing tools as well. +Shadow BShadow B is a small-scale, fixed wing aircraft equipped with a piston engine.Data for Shadow B were provided by its manufacturer, AAI.FLOPS was used to model the Shadow B as closely as possible.FLOPS generated the drag polars, fuel flow rates and climb rates for different phases of flight based on primary input data for Shadow B. The MATLAB-based BADA tool developed at Purdue was used to translate FLOPS output to the required BADA files in the format mandated by EUROCONTROL.The current FLOPS model predicts a maximum take-off gross weight of 593 lb., which is higher than the actual Shadow B gross weight of 467 lb., a difference of approximately 20%.Additionally, FLOPS specifies a cruise Mach number of 0.225 while the actual value is 0.197.Table 17 provides a summary of FLOPS sizing results compared to industry (AAI) data.This model, therefore, is not a perfect representation of Shadow B. However, by using FLOPS' General Aviation module and with the help of correlation factors, it is possible to model an aircraft in the same weight category as that of Shadow B. While matching the exact performance values requires further refinement, the present model appears to be a reasonable basis for this refinement.The .PTF file for Shadow B was shown in Figure 4. Shadow B is equipped with a UEL 741AR74-1102 piston engine.Since all or most of the engine performance details were provided, the .PTF file predicted reasonable values for speed, climb/descent rates and fuel flow. +Summary of BADA Deficiencies and LimitationsBADA deficiencies: None BADA limitations: None.Though the BADA climb/descent schedules were not expected to suite an aircraft as small as the Shadow B, the cruise, climb and descent speeds, fuel flow and climb rates matched manufacturer provided data with reasonable accuracy. +Summary of MACS Deficiencies and LimitationsMACS files were generated directly from the BADA outputs.In addition to filling out the aircraft_specific_model_data.datfile, existing drag models and engine thrust models were mapped.The MACS drag model and engine thrust model used for Shadow B are C172 and O-320-H2AD, respectively.The completed aircraft_specific_model_data.datfile for Shadow B is shown in Figure 18. +Global HawkGlobal Hawk is a medium scale, fixed-wing aircraft equipped with a Rolls-Royce turbofan engine.The aircraft cruises at 31000 ft., with a maximum altitude of 65000 ft., and weighs approximately 26700 lb.The BADA model of Global Hawk was developed using data provided by AAI (collected from Northrop Grumman).The FLOPS model of the Global Hawk is generated by using the built-in Transport Aircraft weight equations, engine deck, and aerodynamic data of FLOPS.The size and propulsion system (e.g.jet) of the Global Hawk aircraft make FLOPS a reasonable choice as a sizing tool.A comparison of sizing results from FLOPS and manufacturer provided data is shown in Table 18.For reasonable estimations of the weights and performance of this aircraft using FLOPS, modifications to the FLOPS built-in weight equations were made as would be appropriate for modeling an unmanned aircraft; weight multipliers for furnishings, passenger compartment, and other amenities were set to zero.Avionics and electrical systems weights were increased to reflect the likelihood of the additional instrumentation carried by the Global Hawk to perform its surveillance mission and to be remotely piloted.Additionally, structural weight equation multipliers were calibrated so as to result in an empty weight that closes matches the published Global Hawk empty weight.FLOPS generated the drag polars, fuel flow rates and climb rates for different phases of flight based on primary input data for Global Hawk.These values are used in the BADA model to generate BADA specific coefficients, which are then used to generate performance characteristics found in the .PTF.Due to their resemblance in design to traditional manned aircraft, generating BADA files for Shadow B and Global Hawk is not complicated.Again, most of the performance characteristics available in the .PTF file matches with the manufacturer provided data with reasonable accuracy. +Summary of BADA deficiencies and limitationsBADA deficiencies: None BADA limitations: None +MACSMACS master files were developed by mapping the BADA files.A new drag model was created for Global Hawk and was mapped as an external file.The following MACS engine thrust model was used for Global Hawk: PW_JT8D-07. +OrbiterThe Orbiter is a small UAS only capable of launch by a slingshot system.Notable features of the aircraft include an aft fuselage propeller electric engine, large swept wings with winglets, and no tail.The engine is an HB2815-2000 electric engine with a two-blade propeller.The empty weight of the aircraft is 12 lb.and the gross weight is 16 lb.The fuselage is 42 in.in length and the wingspan is 86.6 in.A comparison of sizing results from FLOPS and manufacturer provided data is shown in Table 19.The images used in constructing 3D models of Orbiter, and the model generated therefrom, are shown in Figure 19 and Figure 20, respectively.The DATCOM-JSBSim flight modeling tool was used to model Orbiter, from which the BADA files are developed.Orbiter is equipped with an electric engine which posed challenges in accurate calculating fuel flow rates.JSBSim cannot produce fuel flow rate of an electric engine in terms of kilogram per minute.In fact, an electric engine uses batteries as a power source and therefore weight does not change over time.To resolve this matter, fuel flow rate was considered as the current usage rate.Since JSBSim provides throttle usage for each trim state, it was converted into current usage rate in terms of ampere per min.These current usage rates were then converted into equivalent fuel usage in order to represent the aircraft in BADA. +Summary of BADA deficiencies and limitationsBADA deficiencies: Engine type missing.Orbiter is an electric engine and therefore, no fuel flow rates could be provided as mandated by BADA.This calls for a provision for electric engines in the BADA format.BADA limitations: BADA model is currently not defined for electric engines and therefore, the Orbiter BADA files were generated directly from the modeling software, ignoring BADA equations. +MACSThe Orbiter MACS model is not completely developed as MACS is not equipped to handle electric air aircraft.Converting the Orbiter current usage rates to fuel usage rates is not sufficient to complete a MACS engine thrust model.A MACS engine thrust model requires the engine pressure ratio, corrected fuel flow rates etc., to represent a gas engine in its entirety.This calls for a provision to add electric engine capabilities into the motion predictor class of MACS.In addition to the engine thrust file, the drag model of the aircraft is also not available to the level of detail that MACS mandates.Therefore, these fields are left empty in the MACS master file.All fields that are not related to the engine model or drag model are completed using available data from the manufacturers and BADA output files. +AerosondeThe Aerosonde is a small UAV designed for collecting weather data.It is powered by a small piston engine.Notable features of the aircraft include an inverted V-tail at the end of a twin boom.It is also a pusher prop with the engine located behind the wing.The aircraft has an empty weight of 48.9 lb.It has a wingspan of 11 ft.A comparison of sizing results from FLOPS and manufacturer provided data is shown in Table 20.The images used in constructing 3D models of Aerosonde, and the model generated therefrom, are shown in Figure 21 and Figure 22, respectively.The DATCOM-JSBSim flight modeling tool was used to model Aerosonde, from which the BADA files are developed.While running JSBSim, the trim condition was not achieved with the engine model provided by the manufacturer.This may be caused by the lack of propulsion or aerodynamic data.To achieve trim, more powerful engine was used in DATCOM and JSBSim.The excessive thrust input resulted in larger maximum flight path angles and eventually larger rates of climb.More accurate propulsion and aerodynamic information will be able to improve the rate of climb accuracy. +Summary of BADA deficiencies and limitationsBADA deficiencies: None BADA limitations: Ill-suited climb/descent schedules overshoot the speed limits of the aircraft in climb and descent, suggesting modifications that may have to be made in BADA to account for procedures pre-defined by the aircraft manufacturers. +MACSMACS performance files were generated by mapping the BADA files.MACS drag model and engine thrust model were custom made for Aerosonde as the MACS database does not have drag models or thrust models capable of representing an aircraft as light as the Aerosonde. +Predator APredator A is a small-scale, fixed-wing aircraft equipped with a Rotax914 four cylinder piston engine.The aircraft cruises at an altitude of 16000 ft., with maximum altitude at 31000 ft. and weighs approximately 2250 lb.The BADA model of Predator A was developed using data provided by General Atomics (GA).FLOPS piston engine deck was generated using engine data provided by the manufacturer.A comparison of sizing results from FLOPS and manufacturer provided data is shown in Table 21.FLOPS generated values for the drag polar, speed schedules, and climb rates and fuel flow were used in the MATLAB-based BADA model to generate BADA specific coefficients.These coefficients are further used to generate the .PTF file for Predator A.The BADA .PTF file generated for Predator A was found to have several discrepancies in comparison to the manufacturer provided data.The cruise, climb and descent TAS were overpredicted by at least 20% in the .PTF, while the fuel flow rates and climb rates were overpredicted by more than 200% in certain cases.A combination of several problems can be attributed to these discrepancies, such as lack of higher granularity engine thrust data, incompatibilities of BADA climb equations with the aircraft, etc.Additionally, pre-defined procedures set by the manufacturer may alter the performance of the aircraft which may perform differently in different flight profiles.For example, Predator A always descends at a CAS of 75 kts while the FLOPS-BADA combination assumes descent at optimum lift-drag ratio.Modifications along these lines and further investigation into the problem are being conducted at Purdue in order to produce better results. +Summary of BADA deficiencies and limitationsBADA deficiencies: None BADA limitations: Ill-suited climb/descent schedules overshoot the speed limits of the aircraft in climb and descent, suggesting modifications that may have to be made in BADA to account for procedures pre-defined by the aircraft manufacturers +MACSMACS performance files were generated by mapping the BADA files.The following MACS drag model and engine thrust model were used respectively for Predator A: C172 and O-320-H2AD. +Predator BPredator B is a medium-scale, fixed-wing aircraft equipped with a Honeywell TPE331-10YGD turboprop engine.The aircraft cruises at an altitude of 31000 ft., with maximum altitude also at 31000 ft. and weighs approximately 10500 lb.The BADA model of Predator B was developed using data provided by GA.FLOPS model of the Predator B is generated by using the built-in Transport Aircraft weight equations, engine deck, and aerodynamic data of FLOPS.A comparison of sizing results from FLOPS and manufacturer provided data is shown in Table 22.FLOPS generated values for the drag polar, speed schedules, climb rates and fuel flow are used in the MATLAB-based BADA model to generate BADA specific coefficients.These coefficients are further used to generate the .PTF file for Predator B.During BADA production it was identified that the cruise, climb and descent TAS of Predator B were over-predicted by the BADA model due to the stall speed buffer condition employed in BADA.Simulation tools compatible with BADA also apply this limit, making it a hard constraint on the aircraft.Additional discrepancies, if any, are currently being investigated by the manufacturers. +Summary of BADA deficiencies and limitationsBADA deficiencies: None BADA limitations: Stall speed buffer constraints in BADA overshoot the speed of Predator B in cruise, climb and descent.Manufacturer reported cruise speed at 31000 ft. is 160 kts while BADA constraint sets the speed at 209 kts.Further limitations can be identified only after complete validation of the aircraft. +Predator C (Avenger)Avenger is a medium-scale, fixed-wing aircraft equipped with a Pratt and Whitney 545B, high bypass ratio, turbofan engine.The aircraft cruises at an altitude of 40000 ft., with maximum altitude also at 40000 ft. and weighs approximately 15800 lb.The BADA model of Avenger was developed using data provided by GA.FLOPS model of the Avenger is generated by using the built-in Transport Aircraft weight equations, engine deck, and aerodynamic data of FLOPS.A comparison of sizing results from FLOPS and manufacturer provided data is shown in Table 24.FLOPS generated values for the drag polar, speed schedules, climb rates and fuel flow are used in the MATLAB-based BADA model to generate BADA specific coefficients.These coefficients are further used to generate the .PTF file for Avenger.The BADA .PTF file generated for Avenger was found to have several discrepancies in comparison to the manufacturer provided data.The cruise, climb and descent TAS were overpredicted by at least 13% in the .PTF, while the fuel flow rates and climb rates were overpredicted by more than 200% in certain cases.GA reports decreasing fuel flow rates with altitude whereas the BADA model predicts the opposite trend.A combination of several problems can be attributed to these discrepancies, such as, lack of higher granularity engine thrust data, incompatibility of BADA equations with the aircraft etc.Additionally, pre-defined procedures set by the manufacturer may alter the performance of the aircraft which may perform differently in different flight profiles.For example, Avenger always descends at a CAS of 150 kts, while the FLOPS-BADA combination assumes descent at optimum lift-drag ratio.Modifications along these lines and further investigation into the problem are being conducted at Purdue in order to produce better results. +Summary of BADA deficiencies and limitationsBADA deficiencies: None BADA limitations: Ill-suited climb/descent schedules overshoot the speed limits of the aircraft in climb and descent, suggesting modifications that may have to be made in BADA to account for procedures pre-defined by the aircraft manufacturers. +MACSMACS performance files were generated by mapping the BADA files.The following MACS drag model and engine thrust model were used respectively for Avenger: AVEN(created externally and added into the database) and PW_JT8D-07. +Hunter UASHunter UAS is a small-scale, fixed-wing aircraft equipped with two APL heavy fuel engines.The aircraft cruises at an altitude of 18000 ft., with maximum altitude also at 18000 ft. and weighs approximately 1950 lb.The BADA model of Hunter UAS was developed using data provided by AAI.FLOPS piston engine deck was generated using engine data provided by the manufacturer.A comparison of sizing results from FLOPS and manufacturer provided data is shown in Table 25.FLOPS generated values for the drag polar, speed schedule, climb rates and fuel flow are used in the MATLAB-based BADA model to generate BADA specific coefficients.These coefficients are further used to generate the .PTF file for Gray Eagle. +Summary of BADA deficiencies and limitationsBADA deficiencies: None BADA limitations: None +MACSMACS performance files were generated by mapping the BADA files.The following MACS drag model and engine thrust model were used respectively for Hunter UAS: C172 and O-320-H2AD. +Cargo UASThe Cargo UAS aircraft is a medium sized hybrid UAS with a single piston engine at the rear of the fuselage, a rectangular wing planform, and a unique triangular bent tail design.The engine is a UEL 741AR74-1102 piston engine.The empty weight of Cargo UAS is 333 lb. and the gross weight is 467 lb.The fuselage length is 63.1 inches and the wing span is 19.8 feet.A comparison of sizing results from FLOPS and manufacturer provided data is shown in Table 26.The images used in constructing 3D models of Cargo UAS, and the model generated therefrom, are shown in Figure 23 and Figure 24, respectively.The DATCOM-JSBSim flight modeling tool was used to model Cargo UAS, from which the BADA files are developed.Cargo UAS is a hybrid aircraft that uses its rotor for vertical takeoff and landing while it switches to propeller for climb, cruise, and descent segment.It was assumed that only propeller was used for .PTF generation, even for climb and descent close to sea level.Any lift or drag developed by the rotor blades and shaft were neglected in the model and simulation. +Summary of BADA deficiencies and limitationsBADA deficiencies: Aircraft type missing.Cargo UAS is a hybrid air aircraft and therefore, no stall speeds exist during take-off or landing.Additional modes, such as hover, may have to be introduced.BADA limitations: BADA model is currently defined only for fixed-wing aircraft.The Cargo UAS BADA files were generated directly using the modeling software, ignoring equations provided by BADA. +MACSThe Cargo UAS MACS model is not completely developed as MACS is not equipped to handle hybrid air aircraft.Engine thrust file and the drag model of this hybrid air aircraft is not available to the detail that MACS mandates.Therefore, these fields are left empty in the MACS master file.All fields that are not related to the engine model or drag model are completed using available data from the manufacturers and BADA output files. +Fire ScoutFire Scout is a small-scale rotorcraft with a Rolls-Royce 250 C20W turboshaft engine.The empty weight of Fire Scout is 1457 lb. and the gross weight is 3150 lb.The fuselage length is 23.95 feet and the main rotor diameter is 27.5 feet.A comparison of sizing results from FLOPS and manufacturer provided data is shown in Table 27.The .PTF of Fire Scout closely matches the maximum altitude, cruise speed, and rates of climb/descent provided by the manufacturer.Amongst the two rotorcrafts-Fire Scout and NEO S-300 Mk II VTOL (S350)-Fire Scout is perhaps analyzed better by RPAT, mainly due to the larger size of the aircraft and also due to the availability of adequate aircraft specifications from the manufacturer. +Summary of BADA deficiencies and limitationsBADA deficiencies: Aircraft type missing.Since rotorcrafts neither have stall speeds nor drag polars as in the same context as fixed wing aircrafts, some of the blocks in the OPF are not completed.Also, main characteristics of rotorcrafts such as vertical takeoff, land, and hover capabilities cannot be encapsulated in the BADA format.BADA limitations: BADA model is currently not defined for rotorcrafts and therefore, the Fire Scout BADA files were generated directly from the modeling software, ignoring BADA equations. +MACSThe Fire Scout MACS model is not completely developed as MACS is not equipped to handle rotorcrafts.Engine thrust file is not available to the detail that MACS mandates and a drag model cannot be conceived in the same manner as that of aircraft.Therefore, these fields are left empty in the MACS master file.All fields that are not related to the engine model or drag model are completed using available data from the manufacturers and BADA output files. +NEO S-300 Mk II VTOLNEO S-300 Mk II VTOL (S350) is a small-scale rotorcraft with a JETA1 powered single turbine engine.The empty weight of S350 is 187.4 lb. and the gross weight is 330.7 lb.The fuselage length is 10.33 feet and the main rotor diameter is 11.5 feet.A comparison of sizing results from FLOPS and manufacturer provided data is shown in Table 28.The .PTF file of S350 has several mismatches in comparison with the maximum altitude, cruise speed, and rates of climb/descent provided by the manufacturer.The RPAT estimates of fuel flow values for S350 resulted in similar values across different altitudes.This is because the size of S350 is at the lower end of the rotorcraft spectrum. +Summary of BADA deficiencies and limitationsBADA deficiencies: Vehicle type missing.Since rotorcrafts neither have stall speeds nor drag polars as in the same context as fixed wing aircrafts, some of the blocks in the OPF are not completed.Also, main characteristics of rotorcrafts such as vertical takeoff, land, and hover capabilities cannot be encapsulated in the BADA format.BADA limitations: BADA model is currently not defined for rotorcrafts and therefore, the BADA files for S350 were generated directly from the modeling software, ignoring BADA equations. +MACSThe S350 MACS model is not completely developed as MACS is not equipped to handle rotorcrafts.Engine thrust file is not available to the detail that MACS mandates and a drag model cannot be conceived in the same manner as that of aircraft.Therefore, these fields are left empty in the MACS master file.All fields that are not related to the engine model or drag model are completed using available data from the manufacturers and BADA output files. +BADA File ValidationAs mentioned earlier, UAS aircraft were simulated using KTG with the BADA files providing the necessary input.The purpose of these simulations was twofold:  Understand the flight characteristics of the UAS aircraft and identify any anomalies  Submit simulation results to the manufacturers of the UAS and thereby validate the BADA files +Simulation of Shadow B (RQ7B) using KTG +Issues and ResolutionImportant features of Shadow B's flight simulation using KTG are shown in Table 29.Anomalies were observed in the simulation results.For example, the graphs in Figure 25 depict variations in the true airspeed (TAS) of Shadow B with altitude, divided into two phases of the flight: from takeoff at KIAD to cruise altitude, and from cruise altitude to landing at KJFK.TAS increased from 56 kts to about 67 kts during takeoff within a very short altitude, and later to about 71 kts during the climb (identified by the long red-oval).Also, TAS decreased from about 79 kts to 76 kts for a very small change in altitude during descent (identified by the short redoval). +Figure 25. True airspeed (TAS) vs. altitude for RQ7B for flight from KIAD to KJFK +Reason for AnomaliesThe anomalies in Figure 25 were found to be caused by errors in compiling BADA files by the Purdue team.BADA user manual dictates that the flight speed at a given altitude described in the .PTF file should be higher than the stall speeds indicated in the .OPF file by a factor of 1.2 for takeoff and 1.3 for all other segments of the flight-these factors were probably established by airlines to augment safety at flight speeds approaching the stall limits.The different types of stall speeds specified in the .OPF file and the altitudes when they are taken into consideration by KTG are shown in Table 30.The BADA files used in compiling the results in Figure 25 did not correctly take this into consideration and the resulting speed-altitude data in the .PTF file were in conflict with the factors of safety described earlier.The Purdue team was notified of this violation and the BADA files were corrected.The BADA files in Figure 1, Figure 2, Figure 3 and Figure 4 are the corrected versions.However, these criteria affected the way some of the UAS aircraft were modeled, which will be mentioned in later sections of this report. +Simulation results using corrected BADA filesShadow B was simulated using the corrected BADA files (Figure 26, Figure 27, Figure 28 and Figure 29), with the main features of the flight shown in Table 31.The cruise TAS increased to 99 kts (as compared to that in Table 29), which the Purdue team explained as being a result of the factors of safety imposing a higher effective stall speed and causing the aircraft to fly 32.The variation of TAS with altitude is shown in Figure 33.The sharp increase in TAS during climb (red oval in Figure 33) was due to the fact that the airspeed at the corresponding altitude was in conflict with the factor of safety described earlier.For example, the .PTF file for Global Hawk indicates TAS as 124 kts at 2000 ft.(Figure 34), which was less than 1.3 times the cruise stall speed of 107.82 kts from the .OPF file (Table 30 and Figure 35).Since KTG attempts to follow the speed profiles described in the BADA files, TAS increased rapidly in a very short period of time and during a small change in altitude at the beginning of the climb phase.Plan-view of the flight path is shown in Figure 36 and plots describing other aspects of the flight are shown in Figure 37. Cargo UAS (CUAS), Fire Scout (MQ8B) and NEO S-300 Mk II VTOL (S350) are rotorcraft or a hybrid of rotorcraft and conventional aircraft.Therefore, they were not simulated using KTG, and the results from simulating and validating their flight profiles using these files are not presented here.On the other hand, the FAA's William J. Hughes Technical Center (FAA Tech Center) has been developing models to analyze and simulate rotorcraft.Consequently, they were approached to provide technical support in validating the BADA files for the four aforementioned aircraft.However, the timeline of this project was too short to take advantage of the Tech Center's expertise.A collaborative effort between NASA and the FAA Tech Center to develop adequate models for rotorcraft is strongly recommended to fill this gap in knowledge. +Summary of UAS Simulations in KTGResults of UAS flight simulations using KTG are summarized in Table 40.Included in here are four main features of each flight to briefly distinguish the different aircraft: origin and destination airports, target cruise altitude and speed.Also indicated are whether the aircraft reached the target cruise altitude and speed in the simulation, and whether BADA files for each aircraft were validated by its manufacturer.As mentioned earlier, simulation results for each UAS flight were submitted to the corresponding aircraft manufacturer for validation.It should be noted that rotorcraft cannot be simulated in KTG.Hence, the BADA files of Cargo UAS, Fire Scout and NEO S-300 Mk II VTOL were not validated by this approach.As mentioned earlier, the Tech Center was approached to assist in validating BADA files for these aircraft, but the process was not complete within the timeline of this project.Recommendations are made in the latter sections of this report on options to validate these files. +MACS File ValidationMACS files for the twelve UAS aircraft were validated by comparing the simulation results from MACS with those from KTG.The premise to this was that the validation of BADA files by the UAS manufacturer indirectly validated the KTG results. +Issues and ResolutionMACS was developed to simulate manned aircraft.Consequently, there were some issues to be resolved to modify the software and simulate UAS aircraft. +Issue 1: Speed vs. Altitude Constraints in MACSDuring the simulation of Shadow B via MACS the aircraft could not reach its cruise altitude of 8000 ft.Investigation of MACS' software code indicated that an aircraft should have a minimum speed of 100 KCAS when flying between 3500 ft. and 10500 ft. to reach the cruise altitude.Since Shadow B's speed of 80 KCAS at 8000 ft. was less than this minimum speed, it had no vertical speed beyond the altitude of 3500 ft.causing it to not reach cruise altitude.The following modifications were made to MACS' code to address this issue: MACS file modified: commonObjects/PerfDescr.java +Issue 3: Simulation of Rotorcraft and Electric AircraftNo provision was found to configure and simulate rotorcraft flights in MACS.Further, the aircraft model data files for UAS rotorcraft and electric aircraft could not be developed due to the absence of relevant data fields in the files.Consequently, the files for Orbiter (electric aircraft), Cargo UAS, Fire Scout and NEO S-300 Mk II VTOL were incomplete, and hence were not simulated in MACS.No immediate solution was found to address this issue. +Simulation of Shadow B (RQ7B) using MACSImportant features of Shadow B's flight simulation using MACS are shown in Table 41.Unlike KTG, aircraft weight at takeoff and landing are not recorded in MACS and indicated as such.The flight was simulated from KBNA to KATL to prevent MACS from exceeding its memory usage limits and thereby successfully complete the simulation due to the slow speed of Shadow B. The plots in Figure 59 show the variation in true airspeed (TAS) of shadow B with altitude.The plot on the left hand side is from takeoff to cruise altitude and that on the right is from cruise to landing.It is not known what caused the rapid increase in speed during the climb phase, and the jaggedness and the associated increases in speed beyond the cruise speed in the right hand plot.The spikes in airspeed in Figure 60 are a different representation of the jaggedness in Figure 59, and hence could not be explained.Since it is not yet known as to how MACS interprets the files for UAS aircraft, no hypothesis was formed to explain the simulation results.44.It should be noted that, unlike Shadow B and Global Hawk, the flight did not reach cruise altitude and speed.The slow speed of Predator A is a possible reason.However, as mentioned earlier, it is not known as to why MACS cannot simulate a slow flying aircraft.Consequently, simulation results similar to Figure 61 were not compiled for Predator A. Other simulation results are shown in Figure 64.Similar to Shadow B and Global Hawk, the reason for the sharp fluctuations in speed (Figure 64b) is not known.The flight was simulated from KBNA to KATL to prevent MACS from exceeding its memory usage limits and thereby successfully complete the simulation due to the slow speed of Predator A. 45.It should be noted that, unlike Shadow B and Global Hawk, the flight did not reach cruise altitude and speed.While Predator B flies faster than Predator A, it is slower compared to Global Hawk, and this is a possible reason for the unsuccessful simulation.The flight was simulated from KBNA to KATL to prevent MACS from exceeding its memory usage limits and thereby successfully complete the simulation due to the slow speed of Predator B (relative to Global Hawk).However, as mentioned earlier, it is not known as to why MACS cannot simulate a slow flying aircraft.Consequently, simulation results similar to Figure 61 46.It should be noted that, similar to Predator B, the flight did not reach cruise altitude and speed, possibly due to its slow speed compared to Global Hawk.The flight was simulated from KBNA to KATL to prevent MACS from exceeding its memory usage limits and thereby successfully complete the simulation due to the slow speed of Gray Eagle (relative to Global Hawk).However, as mentioned earlier, it is not known as to why MACS cannot simulate a slow flying aircraft.Consequently, simulation results similar to Figure 61 were not compiled for Gray Eagle.Other simulation results are shown in Figure 66.Similar to Shadow B and Global Hawk, the reason for the sharp fluctuations in speed (Figure 66b) is not known.47.It should be noted that, similar to Predator B, the flight did not reach cruise altitude and speed, possibly due to its slow speed compared to Global Hawk.The flight was simulated from KBNA to KATL to prevent MACS from exceeding its memory usage limits and thereby successfully complete the simulation due to the slow speed of Predator C (relative to Global Hawk).However, as mentioned earlier, it is not known as to why MACS cannot simulate a slow flying aircraft.Consequently, simulation results similar to Figure 61 were not compiled for Gray Eagle.Other simulation results are shown in Figure 67.Similar to Shadow B and Global Hawk, the reason for the sharp fluctuations in speed (Figure 67b) is not known.48.It should be noted that, similar to Predator A, the flight did not reach cruise altitude and speed, possibly due to its slow speed compared to Global Hawk.The flight was simulated from KBNA to KATL to prevent MACS from exceeding its memory usage limits and thereby successfully complete the simulation due to the slow speed of Hunter (relative to Global Hawk).However, as mentioned earlier, it is not known as to why MACS cannot simulate a slow flying aircraft.Consequently, simulation results similar to Figure 61 were not compiled for Hunter.Other simulation results are shown in Figure 68.Similar to Shadow B and Global Hawk, the reason for the sharp fluctuations in speed (Figure 68b) is not known.As mentioned earlier, the MACS files for Orbiter (ORBM), Cargo UAS (CUAS), Fire Scout (MQ8B) and NEO S-300 Mk II VTOL (S350) did not truthfully represent the aircraft because: 1) Orbiter is an electric aircraft and cannot be correctly represented within the schema of MACS, and 2) the other three aircraft are either rotorcraft or a hybrid of rotorcraft and conventional aircraft, which cannot be represented in MACS.Therefore, these aircraft were not simulated in MACS and the results from simulating and validating their MACS files are not presented here. +Summary of UAS Simulations in MACSResults of UAS flight simulations using KTG are summarized in Table 49.Included in here are four main features of each flight to briefly distinguish the different aircraft: origin and destination airports, target cruise altitude, and cruise speed reached.Also indicated is the cruise altitude reached by the aircraft in simulation.As mentioned earlier, MACS simulation results for each UAS flight were compared to the corresponding results from KTG for validation, the premise being that the KTG results were validated by the UAS manufacturer.It should be noted that electric aircraft and rotorcraft cannot be simulated in MACS.Hence, the MCAS BADA files of Orbiter (electric aircraft), Cargo UAS, Fire Scout and NEO S-300 Mk II VTOL were not validated by this approach.The FAA's William J. Hughes Technical Center (FAA Tech Center) was contacted for assistance in validating MACS files for these aircraft.However, the process could not be completed during the current project's contract period.However, the issues and difficulties identified in developing these MACS files are discussed in the latter sections of this report, along with recommendations for validating the files. +ACES Simulations for CNS CapabilitiesAs user selectable options in ACES, several Communication system, Navigation system and Surveillance system models are available for use in airspace simulations for experimentation to determine the implications of these real world systems on aircraft operations and airspace concepts.For UAS aircraft, many of these same systems are integrated (or are being considered for integration and use) onboard these unmanned aircraft in varied capacity by the UAS community and are, or may become, integral components of the UAS systems as the future of UAS architectures to enable their use in the NAS progresses.For the UAS in the NAS, Modeling and Simulation effort, twelve UAS aircraft models have been introduced into ACES and have been tested for the ability to configure their systems to use the ACES CNS models.This summary report describes the process for adding these UAS into the latest version of ACES with CNS models, and outlines the steps taken to configure ACES to fly these aircraft with the CNS models.Results of the simulations performed are provided in summary table and comments on the p, improvement. +UAS Aircraft/BADA Data Installation and Preparation for CNS Simulations +Installation of UAS Aircraft Models into ACES and KTGPrior to testing the UAS aircraft with ACES CNS models, databases in ACES and KTG were configured for nine UAS aircraft.This configuration process is explained in detail in Appendix B. +Develop Flight Data SetsSince this was a first-use experience with the UAS aircraft, new Flight Data Sets (FDSs) were defined that were both appropriate to the UAS aircraft characteristics, and also were adequate to exercise the capabilities of the CNS models, especially for communications.The process for defining the FDSs for the simulations used the following information and guidelines:  Speed and cruise altitudes were identified for use in the FDSs from the aircraft .PTF BADA files. Flight route distances were selected (with some experimentation) based on UAS size, weight and aircraft speed characteristics. FDS Flight routes were derived from existing FDSs.The routes used were variations on routes used to represent UAS operations for Homeland Security applications within ZAU Sector in previous UAS work, and were tailored for route length appropriate to the class of UAS that would use the FDS. The FDSs that were created flew UAS flights for gate-to-gate operations between Towered airports to allow for full execution of all communication messages that are part of our communications message sets.(see Note 1)  Airline names used on the FDSs were created to be appropriate to the UAS vendor to help identify the type of UAS flown in a simulation.Aircraft names that were used mapped directly to the designators for the UASs that had been defined in the BADA and ACES database files (i.e.RQ4A, RQ7B, etc…)  FDSs were defined for a single UAS flight each.Note 1: The current architecture defined in ACES has yet to be tailored for UAS operations.Use of UAS flights gate-to-gate is strictly the default scenario for any aircraft, and is planned to be modified for UAS operations.Note 2: Information regarding origin/destination airports, route distances, altitude and speed defined in these FDSs is indicated in the simulation summary chart -Table 50.Summary of results for from ACES simulations to test CNS capabilities of UAS aircraft Testing: Each FDS was tested in independent simulations to verify its applicability prior to applying the CNS equipage.In several instances, initial FDSs were discovered to be defective for the purpose of this testing, because the airports/flight routes that were chosen were all located within the same TRACON where the KTG would alter the flight path to a route that would not allow for the flight duration and airspace transitions that were desired to enable applicability of the full communications message set application for CNS capability modeling.Once this situation was understood, two of the initial FDSs were rebuilt to use separated airports. +Develop CNS Plugin Configuration FilesThe CNS Plugin in ACES allows the user to define the systems that comprise the compliment CNS avionics that are operational in a simulation.The implementation for use of these systems is managed by defining aircraft CNS equipage configuration files for whichever Communication (Comm.),Navigation (Nav.) or Surveillance (Surv.)system the user wishes to have functioning onboard an aircraft during an ACES system.Next, ACES collects and stores those data for analysis and evaluation in an output database.To date, six CNS models are available: +Test ResultsResults from the simulations were very positive, with all of the Comm.and Surv.model simulations completing as expected and generating correct output data.The exception to this was for the use of the Nav.GPS system models, where the simulations would run and indicate a successful completion, but no navigation data was stored, indicating that the Nav.GPS model had not been applied for the flight.Investigating this further it was found that this was also the case for the VOR/DME Nav.model and for simulations that used a standard aircraft with the same results indicating that it had nothing to do with the UAS model.On final investigation, a simulation with the same standard aircraft was run using the MPAST trajectory generator, and the Nav.GPS model performed as expected.The problem has been identified to Intelligent Automation and a fix to correct the KTG interoperability with the Navigation models has been defined but was not able to be implemented for the completion of this testing. +Problems Encountered and Precautions for use of CNS models with UAS Comm.model simulations: FDSs that define flights departing from and arriving at airports located within the same TRACON airspace appear to have their flight path altered to what appears to be a shorter route.This needs to be investigated further to determine just what does happen to the flight path, but this would be problematic for UAS simulations especially for smaller UAS where flight routes are typically of shorter duration and distance. Comm., Nav. or Surv.model simulations: There was one instance where in an airline name was used that did not correlate with an AOC that ACES uses in its AOC XRef file.In this case, communications was set up to use VDL2, however the simulation ran with the Comm.model defaulting to Voice VHF.This is a known problem with the ACES models use, but the remedy to this problem is simply that the user makes sure that all airlines defined in the FDS are common airline names that have an associated AOC.For our further UAS CNS modeling work, we are considering implementing an UAS AOC, where we would use and add as needed, UAS manufacturer airline designators and associated them to this AOC to ensure proper equipage of UAS aircraft, especially for larger simulations. +ConclusionsThe purpose of this project was to provide performance data for twelve UAS aircraft, in formats usable by standard aviation models: BADA and MACS.BADA files for fixed-wing UAS aircraft were developed by modifying a NASA-developed aircraft sizing software called FLOPS.Separate aircraft models were developed to size rotorcraft, hybrid aircraft and electric aircraft.However, the fidelity of the output from these models is limited by the fact that the aircraft data from the UAS manufacturers were not complete and accurate due to proprietary restrictions.Simulations were conducted using KTG for the BADA files, and the MACS software for MACS files.Simulation output from KTG were examined and validated by the UAS manufacturers.However, simulations were not conducted for two rotorcraft and one hybrid aircraft, due to limitations on KTG.Hence, their BADA files were not validated.Similarly, these aircraft cannot be simulated in MACS, and hence, their MACS files were not validated.Furthermore, a number of difficulties and challenges were encountered in simulating the other UAS aircraft in MACS, either due to the lack of format support to represent UAS aircraft data as MACS files or due to limitations on MACS software.Therefore, MACS files of all twelve aircraft were not validated.Recommendations were made to resolve these issues to successfully represent all twelve UAS aircraft in BADA and MACS format.The FAA Tech Center was approached to assist in validating the BADA and MACS files for rotorcraft and hybrid aircraft.However, the effort needed was beyond the scope and timeline of this project and is included as one of the recommendations to extend the scope and benefits of this project.The project also involved simulations to simulate the communication, navigation and surveillance (CNS) capabilities of UAS aircraft.CNS equipage files provided by the UAS manufacturers were used to configure and conduct the simulations in the Airspace Concept Evaluation System (ACES).KTG was the trajectory generator employed in these simulations.The communication and surveillance simulations were successful, whereas the navigation simulations require some modifications to ACES and KTG.This project was focused on producing and validating only BADA and MACS data files for UAS aircraft.However, it is speculated that the challenges encountered in this process and the recommendations to be discussed in the following sections are applicable to almost all other data formats.Therefore, efforts to address these issues will be beneficial to the entire aviation community.13 Recommendations for Future Work +Recommendations to Modify BADA Format for UAS SimulationsAs described earlier, the EUROCONTROL developed the format of BADA files primarily for manned-aircraft.Consequently, many areas and topics were identified that either require modification or new definitions to accommodate UAS aircraft design and operations (Section 4).This section presents some of the important areas in BADA format to be modified for successful simulation of UAS aircraft. +Design-based modificationsSince the current BADA format does not have specific provisions, UAS aircraft have to be represented using the templates of existing manned-aircraft.However, this restricts the number of UAS aircraft that can be represented in the BADA format.In particular, there are no provisions to represent very light aircraft (e.g., Shadow B and Aerosonde), rotorcraft and hybrid aircraft (e.g., Cargo UAS and Hunter), and electric engines (e.g., Orbiter).Furthermore, there are no airline operations for UAS aircraft to compile the .APF BADA file.Since a wide variety of UAS aircraft are being currently developed and operated, the need to update BADA format is not only critical to conducting large-scale simulations of NAS, but also time-sensitive if the FAA has to meet the Congressional mandate of creating necessary framework to operate UAS aircraft in the NAS [12]. +Operations-based changesCurrent BADA format imposes certain restrictions on aircraft operations (e.g., stall speed criteria).These restrictions were formulated based on passenger safety and comfort for manned-aircraft.However, UAS aircraft operate outside the envelope of passenger flights, and hence, should not be subjected to these restrictions.Furthermore, there are no provisions to faithfully represent the flight profiles of UAS rotorcraft and hybrid UAS aircraft in BADA.During discussions with the FAA Tech Center, their experts have voiced similar concerns regarding the current format of BADA in modeling rotorcraft and hybrid aircraft, and expressed interest in future efforts to update the format [13]. +Recommendations to Modify MACS for UAS SimulationsThe challenges and difficulties were encountered in developing MACS files were described in Section 5. Though these initially appear to be different from those encountered for BADA, there are many commonalties between them.For example, 1) both MACS and BADA were not able to represent and simulate rotorcraft and hybrid aircraft, and 2) data and operational rules for existing manned-aircraft were used to model fixed-wing UAS aircraft, leading to similar discrepancies between simulation output and expected aircraft performance.This section presents some of the important areas in MACS to be modified for successful simulation of UAS aircraft. +Aircraft DataFor majority of the twelve UAS aircraft studied in this project, detailed airframe drag data was not available due to the propriety nature of the information.Consequently, data from similarly sized manned-aircraft were substituted for or mapped to UAS aircraft, resulting in many discrepancies between simulation output and expected aircraft performance.Furthermore, electric aircraft, rotorcraft and hybrid aircraft cannot be represented and simulated in MACS (e.g., Orbiter, Cargo UAS, Fire Scout and NEO S-300 Mk II VTOL).The FAA Tech Center was approached to assist in the validation of MACS files for the rotorcraft, but this was beyond the scope and timeline of this project. +Modifications to MACS SoftwareMACS software was found to be consuming large computer memory to simulate slow flying UAS aircraft such as Shadow B. Furthermore, modifications were made to force the software into simulating the cruise segments of flight, which were not successful.Some of these issues might have been resolved by NASA experts but they were not readily available during the period of this project. +Validation of BADA and MACS Files for Rotorcraft and Hybrid AircraftAs mentioned earlier, KTG was used to simulate UAS aircraft based on BADA files, the results from which were validated by the aircraft manufacturers.However, since, KTG cannot simulate rotorcraft and hybrid aircraft, the BADA files for Cargo UAS, Fire Scout and NEO S-300 Mk II VTOL were not validated.Similarly, since MACS cannot simulate these aircraft, the corresponding input files were also not validated.As mentioned earlier, a joint research effort between NASA and the FAA Tech Center to developing rotorcraft models is strongly needed and recommended to leverage the expertise of the two agencies in filling this gap in knowledge. +Other Recommendations Kinematic Trajectory Generator (KTG)KTG has been extensively verified and validated for simulations involving manned aircraft.However, key areas were identified that require modifications to simulate UAS aircraft in KTG. Different types of data are used as input to simulation an aircraft in KTG.These involve the four BADA files described earlier and a file defining the aircraft's control parameters (Appendix B).Therefore, the accuracy of simulation results from KTG is dependent on the accuracy of BADA files and the control parameters file.However, all UAS aircraft simulations presented in this report were conducted using default setting for aircraft control parameters, due to lack of appropriate data.Effort required to compute specific control parameters for each UAS aircraft was beyond the scope of this project, and can be a valuable extension to improve the fidelity of the simulations. The present framework of KTG does not support the simulation of rotorcraft and hybrid aircraft due to lack of appropriate BADA and aircraft control parameters files.Modifications to BADA format based on aforementioned recommendations should be able to address this issue. Another important element currently not available in KTG is the ability to estimate an aircraft's engine thrust, which is essential to simulating rotorcraft and hybrid aircraft. +Airspace Concept Evaluation System (ACES)ACES was used to conduct simulations to evaluate the CNS capabilities, requirements and limitations of UAS aircraft operations.However, similar to KTG, ACES is currently best suited to simulate manned-aircraft, requiring changes to ACES' configuration to conduct these simulations (Appendix B).While these changes addressed a number of difficulties in simulating UAS aircraft, many more remain:  Very small UAS aircraft such as Aerosonde and Orbiter were also simulated using the separation rules for small aircraft category, which may lead to larger separation distances than otherwise necessary.On the other hand, separation criteria for such very small aircraft are non-existent, making this a very important operational issue that needs to be addressed immediately for successful real-world operations of such UAS aircraft. UAS aircraft can have a short range (less than 40 nmi.) due to limitations on actual aircraft range (small fuel tank) or the range of its ground control station imposing line-ofsight restrictions.However, current airspace definitions in ACES did not allow the simulation of such aircraft. +Fleet-level Simulations of UAS AircraftUAS aircraft were simulated in this project only to validate aircraft data (BADA, MACS and CNS equipage), limiting their scope.However, large-scale simulations involving fleets of UAS aircraft in multiple operational regimes are required to thoroughly understand their impact on current-day and future operations in the NAS.Further, such simulations also provide insights into challenges associated with HITL processes such piloting, controlling and managing the UAS traffic.This file contains the flight control parameters for each aircraft type.Due to lack of accurate data for UAS aircraft, the default values specified for existing conventional aircraft were used.For example, to simulate Global Hawk's flight, its control parameters were added to this file in seven separate lines, where "RQ4A" was the aircraft code used to identify Global Hawk.The entries indicated as <> are only shown for clarity and should not be included in the file: <> RQ4A,500,2.25,0,0,0.004,0,0.32,0,0.08,0.000002,0,0.0000081,0.024,0,-0.00000015,0,0<> RQ4A,1000,2.25,0,0,0.004,0,0.32,0,0.08,0.000002,0,0.0000081,0.024,0,-0.00000015,0,0<> RQ4A,2000,2.25,0,0,0.004,0,0.32,0,0.08,0.000002,0,0.0000081,0.024,0,-0.00000015,0,0<> RQ4A,5000,0.225,0,0,0.0004,0,0.032,0,0.008,0.000002,0,0.0000081,0.024,0,-0.00000015,0,0<> RQ4A,10000,0.0225,0,0,0.0004,0,0.032,0,0.008,0.000002,0,0.0000081,0.024,0,-0.00000015,0,0<> RQ4A,20000,0.0225,0,0,0.0004,0,0.032,0,0.008,0.000002,0,0.0000081,0.024,0,-0.00000015,0,0<> RQ4A,30000,0.0225,0,0,0.0004,0,0.032,0,0.008,0.000002,0,0.0000081,0.024,0,-0.00000015,0,0 +Configuration of "MPAS_SYNONYM.LST," "SYNONYM_ALL.LST" and"SYNONYM_ACES_KTG.OLD" These files specify the names of the BADA files to be used for a particular aircraft.For example, to add Global Hawk, use "blank space" to separate entries in each file.Do not use "tabs" for "space."Following the template for the entries of existing aircraft in each file is strongly advised to avoid any errors or misinterpretation of the files by KTG. +Configuration of ACES DatabaseThe UAS aircraft should be added to the table "aircraft_characteristics_ds" in the ACES file "aces_model_input_nodal_model.sql".This file is shown here only as an example.The analyst should use the appropriate ACES database file being used in her/his simulations.This table specifies the aircraft's speed (KCAS) during different phases of flight. +Location of table: Build\modules\acesutilities\data\databaseFigure 1 .1Figure 1.The .APF file for Shadow B (RQ7B).File was compiled by Purdue............................17Figure 2. The .DCT file for Shadow B (RQ7B).File was compiled by IAI...................................18 Figure 3.The .OPF file for Shadow B (RQ7B).File was compiled by Purdue............................19 Figure 4.The .PTF file for Shadow B (RQ7B).File was compiled by Purdue............................20 Figure 5. BADA climb schedules for commercial Jet aircraft .....................................................22 Figure 6.BADA standard airline climb increments for commercial Jet aircraft ...........................22 Figure 7. Convention for MACS-BADA mapping .......................................................................24 Figure 8. Aircraft model data file for Predator B. File produced by Purdue................................25 Figure 9. Airframe drag model data file for Predator B. File produced by Purdue......................26 Figure 10.Snapshot of flight parameters file for Predator B. Speeds are indicated air speeds in knots.File produced by IAI........................................................................................................27 Figure 11.APM generation and validation flowchart .................................................................28 Figure 12.Flowchart representing the BADA generation process using the DATCOM/JSBSim/Flight Sim tool .............................................................................................30 Figure 13.Orthographic projection/picture based modeling ......................................................31 Figure 14.Difficulties of non-orthographic projection picture based modeling ...........................31 Figure 15.Blender 3D Modeling of Cargo UAS .........................................................................32 Figure 16.FlightGear Simulation Testing of Cargo UAS ...........................................................33 Figure 17.Sample power curve ................................................................................................36 Figure 18.Aircraft model data MACS file for Shadow B ............................................................39 Figure 19.Orbiter images used for 3D construction ..................................................................41 Figure 20.Orbiter DATCOM Input Visualization ........................................................................42 Figure 21.Aerosonde images used for 3D model construction .................................................43 Figure 22.Aerosonde DATCOM Input Visualization ..................................................................43 Figure 23.Schematics of Cargo UAS from AAI .........................................................................49 Figure 24.Cargo UAS DATCOM Input Visualization .................................................................49 Figure 25.True airspeed (TAS) vs. altitude for RQ7B for flight from KIAD to KJFK ...................52 Figure 26.Corrected .APF file for Shadow B (RQ7B).File was compiled by Purdue.................53 Figure 27.Corrected .DCT file for Shadow B (RQ7B).File was compiled by IAI.......................53 Figure 28.Corrected .OPF file for Shadow B (RQ7B).File was compiled by Purdue................54 Figure 29.Corrected .PTF file for Shadow B (RQ7B).File was compiled by Purdue.................55 Figure 30.True airspeed (TAS) vs. altitude for Shadow B flight from KIAD to KJFK using corrected BADA files .................................................................................................................56 Figure 31.Plan-view of Shadow B flight path from KIAD to KJFK using corrected BADA files ..56 Figure 32.Details of Shadow B flight from KIAD to KJFK using corrected BADA files ...............57 Figure 33.True airspeed (TAS) vs. altitude for Global Hawk flight from KMSP to KMCO ..........58 Figure 34.The .PTF file for Global Hawk (RQ4A).File was compiled by Purdue......................59 Figure 35.The .OPF file for Global Hawk (RQ4A).File was compiled by Purdue......................60 Figure 36.Plan-view of Global Hawk flight path from KMSP to KMCO ......................................61 Figure 37. Variation in altitude and airspeed (TAS) with time and distance for Global Hawk flight from KMSP to KMCO ................................................................................................................61 Figure 38.True airspeed (TAS) vs. altitude for Orbiter flight from KATL to KBHM .....................62 Figure 39.Plan-view of Orbiter flight path from KATL to KBHM ................................................62 Figure 40.Variation in altitude and airspeed (TAS) with time and distance for Orbiter flight from KATL to KBHM..........................................................................................................................63 Figure 41.True airspeed (TAS) vs. altitude for Aerosonde flight from KATL to KBHM ..............64 Figure 42.Plan-view of Aerosonde flight path from KATL to KBHM ..........................................64 Figure 43.Variation in altitude and airspeed (TAS) with time and distance for Aerosonde flight from KATL to KBHM..................................................................................................................65 +Figure 1.The .APF file for Shadow B (RQ7B).File was compiled by Purdue............................17Figure 2. The .DCT file for Shadow B (RQ7B).File was compiled by IAI...................................18 Figure 3.The .OPF file for Shadow B (RQ7B).File was compiled by Purdue............................19 Figure 4.The .PTF file for Shadow B (RQ7B).File was compiled by Purdue............................20 Figure 5. BADA climb schedules for commercial Jet aircraft .....................................................22 Figure 6.BADA standard airline climb increments for commercial Jet aircraft ...........................22 Figure 7. Convention for MACS-BADA mapping .......................................................................24 Figure 8. Aircraft model data file for Predator B. File produced by Purdue................................25 Figure 9. Airframe drag model data file for Predator B. File produced by Purdue......................26 Figure 10.Snapshot of flight parameters file for Predator B. Speeds are indicated air speeds in knots.File produced by IAI........................................................................................................27 Figure 11.APM generation and validation flowchart .................................................................28 Figure 12.Flowchart representing the BADA generation process using the DATCOM/JSBSim/Flight Sim tool .............................................................................................30 Figure 13.Orthographic projection/picture based modeling ......................................................31 Figure 14.Difficulties of non-orthographic projection picture based modeling ...........................31 Figure 15.Blender 3D Modeling of Cargo UAS .........................................................................32 Figure 16.FlightGear Simulation Testing of Cargo UAS ...........................................................33 Figure 17.Sample power curve ................................................................................................36 Figure 18.Aircraft model data MACS file for Shadow B ............................................................39 Figure 19.Orbiter images used for 3D construction ..................................................................41 Figure 20.Orbiter DATCOM Input Visualization ........................................................................42 Figure 21.Aerosonde images used for 3D model construction .................................................43 Figure 22.Aerosonde DATCOM Input Visualization ..................................................................43 Figure 23.Schematics of Cargo UAS from AAI .........................................................................49 Figure 24.Cargo UAS DATCOM Input Visualization .................................................................49 Figure 25.True airspeed (TAS) vs. altitude for RQ7B for flight from KIAD to KJFK ...................52 Figure 26.Corrected .APF file for Shadow B (RQ7B).File was compiled by Purdue.................53 Figure 27.Corrected .DCT file for Shadow B (RQ7B).File was compiled by IAI.......................53 Figure 28.Corrected .OPF file for Shadow B (RQ7B).File was compiled by Purdue................54 Figure 29.Corrected .PTF file for Shadow B (RQ7B).File was compiled by Purdue.................55 Figure 30.True airspeed (TAS) vs. altitude for Shadow B flight from KIAD to KJFK using corrected BADA files .................................................................................................................56 Figure 31.Plan-view of Shadow B flight path from KIAD to KJFK using corrected BADA files ..56 Figure 32.Details of Shadow B flight from KIAD to KJFK using corrected BADA files ...............57 Figure 33.True airspeed (TAS) vs. altitude for Global Hawk flight from KMSP to KMCO ..........58 Figure 34.The .PTF file for Global Hawk (RQ4A).File was compiled by Purdue......................59 Figure 35.The .OPF file for Global Hawk (RQ4A).File was compiled by Purdue......................60 Figure 36.Plan-view of Global Hawk flight path from KMSP to KMCO ......................................61 Figure 37. Variation in altitude and airspeed (TAS) with time and distance for Global Hawk flight from KMSP to KMCO ................................................................................................................61 Figure 38.True airspeed (TAS) vs. altitude for Orbiter flight from KATL to KBHM .....................62 Figure 39.Plan-view of Orbiter flight path from KATL to KBHM ................................................62 Figure 40.Variation in altitude and airspeed (TAS) with time and distance for Orbiter flight from KATL to KBHM..........................................................................................................................63 Figure 41.True airspeed (TAS) vs. altitude for Aerosonde flight from KATL to KBHM ..............64 Figure 42.Plan-view of Aerosonde flight path from KATL to KBHM ..........................................64 Figure 43.Variation in altitude and airspeed (TAS) with time and distance for Aerosonde flight from KATL to KBHM..................................................................................................................65 +Figure 2 .2Figure 1.The .APF file for Shadow B (RQ7B).File was compiled by Purdue............................17Figure 2. The .DCT file for Shadow B (RQ7B).File was compiled by IAI...................................18 Figure 3.The .OPF file for Shadow B (RQ7B).File was compiled by Purdue............................19 Figure 4.The .PTF file for Shadow B (RQ7B).File was compiled by Purdue............................20 Figure 5. BADA climb schedules for commercial Jet aircraft .....................................................22 Figure 6.BADA standard airline climb increments for commercial Jet aircraft ...........................22 Figure 7. Convention for MACS-BADA mapping .......................................................................24 Figure 8. Aircraft model data file for Predator B. File produced by Purdue................................25 Figure 9. Airframe drag model data file for Predator B. File produced by Purdue......................26 Figure 10.Snapshot of flight parameters file for Predator B. Speeds are indicated air speeds in knots.File produced by IAI........................................................................................................27 Figure 11.APM generation and validation flowchart .................................................................28 Figure 12.Flowchart representing the BADA generation process using the DATCOM/JSBSim/Flight Sim tool .............................................................................................30 Figure 13.Orthographic projection/picture based modeling ......................................................31 Figure 14.Difficulties of non-orthographic projection picture based modeling ...........................31 Figure 15.Blender 3D Modeling of Cargo UAS .........................................................................32 Figure 16.FlightGear Simulation Testing of Cargo UAS ...........................................................33 Figure 17.Sample power curve ................................................................................................36 Figure 18.Aircraft model data MACS file for Shadow B ............................................................39 Figure 19.Orbiter images used for 3D construction ..................................................................41 Figure 20.Orbiter DATCOM Input Visualization ........................................................................42 Figure 21.Aerosonde images used for 3D model construction .................................................43 Figure 22.Aerosonde DATCOM Input Visualization ..................................................................43 Figure 23.Schematics of Cargo UAS from AAI .........................................................................49 Figure 24.Cargo UAS DATCOM Input Visualization .................................................................49 Figure 25.True airspeed (TAS) vs. altitude for RQ7B for flight from KIAD to KJFK ...................52 Figure 26.Corrected .APF file for Shadow B (RQ7B).File was compiled by Purdue.................53 Figure 27.Corrected .DCT file for Shadow B (RQ7B).File was compiled by IAI.......................53 Figure 28.Corrected .OPF file for Shadow B (RQ7B).File was compiled by Purdue................54 Figure 29.Corrected .PTF file for Shadow B (RQ7B).File was compiled by Purdue.................55 Figure 30.True airspeed (TAS) vs. altitude for Shadow B flight from KIAD to KJFK using corrected BADA files .................................................................................................................56 Figure 31.Plan-view of Shadow B flight path from KIAD to KJFK using corrected BADA files ..56 Figure 32.Details of Shadow B flight from KIAD to KJFK using corrected BADA files ...............57 Figure 33.True airspeed (TAS) vs. altitude for Global Hawk flight from KMSP to KMCO ..........58 Figure 34.The .PTF file for Global Hawk (RQ4A).File was compiled by Purdue......................59 Figure 35.The .OPF file for Global Hawk (RQ4A).File was compiled by Purdue......................60 Figure 36.Plan-view of Global Hawk flight path from KMSP to KMCO ......................................61 Figure 37. Variation in altitude and airspeed (TAS) with time and distance for Global Hawk flight from KMSP to KMCO ................................................................................................................61 Figure 38.True airspeed (TAS) vs. altitude for Orbiter flight from KATL to KBHM .....................62 Figure 39.Plan-view of Orbiter flight path from KATL to KBHM ................................................62 Figure 40.Variation in altitude and airspeed (TAS) with time and distance for Orbiter flight from KATL to KBHM..........................................................................................................................63 Figure 41.True airspeed (TAS) vs. altitude for Aerosonde flight from KATL to KBHM ..............64 Figure 42.Plan-view of Aerosonde flight path from KATL to KBHM ..........................................64 Figure 43.Variation in altitude and airspeed (TAS) with time and distance for Aerosonde flight from KATL to KBHM..................................................................................................................65 +) in TRACON airspace by aircraft weight and engine type ......................................................................................................................................... 130 Table 66.Different speed settings of Global Hawk for inclusion in ACES aircraft database.Speeds are Calibrated Airspeed in knots (KCAS)................................................................... 131 +Figure 1 .1Figure 1.The .APF file for Shadow B (RQ7B).File was compiled by Purdue. +Figure 2 .2Figure 2. The .DCT file for Shadow B (RQ7B).File was compiled by IAI. +Figure 3 .3Figure 3.The .OPF file for Shadow B (RQ7B).File was compiled by Purdue. +Figure 4 .4Figure 4.The .PTF file for Shadow B (RQ7B).File was compiled by Purdue. +Figure 5 .Figure 6 .56Figure 5. BADA climb schedules for commercial Jet aircraft +Three files were produced to simulate UAS flight in the Multi-aircraft Control System (MACS):  Aircraft model data file: This file contains an aircraft's description and performance parameters such as the engine type and number of engines, limits on the different operational weights and speeds, and drag model. Airframe drag model data file: This file specifies the lift and drag coefficients, at different +Figure 7 .7Figure 7. Convention for MACS-BADA mapping +Figure 8 .8Figure 8. Aircraft model data file for Predator B. File produced by Purdue. +Figure 9 .9Figure 9. Airframe drag model data file for Predator B. File produced by Purdue. +Figure 10 .10Figure 10.Snapshot of flight parameters file for Predator B. Speeds are indicated air speeds in knots.File produced by IAI. +Figure 1111Figure 11.APM generation and validation flowchart +Figure 12 .12Figure 12. +Figure 12 .12Figure 12.Flowchart representing the BADA generation process using the DATCOM/JSBSim/Flight Sim tool +Figure 14 .14Figure 14.Difficulties of non-orthographic projection picture based modeling +Figure 15 .15Figure 15.Blender 3D Modeling of Cargo UAS +Figure 16 .16Figure 16.FlightGear Simulation Testing of Cargo UAS +22 +Figure 18 .18Figure 18.Aircraft model data MACS file for Shadow B +2 8.8 ft. 2 Figure 19 .2219Figure 19.Orbiter images used for 3D construction +Figure 21 .21Figure 21.Aerosonde images used for 3D model construction +Figure 23 .23Figure 23.Schematics of Cargo UAS from AAI +Altitude from MSL (ft.) Airspeed vs. Altitude (cruise to landing) faster.Airspeed vs. altitude graphs compiled from simulation results with corrected BADA files are shown in Figure 30.Plan-view of the flight path is shown in Figure 31.Graphs describing other aspects of the flight are shown in Figure32.It should be noted that Shadow B's cruise altitude and ceiling were assumed to be equal (18000 ft.MSL) in developing the BADA files.However, commercial aircraft usually cruise at a lower altitude than their ceiling. +Figure 26 .26Figure 26.Corrected .APF file for Shadow B (RQ7B).File was compiled by Purdue. +Figure 27 .27Figure 27.Corrected .DCT file for Shadow B (RQ7B).File was compiled by IAI. +Figure 30 .Figure 32 . 9 . 2303292Figure 30.True airspeed (TAS) vs. altitude for Shadow B flight from KIAD to KJFK using corrected BADA files +Figure 33 .Figure 34 .3334Figure 33.True airspeed (TAS) vs. altitude for Global Hawk flight from KMSP to KMCO +Figure 36 .Figure 37 . 9 . 3363793Figure 36.Plan-view of Global Hawk flight path from KMSP to KMCO +Figure 38 .Figure 41 .Figure 43 .384143Figure 38.True airspeed (TAS) vs. altitude for Orbiter flight from KATL to KBHM +Figure 44 .Figure 46 .4446Figure 44.True airspeed (TAS) vs. altitude for Predator A flight from KATL to KJFK +Figure 47 .Figure 49 .4749Figure 47.True airspeed (TAS) vs. altitude for Predator B flight from KMSP to KMCO +Figure 50 .Figure 52 .5052Figure 50.True airspeed (TAS) vs. altitude for Gray Eagle flight from KATL to KJFK +Figure 53 .Figure 55 .5355Figure 53.True airspeed (TAS) vs. altitude for Predator C flight from KMSP to KMCO +Figure 56 .Figure 58 .5658Figure 56.True airspeed (TAS) vs. altitude for Hunter flight from KATL to KJFK +Figure 59 .Figure 60 .5960Figure 59.True airspeed (TAS) vs. altitude for Shadow B flight simulation using MACS from KMSP to KMCO +Figure 61 .Figure 63 .6163Figure 61.True airspeed (TAS) vs. altitude for Global Hawk flight simulation using MACS from KMSP to KMCO +Figure 64 .64Figure 64.Variation in altitude and airspeed (TAS) with time and distance for Predator A flight simulation using MACS 10.6 Simulation of Predator B (MQ-9) using MACS Important features of Predator B's flight simulation using MACS are shown in Table45.It should be noted that, unlike Shadow B and Global Hawk, the flight did not reach cruise altitude and speed.While Predator B flies faster than Predator A, it is slower compared to Global Hawk, and this is a possible reason for the unsuccessful simulation.The flight was simulated from KBNA to KATL to prevent MACS from exceeding its memory usage limits and thereby successfully complete the simulation due to the slow speed of Predator B (relative to Global Hawk).However, as mentioned earlier, it is not known as to why MACS cannot simulate a slow flying aircraft.Consequently, simulation results similar to Figure61were not compiled for Predator B. Other simulation results are shown in Figure65.Similar to Shadow B and Global Hawk, the reason for the sharp fluctuations in speed (Figure65b) is not known. +were not compiled for Predator B. Other simulation results are shown in Figure 65.Similar to Shadow B and Global Hawk, the reason for the sharp fluctuations in speed (Figure 65b) is not known. +Figure 65 .65Figure 65.Variation in altitude and airspeed (TAS) with time and distance for Predator B flight simulation using MACS 10.7 Simulation of Gray Eagle (MQ1C) using MACS Important features of Gray Eagle's flight simulation using MACS are shown in Table46.It should be noted that, similar to Predator B, the flight did not reach cruise altitude and speed, possibly due to its slow speed compared to Global Hawk.The flight was simulated from KBNA to KATL to prevent MACS from exceeding its memory usage limits and thereby successfully complete the simulation due to the slow speed of Gray Eagle (relative to Global Hawk).However, as mentioned earlier, it is not known as to why MACS cannot simulate a slow flying aircraft.Consequently, simulation results similar to Figure61were not compiled for Gray Eagle.Other simulation results are shown in Figure66.Similar to Shadow B and Global Hawk, the reason for the sharp fluctuations in speed (Figure66b) is not known. +Figure 66 .66Figure 66.Variation in altitude and airspeed (TAS) with time and distance for Gray Eagle flight simulation using MACS 10.8 Simulation of Predator C (AVEN) using MACS Important features of Predator C's flight simulation using MACS are shown in Table47.It should be noted that, similar to Predator B, the flight did not reach cruise altitude and speed, possibly due to its slow speed compared to Global Hawk.The flight was simulated from KBNA to KATL to prevent MACS from exceeding its memory usage limits and thereby successfully complete the simulation due to the slow speed of Predator C (relative to Global Hawk).However, as mentioned earlier, it is not known as to why MACS cannot simulate a slow flying aircraft.Consequently, simulation results similar to Figure61were not compiled for Gray Eagle.Other simulation results are shown in Figure67.Similar to Shadow B and Global Hawk, the reason for the sharp fluctuations in speed (Figure67b) is not known. +a.Figure 67 .67Figure 67.Variation in altitude and airspeed (TAS) with time and distance for Predator C flight simulation using MACS 10.9 Simulation of Hunter (MQ5B) using MACS Important features of Hunter's flight simulation using MACS are shown in Table48.It should be noted that, similar to Predator A, the flight did not reach cruise altitude and speed, possibly due to its slow speed compared to Global Hawk.The flight was simulated from KBNA to KATL to prevent MACS from exceeding its memory usage limits and thereby successfully complete the simulation due to the slow speed of Hunter (relative to Global Hawk).However, as mentioned earlier, it is not known as to why MACS cannot simulate a slow flying aircraft.Consequently, simulation results similar to Figure61were not compiled for Hunter.Other simulation results are shown in Figure68.Similar to Shadow B and Global Hawk, the reason for the sharp fluctuations in speed (Figure68b) is not known. +a.Figure 68 .68Figure 68.Variation in altitude and airspeed (TAS) with time and distance for Hunter flight simulation using MACS +Figure 69 .69Figure 69.Different speeds for Global Hawk (RQ4A) in the BADA file "RQ4A__.PTF".The file shown here is a section of the complete file. +Figure 70 .70Figure 70.Different stall speeds for Global Hawk (RQ4A) in the BADA file "RQ4A__.OPF".The file shown here is a section of the complete file. + + + + + + + + + + + + +Table 1 .1Project Summary .........................................................................................................10 +Table 2 .2Specifications and basic attributes of Shadow B (RQ7B) ............................................11 +Table 3 .3Specifications and basic attributes of Global Hawk (RQ4A) .........................................11 +Table 4 .4Specifications and basic attributes of Aerosonde ........................................................11 +Table 5 .5Specifications and basic attributes of Orbiter ...............................................................12 +Table 6 .6Specifications and basic attributes of Cargo UAS ........................................................12 +Table 7 .7Specifications and basic attributes of NEO S-300 Mk II VTOL .....................................12 +Table 8 .8Specifications and basic attributes of Hunter UAS (MQ5B) .........................................12 +Table 9 .9Specifications and basic attributes of Fire Scout .........................................................12 +Table 10 .10Specifications and basic attributes of Predator A .......................................................13 +Table 11 .11Specifications and basic attributes of Predator B .......................................................13 +Table 12 .12Specifications and basic attributes of Gray Eagle ......................................................13 +Table 13 .13Specifications and basic attributes of DHS Avenger/Predator C ................................14 +Table 14 .14Industry data for Shadow B (RQ7B).Provided by AAI...............................................14 +Table 15 .15Airframe drag model substitutions for UAS aircraft.MACS files for Orbiter, Cargo UAS, Fire Scout and NEO S-300 Mk II VTOL were not simulated.............................................24 +Table 16 .16Summary of the actual engines used and the engine decks used in the project to model BADA and MACS for UAS aircraft ..................................................................................37 Table 17.FLOPS sizing results for Shadow B ...........................................................................38Table 18. FLOPS sizing results for Global Hawk .......................................................................40 Table 19.DATCOM-JSBSim sizing results for Orbiter ...............................................................41 Table 20.DATCOM-JSBSim sizing results for Aerosonde ........................................................43 Table 21.FLOPS sizing results for Predator A ..........................................................................44 Table 22.FLOPS sizing results for Predator B ..........................................................................45 Table 23.FLOPS sizing results for Gray Eagle .........................................................................46 Table 24.FLOPS sizing results for Predator C ..........................................................................47 Table 25.FLOPS sizing results for Hunter UAS ........................................................................48 Table 26.DATCOM-JSBSim sizing results for Cargo UAS .......................................................48 Table 27.RPAT sizing results for Fire Scout .............................................................................50 Table 28.RPAT sizing results for NEO S-300 Mk II VTOL ........................................................50 Table 29.Features of Shadow B flight simulation using KTG ....................................................51 Table 30.Stall speeds and corresponding altitude constraints employed by KTG.Stall speeds are Calibrated Airspeeds (CAS) in knots ...................................................................................52 Table 31.Features of Shadow B flight using corrected BADA files............................................55 Table 32.Results of Global Hawk flight simulation using KTG ..................................................57 Table 33.Features of Orbiter flight simulation using KTG .........................................................62 Table 34.Features of Aerosonde flight simulation using KTG ...................................................63 Table 35.Features of Predator A flight simulation using KTG ...................................................65 Table 36.Features of Predator B flight simulation using KTG ...................................................67 Table 37. Features of Gray Eagle flight simulation using KTG ..................................................69 Table 38.Features of Predator C flight simulation using KTG ...................................................71 Table 39.Features of Hunter flight simulation using KTG .........................................................73 Table 40.Summary of nine UAS flights using KTG.Only origin, destination, cruise altitude and cruise speed are included here.Validation of BADA files implies the aircraft reached target cruise altitude in simulation.......................................................................................................76 Table 41.Features of Shadow B flight simulation using MACS .................................................78 Table 42.Features of Global Hawk flight simulation using MACS .............................................80 Table 43.Features of Aerosonde flight simulation using MACS ................................................80 Table 44.Features of Predator A flight simulation using MACS ................................................81 +Table 45 .45Features of Predator B flight simulation using MACS ................................................82 Table 46.Features of Predator A flight simulation using MACS ................................................83 Table 47.Features of Predator C flight simulation using MACS ................................................84 Table 48.Features of Hunter flight simulation using MACS.......................................................85 Table 49.Summary of nine UAS flight simulations in MACS.Only origin, destination, cruise altitude and cruise speed are included here..............................................................................87 Note 2: Information regarding origin/destination airports, route distances, altitude and speed defined in these FDSs is indicated in the simulation summary chart -Table 50.Summary of results for from ACES simulations to test CNS capabilities of UAS aircraft ...............................88 Table 51.Summary of results for from ACES simulations to test CNS capabilities of UAS aircraft .................................................................................................................................................90 Table 52.Industry data for Shadow B (RQ7B).Provided by AAI...............................................96 Table 53.Industry data for Global Hawk (RQ4A).Provided by AAI...........................................98 Table 54.Industry data for Orbiter.Provided by AAI............................................................... 100 Table 55.Industry data for Aerosonde.Provided by AAI......................................................... 103 Table 56.Industry data for Predator A. Provided by General Atomics..................................... 105 Table 57.Industry data for Predator B. Provided by General Atomics..................................... 108 Table 58.Industry data for Gray Eagle.Provided by General Atomics.................................... 112 Table 59.Industry data for Predator C. Provided by AAI......................................................... 115 Table 60.Industry data for Cargo UAS.Provided by AAI........................................................ 117 Table 61.Industry data for Cargo UAS.Provided by AAI........................................................ 120 Table 62.Industry data for Fire Scout.Provided by AAI.......................................................... 123 Table 63.Industry data for NEO S-300 Mk II VTOL.Provided by AAI..................................... 125 Table 64.Flight crossing altitudes in TRACON airspace by aircraft weight and engine type ... 130 Table 65.Flight Calibrated Airspeed (CAS +Table 1 . Project Summary UAS Aircraft Code in BADA and MACS Files Manuf acturer Industry Data Acquired BADA Delivered BADA Verified MACS Delivered MACS Verified1Shadow BRQ7BAAIYesYesYes*YesYesGlobal HawkRQ4AAAIYesYesYesYesYesOrbiterORBMAAIYesYesYes†YesNo †AerosondeMK47AAIYesYesYes*YesNo †Predator AMQ1BGAYesYesYes*YesFail ‡Predator BMQ-9GAYesYesYesYesFail ‡Gray EagleMQ1CGAYesYesYes*YesFail ‡Predator CAVENGAYesYesYes*YesFail ‡Hunter UASMQ5BAAIYesYesYesYesFail ‡Cargo UASCUASAAIYesYesNo #YesFail ‡Fire ScoutMQ8BAAIYesYesNo #YesNo ##NEO S-300 Mk II VTOLS350AAIYesYesNo #YesNo ##* Aircraft performance altered by BADA stall speed constraints † Aircraft engine profile issues-electric aircraft ‡ Failed to reach designated cruise altitude # Cannot simulate rotorcraft in KTG ## Cannot simulate rotorcraft in MACS +2 Specifications and Basic Attributes of UAS Aircraft 2.1 Eight Aircraft from AAI +Table 2 . Specifications and basic attributes of Shadow B (RQ7B)2Length (ft.)11.2Wingspan (ft.)14.0Max. gross weight (lb.)375Range (nmi.)685 for air aircraft; 27 for controlEndurance (hr.)9Max. altitude (ft.)15000Communication capabilitiesPrimary & secondary datalink, TDMANavigation modesAuto-launch, auto-pilot (altitude, airspeed & heading), fly-to-location, auto-land, flight termination (parachute)SurveillanceATC transponderExample civilian applicationsSurveillance: fuel pipelines, power lines, ports & harbors, and law enforcement +Table 3 . Specifications and basic attributes of Global Hawk (RQ4A)3Length (ft.)44.4Wingspan (ft.)116.2Max. gross weight (lb.)26700Range (nmi.)12000Endurance (hr.)35Max. altitude (ft.)65000Communication capabilitiesKu SATCOM datalink, CDL line-of-sight, UHF SATCOM/LOS, and ATC voiceSurveillanceSynthetic aperture radar, EO NIIRS 6.0, IR NIIRS 5.0Example civilian applicationsAtmospheric research, forest fire monitoring and support, and natural hazard monitoring +Table 4 . Specifications and basic attributes of Aerosonde4Length (ft.)6.9Wingspan (ft.)11.8Max. gross weight (lb.)30Range (nmi.)608Endurance (hr.)10Max. altitude (ft.)15000Communication capabilitiesPrimary & secondary + independent imagery datalinkNavigation modesCloudcap avionics suiteSurveillanceMode 3 IFF transponderExample civilian applications Land survey, ice monitoring, and climate change support +Table 5 . Specifications and basic attributes of Orbiter5Length (ft.)3.2Wingspan (ft.)7.2Max. gross weight (lb.)14.3Range (nmi.)27Endurance (hr.)2-3Max. altitude (ft.)18000Communication capabilitiesOne data uplink and one data downlink channelNavigation modesUMAS avionics for flight control, stabilization, mission control, and payload controlSurveillanceExample civilian applicationsSWAT team monitoring, covert law enforcement and monitoring, and agriculture/animal monitoring +Table 6 . Specifications and basic attributes of Cargo UAS6Length (ft.)38.0 (rotor)Wingspan (ft.)38.0 (wing) -hybridMax. gross weight (lb.)7250Range (nmi.)2800-5500 (based on cargo)Endurance (hr.)Up to 20Max. altitude (ft.)35000Navigation modesAuto-takeoff, auto-land, waypoint, electronic tethering, and auto-trackingExample civilian applications Cargo transport +Table 7 . Specifications and basic attributes of NEO S-300 Mk II VTOL7Length (ft.)LxWxH: 9.0 x 3.1 x 2.8; rotor diameter: 9.8Wingspan (ft.)N/A (rotorcraft)Max. gross weight (lb.)176Range (nmi.)87Endurance (hr.)2Max. altitude (ft.)10000Communication capabilitiesRF Line-of-sight, dedicated datalink for payloadNavigation modesAuto-takeoff, auto-land, waypoint, electronic tethering, and auto-trackingSurveillanceEO/IRExample civilian applications Law enforcement, and search & rescue +Table 8 . Specifications and basic attributes of Hunter UAS (MQ5B)8Length (ft.)23.0Wingspan (ft.)34.25Max. gross weight (lb.)1800Range (nmi.)144Endurance (hr.)21Max. altitude (ft.)22000Communication capabilitiesLDS datalink, UAV airborne relay, and voiceSurveillanceEO/IRExample civilian applicationsSurveillance: fuel pipelines, power lines, ports & harbors, and law enforcement +Table 9 . Specifications and basic attributes of Fire Scout9Length (ft.)23.9 (length); 27.5 (rotor); 4.42 (height)Wingspan (ft.)N/A (rotorcraft)Max. gross weight (lb.)3150 +.2 Four Aircraft from General AtomicsGray Eagle and DHS Avenger/Predator C. Important aircraft specifications and basic attributes of these aircraft are shown in Table10, Table11, Table12and Table13.Manufacturer data for four aircraft were provided by General Atomics (GA): Predator A,Predator B, +Table 10 . Specifications and basic attributes of Predator A10Length (ft.)27.0Wingspan (ft.)55.0Max. gross weight (lb.)2250Range (nmi.)4800Endurance (hr.)40Max. altitude (ft.)25000Communication capabilitiesC-Band line-of-sight, Ku-Band over-the-horizon SATCOM, UHF/VHF voice, communications relayNavigation modesFully autonomousSurveillanceMTS-A EO/IR, Lynx Multi-mode radar, SIGINT/ESM systemExample civilian applicationsCrop and cattle monitoring, ice passage monitoring, national disaster support, and airborne pollution observation +Table 11 . Specifications and basic attributes of Predator B11Length (ft.)36.0Wingspan (ft.)66.0Max. gross weight (lb.)10000Range (nmi.)5700Endurance (hr.)30Max. altitude (ft.)50000Communication capabilitiesC-Band line-of-sight data link control, Ku-Band beyond line-of-sight/SATCOM data link control, communications relayNavigation modesFully autonomousSurveillanceMTS-B EO/IR, Lynx Multi-mode Radar, Multi-mode maritime radar, SIGINT/ESM systemBorder patrol, search and rescue, maritime surveillance, aerialExample civilian applicationsimaging and mapping, and chemical and petroleum spillmonitoring +Table 12 . Specifications and basic attributes of Gray Eagle12Length (ft.)28Wingspan (ft.)56Max. gross weight (lb.)3600Range (nmi.)200Endurance (hr.)30Max. altitude (ft.)29000Communication capabilitiesTCDL line-of-sight satellite communication, TCDL air data relay communications, over-the-horizon Ku-Band SATCOMNavigation modesAuto-takeoff and landing +Table 13 . Specifications and basic attributes of DHS Avenger/Predator C13Length (ft.)44.0Wingspan (ft.)66.0Max. gross weight (lb.)15800Range (nmi.)Endurance (hr.)18Max. altitude (ft.)50000Communication capabilitiesCommunication relayNavigation modesSurveillanceEO/IR, Lynx Multi-mode Radar, SIGINT/ESM SystemEnvironmental monitoring and mapping, in-situ atmosphericExample civilian applicationsresearch, sea-ice observations, crop monitoring, TV signaltransmission, and cell phone signal platform +Table 14 . Industry data for Shadow B (RQ7B). Provided by AAI. Operations Performance Files (OPF)14Design Range685 nmi.Design Endurance9 hr.Basic GeometryWing Aspect Ratio11.1Wing span19.8 ft.Wing taper0.7Fuselage length63.1 in.Fuselage fineness0.181Tail sizeTail Volume Coefficient0.65% (horizontal volume coefficient)Drag PolarsEquation or GraphC D = 0.0497 + C L2 /(pi*0.9*AR) → Wing drag polarMassMax. mass of aircraft333 lb. (Aircraft without fuel. Pop 300 installed) +Table 15 . Airframe drag model substitutions for UAS aircraft. MACS files for Orbiter, Cargo UAS, Fire Scout and NEO S-300 Mk II VTOL were not simulated.15UAS AircraftSubstitution AircraftShadow BCessna 172Global HawkNo substitutionAerosondeNo substitutionOrbiterNot simulatedPredator ACessna 172Predator BNo substitutionGray EagleCessna 172Predator CNo substitutionHunterCessna 172Cargo UASNot simulatedFire ScoutNot simulatedNEO S-300 Mk II VTOLNot simulated +Table 16 . Summary of the actual engines used and the engine decks used in the project to model BADA and MACS for UAS aircraft Aircraft (Engine Type)16Engine NameBADA ModelMACS ModelCommentsShadow B (Piston)UEL 741AR74-1102FLOPS internal piston engineO-320-H2ADEngine data from manufacturers were used to change parameters in FLOPS. No changes made to the MACS modelGlobal Hawk (Jet)Rolls-Royce F137-AD-100AE3007PW_JT8D-07AE3007 mimicked the RR F137 parameters provided by manufacturersPredator A (Piston)Not givenFLOPS internal piston engineO-320-H2ADLack of higher granularity engine data resulted in faulty climb rates and fuel flow ratesPredator B (Turboprop)Honeywell TPE331-10YGDFlops internal turbopropPW118Better thrust model provided by manufacturers used to alter the FLOPS model. Awaiting validationGray Eagle (Piston)Thielert Centurion 2.0L HFEFlops internal piston enginePW_PT6A-34Indicative measures given by manufacturers used to alter FLOPS piston deck. Awaiting validationAvenger (Jet)Pratt & Whitney 545BAE3007PW_JT8D-07Lack of higher granularity engine data resulted in faulty climb rates and fuel flow rates. AE3007 is not suitableHunter UAS (Piston)APL HFEFlops internal piston engineO-320-H2ADIndicative measures given by manufacturers used to alter +Table 17 . FLOPS sizing results for Shadow B17Shadow-BIndustry data from AAI DataFLOPSOperating Empty Weight333 lb.412 lb.Payload Weight60 lb.60 lb.Gross Weight467 lb.593 lb.Max. Operating Mach No.0.1970.225Max. Cruise Speed136 KTAS100 KTASCruise Altitude8000 ft.8000 ft.Reference Wing Area35.41 ft.2 39.22 ft.2 Max.Thrust at Cruise Unknown 287.2 lb. +Table 18 . FLOPS sizing results for Global Hawk18Global HawkIndustry Data from AAIFLOPSOperating Empty Weight9200 lb.9500 lb.Payload Weight2000 lb.2000 lb.Gross Weight26700 lb.27200 lb.Max. Operating Mach No.Unknown0.65Max. Cruise Speed400 KTAS (estimated)343 KTASCruise Altitude31000 ft.31000 ft.Reference Wing Area551.3 ft. 2570.3 ft2Max. Thrust at Cruise7059 lb.7600 lb. +Table 19 . DATCOM-JSBSim sizing results for Orbiter19OrbiterIndustry data from AAIDATCOM-JSBSimOperating Empty Weight12.13 lb.12.13 lb.Payload Weight2.9 lb.2.9 lb.Gross Weight16.5 lb.16.5 lb.Max. Operating Mach No.Unknown0.11Max. Cruise Speed70 KTAS45 KTASCruise Altitude8000 ft.8000 ft.Reference Wing Area8.8 ft. +Table 20 . DATCOM-JSBSim sizing results for Aerosonde20AerosondeIndustry data from AAIDATCOM-JSBSimOperating Empty Weight48.9 lb.48.9 lb.Payload Weight13.3 lb.13.3 lb.Gross Weight75 lb.75 lb.Max. Operating Mach No.Unknown0.12Max. Cruise Speed65 KTAS61 KTASCruise Altitude15000 ft.15000 ft.Reference Wing Area9.67 ft.2 9.67 ft. 2 Max.Thrust at Cruise 4.9 lb.(estimate) 12.4 lb. +Table 21 . FLOPS sizing results for Predator A21Predator AIndustry data from GAFLOPSOperating Empty Weight1665 lb.1745 lb.Payload Weight450 lb.450 lb.Gross Weight2250 lb.2770 lb.Max. Operating Mach No.Unknown0.24Max. Cruise Speed120 KTAS111 KTASCruise Altitude16000 ft.16000 ft.Reference Wing Area132 ft. 2143 ft. 2Max. Thrust at Cruise140 lb.330 lb. +Table 22 . FLOPS sizing results for Predator B22Predator BIndustry data from GAFLOPSOperating Empty Weight4900 lb.4823 lb.Payload Weight4800 lb.4800 lb.Gross Weight10500 lb.10462 lb.Max. Operating Mach No.0.380.38Max. Cruise Speed160 KTAS209 KTASCruise Altitude31000 ft.31000 ft.Reference Wing Area256 ft. 2251 ft. 2Max. Thrust at CruiseUnknown1680 lb. +Table 24 . FLOPS sizing results for Predator C24Predator CIndustry data from GAFLOPSOperating Empty Weight8650 lb.8545 lb.Payload Weight6500 lb.6000 lb.Gross Weight15800 lb.14920 lb.Max. Operating Mach No.0.620.62Max. Cruise Speed400 KTAS331 KTASCruise Altitude40000 ft.40000 ft.Reference Wing Area267 ft. 2243 ft. 2Max. Thrust at Cruise1000 lb.1220 lb. +Table 25 . FLOPS sizing results for Hunter UAS25Hunter UASIndustry data from AAIFLOPSOperating Empty Weight1450 lb.1510 lb.Payload Weight630 lb.650 lb.Gross Weight1950 lb.2090 lb.Max. Operating Mach No.Unknown0.2Max. Cruise Speed120 KTAS119 KTASCruise Altitude18000 ft.18000 ft.Reference Wing Area106 ft. 2111 ft. 2Max. Thrust at CruiseUnknown300 lb. +Table 26 . DATCOM-JSBSim sizing results for Cargo UAS26Cargo UASIndustry data from AAIDATCOM-JSBSimOperating Empty Weight12050 lb.12050 lb.Payload Weight8000 lb.8000 lb.Gross Weight22750 lb.22750 lb.Max. Operating Mach No.Unknown0.40Max. Cruise Speed250 KTAS270 KTAS +Table 27 . RPAT sizing results for Fire Scout27Fire ScoutIndustry data from AAIRPATOperating Empty Weight1457 lb.1510 lb.Payload Weight600 lb.600 lb.Gross Weight3150 lb.3234 lb.Max. Operating Mach No.Unknown0.22Max. Cruise Speed125 KTAS128 KTASCruise Altitude20000 ft.20000 ft.Fuselage Wet Surface Area286 ft. 2291 ft.2 +Table 28 . RPAT sizing results for NEO S-300 Mk II VTOL NEO S-300 Mk II VTOL Industry data from AAI RPAT28Operating Empty Weight187.4 lb.222.6 lb.Payload Weight99.2 lb.99.2 lb.Gross Weight330.7 lb.387.8 lb.2 +Table 29 . Features of Shadow B flight simulation using KTG29OriginKIADDestinationKJFKCruise speed93 KTASCruise altitude8000 ft.Total flight time138 min.Total flight distance201 nmi. +Table 30 . Stall speeds and corresponding altitude constraints employed by KTG. Stall speeds are Calibrated Airspeeds (CAS) in knots Flight phase Altitude constraint Stall speed in .OPF file Buffer factor30Climb< 400 ft.TO1.2400 ft. to 2000 ft.IC1.3> 2000 ft.CR1.3Top of climbNot applicableCR1.3CruiseNot applicableCR1.3Descent≥ 8000 ft.CR1.33000 ft. to 8000 ft.AP1.3< 3000 ft.LD1.3LandingNot applicableLD1.3 +Table 32 . Results of Global Hawk flight simulation using KTG32OriginKMSPDestinationKMCOFlight time217.9 min.Flight distance1167.7 nmi.Cruise speed343 KTASCruise altitude31000 ft.Takeoff mass14203 kgLanding mass10774.93 kgDuration of climb13.9 min.Duration of cruise179.8 min.Duration of descent23.7 min.Duration of landing0.5 min. +Table 33 . Features of Orbiter flight simulation using KTG33OriginKATLDestinationKBHMFlight time177.6 min.Flight distance117.4 nmi.Cruise speed39 KTASCruise altitude8000 ft.Takeoff mass7.5 kgLanding mass7.496 kgDuration of climb8.3 min.Duration of cruise149.9 min.Duration of descent15.8 min.Duration of landing3.6 min. +Table 35 . Features of Predator A flight simulation using KTG35OriginKATLDestinationKJFKFlight time395 min.Flight distance715.3 nmi.Cruise speed111 KTASCruise altitude16000 ft.Takeoff weight1020.5 kgLanding weight870.7 kgDuration of climb12.9 min.Duration of cruise354 min.Duration of descent26.3 min.Duration of landing1.7 min. +Table 36 . Features of Predator B flight simulation using KTG36OriginKMSPDestinationKMCOFlight time350.4 minFlight distance1167.6 nmi.Cruise speed209 KTASCruise altitude31000 ft.Takeoff weight3734.6 kgLanding weight3072.28 kgDuration of climb23.8 min.Duration of cruise292 min.Duration of descent31.7 min.Duration of landing2.9 min. +Table 37 . Features of Gray Eagle flight simulation using KTG37OriginKATLDestinationKJFKFlight time234.4 min.Flight distance714.9 nmi.Cruise speed203 KTASCruise altitude32000 ft.Takeoff weight1620.2 kgLanding weight1542 kgDuration of climb45 min.Duration of cruise136 min.Duration of descent53.8 min.Duration of landing1.5 min. +Table 38 . Features of Predator C flight simulation using KTG38OriginKMSPDestinationKMCOFlight time230.5 min.Flight distance1168.6 nmi.Cruise speed331 KTASCruise altitude40000 ft.Takeoff weight7166.7 kbLanding weight4951 kgDuration of climb19.8 min.Duration of cruise174.8 min.Duration of descent35.1 min.Duration of landing0.8 min. +Table 39 . Features of Hunter flight simulation using KTG39OriginKATLDestinationKJFKFlight time372.7 min.Flight distance715.2 nmi.Cruise speed119 KTASCruise altitude18000 ft.Takeoff weight907.2 kgLanding weight792.28 kgDuration of climb21.2 min.Duration of cruise306.6 min.Duration of descent40 min.Duration of landing4.7 min. +Table 40 . Summary of nine UAS flights using KTG. Only origin, destination, cruise altitude and cruise speed are included here. Validation of BADA files implies the aircraft reached target cruise altitude in simulation.40UASOrigin DestinationTarget Cruise Altitude (ft.)Target Cruise Speed (KTAS)Reached Target Cruise Altitude & SpeedBADA files validated by manufacturerShadow B (RQ7B) KIADKJFK800080YesYesGlobal Hawk (RQ4A)KMSPKMCO31000343YesYesOrbiter (ORBM)KATLKBHM800039YesYesAerosonde (MK47)KATLKBHM800049YesYesPredator A (MQ1B)KATLKJFK16000111YesYesPredator B (MQ-9) KMSPKMCO31000209YesYesGray Eagle (MQ1C)KATLKJFK32000203YesYesPredator C (AVEN)KMSPKMCO40000331YesYesHunter (MQ5B)KATLKJFK18000119YesYesCargo UAS(CUAS)Fire Scout (MQ8B)Rotorcraft cannot be simulated in KTG. Hence, BADA files not validated.NEO S-300 Mk IIVTOL (S350) +Table 41 . Features of Shadow B flight simulation using MACS41OriginKBNADestinationKATLFlight time149.4 min.Flight distance191.5 nmi.Cruise speed86 KTASCruise altitude8000 ft.Takeoff weightNot availableLanding weightNot availableDuration of climb3.6 min.Duration of cruise127.2 min.Duration of descent12 min.Duration of landing6 min. +Table 44 . Features of Predator A flight simulation using MACS44OriginKBNADestinationKATLFlight time100.8 min.Flight distance185.1 nmi.Cruise speed117 KTASCruise altitude16000 ft.Takeoff weightNot availableLanding weightNot availableDuration of climb69.4 min. +Table 45 . Features of Predator B flight simulation using MACS45OriginKBNADestinationKATLFlight time87.7 min.Flight distance181.5 nmi.Cruise speed209 KTASCruise altitude31000 ft.Takeoff weightNot availableLanding weightNot availableDuration of climb31.12 min. +Table 46 . Features of Predator A flight simulation using MACS46OriginKBNADestinationKATLFlight time97.84 min.Flight distance181.9 nmi.Cruise speed203 KTASCruise altitude32000 ft.Takeoff weightNot availableLanding weightNot availableDuration of climb23.47 min.Duration of cruiseDid not reach Cruise +Table 47 . Features of Predator C flight simulation using MACS47OriginKBNADestinationKATLFlight time75.8 min.Flight distance182.1 nmi.Cruise speed331 KTASCruise altitude40000 ft.Takeoff weightNot availableLanding weightNot availableDuration of climb33.1 min.Duration of cruiseDid not reach CruiseDuration of descent29.7 min.Duration of landing12.6 min. +Table 48 . Features of Hunter flight simulation using MACS48OriginKBNADestinationKATLFlight time123.9 min.Flight distance191 nmi.Cruise speed119 KTASCruise altitude18000 ft.Takeoff weightNot availableLanding weightNot availableDuration of climb81.2 min.Duration of cruiseDid not reach CruiseDuration of descent17.4 min.Duration of landing24.82 min. +Table 49 . Summary of nine UAS flight simulations in MACS. Only origin, destination, cruise altitude and cruise speed are included here.49UASOrigin DestinationTarget Cruise Altitude (ft.)Target Cruise Speed (KTAS)Target Cruise Altitude & Reached? SpeedSimilar to KTG? += 0.55 lb./hp-hr.at Sea Level; Static @ 100% RPM BSFC max = 0.55 lb./hp-hr.at Sea Level; Static @ 100% RPM BSFC cruise = 0.55 lb./hp-hr.at Sea Level; Static @ 100% RPMMax. payload Max. fuel weight Flight Envelope Loiter Speed V MO (in CAS or TAS) M MO (Mach Max. Operating) H max Aerodynamics S wet (Total) S wet (Fuselage) S ref Clb. o (Buffet Onset Lift Coeff.) 1.6 13.3 lb. 45 KIAS 65 KTAS N/A 15423 ft. DA (Service ceiling) 5718 in. 2 1642 in. 2 1392 in. 2 Stall Speed (Initial Climb) 35 KIAS Stall Speed (Cruise) 35 KIAS Stall Speed (Take Off) 35 KIAS Stall Speed (Landing) 35 KIAS Stall Speed (Approach) 35 KIAS Engine Thrust Max. Thrust at Climb vs. Height N/A Max. Thrust at Cruise 4.9 lb. (estimated) Max. Thrust at Descent N/A Propulsion Engine 75 HFDI Engine (heavy fuel direct inject JP5/Jp8) Brake Engine Power 4 hp No. of cylinders 1 Baseline Engine Power 6 (derated to 4) hp Critical Turbocharger Altitude N/A Fuel Consumption BSFC min = 0.05247 gal./hr. (estimated) BSFC max = 0.8767 lb./hp-hr. (estimated) BSFC cruise = 0.5973 lb./hp-hr. (estimated) Maximum Engine Crankshaft Speed ECU Limited to 5750 RPM Maximum Propeller Shaft Speed ECU Limited to 5750 RPM Engine displacement 75 cc Equation or Graph Mass Max. mass of aircraft (Empty Weight) 4900 lb. Max. mass of aircraft (Gross Weight) 10500 lb. Max. payload 4800 lb. Max. fuel weight 3764 lb. Flight Envelope Loiter Speed V MO (in CAS or TAS) 230 KIAS or 249 KTAS M MO (Mach Max. Operating) 0.38 H max 30000 ft. at MTOW under ISA conditions Aerodynamics S wet (Total) 529.41 ft. 2 S wet (Fuselage) S ref 256 ft. Stall Speed (Initial Climb) 100 KTAS @ 5000 ft. (MGTOW) Stall Speed (Cruise) 110 KTAS @ 20000 ft. (8000 lb.) Stall Speed (Take Off) 93 KTAS @ Sea Level (MGTOW) Stall Speed (Landing) 70 KTAS @ sea Level (6000 lb.) Stall Speed (Approach) 75 KTAS @ 5000 ft. (6000 lb.) Engine Thrust Max. Thrust at Climb vs. Height Proprietary Max. Thrust at Cruise Proprietary Max. Thrust at Descent Proprietary Propulsion Engine Brake Engine Power N/A No. of cylinders N/A Baseline Engine Power 900 hp @ 100% RPM Critical Turbocharger Altitude N/A Fuel Consumption Max. mass of aircraft (Empty Weight) 8650 lb. Max. mass of aircraft (Gross Weight) 15800 lb. Max. payload 3500 lb. (external); 3000 lb. (internal) Max. fuel weight 9000 lb. Flight Envelope Aerodynamics S wet (Total) 555.4 ft. 2 S wet (Fuselage) S ref 267 ft. Stall Speed (Initial Climb) 105 KIAS Stall Speed (Cruise) 112 KIAS Stall Speed (Take Off) 105 KIAS Stall Speed (Landing) 98 KIAS Stall Speed (Approach) 98 KIAS Engine Thrust Max. Thrust at Climb vs. Height Proprietary Max. Thrust at Cruise 1000 lb. Max. Thrust at Descent Proprietary Propulsion Engine Brake Engine Power No. of cylinders N/A Baseline Engine Power N/A Critical Turbocharger Altitude N/A Fuel Consumption BSFC min = Proprietary BSFC max = Proprietary BSFC cruise = Proprietary Integrated design lift coefficient (for blade) Ground Movement Landing Length 2200 ft. Take Off Length 1275 ft. Width of Runway 100 ft. Aircraft Length 23 ft. BSFC Mass Loiter Speed V MO (in CAS or TAS) Airline Procedures File (APF) 400 KTAS M MO (Mach Max. Operating) Climb Operating Speed 60-80 KIAS 0.62 H max Cruise Operating Speed 80 KIAS 40000 ft. at MTOW under ISA conditions Descent Operating Speed 60-80 KIASEngine compression ratio Engine Envelope Maximum Engine Crankshaft Speed Maximum Propeller Shaft Speed8.9:1 N/A X = 222 mm Y = 336 mm N/A2 Clb.o (Buffet Onset Lift Coeff.)1.23 (no flaps); 1.51 (30° flaps) min 2 Clb.o (Buffet Onset Lift Coeff.)1.26 (Maximum CL with no flaps) 15.10 Cargo UAS (CUAS) +Table 61 . Industry data for Cargo UAS. Provided by AAI.61Basic GeometryOperations Performance File (OPF) Design Range 600 nmi.(with 20 min.reserve) Design Endurance 2.16 hr.(with full cargo load) +MPAS_SYNONYM.LST: No changes are necessary SYNONYM_ALL.LST (a single continuous line):CD -RQ4A Global Hawk UAVNorthrop Grumman RQ4A__RQ4ARQ4ARQ4ARQ4ARQ4ARQ4ARQ4ARQ4ARQ4ARQ4ASYNONYM_ACES_KTG.OLD (a single continuous line):CD * RQ4ANORTHROPGLOBAL HAWK UAVRQ4A__ RQ4A / +Table 64 . Flight crossing altitudes in TRACON airspace by aircraft weight and engine type64Nominal CAS data were derived from the following sources (except as otherwise footnoted): Shen, M.M and Hunter, C.G. "Time to Fly in the DFW Tracon", Seagull TM 92120-03, November, 1992.Shen, M.M., Hunter, C.G. and Sorensen, J.A., "Analysis of Final Approach Spacing Requirements Part II", Seagull TM 92120-02, February, 1992.Hunter, C.G., "Aircraft Flight Dynamics in the Memphis TRACON", Seagull TM 92120-01, January, 1992.Dorsky, S. and Hunter, C.G., "Time to Fly in the Boston TRACON", Seagull TM 91120-01, May, 1991.2.Surrogate CAS data: Same as 2JS3.Surrogate CAS data: Same as 2JL 4. Surrogate CAS data: Same as 3JL 5. Surrogate CAS data: Same as 2TS 6. Surrogate CAS data: Same as 1PS 7. Surrogate CAS data: Same as 4PL 8. Surrogate CAS data: Same as 2PS Following the procedure described earlier, the different speeds of Global Hawk for inclusion in ACES aircraft database yields the values in Table 66.No. of EnginesEngine TypeAircraft Weight CategoryRwy Takeoff Threshold (ft.)Rwy Landing Threshold (ft.)Final Approach Fix (ft.)Cruise Fix (ft.)Arrival Fix (ft.)Departure Fix (ft.)JS00200020001000010000JL00200020001000010000JH00200020001000010000JS00200020001000010000JL00200020001000010000JH00200020001000010000JS00200020001000010000JL00200020001000010000JH00200020001000010000JS00200020001000010000JL00200020001000010000JH00200020001000010000TS001500150080008000TL001500150080008000TH001500150080008000TS001500150080008000TL001500150080008000TH001500150080008000TS001500150080008000TL00150015008000 +Table 66 . Different speed settings of Global Hawk for inclusion in ACES aircraft database. Speeds are Calibrated Airspeed in knots (KCAS).66AIRC RAFT _CHA RACT ERIST ICS_D S_IDAIRC RAF T_TY PE_ CAT EGO RYAIRC RAFT _TYP EEN GIN E_ TY PESEPA RATI ON_C ATEG ORYFINA L_A PPR OAC H_FI XDE PA RT UR E_ FIXAR RIV AL_ FIXCR UIS E_ FIXRUNW AY_LA NDING _THRE SHOL DRUNW AY_TA KEOF F_THR ESHO LDTAK EOF F_ST ALL_ SPE EDLAN DIN G_S TAL L_S PEE D14073748851T/M RQ4ATM123143 155 1551151158376.785931 + + + + +AcknowledgmentsThe research represented in this study was funded by the National Aeronautics and Space Administration under contract NND11AQ74C.We would like to thank the Technical Monitor, Maria Consiglio as well as the project COTR, Eric Mueller, for their technical guidance during the project.The authors acknowledge extremely valuable feedback, testing & analysis suggestions from members of the MACS simulation community as well as the NASA UAS 14 References + + + +MACSMACS performance files were generated by mapping the BADA files.The following MACS drag model and engine thrust model were used respectively for Predator B: MQ-9 (created externally and added into the database) and PW118. +Gray EagleGray Eagle is a small-scale, fixed-wing aircraft equipped with a Thielert Centurion 2.0L heavy fuel engine.The aircraft cruises at an altitude of 32000 ft., with maximum altitude also at 32000 ft. and weighs approximately 3600 lb.The BADA model of Gray Eagle was developed using data provided by GA.FLOPS piston engine deck was generated using engine data provided by the manufacturer.A comparison of sizing results from FLOPS and manufacturer provided data is shown in Table 23.FLOPS generated values for the drag polar, speed schedules, climb rates and fuel flow are used in the MATLAB-based BADA model to generate BADA specific coefficients.These coefficients are further used to generate the .PTF file for Gray Eagle.During BADA production it was identified that the cruise, climb and descent TAS of Gray Eagle were over-predicted by the BADA model due to the stall speed buffer condition employed in BADA.Simulation tools compatible with BADA also apply this limit, making it a hard constraint on the aircraft.Additional discrepancies, if any, are currently being investigated by the manufacturers. +Summary of BADA deficiencies and limitationsBADA deficiencies: None BADA limitations: Stall speed buffer constraints in BADA overshoot the speed of Predator B in cruise, climb and descent.Manufacturer reported cruise speed at 24000 ft. is 140 kts while BADA constraint sets the speed at 177 kts.Further limitations can be identified only after complete validation of the aircraft +MACSMACS performance files were generated by mapping the BADA files.The following MACS drag model and engine thrust model were used respectively for Gray Eagle: C172 and PW_PT6A-34.The method getMinimumSpeed() is invoked by the method getVerticalSpeed() in the file calculators/AltChgCalculator.java to determine the vertical speed at climb.The following is the logic which returns a value of zero for vertical speed: +Issue 2: Simulation of Slow Flying UAS AircraftIt was found that simulation of slow flying UAS aircraft, such as Shadow B and Predator A, in MACS required large computer memory.For example, during the simulation of Predator A from KMSP to KMCO (about 1160 nmi.) at a cruise speed of 93 KCAS and cruise altitude of 16000 ft., resulted in the software's memory usage exceeding 4 GB and crashed the Java Virtual Machine (JVM) since the MACS JVM's maximum memory was set to 4 GB.As a result, the flight was modified to fly from Nashville International Airport (KBNA) to KATL, which are much closer to each other (about 190 nmi.).Even with this short distance, MACS required about 2.5 GB of memory.This issue was also observed when simulating Global Hawk.However, Global Hawk has a higher cruise speed (225 KCAS) compared to Predator A, and MACS was able to complete the simulation before exceeding its memory limits.It is not known as to why MACS cannot successfully simulate a slow flying aircraft, or what modifications are necessary to solve this issue.Therefore, no immediate solution was found to address this issue.Equipage files were created for each UAS aircraft to enable the flights to use each of the Comm., Nav. and Surv.models as onboard and integrated systems.The exceptions to this were: 1) since VOR/DME navigation system will most likely never be used for UAS (currently GPS is the standard due to its technology advantages) no equipage files were created, and therefore, no flights were simulated using VOR/DME model, and 2) it was decided that for the two smallest UAS aircraft (Orbiter and Aerosonde) the equipment size for use of VHF Radios would to be restrictive to ever expect them to be operated onboard those aircraft, and therefore no equipage files were generated (nor flights flown) for these UAS with Voice VHF. +Simulation of Aerosonde (MK47) using MACS +Simulations: Tabulated ResultsForty three simulations were conducted using each of the different Comm., Nav. and Surv.models (mentioned earlier) configured as onboard systems.The results of the simulations are shown in Table 51.Included are data in the FDSs for the simulations such as aircraft names, the cruise speed and altitude, the distance of the flight route, and the origin and destination airports.Also identified are the results of the simulation and a comment column that briefly defines the information that was checked in the output data to verify successful operation of the tested CNS model.The UAS aircraft studied in this project were simulated in KTG.Further, their communication, navigation and surveillance capabilities were simulated in ACES using KTG as the trajectory generator.As mentioned earlier, Cargo UAS, Fire Scout and NEO S-300 Mk II VTOL could not be simulated using ACES and KTG, and hence were excluded from these simulations.This section describes the procedure used to configure KTG and ACES databases to simulate these nine UAS aircraft. +Configuration of KTG DatabaseThe four BADA files described and presented earlier for each UAS aircraft were added to the KTG database folder in ACES: TrajectoryGenerators\ktg\core\data.In addition, the following KTG files were configured to support UAS simulations:The different entries in this table are: AIRCRAFT_CHARACTERISTICS_DS_ID: This is unique number assigned to each aircraft.The simplest way to assign this number would be to continue the sequence in the table. AIRCRAFT_TYPE_CATEGORY: It specifies the number of engines and type, and the aircraft weight category.J = Jet, T = Turboprop, P = Piston.S (small) = up to 12,500 lb.; M (medium) = 12,500 to 41,000 lb.; L (large jet) = 41,000 to 255,000 lb.; H (heavy) = more than 255,000 lb.For example, the Global Hawk has one turbofan engine and belongs to the "M (medium)" weight category.Since ACES does not support the "turbofan" engine type, turboprop (T) was used for Global Hawk.Hence, its entry in this field is "1T/M". AIRCRAFT_TYPE: Aircraft code; RQ4A for Global Hawk. ENGINE_TYPE: J = Jet, T = Turboprop, P = Piston. SEPARATION_CATEGORY: S (small) = up to 12,500 lb.; M (medium) = 12,500 to 41,000 lb.; L (large jet) = 41,000 to 255,000 lb.; H (heavy) = more than 255,000 lb. FINAL_APPROACH_FIX: Speed at final approach fix in KCAS.The values for conventional aircraft are obtained from Table 64 and , based on engine type, number of engines and aircraft weight class.However, due to the large variation in the actual weight of UAS aircraft for the same weight and engine categories, speed corresponding to final-approach-fix's altitude from the .PTF BADA file was used.For example, the Global Hawk belongs to the weight category M and has engine type T, resulting in an altitude of 1500 ft. for its finalapproach-fix.From the .PTF file, this altitude corresponds to 125 KTAS during descent (green-box in Figure 69), since final-approach corresponds to the descent phase of flight.It should be noted that the speeds in .PTF file are in KTAS.Therefore, these were converted to KCAS. DEPARTURE_FIX: Same procedure as above, but using the speed corresponding to departure-fix altitude for a given weight category and engine type in Table 64.For Global Hawk, this is 161 KTAS at 8000 ft.during the climb phase (blue-box in Figure 69). ARRIVAL_FIX: Same procedure as DEPARTURE_FIX.For Global Hawk, this is 174 KTAS at 8000 ft.during the descent phase (blue-dotted-box in Figure 69). CRUISE_FIX: Same value as ARRIVAL_FIX. RUNWAY_LANDING_THRESHOLD: Same procedure as above but corresponding to descent speed at FL0 (Flight Level 0) in the .PTF BADA file.For Global Hawk, this is 115 KCAS (orange-box in Figure 69).It should be noted that at FL0, KTAS and KCAS are equivalent. RUNWAY_TAKEOFF_THRESHOLD: Same procedure as above, but using climb speed at FL0.For Global Hawk, this is 115 KCAS (orange-dotted-box in Figure 69). TAKEOFF_STALL_SPEED: This is indicated in the .OPF BADA file.For Global Hawk, this is 83 KCAS (highlighted with red-box in Figure 70). LANDING_STALL_SPEED: Same procedure as above.For Global Hawk, this is 76.7 KCAS (highlighted with orange-box in Figure 70). + + + + + + + Bio-inspired Predator and Anti-predator Mechanisms for Unmanned Aerial Systems + + Predator B ................................................................................................................. + + 10.2514/6.2022-0274.vid + + + American Institute of Aeronautics and Astronautics (AIAA) + + + Predator B .................................................................................................................. + + + + + Summary of Reported Deficiencies + + KlausRother + + ................................................... + + + 10.1159/000318546 + + + Hereditary and Acquired Complement Deficiencies in Animals and Man + + KARGER + + + + + 6.1 Summary of BADA deficiencies and limitations .................................................... + + + + + How To Increase The Life of Your SSD Drives On Windows 7 + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + 10.4016/26630.01 + + + SciVee, Inc + + + 6.2 MACS .................................................................................................................. + + + + + MQ-1C Gray Eagle Unmanned Aircraft System (MQ-1C Gray Eagle) + + TimothyRBaxter + + 7 Gray Eagle ................................................................................................................. + + + 10.21236/ada613350 + + + Defense Technical Information Center + + + 7 Gray Eagle .................................................................................................................. + + + + + Summary of Reported Deficiencies + + KlausRother + + ................................................... + + + 10.1159/000318546 + + + Hereditary and Acquired Complement Deficiencies in Animals and Man + + KARGER + + + + + 7.1 Summary of BADA deficiencies and limitations .................................................... + + + + + How To Increase The Life of Your SSD Drives On Windows 7 + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + 10.4016/26630.01 + + + SciVee, Inc + + + 7.2 MACS .................................................................................................................. + + + + + LVIII the avenger + + AlexandreDumas + + 8 Predator C (Avenger) ................................................................................................. + + + 10.1093/owc/9780199537266.003.0059 + + + Twenty Years After + + Oxford University Press + + + + 8 Predator C (Avenger) .................................................................................................. + + + + + Summary of Reported Deficiencies + + KlausRother + + ................................................... + + + 10.1159/000318546 + + + Hereditary and Acquired Complement Deficiencies in Animals and Man + + KARGER + + + + + 8.1 Summary of BADA deficiencies and limitations .................................................... + + + + + How To Increase The Life of Your SSD Drives On Windows 7 + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + 10.4016/26630.01 + + + SciVee, Inc + + + 8.2 MACS .................................................................................................................. + + + + + Service-Oriented Separation Assurance for Small UAS Traffic Management + + GeorgeHunter + + 9 Hunter UAS................................................................................................................ + + + + PengWei + + 9 Hunter UAS................................................................................................................ + + + 10.1109/icnsurv.2019.8735165 + + + 2019 Integrated Communications, Navigation and Surveillance Conference (ICNS) + + IEEE + + + + 9 Hunter UAS................................................................................................................. + + + + + Summary of Reported Deficiencies + + KlausRother + + ................................................... + + + 10.1159/000318546 + + + Hereditary and Acquired Complement Deficiencies in Animals and Man + + KARGER + + + + + 9.1 Summary of BADA deficiencies and limitations .................................................... + + + + + How To Increase The Life of Your SSD Drives On Windows 7 + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + + JakieMacsJakie Macs + + 2 MACS ................................................................................................................. + + + 10.4016/26630.01 + + + SciVee, Inc + + + 9.2 MACS .................................................................................................................. + + + + + Aircraft Cargo Systems - Missing Restraint Limitations Layouts + 10.4271/arp5492a + + null + SAE International + + + Cargo UAS ................................................................................................................. 8.10.1 Summary of BADA deficiencies and limitations ............................................. + + + + + How To Increase The Life of Your SSD Drives On Windows 7 + + JakieMacsJakie Macs + + 10.2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 10.2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 10.2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 10.2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 10.2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 10.2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 10.2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 10.2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 10.2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 10.2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 10.2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 10.2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 10.2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 10.2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 10.2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 10.2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 10.2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 10.2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 10.2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 10.2 MACS .......................................................................................................... + + + 10.4016/26630.01 + + + SciVee, Inc + + + 10.2 MACS ........................................................................................................... + + + + + MQ-8 Fire Scout Unmanned Aircraft System (MQ-8 Fire Scout) + + JeffreyDodge + + Fire Scout ................................................................................................................... 8.11.1 Summary of BADA deficiencies and limitations ............................................ + + + 10.21236/ad1019503 + + + Defense Technical Information Center + + + Fire Scout ................................................................................................................... 8.11.1 Summary of BADA deficiencies and limitations ............................................. + + + + + How To Increase The Life of Your SSD Drives On Windows 7 + + JakieMacsJakie Macs + + 2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 2 MACS .......................................................................................................... + + + 10.4016/26630.01 + + + SciVee, Inc + + + 11.2 MACS ........................................................................................................... + + + + + L. Müller's Nonius, Part II - Nonii Marcelli Compendiosa Doctrina, Pars II. Emendavit Lucianus Muller. 12 Mk. + + JHOnions + + 12 NEO S-300 Mk II VTOL .............................................................................................. 8.12.1 Summary of BADA deficiencies and limitations ............................................ + + + 10.1017/s0009840x00195423 + + + The Classical Review + The Class. Rev. + 0009-840X + 1464-3561 + + 3 + 7 + + + Cambridge University Press (CUP) + + + 12 NEO S-300 Mk II VTOL .............................................................................................. 8.12.1 Summary of BADA deficiencies and limitations ............................................. + + + + + How To Increase The Life of Your SSD Drives On Windows 7 + + JakieMacsJakie Macs + + 2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 2 MACS .......................................................................................................... + + + + JakieMacsJakie Macs + + 2 MACS .......................................................................................................... + + + 10.4016/26630.01 + + + SciVee, Inc + + + 12.2 MACS ........................................................................................................... + + + + + Supplementary file 1. Validation statistics for FimA model + + BADA File Validation .................................................................................................... + + 10.7554/elife.31662.019 + + null + eLife Sciences Publications, Ltd + + + 9 BADA File Validation ..................................................................................................... + + + + + Development of the Seasonal Korean Aviation Turbulence Guidance (KTG) System Using the Regional Unified Model of the Korea Meteorological Administration (KMA) + + Dan-BiLee + + RQ7B) using KTG .............................................................. + + + + Hye-YeongChun + + RQ7B) using KTG .............................................................. + + + 10.14191/atmos.2014.24.2.235 + + + Atmosphere + Atmosphere + 1598-3560 + + 24 + 2 + + + Korean Meteorological Society + + + 9.1 Simulation of Shadow B (RQ7B) using KTG ............................................................... + + + + + Part III The EU Resolution Regime, 11 Institutional and Cross-Border Issues + + GleesonSimon + + 1 Issues and Resolution ......................................................................................... + + + + GuynnRandall + + 1 Issues and Resolution ......................................................................................... + + + 10.1093/law/9780199698011.003.0011 + + + Oxford University Press + + + 1.1 Issues and Resolution .......................................................................................... + + + + + Anomalies of Choice and Reason + + LeightonVaughan Williams + + 2 Reason for Anomalies ......................................................................................... + + + 10.1201/9781003083610-6 + + + Probability, Choice, and Reason + + Chapman and Hall/CRC + + + + + 1.2 Reason for Anomalies .......................................................................................... + + + + + Author Index– Autogenerate form corrected SGMl files + + ..................................................... + + 10.1016/s1007-5704(06)00106-7 + + + Communications in Nonlinear Science and Numerical Simulation + Communications in Nonlinear Science and Numerical Simulation + 1007-5704 + + 11 + 8 + VII + + Elsevier BV + + + 1.3 Simulation results using corrected BADA files...................................................... + + + + + Development of the Seasonal Korean Aviation Turbulence Guidance (KTG) System Using the Regional Unified Model of the Korea Meteorological Administration (KMA) + + Dan-BiLee + + RQ4A) using KTG .......................................................... + + + + Hye-YeongChun + + RQ4A) using KTG .......................................................... + + + 10.14191/atmos.2014.24.2.235 + + + Atmosphere + Atmosphere + 1598-3560 + + 24 + 2 + + + Korean Meteorological Society + + + 9.2 Simulation of Global Hawk (RQ4A) using KTG ........................................................... + + + + + Development of the Seasonal Korean Aviation Turbulence Guidance (KTG) System Using the Regional Unified Model of the Korea Meteorological Administration (KMA) + + Dan-BiLee + + ORBM) using KTG ................................................................... + + + + Hye-YeongChun + + ORBM) using KTG ................................................................... + + + 10.14191/atmos.2014.24.2.235 + + + Atmosphere + Atmosphere + 1598-3560 + + 24 + 2 + + + Korean Meteorological Society + + + 9.3 Simulation of Orbiter (ORBM) using KTG .................................................................... + + + + + GPS Remote Sensing Measurements Using Aerosonde UAV + + MichaelGrant + + MK47) using KTG .............................................................. + + + + StephenKatzberg + + MK47) using KTG .............................................................. + + + + RolandLawrence + + MK47) using KTG .............................................................. + + + 10.2514/6.2005-7005 + + + Infotech@Aerospace + + American Institute of Aeronautics and Astronautics + + + + 9.4 Simulation of Aerosonde (MK47) using KTG ............................................................... + + + + + Development of the Seasonal Korean Aviation Turbulence Guidance (KTG) System Using the Regional Unified Model of the Korea Meteorological Administration (KMA) + + Dan-BiLee + + MQ1B) using KTG ............................................................. + + + + Hye-YeongChun + + MQ1B) using KTG ............................................................. + + + 10.14191/atmos.2014.24.2.235 + + + Atmosphere + Atmosphere + 1598-3560 + + 24 + 2 + + + Korean Meteorological Society + + + 9.5 Simulation of Predator A (MQ1B) using KTG .............................................................. + + + + + Important and Critical Psychological Attributes of USAF MQ-1 Predator and MQ-9 Reaper Pilots According to Subject Matter Experts + + WayneChappelle + + MQ-9) using KTG .............................................................. + + + + KentMcdonald + + MQ-9) using KTG .............................................................. + + + + KatharineMcmillan + + MQ-9) using KTG .............................................................. + + + 10.21236/ada545552 + + + Defense Technical Information Center + + + 9.6 Simulation of Predator B (MQ-9) using KTG ............................................................... + + + + + MQ-1C Gray Eagle Unmanned Aircraft System (MQ-1C Gray Eagle) + + TimothyRBaxter + + MQ1C) using KTG ............................................................ + + + 10.21236/ada613350 + + + Defense Technical Information Center + + + 9.7 Simulation of Gray Eagle (MQ1C) using KTG ............................................................. + + + + + Development of the Seasonal Korean Aviation Turbulence Guidance (KTG) System Using the Regional Unified Model of the Korea Meteorological Administration (KMA) + + Dan-BiLee + + AVEN) using KTG ............................................................. + + + + Hye-YeongChun + + AVEN) using KTG ............................................................. + + + 10.14191/atmos.2014.24.2.235 + + + Atmosphere + Atmosphere + 1598-3560 + + 24 + 2 + + + Korean Meteorological Society + + + 9.8 Simulation of Predator C (AVEN) using KTG .............................................................. + + + + + Development of the Seasonal Korean Aviation Turbulence Guidance (KTG) System Using the Regional Unified Model of the Korea Meteorological Administration (KMA) + + Dan-BiLee + + MQ5B) using KTG ................................................................... + + + + Hye-YeongChun + + MQ5B) using KTG ................................................................... + + + 10.14191/atmos.2014.24.2.235 + + + Atmosphere + Atmosphere + 1598-3560 + + 24 + 2 + + + Korean Meteorological Society + + + Simulation of Hunter + 9 Simulation of Hunter (MQ5B) using KTG .................................................................... + + + + + MQ-8 Fire Scout Unmanned Aircraft System (MQ-8 Fire Scout) + + JeffreyDodge + + 10.21236/ad1019503 + NEO S-300 + + + Defense Technical Information Center + + + 10 Simulation of BADA Files for Cargo UAS (CUAS), Fire Scout (MQ8B) and NEO S-300 + + + + + A High-Speed, High-Efficiency VTOL Concept Using CoFlow Jet Airfoil + + IIMk + + S350) using KTG............................................................................................... + + + + Vtol + + S350) using KTG............................................................................................... + + + 10.2514/6.2020-2792.vid + + + American Institute of Aeronautics and Astronautics (AIAA) + + + Mk II VTOL (S350) using KTG................................................................................................ + + + + + Supplementary file 1. Summary of MD simulations. + + 11 Summary of UAS Simulations in KTG ......................................................................... 10 MACS File Validation .................................................................................................... + + 10.7554/elife.25850.017 + + null + eLife Sciences Publications, Ltd + + + 11 Summary of UAS Simulations in KTG ......................................................................... 10 MACS File Validation ..................................................................................................... + + + + + Low Altitude, High Speed Personnel Parachuting: Medical and Physiological Issues + + DavidJWehrly + + 1 Issues and Resolution ................................................................................................. ; Altitude Constraints in MACS ......................................... + + + 10.21236/ada181199 + + + Defense Technical Information Center + + + 1 Issues and Resolution ................................................................................................. 10.1.1 Issue 1: Speed vs. Altitude Constraints in MACS .......................................... + + + + + Wake vortex encounter modeling and simulation of a small flying-wing UAS + + UAS Aircraft ........................................... + + 10.2514/6.2022-4066.vid + + + American Institute of Aeronautics and Astronautics (AIAA) + + + 1.2 Issue 2: Simulation of Slow Flying UAS Aircraft ............................................ + + + + + Aircraft and Rotorcraft Flight Simulation Using the Julia Language + + UmbertoSaetti + + Electric Aircraft................................... + + + + JosephFHorn + + Electric Aircraft................................... + + + 10.2514/6.2022-2354 + + + AIAA SCITECH 2022 Forum + + American Institute of Aeronautics and Astronautics + + + + 1.3 Issue 3: Simulation of Rotorcraft and Electric Aircraft.................................... + + + + + How To Increase The Life of Your SSD Drives On Windows 7 + + JakieMacsJakie Macs + + 2 Simulation of Shadow B (RQ7B) using MACS ........................................................... + + + + JakieMacsJakie Macs + + 2 Simulation of Shadow B (RQ7B) using MACS ........................................................... + + + + JakieMacsJakie Macs + + 2 Simulation of Shadow B (RQ7B) using MACS ........................................................... + + + + JakieMacsJakie Macs + + 2 Simulation of Shadow B (RQ7B) using MACS ........................................................... + + + + JakieMacsJakie Macs + + 2 Simulation of Shadow B (RQ7B) using MACS ........................................................... + + + + JakieMacsJakie Macs + + 2 Simulation of Shadow B (RQ7B) using MACS ........................................................... + + + + JakieMacsJakie Macs + + 2 Simulation of Shadow B (RQ7B) using MACS ........................................................... + + + + JakieMacsJakie Macs + + 2 Simulation of Shadow B (RQ7B) using MACS ........................................................... + + + + JakieMacsJakie Macs + + 2 Simulation of Shadow B (RQ7B) using MACS ........................................................... + + + + JakieMacsJakie Macs + + 2 Simulation of Shadow B (RQ7B) using MACS ........................................................... + + + + JakieMacsJakie Macs + + 2 Simulation of Shadow B (RQ7B) using MACS ........................................................... + + + + JakieMacsJakie Macs + + 2 Simulation of Shadow B (RQ7B) using MACS ........................................................... + + + + JakieMacsJakie Macs + + 2 Simulation of Shadow B (RQ7B) using MACS ........................................................... + + + + JakieMacsJakie Macs + + 2 Simulation of Shadow B (RQ7B) using MACS ........................................................... + + + + JakieMacsJakie Macs + + 2 Simulation of Shadow B (RQ7B) using MACS ........................................................... + + + + JakieMacsJakie Macs + + 2 Simulation of Shadow B (RQ7B) using MACS ........................................................... + + + + JakieMacsJakie Macs + + 2 Simulation of Shadow B (RQ7B) using MACS ........................................................... + + + + JakieMacsJakie Macs + + 2 Simulation of Shadow B (RQ7B) using MACS ........................................................... + + + + JakieMacsJakie Macs + + 2 Simulation of Shadow B (RQ7B) using MACS ........................................................... + + + + JakieMacsJakie Macs + + 2 Simulation of Shadow B (RQ7B) using MACS ........................................................... + + + 10.4016/26630.01 + + + SciVee, Inc + + + 2 Simulation of Shadow B (RQ7B) using MACS ............................................................ + + + + + How To Increase The Life of Your SSD Drives On Windows 7 + + JakieMacsJakie Macs + + RQ4A) using MACS ....................................................... + + + + JakieMacsJakie Macs + + RQ4A) using MACS ....................................................... + + + + JakieMacsJakie Macs + + RQ4A) using MACS ....................................................... + + + + JakieMacsJakie Macs + + RQ4A) using MACS ....................................................... + + + + JakieMacsJakie Macs + + RQ4A) using MACS ....................................................... + + + + JakieMacsJakie Macs + + RQ4A) using MACS ....................................................... + + + + JakieMacsJakie Macs + + RQ4A) using MACS ....................................................... + + + + JakieMacsJakie Macs + + RQ4A) using MACS ....................................................... + + + + JakieMacsJakie Macs + + RQ4A) using MACS ....................................................... + + + + JakieMacsJakie Macs + + RQ4A) using MACS ....................................................... + + + + JakieMacsJakie Macs + + RQ4A) using MACS ....................................................... + + + + JakieMacsJakie Macs + + RQ4A) using MACS ....................................................... + + + + JakieMacsJakie Macs + + RQ4A) using MACS ....................................................... + + + + JakieMacsJakie Macs + + RQ4A) using MACS ....................................................... + + + + JakieMacsJakie Macs + + RQ4A) using MACS ....................................................... + + + + JakieMacsJakie Macs + + RQ4A) using MACS ....................................................... + + + + JakieMacsJakie Macs + + RQ4A) using MACS ....................................................... + + + + JakieMacsJakie Macs + + RQ4A) using MACS ....................................................... + + + + JakieMacsJakie Macs + + RQ4A) using MACS ....................................................... + + + + JakieMacsJakie Macs + + RQ4A) using MACS ....................................................... + + + 10.4016/26630.01 + + + SciVee, Inc + + + 3 Simulation of Global Hawk (RQ4A) using MACS ........................................................ + + + + + GPS Remote Sensing Measurements Using Aerosonde UAV + + MichaelGrant + + 4 Simulation of Aerosonde (MK47) using MACS ........................................................... + + + + StephenKatzberg + + 4 Simulation of Aerosonde (MK47) using MACS ........................................................... + + + + RolandLawrence + + 4 Simulation of Aerosonde (MK47) using MACS ........................................................... + + + 10.2514/6.2005-7005 + + + Infotech@Aerospace + + American Institute of Aeronautics and Astronautics + + + + 4 Simulation of Aerosonde (MK47) using MACS ............................................................ + + + + + How To Increase The Life of Your SSD Drives On Windows 7 + + JakieMacsJakie Macs + + MQ1B) using MACS .......................................................... + + + + JakieMacsJakie Macs + + MQ1B) using MACS .......................................................... + + + + JakieMacsJakie Macs + + MQ1B) using MACS .......................................................... + + + + JakieMacsJakie Macs + + MQ1B) using MACS .......................................................... + + + + JakieMacsJakie Macs + + MQ1B) using MACS .......................................................... + + + + JakieMacsJakie Macs + + MQ1B) using MACS .......................................................... + + + + JakieMacsJakie Macs + + MQ1B) using MACS .......................................................... + + + + JakieMacsJakie Macs + + MQ1B) using MACS .......................................................... + + + + JakieMacsJakie Macs + + MQ1B) using MACS .......................................................... + + + + JakieMacsJakie Macs + + MQ1B) using MACS .......................................................... + + + + JakieMacsJakie Macs + + MQ1B) using MACS .......................................................... + + + + JakieMacsJakie Macs + + MQ1B) using MACS .......................................................... + + + + JakieMacsJakie Macs + + MQ1B) using MACS .......................................................... + + + + JakieMacsJakie Macs + + MQ1B) using MACS .......................................................... + + + + JakieMacsJakie Macs + + MQ1B) using MACS .......................................................... + + + + JakieMacsJakie Macs + + MQ1B) using MACS .......................................................... + + + + JakieMacsJakie Macs + + MQ1B) using MACS .......................................................... + + + + JakieMacsJakie Macs + + MQ1B) using MACS .......................................................... + + + + JakieMacsJakie Macs + + MQ1B) using MACS .......................................................... + + + + JakieMacsJakie Macs + + MQ1B) using MACS .......................................................... + + + 10.4016/26630.01 + + + Simulation of Predator A + + + SciVee, Inc + + + 5 Simulation of Predator A (MQ1B) using MACS ........................................................... + + + + + Important and Critical Psychological Attributes of USAF MQ-1 Predator and MQ-9 Reaper Pilots According to Subject Matter Experts + + WayneChappelle + + MQ-9) using MACS ........................................................... + + + + KentMcdonald + + MQ-9) using MACS ........................................................... + + + + KatharineMcmillan + + MQ-9) using MACS ........................................................... + + + 10.21236/ada545552 + + + Simulation of Predator B + + Defense Technical Information Center + + + + 6 Simulation of Predator B (MQ-9) using MACS ............................................................ + + + + + MQ-1C Gray Eagle Unmanned Aircraft System (MQ-1C Gray Eagle) + + TimothyRBaxter + + MQ1C) using MACS ......................................................... + + + 10.21236/ada613350 + + + Defense Technical Information Center + + + Simulation of Gray Eagle + 7 Simulation of Gray Eagle (MQ1C) using MACS .......................................................... + + + + + How To Increase The Life of Your SSD Drives On Windows 7 + + JakieMacsJakie Macs + + 8 Simulation of Predator C (AVEN) using MACS .......................................................... + + + + JakieMacsJakie Macs + + 8 Simulation of Predator C (AVEN) using MACS .......................................................... + + + + JakieMacsJakie Macs + + 8 Simulation of Predator C (AVEN) using MACS .......................................................... + + + + JakieMacsJakie Macs + + 8 Simulation of Predator C (AVEN) using MACS .......................................................... + + + + JakieMacsJakie Macs + + 8 Simulation of Predator C (AVEN) using MACS .......................................................... + + + + JakieMacsJakie Macs + + 8 Simulation of Predator C (AVEN) using MACS .......................................................... + + + + JakieMacsJakie Macs + + 8 Simulation of Predator C (AVEN) using MACS .......................................................... + + + + JakieMacsJakie Macs + + 8 Simulation of Predator C (AVEN) using MACS .......................................................... + + + + JakieMacsJakie Macs + + 8 Simulation of Predator C (AVEN) using MACS .......................................................... + + + + JakieMacsJakie Macs + + 8 Simulation of Predator C (AVEN) using MACS .......................................................... + + + + JakieMacsJakie Macs + + 8 Simulation of Predator C (AVEN) using MACS .......................................................... + + + + JakieMacsJakie Macs + + 8 Simulation of Predator C (AVEN) using MACS .......................................................... + + + + JakieMacsJakie Macs + + 8 Simulation of Predator C (AVEN) using MACS .......................................................... + + + + JakieMacsJakie Macs + + 8 Simulation of Predator C (AVEN) using MACS .......................................................... + + + + JakieMacsJakie Macs + + 8 Simulation of Predator C (AVEN) using MACS .......................................................... + + + + JakieMacsJakie Macs + + 8 Simulation of Predator C (AVEN) using MACS .......................................................... + + + + JakieMacsJakie Macs + + 8 Simulation of Predator C (AVEN) using MACS .......................................................... + + + + JakieMacsJakie Macs + + 8 Simulation of Predator C (AVEN) using MACS .......................................................... + + + + JakieMacsJakie Macs + + 8 Simulation of Predator C (AVEN) using MACS .......................................................... + + + + JakieMacsJakie Macs + + 8 Simulation of Predator C (AVEN) using MACS .......................................................... + + + 10.4016/26630.01 + + + SciVee, Inc + + + 8 Simulation of Predator C (AVEN) using MACS ........................................................... + + + + + Source code 1. Code used to analyze raw sequencing files using the programs STAR, Bowtie2, MACS, and Homer. + + MQ5B) using MACS ............................................................... ; MQ8B) and NEO S-300 Mk II VTOL (S350) using MACS .................................................................. + + 10.7554/elife.40167.016 + + + 10.10 Simulation of BADA Files for Orbiter (ORBM), Cargo UAS (CUAS), Fire Scout + + eLife Sciences Publications, Ltd + null + + + Simulation of Hunter + 9 Simulation of Hunter (MQ5B) using MACS ................................................................. 10.10 Simulation of BADA Files for Orbiter (ORBM), Cargo UAS (CUAS), Fire Scout (MQ8B) and NEO S-300 Mk II VTOL (S350) using MACS ................................................................... + + + + + On the Accuracy of Flexible Antennas Simulations + + SimaNoghanian + + 11 Summary of UAS Simulations in MACS ...................................................................... 11 ACES Simulations for CNS Capabilities ....................................................................... + + + + MichaelGriesi + + 11 Summary of UAS Simulations in MACS ...................................................................... 11 ACES Simulations for CNS Capabilities ....................................................................... + + + 10.47037/2020.aces.j.351175 + + + Applied Computational Electromagnetics Society + ACES + 1054-4887 + + 35 + 11 + + + River Publishers + + + 11 Summary of UAS Simulations in MACS ...................................................................... 11 ACES Simulations for CNS Capabilities ........................................................................ + + + + + FAA Unmanned Aircraft Systems (UAS) Sighting Reports: A Preliminary Survey + + UAS Aircraft/BADA Data Installation and Preparation for CNS Simulations ................ 11.1.1 Installation of UAS Aircraft Models into ACES and KTG .............................. + + 10.2514/6.2023-4099.vid + + + American Institute of Aeronautics and Astronautics (AIAA) + + + 1 UAS Aircraft/BADA Data Installation and Preparation for CNS Simulations ................ 11.1.1 Installation of UAS Aircraft Models into ACES and KTG ............................... + + + + + Researching Data Sets to Develop State Library Standards + + LesleyS JFarmer + + 2 Develop Flight Data Sets ............................................................................. + + + 10.29173/iasl7765 + + + IASL Annual Conference Proceedings + iasl + 2562-8372 + + + University of Alberta Libraries + + + 1.2 Develop Flight Data Sets .............................................................................. + + + + + FastTest Plugin: a New Plugin to Generate Moodle Quiz XML Files + + MilagrosHuerta + 0000-0001-5805-4886 + + 3 Develop CNS Plugin Configuration Files ...................................................... + + + + ManuelAlejandroFernandez-Ruiz + + 3 Develop CNS Plugin Configuration Files ...................................................... + + + 10.20944/preprints202202.0282.v1 + + + MDPI AG + + + 1.3 Develop CNS Plugin Configuration Files ....................................................... + + + + + Navy-wide Personnel Survey (NPS) 2003: Tabulated Results + + KimberlyPWhittam + + 2 Simulations: Tabulated Results .................................................................................. + + + + JessicaBJanega + + 2 Simulations: Tabulated Results .................................................................................. + + + 10.21236/ada441225 + + + Defense Technical Information Center + + + 2 Simulations: Tabulated Results ................................................................................... + + + + + Poster session 1: Test results: Mammalian cells; Test results: Drosophila & plants; Test results: Micronucleus/cytogenetics; Test results: Comet assay; Test results: Battery of assays; Test analysis: Risk assessment; Test development + + Test Results ............................................................................................................... + + 10.1002/(sici)1098-2280(1998)29+<37::aid-em5>3.0.co;2-f + + + Environmental and Molecular Mutagenesis + Environ. Mol. Mutagen. + 0893-6692 + 1098-2280 + + 31 + S29 + + + Wiley + + + Test Results ................................................................................................................ + + + + + On the use of the variance in resolving two practical problems often encountered in input-output analysis + + SDGerking + + ................ + + + 10.1007/978-1-4613-4362-2_4 + + + Estimation of stochastic input-output models + + Springer US + + + + + 4 Problems Encountered and Precautions for use of CNS models with UAS ................. + + + + + Conclusions and Recommendations for Future Work + 10.6027/9789289328692-9-en + + + Nordic Council of Ministers + + + Conclusions ................................................................................................................... 13 Recommendations for Future Work ............................................................................... + + + + + Small-Format Aerial Photography and UAS Imagery + + 1 Recommendations to Modify BADA Format for UAS Simulations .............................. + + 10.1016/c2016-0-03506-4 + + + Elsevier + + + 1 Recommendations to Modify BADA Format for UAS Simulations ............................... + + + + + How To Increase The Life of Your SSD Drives On Windows 7 + + JakieMacsJakie Macs + + 2 Recommendations to Modify MACS for UAS Simulations .......................................... + + + + JakieMacsJakie Macs + + 2 Recommendations to Modify MACS for UAS Simulations .......................................... + + + + JakieMacsJakie Macs + + 2 Recommendations to Modify MACS for UAS Simulations .......................................... + + + + JakieMacsJakie Macs + + 2 Recommendations to Modify MACS for UAS Simulations .......................................... + + + + JakieMacsJakie Macs + + 2 Recommendations to Modify MACS for UAS Simulations .......................................... + + + + JakieMacsJakie Macs + + 2 Recommendations to Modify MACS for UAS Simulations .......................................... + + + + JakieMacsJakie Macs + + 2 Recommendations to Modify MACS for UAS Simulations .......................................... + + + + JakieMacsJakie Macs + + 2 Recommendations to Modify MACS for UAS Simulations .......................................... + + + + JakieMacsJakie Macs + + 2 Recommendations to Modify MACS for UAS Simulations .......................................... + + + + JakieMacsJakie Macs + + 2 Recommendations to Modify MACS for UAS Simulations .......................................... + + + + JakieMacsJakie Macs + + 2 Recommendations to Modify MACS for UAS Simulations .......................................... + + + + JakieMacsJakie Macs + + 2 Recommendations to Modify MACS for UAS Simulations .......................................... + + + + JakieMacsJakie Macs + + 2 Recommendations to Modify MACS for UAS Simulations .......................................... + + + + JakieMacsJakie Macs + + 2 Recommendations to Modify MACS for UAS Simulations .......................................... + + + + JakieMacsJakie Macs + + 2 Recommendations to Modify MACS for UAS Simulations .......................................... + + + + JakieMacsJakie Macs + + 2 Recommendations to Modify MACS for UAS Simulations .......................................... + + + + JakieMacsJakie Macs + + 2 Recommendations to Modify MACS for UAS Simulations .......................................... + + + + JakieMacsJakie Macs + + 2 Recommendations to Modify MACS for UAS Simulations .......................................... + + + + JakieMacsJakie Macs + + 2 Recommendations to Modify MACS for UAS Simulations .......................................... + + + + JakieMacsJakie Macs + + 2 Recommendations to Modify MACS for UAS Simulations .......................................... + + + 10.4016/26630.01 + + + SciVee, Inc + + + 2 Recommendations to Modify MACS for UAS Simulations ........................................... + + + + + Validation of a Rotorcraft Mathematical Model for Autogyro Simulation + + SSHouston + + Rotorcraft and Hybrid Aircraft ...................... + + + 10.2514/2.2640 + + + Journal of Aircraft + Journal of Aircraft + 0021-8669 + 1533-3868 + + 37 + 3 + + + American Institute of Aeronautics and Astronautics (AIAA) + + + 3 Validation of BADA and MACS Files for Rotorcraft and Hybrid Aircraft ....................... + + + + + Are Analysts’ Recommendations for Other Investment Banks Biased? + + ErikDevos + + Other Recommendations ........................................................................................... + + + 10.1111/fima.12031 + + + Financial Management + Financial Management + 0046-3892 + + 43 + 2 + + + Wiley + + + Other Recommendations ............................................................................................ + + + + + Current References + + References ................................................................................................................... + + 10.1109/9780470547038.ch2 + + + Voltage References + + IEEE + + + + References .................................................................................................................... + + + + + Appendix A: UAS Organisations + + Appendix A: Industry Data of UAS Aircraft .................................................................... + + 10.1002/9780470664797.app1 + + + Unmanned Aircraft Systems + + John Wiley & Sons, Ltd + + + + + Appendix A: Industry Data of UAS Aircraft ..................................................................... + + + + + Joe Shadow Nursery Company [price list] / + + AJShadow + + RQ7B) ..................................................................................................... + + + + Joe.Shadow + + RQ7B) ..................................................................................................... + + + + WAShadow + + RQ7B) ..................................................................................................... + + + 10.5962/bhl.title.136604 + + + Joe Shadow Nursery Co., Inc., + + + Shadow B (RQ7B) ...................................................................................................... + + + + + Global Hawk Systems Engineering. Case Study + + BillKinzig + + 2 Global Hawk (RQ4A) ................................................................................................. + + + 10.21236/ada538761 + + + Defense Technical Information Center + + + 2 Global Hawk (RQ4A) .................................................................................................. + + + + + MAG Solar Orbiter magnetometer + 10.5270/esa-ux7y320 + + null + European Space Agency + + + 3 Orbiter (ORBM) ......................................................................................................... + + + + + Aerosonde Technical Development + + DSEberhardt + + 4 Aerosonde (MK47) ................................................................................................... + + + + JurisVagners + + 4 Aerosonde (MK47) ................................................................................................... + + + + EliLivne + + 4 Aerosonde (MK47) ................................................................................................... + + + + Uy-LoiLy + + 4 Aerosonde (MK47) ................................................................................................... + + + 10.21236/ada630376 + + + Defense Technical Information Center + + + 4 Aerosonde (MK47) .................................................................................................... + + + + + Bio-inspired Predator and Anti-predator Mechanisms for Unmanned Aerial Systems + + APredator + + MQ1B) .................................................................................................. + + + 10.2514/6.2022-0274.vid + + + American Institute of Aeronautics and Astronautics (AIAA) + + + Predator A (MQ1B) ................................................................................................... + + + + + + + BPredator + + MQ-9) ................................................................................................... + + + + + Predator B (MQ-9) .................................................................................................... + + + + + + + GrayEagle + + MQ1C) ................................................................................................. + + + + + Gray Eagle (MQ1C) .................................................................................................. + + + + + Reliability Achievement + + TerjeAven + + Predator C (AVEN) .................................................................................................. + + + 10.1201/9781351076340 + + + Chapman and Hall/CRC + + + Predator C (AVEN) ................................................................................................... + + + + + Hunter, John Alexander (1882–1963), hunter + + MartinBooth + + MQ5B) ........................................................................................................ + + + 10.1093/ref:odnb/39753 + + + Oxford University Press + + + 9 Hunter (MQ5B) ......................................................................................................... + + + + + Vertiports - Infrastructure and equipment for vertical take-off and landing (VTOL) of electrically powered cargo unmanned aircraft systems (UAS) + + Cargo UAS (CUAS) ................................................................................................. + + 10.3403/30426830u + + null + BSI British Standards + + + Cargo UAS (CUAS) .................................................................................................. + + + + + MQ-8 Fire Scout Unmanned Aircraft System (MQ-8 Fire Scout) + + JeffreyDodge + + MQ8B) ................................................................................................... + + + 10.21236/ad1019503 + + + 11 Fire Scout + + Defense Technical Information Center + + + + 11 Fire Scout (MQ8B) .................................................................................................... + + + + + L. Müller's Nonius, Part II - Nonii Marcelli Compendiosa Doctrina, Pars II. Emendavit Lucianus Muller. 12 Mk. + + JHOnions + + S350) ................................................................................ + + + 10.1017/s0009840x00195423 + 12 NEO S-300 Mk II VTOL + + + The Classical Review + The Class. Rev. + 0009-840X + 1464-3561 + + 3 + 7 + + + Cambridge University Press (CUP) + + + 12 NEO S-300 Mk II VTOL (S350) ................................................................................. + + + + + Appendix A: UAS Organisations + + KTG to Simulate UAS Aircraft ..................... + + 10.1002/9780470664797.app1 + + + Unmanned Aircraft Systems + + John Wiley & Sons, Ltd + + + + + Appendix B: Configuration of ACES and KTG to Simulate UAS Aircraft ...................... + + + + + Introduction + + JonathanWolff + + 1 Introduction .............................................................................................................. + + + 10.1093/hepl/9780199658015.003.0001 + + + An Introduction to Political Philosophy + + Oxford University Press + + + + 1 Introduction ............................................................................................................... + + + + + eLongated Double Wing (LDW) Aircraft Configuration + + 2 Configuration of KTG Database ................................................................................ 16.2.1 Configuration of "aircraft_control_gain.csv" ................................................ + + 10.2514/6.2022-3517.vid + + + American Institute of Aeronautics and Astronautics (AIAA) + + + 2 Configuration of KTG Database ................................................................................ 16.2.1 Configuration of "aircraft_control_gain.csv" ................................................. + + + + + synonym + + Mpas_Synonym + + "SYNONYM_ACES_KTG.OLD" .................................................................................. + + + + Lst + + "SYNONYM_ACES_KTG.OLD" .................................................................................. + + + 10.1007/springerreference_26234 + + + SYNONYM_ALL.LST" and + + Springer-Verlag + null + + + 2.2 Configuration of "MPAS_SYNONYM.LST," "SYNONYM_ALL.LST" and "SYNONYM_ACES_KTG.OLD" ................................................................................... + + + + + A Gain-enhanced Dual-band Microstrip Antenna using Metasurface as Superstrate Configuration + + HuqiangTian + + 3 Configuration of ACES Database ............................................................................. + + + + JunlinWang + + 3 Configuration of ACES Database ............................................................................. + + + + DingHan + + 3 Configuration of ACES Database ............................................................................. + + + + XinWang + + 3 Configuration of ACES Database ............................................................................. + + + 10.13052/2021.aces.j.361210 + + + The Applied Computational Electromagnetics Society Journal (ACES) + ACES Journal + 1054-4887 + 1943-5711 + + + River Publishers + + + 3 Configuration of ACES Database .............................................................................. + + + + + Thrust at Cruise 11350 lb + + Max + + + 12450 lb + + + Max. Thrust at Cruise 11350 lb. 12450 lb. + + + + + Base of Aircraft Database (BADA) version 3.9 + + ExperimentalEurocontrol + + + Centre + + + + April 2011 + + + EUROCONTROL Experimental Centre, "Base of Aircraft Database (BADA) version 3.9," April 2011. [Online]. Available: http://www.eurocontrol.int/eec/gallery/content/public/document/eec/other_document/2011/E EC-Technical-Report-110308-08.pdf. + + + + + The Airspace Operations Laboratory (AOL) at NASA Ames Research Center + + ThomasPrevot + + + NancySmith + + + EverettPalmer + + + JoeyMercer + + + PaulLee + + + JeffreyHomola + + + ToddCallantine + + 10.2514/6.2006-6112 + + + + AIAA Modeling and Simulation Technologies Conference and Exhibit + + American Institute of Aeronautics and Astronautics + June 2011 + + + NASA + Airspace Operations Laboratory (AOL), "The Multi Aircraft Control System (MACS)," NASA, [Online]. Available: http://hsi.arc.nasa.gov/groups/AOL/technologies/macs.php. [Accessed June 2011]. + + + + + KTG: A Fast-Time Kinematic Trajectory Generator for Modeling and Simulation of ATM Automation Concepts and NAS-wide System Level Analysis + + YingchuanZhang + + + GoutamSatapathy + + + VikramManikonda + + + NikhilNigam + + 10.2514/6.2010-8365 + + + AIAA Modeling and Simulation Technologies Conference + Toronto, Canada + + American Institute of Aeronautics and Astronautics + 2010 + + + Y. Zhang, G. Satapathy, V. Manikonda and N. Nigam, "KTG: A Fast-time Kinematic Trajectory Generator for Modeling and Simulation of ATM Automation Concepts and NAS- wide System Level Analysis," in AIAA Modeling and Simulation Technologies Conference, Toronto, Canada, 2010. + + + + + Studying NextGen Concepts with the Multi-Aircraft Control System + + JoeyMercer + + + ThomasPrevôt + + + RichardJacoby + + + AlbertGlobus + + + JeffreyHomola + + 10.2514/6.2008-7026 + + + AIAA Modeling and Simulation Technologies Conference and Exhibit + Honolulu, HI + + American Institute of Aeronautics and Astronautics + 2008 + + + J. Mercer, T. Prevot, R. Jacoby, A. Globus and J. Homola, "Studying Nextgen Concepts with Multi Aircraft Control System," in Modeling and Simulation Technologies Conference, Honolulu, HI, 2008. + + + + + The USAF Stability and Control DATCOM + + McdonnellDouglas + + + AstronauticsCompany + + + + 1979. 9d0267197b8160bffff8ab6ffffe415 + + + McDonnell Douglas Astronautics Company, "The USAF Stability and Control DATCOM," 1979. [Online]. Available: http://www.holycows.net/datcom/media/e9d0267197b8160bffff8ab6ffffe415.zip. + + + + + JSBSim: An Open Source Flight Dynamics Model in C++ + + JonBerndt + + 10.2514/6.2004-4923 + + + + AIAA Modeling and Simulation Technologies Conference and Exhibit + + American Institute of Aeronautics and Astronautics + August 2012 + + + J. Berndt, "JSBSim Reference Manual," [Online]. Available: http://jsbsim.sourceforge.net/JSBSimReferenceManual.pdf. [Accessed 28 August 2012]. + + + + + Blender Overview + + Blender Foundation + + 10.1007/978-1-4302-1977-4_1 + + + + Foundation Blender Compositing + + Apress + August 2012 + + + + Blender Foundation, "Blender Website," [Online]. Available: http://www.blender.org/. [Accessed 28 August 2012]. + + + + + Author response image 3. + 10.7554/elife.03700.025 + + + August 2012 + eLife Sciences Publications, Ltd + + + JSBSim + "Aeromatic 2 Website," JSBSim, [Online]. Available: http://jsbsim.sourceforge.net/aeromatic2.html. [Accessed 28 August 2012]. + + + + + Open-Source Visualization of Reusable Rockets Motion: Approaching Simulink - FlightGear Co-simulation + + Flightgear + + 10.2514/6.2021-0410.vid + + + August 2012 + American Institute of Aeronautics and Astronautics (AIAA) + + + FlightGear, "FlightGear Website," [Online]. Available: http://www.flightgear.org/. [Accessed 28 August 2012]. + + + + + Helicopter Performance, Stability, and Control + + RWProuty + + + 1986 + PWS Engineering + Boston, MA + + + R. W. Prouty, Helicopter Performance, Stability, and Control, Boston, MA: PWS Engineering, 1986. + + + + + Helicopter Sizing by Statistics + + OmriRand + + + VladimirKhromov + + 10.4050/jahs.49.300 + + + Journal of the American Helicopter Society + j am helicopter soc + 2161-6027 + + 49 + 3 + + 2004 + American Helicopter Society + + + O. Rand and V. Khromov, "Helicopter Sizing by Statistics," Journal of the American Helicopter Society, vol. 49, no. 3, pp. 300-317, 2004. + + + + + Private Communication with FAA William J. Hughes Technical Center + + MKonyak + + + August 16, 2012 + Atlantic City, NJ + + + Konyak, M., Private Communication with FAA William J. Hughes Technical Center, Atlantic City, NJ, August 16, 2012. + + + + + BADA: An advanced aircraft performance model for present and future ATM systems + + AngelaNuic + + + DamirPoles + + + VincentMouillet + + 10.1002/acs.1176 + + + International Journal of Adaptive Control and Signal Processing + Int. J. Adapt. Control Signal Process. + 0890-6327 + + 24 + 10 + + 2010 + Wiley + + + A. Nuic, P. Damir and V. Mouillet, "BADA: An Advanced Aircraft Performance Model for Present and Future ATM Systems," International Journal of Adaptive Control and Signal Processing, pp. 850-866, 2010. + + + + + Appendix A: UAS Organisations + 10.1002/9780470664797.app1 + + + Unmanned Aircraft Systems + + John Wiley & Sons, Ltd + + + + + Appendix A: Industry Data of UAS Aircraft + + + + + REEFS AND BIOACCUMULATIONS IN THE MIOCENE DEPOSITS OF THE NORTH CROATIAN BASIN – AMAZING DIVERSITY YET TO BE DESCRIBED + + JasenkaSremac + + + MarijaBošnjak Makovec + + + DavorVrsaljko + + + BojanKaraica + + + KristinaTripalo + + + KarmenFio Firi + + + AnaMajstorović Bušić + + + TihomirMarjanac + + 10.17794/rgn.2016.1.2 + + + Rudarsko-geološko-naftni zbornik + MGPB + 0353-4529 + 1849-0409 + + 31 + 1 + + + Faculty of Mining, Geology and Petroleum Engineering + + + 575 lb + Design Endurance 24 hr. (575 lb. of fuel and loiter at 10000 ft.) + + + + + + diff --git a/file781.txt b/file781.txt new file mode 100644 index 0000000000000000000000000000000000000000..ec68e4cc38c346d5d83db12e267d5f29d96eeba8 --- /dev/null +++ b/file781.txt @@ -0,0 +1,886 @@ + + + + +SummaryThe Center-TRACON Automation System (CTAS), under development at the Ames Research Center, is designed to assist controllers with the management and control of air traffic in the extended terminal area.The Langley Research Center is participating in a joint program with Ames to investigate the issues of and develop systems and procedures for the integration of CTAS and airborne automation systems.A central issue in this research is the accuracy of the CTAS trajectory prediction process and compatibility with airborne Flight Management Systems for the scheduling and control of arrival traffic.Two flight experiments were conducted (Phase I in October 1992 and Phase II in September 1994) at Denver to evaluate the accuracy of the CTAS trajectory prediction process during the en route arrival phase of flight.The Transport Systems Research Vehicle (TSRV) Boeing 737 airplane based at the Langley Research Center flew a combined total of 57 arrival trajectories from cruise altitude to a terminalarea metering fix while following CTAS descent clearance advisories.Actual trajectories of the airplane were compared with the trajectories predicted by the CTAS trajectory synthesis algorithms and airplane Flight Management System.Trajectory prediction accuracy was evaluated over several levels of cockpit automation, which ranged from a conventional cockpit to a performance-based vertical navigation (VNAV) Flight Management System.Error sources and their magnitudes were identified and measured from the flight data.The CTAS descent advisor was found to provide a reasonable prediction of metering fix arrival time performance during these tests.Overall arrival time errors (Mean + Standard deviation) were measured to be approximately 24 sec during Phase I and 15 sec during Phase II.The major source of error during these tests was found to be the predicted winds aloft used by CTAS.Position and velocity estimates of the airplane provided to CTAS by the Air Traffic Control (ATC) Host radar tracker were found to be a relatively insignificant error source.Airplane performance modeling errors within CTAS were found to not significantly affect arrival time errors when the constrained descent procedures were used.The most significant effect related to the flight guidance was observed to be the cross-track and turn-overshoot errors associated with conventional VOR (very high frequency omnidirectional radio range) guidance.Lateral navigation (LNAV) guidance significantly reduced both the cross-track and turn-overshoot errors.Pilot procedures and VNAV guidance were found to significantly reduce the vertical profile errors associated with atmospheric and airplane performance model errors. +IntroductionSince 1989, a joint program has been underway between the Ames Research Center and the Langley Research Center to investigate the issues of and develop systems for the integration of Air Traffic Control (ATC) and airborne automation systems.Ames has developed the Center-TRACON Automation System (CTAS), a ground-based ATC automation system designed to assist controllers in the efficient handling of traffic of all types and capabilities (ref.1).This system has the ability to accurately predict airplane trajectories and determine effective advisories to assist the controller in managing traffic.Langley has been conducting and sponsoring research on flight operations and Flight Management Systems (FMSs) of advanced transport airplanes for a number of years.During the course of this joint research, operational issues have been a primary concern; these include the practical integration of Flight Management System concepts to permit fuel efficient operations in a time-based ATC environment.The primary focus has been on the transition from en route cruise to the arrival phase of flight because of the significant impact of terminal area constraints on the en route trajectory.Concepts for airplane-ATC automation integration were evaluated in two real-time pilotedcockpit ATC simulations described in references 2 through 5. Early studies focused on the development and evaluation of automation functions and procedures for integrating CTAS, FMS, and data-link systems in the extended terminal area.The emphasis was on time-based traffic management, long lead-time (approximately 20 min) conflict prediction, and efficient conflict resolution in the en route and arrival phases of flight.A central issue to integration of FMS and ATC automation is the accuracy of the trajectory prediction process used by each system.CTAS uses trajectory predictions of each airplane to schedule arrivals, ensure conflict-free trajectories, and provide suggested speed, altitude, and routing clearances to maximize throughput with minimum deviation from user preferences.Airborne FMS trajectory predictions are used to provide economical flight profiles which satisfy airplane performance restrictions while adhering to operational constraints.Early piloted-simulation testing of CTAS trajectories with airline flight crews demonstrated favorable results in terms of arrival time accuracy at a terminalarea metering fix (refs. 6 and 7).These tests, however, evaluated CTAS trajectory predictions based on ideal knowledge of airplane state, airplane performance, and atmospheric characteristics (winds and temperatures aloft).The next step was to evaluate CTAS trajectory prediction accuracy under realistic field conditions including the errors associated with radar tracking, airplane performance modeling, and atmospheric modeling.The establishment of CTAS field sites at several FAA ATC facilities provided an opportunity to exercise CTAS under actual traffic and weather conditions.However, accurate airplane and atmospheric state information was not available for trajectory prediction validation.Following the initial fielding of CTAS at the Denver Air Route Traffic Control Center (ARTCC or Center), it was recognized that the Transport Systems Research Vehicle (TSRV) Boeing 737 airplane based at Langley Research Center could be used for actual flight test verification of the CTAS trajectory prediction process.Use of the TSRV airplane provided several advantages including the opportunity to exercise CTAS clearance advisories (with minimum impact on the airspace users), a platform for the accurate measurement of actual airplane and atmospheric state, and the ability to evaluate new cockpit procedures in a flight environment.Ames began conducting field tests of the descent advisor (DA) portion of CTAS in 1992.Designed for Center airspace, DA provides clearance advisories for traffic management restrictions (e.g., metering) while assisting the controller with the detection and resolution of conflicts between airplanes in all phases of flight (ascent, cruise, and descent).The primary goal of these tests was to evaluate the accuracy of the CTAS trajectory prediction process for the en route arrival phase of flight.Two TSRV flight experiments were conducted: Phase I in October 1992 and Phase II in September 1994.This report describes both phases and presents results in terms of the trajectory prediction accuracy and the sources and magnitudes of trajectory prediction errors.Although the combined flight test data set is not large enough to be statistically significant, the data do provide insight into the size and impact of errors associated with trajectory prediction under realworld operating conditions.These data can be used as input and validation for trajectory sensitivity studies to determine the statistical representation of errors (refs. 8, 9, and 10).The results of such studies can be used to guide improvements to prediction algorithms and data sources (e.g., prediction of atmospheric characteristics and airplane tracking), determine the appropriate buffers for conflict prediction, and develop trajectory prediction error models for real-time analysis of conflict probability. +BackgroundCapacity and efficiency improvements in the national airspace system are needed to cope with increased traffic demand and ensure the economic viability of the air transportation industry.Airborne flight management systems have been developed to provide cost-efficient flight guidance for individual airplane operations.Air traffic control automation tools (decision support tools) are currently being designed to assist controllers in achieving greater efficiency with current ATC procedures as well as enable the introduction of new, more efficient procedures.Such tools include conflict prediction and resolution tools, for allowing more user-preferred flight paths, and time-based traffic management tools for minimizing delay.Both the FMS and ATC automation systems share the common need for accurate prediction of airplane flight trajectories in order to achieve their respective performance goals.The focus of this publication is on the CTAS trajectory prediction process, with reference and comparison with airborne FMS as deemed appropriate. +Center-TRACON Automation SystemCTAS is an integrated system comprised of three tools that provide computer-generated advisories for both en route (Center) and terminal (TRACON) controllers (ref.1).The three tools include the Traffic Management Advisor (TMA), the Descent Advisor (DA), and the Final Approach Spacing Tool (FAST).These tools are designed to assist controllers in achieving greater efficiency in the management and control of arrival traffic in the extended terminal area as well as assist in the conflict prediction and resolution of traffic along airway and user-preferred trajectories.As flights approach their destination (e.g., within 200 n.mi.), DA predicts the trajectories of airplanes in Center airspace.The TMA then generates sequences and schedules for arriving flights including those that originate from nearby feeder airports.DA iterates on speed profile, in addition to path and altitude, to provide the Center controller with clearance advisories that meet the TMA schedule with fuel-efficient cruise and descent profiles.DA conflict prediction and resolution tools assist the controller in separating traffic in all en route phases of flight (climb, cruise, and descent) while minimizing clearance changes.As airplanes enter the terminal area, FAST updates the sequences and schedules and provides TRACON controllers with advisories for runway assignment, sequence, headings, and speeds to optimize the delivery of airplanes to the runways. +CTAS Trajectory Prediction ProcessThe trajectory prediction process is the foundation of CTAS.Because it has been developed from an airborne FMS concept, the CTAS trajectory prediction process is similar in many ways to that employed for an FMS.Whereas an FMS application tends to focus on trajectory optimization for a single airplane, the ATC application must also consider the interrelationships of trajectories of multiple flights.The ATC application goes beyond the single focus of required time of arrival (RTA) for time-based traffic management and must consider separation between neighboring flights along entire trajectories not just at procedurally controlled focal points such as a metering fix.The task of reliable conflict prediction along random 4D trajectories is critical to achieving the benefits associated with the "free-flight" concept (ref. 11).The effectiveness and efficiency of conflict resolution actions depend on the accuracy of the trajectory predictions used for conflict detection.CTAS trajectory synthesis begins with the trajectory initial condition and a series of flight path constraints.The initial condition (position, altitude, and velocity) is based on airplane track (radar or airplane reported) or flight plan data.The set of flight path constraints is based on a series of waypoints and segments which define the bounds of a horizontal path to the runway or trajectory end-point.The horizontal path prediction is based on the current state of the airplane, flight plan, airspace procedures, and heuristics which relate the current state of the airplane to the flight plan and local ATC procedures.For exceptional cases where the CTAS heuristics do not match controller intent, the controller may update the CTAS path prediction with quick keyboard and graphical inputs that are separate from the formal Host flight plan amendments.The waypoint constraints, generated to comply with ATC procedures as defined in a CTAS navigation database, may include altitude, airspeed, course, and/ or time.CTAS trajectories are synthesized in two steps.First, a horizontal ground track is generated by curve fitting the waypoints with a series of straight-line and circular-turn segments.The waypoints are designated as either "fly-by," or "fly-over" based on the CTAS navigational database adapted for a particular airspace.The turn segments are based on a parameterized bank angle and an estimated ground speed.This ground speed is computed from an airspeed profile and a wind estimate along a simple kinematic altitude profile.The airspeed profile is either inferred from a combination of flight plan, controller input, and the CTAS database or selected for time-control iteration.Second, the altitude and time profiles are computed by integrating a set of simplified point-mass equations of motion along the established ground track.Within Center airspace, a detailed set of airplane performance models is used to determine thrust, drag, and speed envelope as a function of airplane type.The atmosphere is modeled with a three-dimensional grid of wind, temperature, and pressure (ref.12).A detailed description of the CTAS trajectory synthesis process is presented in references 13 and 14. +Error SourcesTrajectory prediction accuracy is the key for creating effective and efficient ATC advisories.Errors refer to the difference between the predicted and actual airplane state along a flight path.Error sources include the estimation of an airplane state (position and velocity) for initializing a trajectory prediction, trajectory modeling, and clearance conformance.Trajectory modeling includes airplane performance (e.g., thrust, drag, weight), flight procedures, atmospheric characteristics (e.g., wind and temperature aloft), and trajectory generation algorithms.Although both CTAS and FMS are subject to errors, differences between the two systems depend on the environment and application.If the basic trajectory generation algorithms are assumed similar, the differences between FMS and CTAS predictions are primarily due to differences in the sensors and modeling databases used by either system.Whereas the most accurate sensors for determining airplane position and velocity are available to the FMS, ATC systems are currently dependent on less-accurate radar tracking.As for winds and temperature, FMS-equipped airplanes typically have the most accurate data at the current position of the airplane whereas ATC systems have access to the latest prediction over the future flight path, particularly the descent profile.Most FMS systems allow the flight crew to enter forecast winds and temperatures at each waypoint along a flight plan, as well as at several altitudes spanning the descent profile.A few newer airplanes support automatic uplink of these winds and temperatures; however, such data are rarely updated in flight and may be 3 to 6 hr old upon entry.Regarding airplane performance modeling, most FMS systems have extensive performance data which may be "tuned" to the airframe and engine.In comparison, ATC systems must rely on engineering data when available or synthesized data when they are not.Given the current FAA flight plan procedures, ATC systems must estimate weight (usually known to the FMS) and must categorize airplanes within FAA designated types.Many of the differences between CTAS and a particular FMS may be mitigated through the use of data exchange to provide increased precision between the air and ground computations as well as an overall increase in trajectory prediction accuracy (ref.15). +Experiment Design +ObjectiveThe primary objective of the flight tests was the evaluation of CTAS trajectory prediction accuracy for the en route arrival phase of flight, including identification and measurement of significant potential error sources.Secondary objectives included investigation of flight procedures as well as the application of cock-pit automation tools for improving flight precision in descent. +Phase IPhase I, October 1992, focused on straight-path descents with an emphasis on the analysis of modeling errors.In addition, the basic descent procedures tested in simulation would be used for the first time in a flight environment.Flight-idle descent procedures were used to isolate modeling errors, and "constrained" descents were flown to investigate flight procedures for efficient vertical profile control to a required altitude and speed at a fix.Constraineddescent procedures were evaluated with and without cockpit automation for visualizing the bottom-ofdescent crossing restriction.A limited FMS capability, consisting of lateral navigation (LNAV) and guidance along the straight path and navigation map display of range to intercept of a selected altitude, was used for the cockpit automation in Phase I. +Phase IIThe primary objective of Phase II, September 1994, was to evaluate CTAS trajectory prediction accuracy along a more complex arrival route with expanded flight procedures and a wider range of FMS capability for LNAV and performance-based vertical navigation (VNAV).The arrival route was chosen to provide a large turn during the middle of the descent.Previous simulation testing at Ames (ref.6) had shown that pilots without LNAV exhibit a tendency to overshoot the turn and subsequently fly a longer than predicted path.Imprecision in the pilot overshoot presents an additional challenge in accurately predicting the lateral path of a conventionally equipped airplane.The intent was to determine whether the lateral errors observed in the earlier simulation tests and the vertical errors observed in Phase I could be reduced by improved piloting procedures and what additional improvement could be gained by utilizing FMS LNAV and VNAV capability.A secondary objective of Phase II was to sample actual atmospheric conditions for comparison with the CTAS model along the arrival test route as well as at additional locations in the test airport vicinity. +ApproachThe test was designed to expose DA to realistic modeling errors under field conditions with minimum impact on the ATC facility and commercial flight operations.During both test phases, the TSRV was operated on an arrival flight plan tailored to replicate a typical commercial airline arrival at Denver.Each test flight consisted of several test runs conducted by using a closed-circuit routing designed to maximize the amount of data collected on a given flight.The TSRV was flown from both the forward flight deck, representing a conventionally equipped airplane (e.g., Boeing 737-200, Boeing 727-200, McDonnell Douglas DC-9/MD-80), and the research flight deck, representing an FMS-equipped airplane (e.g., Boeing 737-400, Boeing 757/767).Test runs were conducted during low traffic periods to minimize the impact on commercial flight operations and to allow the TSRV to conduct uninterrupted descents.Although interruptions commonly occur as a part of normal ATC operations, isolating the TSRV was desirable to enable identification and measurement of trajectory prediction error sources.CTAS was operated by a test engineer due to the absence, at that time, of an FAA-approved CTAS interface for the radar controllers.The approach was for the TSRV pilot and controller to coordinate pilot discretion (PD) descents while the CTAS operator relayed the DA advisories to the TSRV over a dedicated (non-ATC) frequency.CTAS was operated with data sources that represent the quality of data available to a current operational system.Airplane track and flight plan data were obtained by CTAS through established operational interfaces to the ATC Host computer.For the TSRV airplane, CTAS used manufacturer's performance data.The performance data included drag, thrust, and fuel consumption as a function of airplane and atmospheric state.Atmospheric data (winds and temperature aloft) were obtained from the National Oceanographic and Atmospheric Administration (NOAA) Mesoscale Analysis and Prediction System (MAPS) (ref.16).MAPS is the research prototype version of the Rapid Update Cycle (ref.17) operated by the National Center for Environmental Prediction (NCEP), formerly the National Meteorological Center (NMC).For Phase II, the TSRV FMS used data from different sources than CTAS, which were also the most accurate sources of data available.Airplane state data were taken directly from airplane measurements, atmospheric data were entered into the FMS by hand based on the measurements of previous runs, and the performance data were based on data from earlier flight tests.These differences in input data between CTAS and the TSRV FMS were used to ensure differences in the respective trajectory predictions.This approach provided two advantages:1.It would highlight the potential differences between CTAS and FMS trajectories under operational conditions 2. It would provide insight into the sensitivity of trajectory prediction accuracy to the accuracy of these data sources Airplane state and observed atmospheric data were recorded onboard the TSRV airplane for postflight comparison with the real-time CTAS trajectory predictions, airplane track, and MAPS data.Throughout this report, the term "actual" refers to the measurements made onboard the TSRV airplane with the Global Positioning System (GPS) navigation system. +Flight Test Area +Phase IThe area of test operations for Phase I, including the nominal flight path of the airplane, is shown in figure 1.The test was confined to one area (group of sectors) within Denver Center and primarily involved two radar sectors.The high altitude sector 9 (HA9) sets the sequence of arrivals from the northeast and controls the airspace including flight level (FL) 240 and above.Arriving flights are typically handed off to the low altitude sector 15 (LA15) for metering into the Denver TRACON via the KEANN metering fix.A flight plan was developed, with the assistance of the Denver Center and TRACON controllers, to allow for a closed-circuit routing using jet route 10 (J10) for the test runs and the airspace southeast of J10 for climb out and prerun maneuvering.The nominal plan was to depart from Denver Stapleton International Airport, proceed direct to AKO (Akron VOR station), direct to LEWEL, direct to PONNY, direct to Denver Airport.The test run was conducted between the initial point (IP) at PONNY and the TRACON boundary at KEANN.The actual flight path between AKO and PONNY varied from run to run, depending on the climb performance of the TSRV and traffic conditions, to enable the TSRV to be stabilized in cruise at the IP.Descents were initiated from FL350 with a metering fix crossing condition at KEANN of FL170 at or below 250 KCAS.Pressure altitude was used throughout the descent to remove the step change in altitude effect from the data analysis for this test phase.After crossing KEANN, the airplane would either climb eastbound for another run or return to Denver for landing. +Phase IIFigure 2 illustrates the Phase II area of test operations along with the nominal flight path.This test was conducted primarily in the northwest area.The high altitude sector HA14 sets the sequence of arrivals from the northwest and controls the airspace including FL240 and above.Arriving flights are typically handed off to the low altitude sector LA13 for metering into the Denver TRACON via the DRAKO metering fix.In Phase II, the primary test runs were flown along J56 with the airspace to the south used for climb out and prerun maneuvering.The descents were initiated from FL330 with a metering fix crossing condition at DRAKO of 17000 feet at or below 250 KCAS.During Phase II, the proper altimeter setting was used to determine metering fix crossing altitude.The initial point for the primary test runs was at CHE (Hayden) VOR.A second route, beginning at IP2, joined the arrival traffic inbound to the KEANN metering fix.This second route was used to obtain additional atmospheric data with the TSRV from a different quadrant.Runs conducted along this secondary route were not used to complete the primary test matrix of descent trajectory cases. +Research SystemThe primary equipment used for these tests consisted of the TSRV airplane operating in the Denver terminal area and the CTAS field system on the ground at Denver Center.In addition to standard twoway voice communication between the pilots and ATC, a dedicated frequency was used to support two-way voice communication between the TSRV airplane and the CTAS ground station. +TSRV AirplaneThe airplane used in these tests was the TSRV airplane, a modified Boeing 737-100 (fig.3).The TSRV is a flying laboratory equipped with a research flight deck (RFD) located in the cabin behind the conventional forward flight deck (FFD), as shown in the cutaway model of the airplane in figure 4. The interior of the RFD is a full-size flight deck that features eight 8by 8-in.flight-quality, color CRT displays and sidestick flight controllers (fig.5).Experimental systems used in the RFD consist of an electronic flight display system, a digital fly-by-wire flight control and flight guidance system, and an advanced area navigation system with GPS sensor inputs.The airplane may be flown from either the RFD or FFD.The TSRV airplane was equipped with a fully capable four-dimensional (4D) navigation and guidance system developed during the mid 1970's in support of the Terminal Configured Vehicle Program (ref.18).This baseline system, however, did not incorporate performance management features necessary for computation of vertical trajectories.Ground speeds and altitudes were required inputs to each waypoint in the guidance buffer of the flight management computer.The system also lacked the flexibility of flight plan generation and modification found in current commercial flight management systems.The system was upgraded in the late 1980's to incorporate modern control display units, as illustrated in figure 6.At the same time, expanded lateral flight plan generation capability was added which closely approximated the functionality of commercial flight management systems.In addition to the lateral navigation features, the navigation display included a rangealtitude arc for displaying the predicted intercept of a desired altitude.This capability was used during Phase I.For Phase II, the capability was added to compute vertical trajectories and provide vertical guidance similar to the commercial Boeing 737-300 commercial systems.This was accomplished with the NASAdeveloped profile generation algorithm (PGA4D) described in references 2 and 4. The time-control (4D) mode was not implemented for this test.In addition, the range-altitude arc was augmented with the capability to display the projected altitude intercept along a curved path, as shown in figure 7.Selection of flight guidance and control modes in the RFD are made through the mode control panel (MCP) located in the center of the glare shield (fig.8).A description of the MCP and baseline guidance modes available in the RFD may be found in reference 19. +CTAS SystemFigure 9 illustrates the test setup within the Denver Center.The CTAS station, located adjacent to the Traffic Management Unit (TMU) on the control room floor, was comprised of a distributed network of Sun Microsystems Sparc-10 workstations.Real-time updates of radar track and flight plan data for arrivals were received from the FAA Host computer via a oneway (Host-to-CTAS) interface.Radar track data (position, mode-C altitude, and velocity) were nominally updated by the Host computer on a 12-sec cycle.MAPS forecasts of winds and temperatures aloft were received from NOAA on a 3-hr update cycle.These forecast updates were received (and used) by CTAS approximately 30 min prior to the forecast period.Host track data were displayed on a CTAS plan view graphical user interface (PGUI) with DA advisory data superimposed on the display in both tabular and color graphical form (ref. 20).For the purposes of these flight tests, the descent speed profiles for the TSRV airplane were selected from a test matrix to provide a controlled set of speed profile conditions to support the analysis of trajectory prediction accuracy.The test matrix speed profiles were input to DA for each run and used to compute a top-of-descent (TOD) clearance advisory.Additional DA functionality, including advisories for cruise speed, cruise altitude, direct headings, delay vectors, and conflict detection and resolution, was not evaluated in these tests.Prior to both Phases, the CTAS/DA trajectory calculations were validated against the FMS/PGA4D calculations.The validation was based on running a series of trajectory predictions, over a range of speed profiles, for a common set of input data (atmospheric conditions and performance data).The comparison trajectories were based on a nominal flight distance of 100 n.mi.with descents that were on par with those to be explored in the flight tests.Results indicated that the two systems produced comparable trajectories with no more difference than 1 n.mi. in top of descent and 2 sec in arrival time. +Test ProceduresThe test procedures used during both Phases were essentially the same.TSRV flights were coordinated with Denver traffic management to allow multiple descent runs during low traffic periods.A list of desired test conditions (including speed profile and cockpit procedure) was prepared prior to each flight.The desired test condition for each run was chosen during the climb phase of the run.Selection of this test condition was a function of the traffic situation, performance capability of the airplane, fuel status, and test matrix completion.During Phase I, the DA conflict probe was used by the test engineer to shadow the arrival traffic and determine which test conditions would allow for an uninterrupted descent.The high altitude controller would then issue radar vectors to the TSRV, prior to the IP, to allow a pilot discretion descent without traffic conflicts.A traffic management controller coordinated test activities between the CTAS station and each participating radar sector.The CTAS test engineer monitored the progress of the TSRV airplane on the DA PGUI.After the airplane crossed the IP, the TSRV test engineer would report the CAS, ground speed, and measured wind for comparison with the test condition and CTAS estimates of the same variables.When the airplane was stable at the desired cruise speed, the CTAS engineer would relay the approximate TOD to the TSRV engineer and high altitude controller.When the airplane was nominally within 20 to 50 n.mi. of the TOD, the CTAS trajectory was recorded and final TOD location transmitted to the TSRV engineer.With the PD descent clearance issued, the TSRV engineer would relay the TOD to the flight crew to simulate the controller's issuance of a DA-based descent clearance.Airborne measurements of actual airplane and atmospheric state were recorded automatically on the TSRV.The flight crew onboard the TSRV airplane consisted of two pilots in the FFD and a single pilot in the left seat of the RFD.The right seat of the RFD was occupied by the TSRV test engineer.All normal ATC communications were handled by the FFD pilots.Communication with the CTAS workstation was handled by the TSRV test engineer.Voice communications to both ATC and CTAS could be monitored by all pilots.Each test condition specified whether the run would be flown from the FFD or the RFD.Prior to reaching the IP waypoint, the flight crew in the appropriate cockpit would assume control of the airplane.All FFD test runs were flown manually by the pilots without the use of autopilot or autothrottle.The RFD pilot used manual control during Phase I and autopilot during Phase II.The pilot began each run by establishing the airplane in level cruise at the appropriate altitude and speed for the test condition.Prior to top of descent, the pilot was advised by the TSRV engineer of the desired TOD in terms of DME distance from the Denver VOR.The pilot would monitor DME distance and initiate descent upon reaching the specified range to Denver.The pilot conducted the descent by using the profile descent tracking procedures specified by the test condition (defined later).The test run ended when the airplane reached the final altitude and speed and crossed the MF waypoint (KEANN or DRAKO). +Data RecordingTwo primary sets of data parameters were collected during these tests:1. Measured conditions, such as airplane state and atmospheric data 2. Predicted conditions, such as trajectory predictions from CTAS DA and the airplane FMS as well as predicted atmospheric conditions Data recording onboard the airplane and at the CTAS workstations was tagged to Universal Time (UTC) for postflight correlation. +Measured DataThe TSRV sensors provided airplane state data, such as position (latitude and longitude), airspeed, ground speed, altitude, body angles, and accelerations.Wind speed and wind direction were computed in real time based on airspeed, ground speed, and body angles.Atmospheric temperature measurement was also provided by the TSRV air data system.Most parameters were updated and recorded at a rate of 20 Hz but were averaged over 1 sec in postprocessing.Airplane tracking data, including position (x,y coordinates in the Denver Center reference frame), mode-C altitude, track angle, and ground speed, were obtained from the Denver Center Host computer with an approximate update rate of one track report every 12 sec (ref.21).Radar track position data were provided to CTAS in the Denver Center reference frame, a stereographic coordinate system with the origin approximately 700 n.mi.southwest of the Denver airport.For the purposes of comparison, TSRV position data were converted to the Denver Center reference frame. +Predicted DataTrajectory predictions were computed and recorded by the CTAS DA for all test runs during both Phases.In addition, the TSRV FMS computed and recorded predicted trajectories for Phase II (FMS predictions were not available in Phase I).Both sources of trajectory predictions provided point-to-point fourdimensional trajectories for each descent from the initial position of the airplane up to and including the metering fix location.CTAS received and recorded the 3-hr MAPS forecast on a 3-hr update cycle.This forecast was received approximately 30 min prior to the forecast period and was based on an analysis of the atmosphere during the preceding period.CTAS obtained the predicted winds and temperature along a flight path by interpolating within the MAPS data grid. +Test ConditionsThe test conditions employed in both tests were designed to provide a reasonable representation of commercial airline jet transport descents as anticipated in a CTAS Descent Advisor operational environment.Cockpit automation and the corresponding pilot procedures were studied to investigate their impact on the descent trajectory.The NASA test pilots were instructed to fly the descents as precisely as possible in order to minimize pilot-induced variations in the descent profiles.The goal was to emphasize the differences between the systems (and associated procedures). +Phase ITwo specific types of descent procedures were used in Phase I: (1) idle descents, in which idle thrust was used from TOD to BOD altitude and metering fix crossing speed and (2) constrained descents in which the pilot employed thrust and/or speed brake during the descent in order to achieve BOD altitude and airspeed as closely as possible to the metering fix location.The purpose of the idle descent procedure was to provide a direct measurement of the trajectory prediction accuracy of CTAS, which utilized an idle descent model in the trajectory predictions for this test.Operational versions of CTAS are anticipated to use a nearidle thrust model for descent trajectory predictions to match the procedures related to individual airplane performance types and operating conditions.The constrained descent procedure represented a more realistic procedure in which the pilot adjusts the altitude profile in descent to achieve the desired crossing conditions (speed and altitude) at a waypoint assigned by ATC.This procedure has the added benefit of mitigating the impact of trajectory prediction errors by closing the loop on the vertical profile.The idle and constrained descents were flown from both the FFD and RFD.All descents were flown manually since the TSRV was not equipped with autopilot functions which held airspeed by using pitch control.The specific procedures used are detailed in the following paragraphs. +Idle DescentThe pilot procedures for idle descents were essentially the same for both the FFD and RFD.The pilot would begin the idle descent procedure when the airplane reached the CTAS-specified TOD point.This point was identified as a DME distance from the Denver VOR.Following TOD, the pilot flew one of three vertical profile types, depending on speed (fig.10).If the descent CAS was less than or equal to the cruise CAS, the pilot flew a slow descent profile (fig.10(a)).At TOD, the pilot would immediately retard the throttle to idle and decelerate in level flight.Once the descent speed was achieved, the pilot initiated a descent while using pitch control to maintain airspeed.If the descent Mach was equal to the cruise Mach, the pilot flew a nominal descent profile (fig.10(b)).At TOD, the pilot would immediately retard the throttle to idle and initiate a descent while using pitch control to maintain the Mach/CAS speed schedule.If the descent Mach was greater than the cruise Mach, the pilot flew a fast descent profile (fig.10(c)).At TOD, the pilot would immediately initiate a descent (nominally 3000 ft/min) while maintaining cruise thrust to accelerate to the descent Mach.Once the descent Mach was achieved, the pilot would retard the throttle to idle while using pitch control to maintain the Mach/ CAS speed schedule.As the airplane approached the metering fix crossing altitude, the pilot would initiate a level-off deceleration segment, depending on the descent speed and metering fix crossing speed.If the speeds required a deceleration, the pilot maintained idle throttle until the airplane approached the metering fix speed and then increased throttle as necessary to maintain speed until crossing the metering fix.If no deceleration was necessary, the pilot increased throttle as necessary to level off and maintain the descent speed until crossing the metering fix. +Constrained DescentThe pilot procedures for the constrained descents were the same as for the idle descents up to the constant CAS segment of the descent.Once the constant CAS segment was established, the pilot would adjust the descent angle to achieve a BOD point which was just prior to the metering fix.The BOD location was chosen by a rule of thumb, to allow 1 n.mi. of deceleration distance for each 10 knots of speed reduction required to achieve the assigned crossing speed at the metering fix.The RFD pilot used the range-altitude arc on the navigation display to target the desired BOD point (fig.7).This arc showed the range at which the airplane would reach the altitude selected on the mode control panel at the current inertial flight path angle of the airplane.The pilot would then adjust throttle and/ or speed brake to hold the descent CAS while targeting the desired BOD location.The FFD pilot procedures for constrained descents were somewhat more complex than the RFD procedures since the FFD pilots had no direct indication of the range at which they would reach the BOD altitude.Commercial crews typically use the 3:1 rule of thumb to plan 3 n.mi. of descent path for every 1000 ft of descent.This rule works well in terms of workload and fuel efficiency for a small range of descent speeds which vary as a function of airplane type, weight, and atmosphere.However, for the CTAS application, it is desirable for ATC to specify descent speed to allow for safe and efficient merging of arrivals.Under these conditions, it is desirable to allow the flight path (e.g., TOD) to vary as a function of descent speed, type, and atmosphere, much like an FMS would.For fuelefficient descents, the TOD and flight path angle may vary as much as 30 to 40 percent over the speed envelope of typical jet transport types.The challenge is for the pilot to maintain a situational awareness of vertical profile progress.Paper charts and a custom-programmed hand calculator were provided to the FFD pilots to assist in the constrained descents.The charts provided tables of DME distance, altitude, and corresponding flight path angles for each of the descent speed conditions in the test.The pilots would determine the required flight path angle to achieve BOD altitude by noting their altitude and DME distance when the airplane reached the target descent CAS.With this flight path angle as a reference, the pilots could then determine the proper altitude at a given DME distance or conversely the proper DME distance at a given altitude needed to maintain the correct decent angle.The descent rate could then be adjusted with throttle or speed brake, depending on whether the airplane was below or above the desired altitude.The programmed hand calculator provided the same information.Both the charts and calculator were developed during local flight testing of the descent procedures as aids for the NASA test pilots.They were not intended to represent operational techniques for airline pilots to use for CTAS descent advisories.Such operational procedures would require careful development and testing with actual airline crews. +Test MatrixThe test matrix for Phase I, given in table 1, was defined to evaluate CTAS trajectory prediction accuracy over two primary test variables: speed profile and pilot procedure.Seven speed profiles were selected to exercise the nominal speed envelope of the TSRV while generating a representative set of constant-speed and variable-speed trajectory segments.This approach was used to generate a balanced set of trajectory cases for analysis of prediction accuracy as well as a broad data set for evaluating the TSRV performance characteristics.Each of the seven speed profiles was flown by using the idle-thrust descent procedure.The first three speed profile cases were repeated with the constrained descent procedures from both the FFD and RFD.The goal was to complete two runs for each of the 13 conditions combining speed profile and pilot procedures. +Phase IITest conditions for Phase II were designed to expand on Phase I with an emphasis on evaluating how to best utilize current FMS capabilities for constrained descents within a CTAS environment.Descents with turns were of particular interest due to the increased complexity of lateral and vertical profile tracking.Three different levels of FMS automation were chosen to represent a cross section of FMS automation capabilities available within the current commercial fleet.These levels represent 1. Conventional airplanes (without FMS) 2. FMS-equipped airplanes with VNAV capability +FMS-equipped airplanes with range-altitude arc capabilityThese levels of FMS automation were simulated by restricting the usage of the FMS on the TSRV at the defined levels.Four sets of pilot procedures were developed for the TSRV to take advantage of these levels of FMS automation.These procedures included 1. Conventional non-FMS 2. Conventional FMS (using FMS TOD) +FMS with CTAS TOD +Range-altitude arcThe TSRV pilot procedures were not intended as exact prototypes for operational use because of the significant differences in the TSRV FMS, pilot interface devices (mode control panel, CDUs, and side-stick flight controllers), and flight control mode (velocity control stick steering) compared with typical commercial equipment.Instead, the procedures were designed to mimic as closely as possible the techniques proposed for use by airline flight crews following CTAS descent advisories.A focused investigation of operational procedures and flight crew human factors was beyond the scope of this test.However, an evaluation of pilot procedures involving commercial airline flights was conducted in parallel with this test phase (ref.22).The test conditions flown in the RFD required significant preparation and pilot training.The RFD mode control panel was designed many years before the development of the performance-based VNAV systems which are common on modern commercial flight decks.The TSRV system is highly flexible, however, and techniques were devised to closely approximate the commercial FMS modes.Flight cards were developed for each test condition with an event sequence of TSRV-specific procedures to be followed in order to mimic the desired commercial FMS functionality.The exact procedures and flight cards used in the test are described in the following sections. +Conventional non-FMSThese conventional non-FMS procedures were designed to represent airplanes which are not equipped with flight management systems.They were flown by the pilots in the FFD.One pilot was designated as the flying pilot and manually flew the airplane from the IP to the metering fix.The other pilot in the FFD handled the nonflying duties, including communication with ATC and the TSRV and CTAS test engineers.A TSRV test engineer (or observer) was located in the jump seat behind the FFD to observe and assist in communication.The flying pilot established the airplane on the inbound leg of the flight plan at the desired cruise altitude and speed prior to crossing the IP.Conventional VOR guidance was used for lateral tracking of the flight plan route.The pilot maintained altitude and speed up to the CTAS TOD.The CTAS TOD was identified as a DME distance to a reference VOR station (DEN).The nonflying pilot tuned a navigation radio to the appropriate station and monitored the DME distance.The flying pilot was instructed to begin the descent procedure within 0.1 n.mi. of the CTAS-specified DME range.At the top of descent, the flying pilot would initiate the descent by retarding the throttle smoothly to idle.If the descent speed was less than cruise speed, the pilot would decelerate in level flight to achieve the desired descent speed.The flying pilot flew the remainder of the descent by using pitch to hold the Mach/CAS speed schedule.Prior to crossing 18000 ft, the altimeter setting was changed to the local altimeter setting.The pilots were instructed to target their BOD to be just prior to crossing the metering fix.Throttle and/or speed brake were used to adjust the descent rate in order to reach BOD with just enough distance to decelerate from the descent CAS to the crossing speed of 250 knots at the metering fix. +Conventional FMSThese conventional FMS descent procedures were designed to utilize the VNAV capability of FMSequipped airplanes to generate and fly a VNAV profile, including TOD, based on the CTAS-assigned descent speed profile.They were flown from the RFD by a NASA test pilot with the assistance of the TSRV test engineer acting as the nonflying pilot.All RFD test runs were flown by using autopilot for lateral tracking of the FMS flight plan in order to provide consistent performance for comparison with CTAS horizontal path predictions.The appropriate flight plan (company route) and prestored approach were entered into the CDU prior to reaching the IP for the test scenario.Measured wind speed, wind direction, and static air temperature were hand recorded at intervals of 4000-ft altitude from 17000 to 33000 ft during the initial climb and on each subsequent descent.The latest data were manually entered into the descent wind page of the CDU for use in the FMS trajectory prediction.(This approach enabled using the FMS prediction to represent the ideal case of minimum modeling error for trajectory prediction, airborne or ground based.)Cruise speed (Mach = 0.72 or 0.76, depending on test condition) was entered as the selected speed on the CRUISE CDU page, and the EXECUTE button pressed to activate the flight plan.The airplane was stabilized at cruise altitude and speed prior to crossing the IP.After crossing the IP, the appropriate test card shown in figure 11 was used to specify the sequence of activities in the RFD.As shown on the card, there were six key events which required specific actions by the pilot and test engineer.The test engineer would monitor the events and call out the activities.The pilot would cross-check and confirm the activities.Typically the test engineer would perform the activities which required CDU entries and the pilot would handle mode control panel, throttle, and flight controller inputs.The test engineer would also handle some mode control panel entries at the request of the pilot.The first event was after the IP and prior to receiving the CTAS descent advisory clearance.The crew verified that the airplane was level at the correct cruise altitude and speed and on path.The mode control panel was set to indicate AUTO, ALT, HOR PATH, and CAS ENG selected.This indicated autopilot engaged with pitch control holding altitude, roll control following the programmed flight plan horizontal path, and throttle holding airspeed.After receiving the CTAS descent advisor clearance from the CTAS test engineer, the TSRV test engineer would select the LEGS page on the CDU to verify the proper crossing restrictions at DRAKO, enter the appropriate descent speed on the DESCENT page, and press EXECUTE to generate an updated trajectory.The CTAS TOD DME distance was entered on the CDU FIX page to display a circle with that radius around the reference VOR.The TSRV test engineer noted the discrepancy, if any, between the CTAS TOD and that computed by the FMS.The MCP altitude was then set to 17000 ft, the crossing restriction at the metering fix.At approximately 10 mi from the FMS TOD point, the autothrottle was disengaged and the DESCENT page was selected on the CDU in preparation for the descent.Upon reaching the FMS TOD, the pilot would bring the throttle to idle and set the MCP CAS to the test condition descent CAS.The autopilot would pitch the airplane to follow the programmed descent path.During the descent, the pilot would use throttle to hold airspeed to within 5 to 10 knots of the desired descent speed schedule.If the airplane speed increased to more than 5 knots above the desired speed, the RFD pilot would request the FFD pilot to deploy speed brakes to slow the airplane.This was necessary since the TSRV RFD did not have direct speed brake controls.The final event occurred near the bottom of descent.Altimeter setting was changed to the local pressure prior to crossing 19000 ft, MCP CAS was set to the metering fix crossing speed (if necessary), and autopilot disengaged prior to 18000 ft.The pilot would then manually level the airplane at 17000 ft and adjust throttle to cross at the desired airspeed. +FMS With CTAS TODThe FMS with CTAS TOD procedures were an extension of the FMS VNAV procedures with the airplane now restricted to initiate descent at the CTASspecified point rather than the FMS-computed point.The primary advantage of the CTAS TOD procedure is that it establishes a predictable TOD for the controller to plan for separation with minimum workload.Four flight cards were prepared to account for the possible situations which could be encountered in the test.These situations were The procedures used for all four situations were the same as the conventional FMS procedures up to the point where the CTAS TOD DME distance was entered into the CDU FIX page.At 10 mi from the CTAS TOD (event 3 on the test card), the pilot would select FPA mode (flight path angle hold) for the autopilot.This selection prevented the autopilot from descending at the FMS TOD and allowed a manually selected descent at the CTAS TOD.Upon reaching the CTAS TOD, the pilot would execute the following descent procedures: CTAS TOD prior to FMS TOD: If a deceleration was required, the throttle would be set to idle and cruise altitude maintained until the descent speed was achieved.A descent angle of -1.5°(adjusted to provide a descent rate approximately 1000 to 1500 ft/min) was set in the MCP to initiate descent and capture the FMS VNAV path from below.Throttle was then used to maintain the descent speed schedule.Once the FMS-computed TOD was crossed, vertical path guidance was selected by pressing VERT PATH on the MCP.The desired FPA was reset to the appropriate value to continue a descent rate of 1500 ft/min until the vertical path was captured.The rest of the descent was flown the same as described for the conventional FMS case. +CTAS TOD after FMS TOD:Throttles were retarded to idle and descent initiated by using the MCP FPA mode.Deceleration to descent speed, if necessary, was done in level flight.Initial target descent angles of between -3° and -6° were selected, based on the descent speed, to capture the FMS VNAV path from above.VERT PATH was then selected to arm vertical path guidance.Descent angle was adjusted as necessary to maintain a reasonable closure on the programmed vertical path.Speed brakes were deployed as necessary to maintain descent speed.Upon capture of the FMS descent path, the speed brakes were retracted and the remainder of the descent was flown the same as described for the conventional FMS. +Range-Altitude ArcThe range-altitude arc conditions were designed to represent descents which do not require FMS VNAV to achieve the proper BOD.Instead, the so-called range-altitude arc would be used to target BOD, with CTAS providing the TOD.The goal was to explore the feasibility of a simple alternative to VNAV for improving the precision of vertical profile conformance.Figure 13 shows the flight cards used for these procedures.These procedures were similar to the constrained descents flown from the RFD during the Phase I flights.During this test, however, the range-altitude arc was modified to show the projected range along the FMS lateral path at which the airplane would reach the MCP altitude (fig.7) in addition to the straight-line distance.This modification allowed the pilot to more accurately target the proper BOD location during the early stages of the descent.Also for this test, the RFD pilot had the FMS-computed TOD to assist in determining the possible throttle and/or speed brake control activity needed during the descent.An early descent would generally require throttle, whereas a late descent would need some speed brake.As seen in the flight cards, the procedures for early and late descents were identical, with only the wording in step 5 modified to indicate the expected primary speed control device. +Test MatrixThe Phase II test matrix, as in Phase I, was based on two primary test variables: speed profile and pilot procedure.Table 2 presents the 12 conditions defined by the combination of 3 speed profiles and 4 procedures.The goal was to complete two runs of each of the 12 conditions combining speed profile and pilot procedures.In addition, as time permitted, several flights into the northeastern arrival gate (KEANN) at Denver were conducted to collect atmospheric data away from the Rocky Mountains. +Results and DiscussionThe TSRV Boeing 737 airplane was deployed on two separate occasions to Denver Stapleton International Airport for these tests.During each deployment, the airplane conducted multiple descents from cruise altitude into the Denver terminal area while the CTAS field system at Denver ARTCC provided real-time descent advisories.Phase I included 23 descent runs conducted during 7 flights over a period of 1 week in October 1992.Nine runs were conducted during two night flights, and the rest were day flights.Three additional runs were excluded from the analysis due to experimental system errors encountered while conducting the runs.Table 3 provides a summary of the test conditions completed for Phase I.Weather conditions during Phase I were generally good, with no adverse conditions encountered which delayed or canceled a planned flight.The most significant weather events encountered were strong jet stream winds during the two night flights (R679 and R680), with pronounced wind gradients during descent.The impact of these winds is discussed in section 6.1.3.Phase II included 25 descents conducted during 9 daylight flights over a period of 1 week in September 1994.Four additional runs were conducted to collect atmospheric and radar tracking data in another area and one additional run was conducted to investigate a mid-descent correction in speed profile.An additional six descent runs were initiated but aborted because of experimental system errors and ATC interruptions encountered in conducting the runs.Table 4 provides a summary of the test conditions completed for Phase II.A variety of weather conditions were encountered during Phase II.Light winds and stable atmospheric conditions prevailed for the first 2 days (flight R728 and R729).Convective buildups and slightly stronger winds were encountered during flight R730, with storm cells and light rain near the turn at ESTUS during descent.On flight R732, a frontal passage, associated with a brief snow storm in the Colorado area, provided strong and variable winds aloft and forced early termination of the flight.The following day (flight R733) was clear with strong, steady northerly winds at all altitudes.High pressure dominated the area throughout the test period with altimeter setting above standard each day.The analysis of the results from these flight tests is divided into four major sections.First, the trajectory prediction error sources encountered during the test are examined.Second, the actual flight trajectories are compared with the CTAS predictions to determine the overall accuracy.Third, a sensitivity analysis of the modeling error sources is performed to identify their contributions to both metering fix arrival time and vertical trajectory errors.The sensitivity analysis involved recomputing the idle descent trajectories of Phase I by using combinations of updated performance and atmospheric models using both the CTAS trajectory synthesis program and the TSRV flight management profile generation algorithms.Finally, the error sources and their impact on trajectory prediction accuracy are summarized.An additional source of error, in section 6.1.5,also affected test results.Unlike the four basic error sources, these errors were due to problems uniquely attributable to the experimental nature of the CTAS field system used for these tests. +Error Sources +Radar Tracking ErrorsUntil more accurate track data become available (via airplane data link reports or improved radar tracking algorithms), CTAS will depend on FAA Host radar track data to initialize trajectory predictions.The track data provide the airplane position, altitude (mode-C), and inertial velocity (ground speed and track angle).Errors in the current radar tracking system translate directly into initial condition errors for CTAS.Determination of the nature and magnitude of the radar tracking errors is therefore of significant importance to the CTAS project as well as other ground-based trajectory prediction tools.Actual airplane state conditions, as measured by the TSRV during these flight tests, were compared with the ATC radar track data provided to CTAS from the ATC Host computer.During Phase I, TSRV data were only recorded during the actual test runs; this limited the data to nonturning conditions in which the airplane was heading directly toward Denver.During Phase II, TSRV data were recorded continuously throughout each flight; this allowed a more comprehensive analysis of radar tracking errors under conditions that included climbing, descending, turning, and accelerating segments of flight.Errors in radar track to TSRV flight data are presented in three tables.Errors are expressed as airplane measurements minus radar track.Table 5 presents the summary of radar tracking errors for both Phases at the initial and final conditions used for the CTAS trajectory predictions.These differences represent the sole contribution of radar tracking errors to the CTAS predictions evaluated in these tests.Tables 6 and7 present similar data for position and velocity, respectively, based on the entire set of flight data collected during Phase II.These data represent the potential errors that affect trajectory prediction and conformance monitoring in en route airspace.Table 5 presents both the velocity and position errors at the initial and final conditions associated with the CTAS predictions in these tests.The initial condition errors (Mean + Standard deviation) for both Phases were less than 10 knots in ground speed and 8°i n track angle.Although these errors are small for the Host track data (typical of level unaccelerated flight at cruise), the ground speed error provides a direct contribution to CTAS accuracy.An error of 10 knots for a typical jet airplane operating at a ground speed of 450 knots translates into an error of 18 sec for every 100 n.mi. of cruise.The final condition (metering fix) velocity errors listed in table 5(b) do not affect the accuracy of CTAS but are indicative of the tracker accuracy during level-flight deceleration segments.Particularly notable are the ground speed errors which were due to the transients in velocity associated with the descent and level-off deceleration to the metering fix.The position error shown in table 5 was the absolute range difference from the GPS-measured location of the airplane to the radar tracked position of the airplane.The along-track error is the projection of the position error along the instantaneous track angle of the airplane.The cross-track error is the component of the position error normal to the airplane track angle.As seen in the table, nearly all the position error was contained in the along-track error component.An "equivalent" time error was computed by dividing the along-track error component by the airplane ground speed at that position.Essentially, the radar-tracked position of the airplane was lagging the actual airplane position by this equivalent time error.The position errors in table 5(a) add a direct contribution to CTAS trajectory prediction error, whereas the errors in table 5(b) represent the errors that would be included if the Host tracker was used to measure the end-point accuracy of the trajectory prediction.From a controller's point of view, the mean along-track errors would essentially cancel themselves while the variation will most likely introduce some error.From an air-ground integration (trajectory exchange) point of view, both the mean and variation in along-track error will affect trajectory prediction accuracy if not accounted for.Some of the equivalent along-track time error is attributed to the lack of a time stamp on the track data received from the Host computer.CTAS processing must assign its own time stamp based on the time of receipt.Since the Host transmits track data to CTAS in batches, the CTAS time stamp estimate may be off by as much as one update period (approximately 12 sec).The data in table 5 were generated based on the initial and final conditions of the test runs listed in tables 3 and 4. A summary of all radar-track position errors from the Phase II flights is given in table 6.As seen in table 6, the track position errors were extremely consistent throughout all the flights.The average along-track error of about 0.7 n.mi.was slightly less than recorded at the CTAS initial condition point because it includes flight at all altitudes and speeds.The CTAS initial conditions were recorded at cruise altitude with the highest ground speeds resulting in larger along-track errors.The along-track error of 6 to 7 sec was consistent for all conditions.The cross-track error was also consistent for all conditions and was relatively insignificant.Table 7 presents the ground speed and track angle errors associated with level flight, altitude change, and turning segments for all data collected in Phase II.The turning segments are further divided into turn and postturn segments.Turn segments are defined as a segment where the actual airplane turn rate exceeds 0.5 deg/sec.Postturn segments are defined as segments which immediately follow a turn segment and continue until the radar tracking ground speed error falls below a value of 10 knots.The altitude change segments are defined by segments involving ascent and descent rates greater than 100 ft/min and not in a turning segment.Level flight segments are defined as everything else (constant altitude and not in a turning segment).For level flight segments, for which the CTAS initial conditions were a subset, the mean ground speed error was approximately 2 knots with a standard deviation of about 12 knots.These segments included level, unaccelerated flight, as well as level acceleration and deceleration segments.The differences between these level flight data and the ground speed errors in table 5 were caused by several factors.Table 5 included a very small subset of the data in table 7 (less than 4 percent).Table 5(a) represents unaccelerated flight, whereas table 5(b) represents level deceleration segments at the peak of the deceleration transient in radar track ground speed.Comparatively, the ground speed errors during altitude change segments were nearly the same as the errors for level flight segments.For turning segments, ground speed errors were substantially greater, with the tracker ground speed less than actual ground speed.The mean error was 37 knots during the actual turn with a standard deviation of 59 knots.During the postturn segments, the error was observed to be significantly greater in mean with about the same variation.The larger postturn mean error was caused in part by the segment definition as well as the characteristics of the tracker.By definition, the postturn segment included ground-speed errors of at least 10 knots (the 10-knot criterion was considered reasonable in order to separate the relatively large turn-induced errors from the normal variation experienced in level flight).Regarding tracker characteristics, the initial error growth lags the actual start of the turn and the maximum error tends to occur just after the actual turn is completed.Both these lags tend to reduce the mean error measured during the turn compared with the mean error in the postturn.The length of the postturn segment was observed to be quite variable and dependent on the size of the turn, magnitude of the ground speed error, and acceleration rate of the airplane following the turn.For the data shown in table 7, there were 45 turns analyzed, with turn angles ranging from 3° to 305°.Mean turn angle was 68° with the average length of the postturn segment being 93 sec.In comparison with the position errors, velocity errors may have a greater impact on trajectory prediction accuracy, particularly for cruise flight where the track velocity is used to infer the velocity for that segment of the trajectory.For example, each 15 knots of error results in an along-track prediction error growth rate of 0.25 n.mi./min (5 n.mi.for a 20-min prediction).Controllers, who accept these velocity anomalies as a part of their job, have learned to anticipate and filter out the errors from their decision making and/or provide larger separation buffers to protect against anomalies.To the extent that these anomalies may be reduced or filtered, automation may be able to lead to a reduction in excess separation buffers.With regard to track angle errors for both level flight and altitude change segments, the track angle errors exhibited a negligible mean with a standard deviation of about 5°.For turning segments, the angle error was substantially greater as was seen for the ground speed error.During both turn and postturn segments, the mean error was observed to be approximately 5° with a standard deviation of 28° and 13°, respectively.The difference in variations is explained by the observation that the track angle error tended to die off before the ground speed error did.Because the postturn segments were defined based on ground speed error, the track angle computation included a considerable number of data points with relatively little error. +Airplane Performance Model ErrorsThe CTAS trajectory synthesis algorithms use detailed models of airplane drag and idle thrust to compute descent trajectories.Drag is represented by high-speed drag polars providing drag coefficient as a function of lift coefficient and Mach number.Thrust is modeled as a function of engine setting, Mach number, altitude, and temperature.For this test and airplane type, the CTAS descent prediction was nominally based on an idle-thrust engine setting.Langley has developed performance models for the Boeing 737-100 airplane suitable for use in trajectory generation programs for airborne flight management systems.These models are based on manufacturer's performance data for the generic Boeing 737-100 airplane.These models were used to generate data tables of drag coefficient and thrust for use by the CTAS trajectory synthesis program.The performance of the TSRV airplane was known to differ from that expected from the generic data.The airplane was the original prototype for the Boeing 737-100 series of jet transports and was well over 20 years old at the time of these tests.In addition, this airplane has numerous external antennas and exposed rivets on the fuselage which were not present during the original performance testing by the manufacturer.Langley had previously developed adjustments to the baseline Boeing 737-100 performance for use in the airborne flight management system to account for the degraded performance of the airplane.These adjustments were not included in the data used by CTAS during the flight test experiment.These adjustments were excluded from CTAS in order to introduce performance-model error into the test.Operational airplanes, of the same type, are expected to vary in actual performance due to age as well as equipment variation (e.g., power plants, antennas, and airframe modifications).The stabilized cruise and descent conditions flown in Phase I were used to refine the performance model of the airplane to reflect the actual performance measured during the test.Data tables were then generated by this revised performance model for use in the sensitivity studies described later in this report.The appendix describes the methods used in updating the airplane performance model and presents the resulting modifications made to the thrust and drag models.The actual TSRV drag differed from the manufacturer's performance data by approximately 11 percent (greater).The idle thrust also differed with a variation over altitude.The combined effect on the descent performance of the airplane was, on the average, a 5-percent lower value of net TMD, which resulted in a 5-percent increase in descent rate.These updated performance data were the basis for the FMS computations in Phase II.In addition to thrust and drag, CTAS estimates the airplane weight to evaluate the point mass equations of motion for the vertical profile calculations.CTAS is capable of estimating the weight of individual flights as a function of time based on knowledge of a reference weight (e.g., takeoff gross weight) and fuel-burn estimation.It is anticipated that the reference in-flight weight could be made available to CTAS via a new field in the files flight plan or by data link.Until the FAA infrastructure is in place to supply a reference weight, CTAS relies on an estimated weight as a function of airplane type and phase of flight.For descents, a typical descent weight is used for descent calculations.For the flight tests, a typical descent weight of 85000 lb was used for all runs.For the Phase I idle runs, the average weight of the TSRV was 83560 lb with a standard deviation of 4380 lb. +Atmospheric Modeling ErrorsCTAS trajectory prediction accuracy depends, in large part, on the accuracy of the atmospheric model data it receives from external sources such as MAPS.Atmospheric characteristics (winds and temperature), as a function of position and altitude, affect CTAS trajectories in several ways.Winds aloft form the basis of predicting the ground speed profile, as a function of airspeed and path, as well as estimating airspeed from radar-based ground speed.Wind gradient, with respect to altitude, can also have a significant influence on rate of ascent and descent.Temperature profiles and altimeter setting are used to determine geometric altitude, as a function of pressure altitude and position, to provide an inertial basis for integrating the point mass equations of motion over ascent and descent segments.Temperature is also used to correct performance data for nonstandard temperatures and convert between TAS and Mach/CAS.Atmospheric modeling errors were determined by comparing the airplane measurements of winds and temperature with the CTAS interpolated model values at specific altitudes along the predicted descent trajectory.Figure 14 summarizes the altitude profile of air temperature with measurements and corresponding model errors for all flights in both Phases.These data are presented in pressure altitude intervals of 2000 ft in terms of the mean value and standard deviation for each Phase.The temperature profiles were similar for both Phases.Compared with the standard atmosphere, the profiles tended to be warmer with a greater gradient (lapse rate) in temperature with altitude.The mean temperatures ranged from 8° to 9°C above standard at the lower altitudes (17000 ft) to approximately 2°C above standard at cruise altitude.The mean errors tended to be within 3°C for Phase I, with greater accuracy at the lower altitudes, whereas the errors in Phase II were within 1°C.These temperature errors, although only representative of a small sample of realistic atmospheres, were considered to have a negligible effect on the trajectory prediction accuracy results.The data are presented in terms of the mean and standard deviation of the wind, at common altitudes, over each descent run of a particular flight.The cruise altitude data are presented slightly differently for each Phase.For Phase I, a single data point (mean and standard deviation) is presented at cruise altitude based on the mean wind over the cruise segment of each run.The average length of the Phase I cruise segments was 9.8 n.mi.with a standard deviation of 6.5 n.mi.For Phase II, the cruise winds are presented at three positions corresponding to the analysis gates introduced in section 6.2.These data points include the initial condition, TOD, and a position in the middle of cruise.The average length of the Phase II cruise segments was 21.3 n.mi.with a standard deviation of 7.0 n.mi.The variation in measurement (between and within the Phase II cruise data points) may be due to several factors that include variation in wind with position, variation in wind with time (at a position), and measurement error.Airborne measurements of wind tend to be more accurate in the along-track component and during steady-state (nonturning) flight.Figures 17 and18 present the differences between measured winds and the CTAS model winds for Phases I and II.These data include the along-track component of the wind error to better illustrate the wind contribution to trajectory prediction errors.In some flights (figs.17(c), 17(e), and 18(e)), the alongtrack wind-error component was relatively small compared with the total wind error.In particular, flight 732 (fig.18(e)) experienced a total wind error greater than 60 knots at the higher altitudes with negligible alongtrack wind errors.The unusually large variation in along-track error at cruise altitude in flight 729 is due to the CTAS interpolation error described in section 6.1.5.A composite of all wind errors for Phases I and II is shown in figure 19.Although the mean errors tend to indicate that CTAS/MAPS does a better job of predicting the winds along the descent at lower altitude than at cruise, the variations are relatively large.These variations, coupled with a relatively small data set representing a few atmospheric conditions, make it difficult to interpret atmospheric prediction performance.Several of the Phase II runs were analyzed further to determine what errors, if any, were contributed by the CTAS processing of MAPS data (ref.23).Results indicated that although CTAS processing of MAPS data contributed a measurable amount of error, the errors in the MAPS data (compared with the TSRV measurements) were substantially greater.For example, analysis of flights 729, 730, and 732 indicate that the CTAS-processed winds had a combined rootmean-square (rms) wind error of 21 knots compared with 18 knots for the actual MAPS data.Figure 20 shows the differences between measured winds and those entered into the FMS during Phase II.These data are used to support the analysis of the TSRV FMS-based trajectory predictions in section 6.2. +Pilot ConformanceThe pilot conformance errors are related to the accuracy of the pilot's tracking (manually or automatically) of the clearance speed, TOD, and course.The TSRV airplane was flown by NASA pilots who were instructed to fly as accurately as possible in order to minimize piloting errors and isolate the effect of the other error sources.Table 8 presents the overall pilotinduced speed errors for both Phases I and II.The data represent the mean and standard deviation of speed error sampled at a rate of once per second throughout the cruise, constant Mach descent, and constant CAS descent segments for the FFD and RFD runs.As seen in the table, the pilots were able to follow the CTAS speed schedule with a high degree of accuracy and effectively eliminate speed conformance error from the flight data analysis.Extension of the results in this paper to commercial flight operations should consider the variation with which line pilots would maintain speed.With regard to TOD, the pilots were careful to initiate the descent procedure no sooner than and within 1 n.mi. of the CTAS TOD advisory.The measurement of actual TOD errors is presented in section 6.2.3.Lateral-path errors (cross track and along track) were not a factor for the straight-path descents in Phase I.For Phase II however, the runs involving conventional VOR radial tracking experienced lateralpath deviations which made a significant contribution to the trajectory prediction error.During these runs, the pilots tracked the radials as precisely as possible and were generally within one needle width of the outbound radial from CHE. Lateral-path deviations of greater than a mile occurred during and after the turn inbound to DEN even though the pilots were using the flight director and course deviation indicator (CDI) to their best advantage.Although no data were recorded on CDI deflection, cross-track error was recorded and is examined in section 6.2.1 as part of the trajectory prediction error analysis. +Experimental System ErrorsThe experimental system errors were introduced during the tests but are not representative of operational errors faced by CTAS.Where possible, corrections for these errors were introduced into the analysis.These errors, and the associated corrections applied to the data, are described in the following paragraphs.During Phase I, three CTAS trajectory predictions were not recorded and had to be regenerated based on the recorded track of the airplane.The recomputed trajectories produced TOD advisories which were within 0.5 n.mi. of the original descent advisory given to the airplane.This difference was considered to have a negligible effect on the Phase I results.The absolute time profile, however, could not be reproduced for the regenerated trajectory data because of limitations in the regeneration technique which was used.In order to properly account for initialization errors, the recomputed trajectories were combined with the actual radar tracking data to determine the initial condition which would have produced the resultant descent advisory.This determination was done by computing the distance to the Denver VOR for each radar tracking point during a test run.The trajectory range from the CTASpredicted trajectory was then used to interpolate on the radar track data to determine the time at which the airplane was at this range according to the radar data.This time was then used as the initial condition for the CTAS prediction.A second problem, affecting all Phase I runs, involved the computation of wind gradient and its effect on the descent rate prediction.A new atmospheric data interpolation scheme was introduced into CTAS just prior to Phase I and the wind gradient computation was inadvertently switched off.This problem, detected in posttest analysis, was corrected prior to Phase II.The impact of this problem was analyzed by using a stand-alone version of the CTAS trajectory generator.A series of descent trajectories were generated with and without the wind gradient computation for a Boeing 737 airplane model.This series of trajectories included along-track wind gradients ranging from 0 to 4 knots/1000 ft.In general, each 1 knot/1000 ft of wind gradient (along track) contributes approximately 3.5 percent to the descent rate.During Phase II a different problem was encountered.Following completion of the flight testing, it was discovered that a change to the Denver radar coordinate system had been implemented in the ATC radar tracking data which had not been added to the CTAS software used during the test.The result was a systematic error of approximately 1.5 n.mi. to the initial conditions used by CTAS.In order to compensate for this error, the TSRV flight data were converted to both the CTAS and Denver ATC radar coordinate systems during data analysis.Radar tracking and lateral-path errors were calculated with the Denver ATC radar coordinates.Comparison with CTAS vertical trajectory prediction was done with the CTAS coordinate system.CTAS initial condition errors for Phase II could not be precisely determined due to the error introduced by the coordinate system difference between CTAS and the ATC radar tracker.Correcting the TSRV flight data to the CTAS coordinates resulted in a lateral offset at the beginning of the trajectory.This offset was an artifact of the coordinate system error and not indicative of the CTAS prediction process under normal conditions.In order to compare CTAS and flight vertical trajectories, the small offset in lateral path was ignored, and vertical trajectory parameters were compared solely based on distance to go along their respective paths.The initial condition errors were assumed to be zero for the trajectory comparisons.An approximation of initial condition errors for Phase II was determined from the comparison of flight and radar tracking data, as described in section 6.2.An error in the initial conditions for a few of the runs in Phase II was introduced by a CTAS software error in the interpolation of the atmospheric model data.This error resulted in an incorrect initial ground speed calculation.The initial cruise airspeed was determined correctly from radar tracking ground speed and atmospheric data models.The cruise trajectory is generated based on either holding the initial cruise airspeed constant or accelerating to an "advisory" airspeed to be held constant.By holding the cruise airspeed constant, CTAS correctly predicts the variations in ground speed caused by variations in wind and course.During cruise trajectory integration, however, the interpolation error resulted in a predicted ground speed that differed from the radar track value at the initial condition.Only the first three runs during flight 729 were affected by this error.An additional systematic error, related to the definition of the metering fix crossing altitude, was introduced into Phase II runs.Although the descents are initiated at flight level altitudes, the bottom of descent is defined by an indicated altitude based on the local altimeter setting correction.For the purposes of this test, the altimeter correction was applied manually.(CTAS software and interface for automatic collection and processing of the local altimeter setting were not available in time for this test.)The correction was applied in the opposite sense throughout the test and the error was not discovered until after the test was completed. +CTAS Trajectory Prediction AccuracyThe trajectory accuracy analysis is based on a comparison between the CTAS-predicted trajectories and TSRV-measured flight trajectories.The analysis is facilitated by the decomposition of the 4D trajectory into five component 2D profiles that are Cross-track profile Along-track profile Altitude profile +Speed profile +Time profileComparisons are accomplished by correlating the profile parameters (e.g., distance flown, speed, altitude, and time) to a common reference path defined by the predicted trajectory.The profile decomposition facilitates the identification of the primary error sources affecting each profile parameter and provides insight into the influence of errors in one profile parameter on another.Analysis of the Phase II runs includes a similar comparison between the onboard TSRV FMS predictions and the measured trajectories flown.The TSRV FMS predictions, based on an updated performance model and atmospheric observations, represent the case of minimal modeling error.Because both TSRV and CTAS predictions result in nearly the same trajectories given the same model data, this approach provides insight into the sources of errors affecting the CTAS trajectories and the potential differences between airborne and ground-based predictions.The comparison of flight and trajectory prediction data (CTAS and FMS) involved a multistep process.First, the flight and FMS prediction data were converted from latitude and longitude to the Denver Center radar-track reference frame used by CTAS.Next, radar tracking errors, which introduced initialization errors to the CTAS prediction process, were quantified (table 5).The actual trajectories were then adjusted to common initialization conditions (position and time) to isolate the errors introduced by other elements of the trajectory prediction process.Finally, the trajectory comparisons were accomplished by referencing the trajectory parameters to a common alongtrack range based on the predicted trajectory.Phase I trajectories were flown direct to the metering fix (KEANN) along a straight-line route.The distance to go to KEANN was therefore used as the common reference for trajectory comparison.The Phase II route involved a more complex path with a turn during the middle of the descent.The FMS-and CTAScomputed lateral paths were nearly the same, with only a small discrepancy at the initial condition (IC) caused by the coordinate system transformation problem described in section 6.1.5.This error, along with the turn radius differences between CTAS and the FMS lateral paths, was found to contribute no more than 0.1 n.m. difference in the calculated distance along the path.The respective range along the reference CTAS and FMS lateral paths was therefore used as the common reference for comparing trajectory parameters for the Phase II data.Differences between the actual and predicted trajectories were computed at specific locations (gates) along the flight path.The analysis gates were defined as reference positions along the predicted path (CTAS or FMS) which vary with the geometry of each trajectory altitude profile.The gates were defined at fixed geographic locations, vertical profile transitions, and at even increments of pressure altitude.Figure 21 illustrates the analysis gates for both Phases I and II.During Phase I, the airplane was stabilized (constant altitude, heading, and speed) in cruise at the PONNY intersection.The initial condition gate (IC in fig.21(a)) was the point at which the final CTASpredicted trajectory was computed.This point varied from run to run.The top-of-descent gate (TODG) was defined as the final point at cruise altitude of the predicted trajectory.TODG represents the same point as TOD except when the airplane must decelerate to its descent speed (the difference being equivalent to the deceleration distance).TODG was chosen for analysis to provide a consistent comparison between runs.The bottom-of-descent gate (BODG) was defined as the point where the predicted trajectory reached the altitude constraint for crossing the metering fix.The trajectory ended at the metering fix (KEANN in fig.21(a)).For Phase II, the airplane was stabilized inbound at the Hayden VOR (CHE in fig.21(b)).The IC was chosen to be the location of either the final CTAS or FMS prediction, whichever was later.An additional analysis gate at the GOULL intersection during the cruise portion of the run was included for Phase II.The TODG and altitude gates were defined the same as Phase I for the CTAS comparisons but were referenced to the FMS predicted trajectory for the FMS comparisons.There was no BODG for Phase II, since analysis of errors at BOD was not significantly different than at the metering fix.The Phase II trajectories ended at the DRAKO metering fix.The ground tracks are presented in terms of the Denver Center x,y coordinate system which corresponds to true east and north, respectively.The TSRV flight data, CTAS predictions, and FMS predictions (Phase II only) were interpolated to provide data corresponding to the gate locations.The following sections summarize the results of the trajectory analysis in terms of the cross-track, along-track, altitude, speed, and time profiles.The cross-track and along-track analyses presented herein focus on Phase II.The straight path utilized in Phase I essentially negated the influence of cross-track errors on the CTAS trajectory prediction accuracy.The turn within the descent of the Phase II path was designed to emphasize the potential influence of cross-track and along-track path errors on trajectory prediction accuracy. +Cross-Track ProfileFigure 22 shows a summary of lateral cross-track error for Phase II at each trajectory analysis gate as a function of FMS automation level.The three levels for which LNAV was used for lateral guidance (FMS TOD, CTAS TOD, and ND arc) exhibited essentially no cross-track error, as might be expected.The non-FMS runs, however, showed an average offset of approximately 5000 ft left of desired course during the run prior to the turn that increased to an average 13000 ft left of desired course following the turn (which was to the right).Figure 23 illustrates the ground track of the non-FMS runs conducted during flight 729.The left offset during the preturn segment was well within the expected navigational accuracy of VOR-based airways.Pilot comments indicated that the predominant tailwind changing to a crosswind following the turn encountered along this route contributed to the inbound course overshoot.The largest error occurred during run 3 of flight 729 (fig.23) when the pilot followed flight director commands throughout the turn (by keeping the lateral flight director command bar centered) and did not attempt to adjust for the indicated overshoot on the CDI.Pilot comments indicated that most pilots would wait for the flight director cue to initiate the turn; however, they tended to apply additional correction back to the desired course once the overshoot occurred. +Along-Track ProfileThe effect of the VOR-radial offset and turn overshoot on the distance flown is shown in figure 24.The actual distance flown by the airplane was compared with the predicted distance flown at each analysis gate.The distance flown during the non-FMS runs was, on average, 1.3 n.mi.greater than predicted, with a standard deviation of 1.1 n.mi.This increased range occurs at the turn, which typically happened between the FL250 and FL210 analysis gates.Anticipation of the overshoot and initiating the turn earlier than indicated by the flight director could reduce this error.The CTAS path generation could be modified to remove the mean contribution of the overshoot phenomenon by modeling the overshoot as a function of turn angle.However, trajectory prediction errors due to variations in pilot navigation error can only be reduced by improving the precision with which pilots navigate. +Altitude ProfileFigure 25 presents the altitude error, for Phase I, between the idle and constrained descent procedures flown from the RFD.The constrained procedures result in a significant reduction in altitude error (both mean and variation) over the idle procedure.Both procedures behave similarly in the initial stages of the descent, by first exhibiting a slight positive altitude error followed by an increasingly negative (below path) error.The initial error is due to the unmodeled (within CTAS) segment at the TOD related to the pilot response and throttle reduction as well as the rounding off to the nearest nautical mile of the CTAS TOD advisory from the reference fix.The airplane then descends at a higher than predicted rate (about 15 percent), primarily due to two factors: performance modeling and wind gradient effects.The performance modeling errors described previously account for a descent rate error of approximately 5 percent.The along-track wind gradient, which averaged approximately 2 knots/1000 ft over the Phase I idle runs, accounts for a descent rate error of about 7 percent.The sensitivity of descent rate error to unmodeled wind gradient was determined through a series of fast-time trajectory simulations.CTAS was used to generate a set of descent trajectories for a Boeing 737 airplane with a standard atmosphere, nominal weight (85000 lb), and a descent from FL350 to 10000 ft at 0.72 Mach/280 KCAS.Trajectories were generated with an along-track headwind gradient which varied between 0 and 4 knots/1000 ft in 1-knot increments.Weight errors contributed little, if any, effect on the altitude profile accuracy in descent for the airplane and conditions tested (weight would have a significant effect on climb profile accuracy).The mean descent rate error due to weight was slightly less than 1 percent (actual steeper than predicted).After the Mach/CAS transition point, the altitude error continues to increase for the idle descent conditions until the pilot begins to level off at the crossing altitude.The largest errors occur as the airplane levels off with a mean altitude error of just over 1500 ft plus a standard deviation of 900 ft.For the constrained conditions, however, the growth in altitude error is arrested midway in the descent as the pilot initiated corrections during the constant CAS portion of the descent.The constrained procedures reduced the maximum mean error in altitude by nearly 800 ft and the standard deviation by 400 ft.Although modeling errors reduce the efficiency of the planned descent profile, the pilot procedure serves as a useful tool to minimize the associated trajectory prediction errors.The altitude error results from Phase II were more complex, as shown in figure 26 for the CTAS predictions.The ND arc runs, which were nearly the same procedures as the constrained descent runs of Phase I, exhibited the same characteristics of increasingly negative altitude errors (below the predicted path) correcting back toward zero error midway through the descent.The non-FMS runs, however, showed a strong increase in negative altitude error near the bottom of descent.This result was caused by the longer distance flown during the non-FMS runs which masked the altitude error until after the turn (at approximately FL210).Each nautical mile of extra distance flown contributes approximately 300 ft of altitude error (below path).The FMS runs, using both CTAS TOD and FMS TOD, had a more positive altitude error due to the general tendency of the FMS path to be steeper than the CTAS path (resulting in a later TOD).In comparing the CTAS and FMS TOD runs, relatively large errors are associated with the CTAS TOD runs.These larger errors were not caused by the CTAS TOD procedures per se but were because of the small number of runs flown.In fact, the CTAS TOD procedure reduces the altitude error at the top by initiating the descent at the CTAS TOD.After capturing the FMS path within the first 1000 ft of descent, the remainder of the descent was an exact duplicate of the FMS procedure at all gates from FL310 to DRAKO.The larger errors associated with using the CTAS TOD was a random phenomenon attributable to variations in the atmospheric prediction errors.All Phase II runs show a small negative (below predicted path) altitude error at the metering fix.This anomaly, due to the altimeter setting error described earlier, actually introduced a bias in each descent trajectory equivalent to the final error.The most significant influence of altitude profile error is the impact on the top of descent point.Table 9 presents the along-track error of the TOD event for Phase II.These data present the differences between the measured airplane TOD and the CTAS prediction.A positive error indicates the airplane descended later than the prediction.This convention was used to facilitate comparison between results from these flight tests and from later field trials involving commercial flights.As seen in table 9, those procedures which actively used the CTAS TOD for descent guidance exhibited a mean error of about 1 n.mi.with a standard deviation of another mile.Most of this error was due to time required for the reduction of throttle (not modeled within CTAS) and rounding off in the TOD advisory issued to the pilot.By comparison, the FMS TOD procedure had a mean error of 2.5 n.mi.with a standard deviation of 2.8 n.mi.This larger error reflects the differences in TOD computed by the FMS compared with that computed by CTAS.A comparison of the difference between FMS and CTAS TOD predictions for all Phase II runs revealed a mean error of 3.8 n.mi.with a standard deviation of 3.4 n.mi.The largest differences in FMS versus CTAS TOD actually occurred during the ND arc and CTAS TOD procedure cases.These results are consistent with the altitude errors shown in figure 26.Altitude errors from the FMS-predicted vertical profile were also computed for the Phase II test (fig.27).The ND arc and non-FMS runs were excluded from this analysis because those procedures did not follow the FMS path.As expected, the FMS TOD and CTAS TOD runs exhibited very little error as the procedures called for the pilot to fly the FMSgenerated altitude profile.The slight negative error of about 300 ft at FL190 and DRAKO for all runs was caused by the lack of an altimeter setting correction within the FMS path generation.The flight crew entered the altimeter setting prior to reaching FL190 and flew the airplane to a barometric altitude of 17000 ft as required.The only substantial difference between the two procedures was the difference in TOD which was caused by differences in model data (atmosphere and performance). +Speed ProfileErrors in the CTAS prediction of a ground speed profile depend on (1) piloting conformance to speed schedule, (2) errors in the altitude profile which result in true airspeed errors at the correct Mach/CAS speeds, (3) errors in the predicted wind and temperature aloft which result in ground speed errors at the correct Mach/CAS and altitude, and (4) ATC radar tracking errors which result in incorrect initial condition ground speed.For this test, pilot conformance errors with the speed schedule were negligible as described in section 6.1.4.The effects of altitude profile errors, atmospheric modeling errors, and ATC radar tracking errors on the speed profile can be observed by determining speed errors along the predicted path at common range locations.The Phase I test results exhibit altitude error effects induced by the idle versus constrained descents as discussed previously.Phase II attempted to minimize altitude errors by using various vertical guidance techniques.Radar tracking and atmospheric modeling errors were encountered to differing degrees in both tests.Figure 28 presents the ground speed, true airspeed, and calibrated airspeed errors at the trajectory analysis gates for the Phase I flight test.The IC errors from the radar tracker were on the order of about 7 knots standard deviation with negligible mean error throughout cruise (IC to TODG).This result is consistent with the raw radar ground speed in table 5.In comparison, a true airspeed error of about 12 knots mean with about 12 knots standard deviation is seen at the IC.Since CTAS estimates true (and calibrated) airspeed at the IC based on radar-tracked ground speed and atmospheric wind and temperature models, the additional true airspeed error is induced by errors in the atmospheric model.CTAS uses this estimated cruise true airspeed in conjunction with the atmosphere model to predict the ground speed for the rest of the cruise segment.For the descent prediction, CTAS uses the scheduled descent Mach/CAS (with an appropriate acceleration or deceleration from the computed cruise speed) to predict true airspeed.At the first trajectory gate past TOD (FL330 in fig.21(a)), the initial true and calibrated airspeed errors are shifted toward zero with the ground speed error exhibiting a comparable shift in mean error to approximately -10 knots.Altitude variations during the constant Mach descent segments (FL330 through FL250) produced true airspeed (and calibrated airspeed) errors even though the airplane flew the Mach schedule precisely.The calibrated airspeed error at the FL230 and FL210 gates, where all runs were at the scheduled descent CAS, is reduced to the level of piloting accuracy presented in table 8.The true airspeed error is shifted by 5 to 10 knots slower than predicted primarily because of the mean altitude error of 500 to 1500 ft below the predicted altitude as shown in figure 25 (true airspeed changes by approximately 6 knots for each 1000 ft of altitude change at the same calibrated airspeed for these test conditions).The idle descent procedures required the pilot to slow to the metering fix crossing speed before bringing the throttles up to hold speed and altitude.As a result, the true airspeed error at predicted BOD was seen to be an average of nearly 30 knots slow for the idle descents, even though the altitude error was insignificant at that point.In contrast, the constrained descent procedures resulted in a significant reduction in the airspeed errors at the BODG.Overall, the ground speed error essentially tracked the true airspeed error due to the negligible mean wind error during descent as illustrated in figure 19(a).The speed error results from Phase II for the CTAS trajectory predictions are presented in figure 29.In comparison with the constrained descents of Phase I, the ground speed errors appeared greater in Phase II.The mean ground speed errors during cruise (IC through TODG) were significantly greater than Phase I, with mean errors between 10 and 30 knots at TODG.Five knots of this error is due to the initial condition ground speed error from radar tracking (table 5), and some of the error growth in the cruise segment is attributed to a variation in the wind modeling error along the cruise path.However, a significant portion of the mean error (and variation) in cruise was due to the three non-FMS runs within flight 729 which experienced the wind interpolation error discussed in section 6.1.5.For the descent segment, all of which were constrained in Phase II, a much more uniform calibrated airspeed error distribution is observed throughout the descent (figs.29(b) and (c)).The true airspeed errors followed the calibrated errors closely with only slight difference in mean error (5 knots in some cases at lower altitude) caused primarily by small errors in the altitude profile (fig.26).The somewhat larger variation in true airspeed error was further attributed to small errors (typically less than 3 knots) that were induced by variations in the atmospheric pressure and geometric altitude tables used by CTAS.The value of atmospheric pressure determined from these tables at a given geometric altitude was used by CTAS for the calculation of true airspeed for a given calibrated airspeed.These tables were constructed based on MAPS weather models for each test run and at times did not accurately represent the correlation of atmospheric pressure to pressure altitude.This minor problem has subsequently been corrected in the CTAS airspeed conversion routines.The relatively larger ground speed errors (both mean and variation) were directly attributable to the wind error as illustrated in figure 19(b).The differences in ground speed errors between procedures (e.g., non-FMS versus FMS TOD) were not due to the procedures themselves but to the large variation in wind errors from flight to flight as shown in figure 18.Speed errors for the FMS-predicted paths of Phase II are presented in figure 30.The ND arc and non-FMS runs were excluded from this analysis because those procedures did not follow the FMS path.As expected, the ground speed errors in cruise were significantly better than for the CTAS predictions.The relatively large increase in variation at the FL250 and FL230 gates was attributed to a ground speed interpolation anomaly during the turn. +Time ProfileThe ultimate output of the CTAS trajectory prediction process is the time profile along the predicted path.CTAS sequences and schedules airplanes based on the predicted time of arrival at traffic merge points (e.g., common metering fix, approach segment, or runway).Furthermore, the time profile forms the basis of conflict probing along the trajectory.Knowledge of trajectory prediction accuracy may be used to scale separation buffers and determine conflict probability.Smaller time errors can allow smaller separation buffers and permit higher terminal arrival capacity or more efficiency at the same capacity.The analysis of the time errors from these flight tests focuses on the basic trajectory prediction results based on the comparison of CTAS predictions with TSRV-measured position.ATC radar position errors, as well as the coordinate system errors, are explicitly removed from the analysis.Final application of these time error results, such as the sizing of separation buffers or calculation of conflict probability, must account for ATC radar position errors.A key output of the CTAS Descent Advisor trajectory prediction is the time of arrival at the metering fix.Table 10 summarizes the time-of-arrival accuracy results from the Phase I flight test for the idle and constrained descent runs.The arrival time error (Mean + Standard deviation) for all runs (idle and constrained procedures) was less than 25 sec.However, a significant difference in results existed between the procedures.The constrained procedures were expected to be more accurate because the procedure would reduce speed profile errors by mitigating the effect of modeling errors on the vertical profile as evidenced by figure 28.The RFD constrained cases did result in a 40-percent reduction in mean error (and a 33-percent reduction in std.dev.) compared with idle.However, the FFD constrained cases resulted in similar mean error with a 50-percent increase in standard deviation.This anomaly in the FFD constrained cases is attributed to two factors.First the number of FFD constrained runs was significantly smaller, and second, it was difficult for the research pilots to interpret vertical profile progress with the conventional instrumentation of the FFD cockpit.The lessons learned in Phase I led to improvements in the Phase II pilot procedures and training which supported a more comprehensive study of conventional cockpit (non-FMS) cases within Phase II.Figure 31 illustrates the trends in time profile error that lead to the differences in results between the idle and constrained procedures.In comparing the error growth between procedures, the time error is nearly the same up to the FL190 gate.Below the FL190 gate, the growth of time error for the idle cases increases dramatically as the airplane reaches its clearance altitude early and initiates deceleration.These characteristics are clearly illustrated in the altitude profile errors of figure 25 and the airspeed profiles of figure 28.Comparatively, the constrained procedures reduce the altitude error leading to early deceleration.This "additional" time error associated with the idle descent procedure could be largely eliminated by procedures which require the pilot to maintain descent speed until it is necessary to decelerate for a crossing restriction.The most efficient method to accomplish such a procedure is for the pilot to adjust the vertical profile to target an appropriate bottom of descent.Cockpit automation such as VNAV guidance and/or range-altitude arcs provides valuable assistance to visualize and control the vertical profile, particularly for off-airway navigation.The trajectory prediction results for Phase II included comparisons of actual time profiles with both CTAS-predicted and FMS-predicted trajectories.The CTAS predictions provide a measure of trajectory prediction accuracy using CTAS (atmospheric and performance) models and radar ground speed, whereas the FMS predictions provide a similar measure using the actual airplane performance, measured atmospheric conditions, and actual ground speed.Caution is advised when comparing these CTAS and FMS results because of the influence of the pilot procedures on the actual trajectories flown.In all but the FMS TOD cases, the pilots used the CTAS TOD location for descent, whereas the FMS trajectories are all based on the FMS TOD.In addition, the extremely small number of test cases (no more than 6 for each condition) precludes any statistically significant analysis.Table 11 summarizes the error results at the metering fix arrival time using the CTAS trajectory predictions for Phase II.An interesting comparison may be made between the CTAS arrival time results of Phases I and II.A comparison of tables 10 and 11 shows a general shift in the mean arrival time error.In general, the airplane arrived later than predicted in Phase I compared with Phase II where the airplane arrived earlier than predicted.This general shift is attributed to the effect of wind modeling errors and flight path orientation.Although the winds were generally out of the west and stronger than predicted for both Phases, the mean along-track wind error differed between the two Phases (fig.19) because of the nearly opposite course orientation.The Phase I course was generally into the wind and resulted in the airplane flying a slower ground speed than predicted, whereas the Phase II course was with the wind and resulted in the airplane flying faster than predicted.This comparison underscores the influence of the wind-error field on conflict prediction accuracy, namely that two crossing trajectories may share the same wind field, but the net effect of the wind error on each trajectory varies with its orientation.For the Phase II data alone, the comparison between the non-FMS and FMS-related runs was unexpected.In particular, the non-FMS runs were expected to result in a greater time error (mean and standard deviation) than FMS-related runs due to the advantages of FMS guidance.Further analysis of the time errors, in terms of their growth along the path (fig.32), revealed several interesting characteristics that were a direct result of the small and unique sample of data taken.For the non-FMS runs, the mean time error had built up to about -15 sec at FL250 due to the large ground speed errors seen in figure 29(a).Following the turn, however, the time error reversed and ended with a mean error of +2 sec.The wind errors in the CTAS prediction were therefore compensated by the longer distance flown in the non-FMS runs to end with a coincidentally small time error at the metering fix.To quantify the effect of the longer distance flown by the non-FMS runs, the arrival times were adjusted to remove the time associated with the longer distance flown.This adjustment provides for a more consistent comparison with the other runs which used FMS guidance to fly the lateral path.The adjustment was computed for each run based on the excess distance flown and the ground speed of the airplane at FL190.The result was a mean arrival time error of -11.0 sec with a standard deviation of 15.5 sec.These adjusted time errors clearly show the overriding effect of wind error on the arrival time performance during this test.Conversely, had the wind errors been less (or more consistent), the CTAS TOD and FMS TOD conditions would have achieved the best arrival time results.The ND arc would have been only slightly worse due to the tendency of the airplane to fly lower than predicted resulting in a slightly lower TAS.In addition, the seemingly lower standard deviation of time error for the nonadjusted non-FMS cases (shown in table 11), was because of a favorable coupling of the time error due to wind and that due to the longer distance flown.Removing the effect of longer distance increased the standard deviation from 8.7 to 15.5 sec, which is more in line with the other cases.Table 12 presents the arrival time accuracy based on the TSRV FMS predictions for the two VNAV procedures flown (the non-FMS and ND arc did not follow the FMS VNAV path).These data illustrate the arrival time differences between the CTAS and TSRV FMS predictions.The primary factor contributing to these differences between the FMS and CTAS trajectory predictions was the source of wind data.CTAS used wind data from the NOAA MAPS model, whereas the FMS used winds entered manually during the flight, as discussed in the section "Test Procedures."The FMS-entered winds came from hand recording the winds on the previous descent and, in general, were more accurate than the CTAS winds.Figures 33 and34 present a summary of along-track wind errors for the CTAS and FMS predictions for each of the guidance conditions.Comparison of figure 33(a) with 34(a) clearly shows the lower mean wind error corresponding to the FMS prediction cases.As a result, the mean time error for the FMS predictions was coincidentally the smallest.In addition, the mean time errors for the various guidance conditions are seen to follow the mean wind errors for the CTAS prediction cases (when adjusted to the same distance flown).For the FMS predictions, the variation in wind error was observed to be greater for the FMS TOD guidance cases with a resulting higher variation in arrival time error. +Sensitivity AnalysisThe effects of airplane performance and atmospheric modeling errors on the time profile predictions were examined by using the stand-alone version of the airborne FMS PGA4D trajectory generation program.This analysis was applied to the Phase I idle conditions in an effort to relate the sensitivity analysis to real-world measurements and to identify the contributions of the dominant trajectory prediction error sources.This analysis is restricted to the straightpath idle cases.The straight path is necessary to isolate navigation (overshoot) errors from the remaining sources.The idle cases are necessary to remove the influence of pilot variations in thrust-drag management.Two executable versions of the program were created for this analysis.The first version contained the airplane performance model representative of a baseline Boeing 737-100, the same as that used by the CTAS trajectory generation program in the flight tests.The second version contained the performance model of the TSRV airplane as modified in the appendix.A simple straight-line route consisting of a starting point at the PONNY waypoint and ending at the KEANN metering fix (fig. 1) was used for the vertical trajectory generation.Initial and final conditions (altitude, calibrated airspeed, and true track angle) were created to represent each of the idle descent test runs of flights 679 and 680.Two sets of weather data (wind speed, wind direction, and air temperature) were created for each test run.The first set used the weather data recorded by the airplane at pressure altitude steps of 500 ft from top of descent down to the metering fix altitude of 17000 ft.The second set used the CTAS MAPS weather model with wind and temperature values interpolated at the same horizontal location and pressure altitude as was used for the first data set.Four unique combinations of airplane performance and weather models were used to generate trajectories for comparison, as shown in table 13.Trajectories were generated for each test condition from flights 679 and 680 by using each of the four combinations of performance and weather models.The trajectories generated with the baseline set were used as the references for the trajectory error comparisons.The primary parameter for comparison was time of arrival at the final range of the reference trajectory with TOD assumed to begin at the reference trajectory TOD range.If the test trajectory ended before the reference trajectory final range, the test trajectory final point was extrapolated by assuming constant altitude and ground speed to determine the time of arrival at the reference trajectory end condition.Similarly, if the test trajectory continued past the end of the reference trajectory, the arrival time was computed by linearly interpolating on the range corresponding to the reference trajectory final condition.This method for finding arrival time matched the way the idle descents were flown in Phase I. Time errors were then computed by subtracting the test trajectory arrival time from the reference trajectory arrival time for each test condition and model combination.A summary of the time error results is given in table 14.As seen in table 14, the inclusion of both the performance model and weather model revisions in the idle descent trajectory generation resulted in time errors nearly the same as those measured in Phase I, as shown in table 10.The performance model alone accounted for approximately one third of the mean time error with little variation.The weather model accounted for slightly more than two thirds of the total mean time error and nearly all the variation.The constrained procedures would reduce most of the mean error due to performance modeling and a part of the mean error due to the wind model by eliminating the early slow-down at BOD. +Qualitative Impact of Error SourcesThis section summarizes, based on the flight test data analysis, the impact of trajectory prediction error sources.Although not a comprehensive statistical analysis, the discussion indicates the potential impact on trajectory prediction accuracy as well as the flyability and efficiency of CTAS descent advisories.Individual error sources are ranked in terms of their potential time-error impact on CTAS clearance advisories for constrained descents.The rankings are defined as follows based on a 10-min prediction horizon:Primary >10 sec impact Secondary 5-10 sec impact Minimal <5 sec impactThe impact on lateral and vertical profile accuracy is also summarized.Where applicable, the discussion is extended to cover other trajectory segments such as ascents, en route cruise, and unconstrained descents.For active CTAS applications (e.g., time-based clearance advisories for speed, TOD, and routing), trajectory prediction accuracy is primarily affected by errors in winds, tracking, and pilot conformance.In addition to accuracy, another important factor is the flyability and efficiency of the CTAS TOD advisory.This factor is primarily affected by performance modeling as well as atmospheric modeling.The constrained pilot procedure for a CTAS-based clearance, like a VNAV profile, calls for the pilot to add thrust or drag to correct for altitude profile errors.The magnitude and sense of these corrections directly affect the flyability and fuel efficiency of the profile.The need to add drag on descent is often considered unacceptable for passenger comfort, and for most transport airplanes, drag devices lack effectiveness.The need to add drag or thrust indicates a waste of fuel relative to the optimum profile.Atmospheric errors are of a random nature depending on the atmospheric field, model performance, and route of flight.To ensure flyability in the presence of all errors, the performance models and pilot procedure may need to include buffers.Proper procedures will improve accuracy in the presence of modeling errors, at a cost in efficiency, and will minimize workload. +Radar Track +Position.Along-track errors were found to be of Secondary impact.The measured along-track error was generally consistent over all Phase II flights with the track position trailing the actual position by 6.3 ± 3.4 sec (Mean ± Standard deviation).Much, if not most of this error may be corrected by a Host track time stamp that is not currently provided to CTAS.If all flights are tracked by radar, the contribution of the mean along-track error tends to cancel when any two trajectories are compared for separation.However, if tracking sources are mixed (e.g., some airplanes tracked by radar, some by automatic dependent surveillance (ADS)), the mean error of the radar-tracked flight would contribute to the conflict prediction error.The mean along-track error would also reveal itself when radar-tracked airplanes are compared with airplanes operating to RTA.Cross-track errors were found to have a Minimal impact both in terms of cross-track position as well as their contribution of error to the prediction of alongtrack position.(Actual cross-track error, due to pilot navigation, is addressed later in section 6.4.4.) 6.4.1.2.Speed.Ground speed errors were found to have a Minimal impact on trajectory segments with speed clearances such as CTAS descent advisories.CTAS descents (as well as ascents and future cruise segments) are predicted by combining the winds along the path with an estimated airspeed based on clearance, flight plan, or file-based user preference.The only impact on accuracy is caused by the influence of ground speed (and atmospheric model) in estimating the airspeed prior to acceleration to the cleared airspeed.Ground speed errors would, however, have a Primary impact on the prediction accuracy of "openloop" trajectory segments (i.e., those segments for which speed is inferred from the observed ground speed as opposed to an advisory or clearance airspeed).Although the flight test runs experienced a smaller ground speed error, the measured standard deviation of speed error in level cruise was 13 knots (3 percent for an airplane at 420 knots or about an 18-sec error for a 10-min prediction).During turning maneuvers, the tracker lagged the airplane with substantially greater errors (exceeding 100 knots in many cases).Clearly, the raw tracker data are not good enough during these transients (maneuvers) to support a passive en route conflict probe.Some sort of filtering, or additional data, would be needed to supplement the Host track data during transient maneuvers.One example of a filter, short of an advanced tracking algorithm, would be to simply ignore changes in ground speed during transient periods (e.g., turns) with a lag of 1 to 3 cycles to allow for the positive identification of the transient. +Track angle.For many cases, the impact of track-angle errors may be mitigated by path generation algorithms which correlate airplane position with the planned route of flight.For other cases, such as vectoring, open-loop pilot maneuvering (e.g., thunderstorm avoidance), and turns, the impact of track-angle errors may be significant.During vectors, track-angle errors may have a Primary impact on accuracy if the track angle is used to project the future path of the airplane.Track errors may have a substantial impact on the predicted path and time to fly depending on navigation geometry.As with ground speed, some sort of filtering or additional data are needed to supplement the Host track data during turn transients, particularly if the data are to be used for monitoring of clearance conformance.For vectors, much of the error may be reduced by providing the ATM automation with an input of the heading clearance to damp out the error in projected heading. +Atmospheric Model +Wind component along path.Wind errors were found to have a Primary impact on trajectory segments based on speed clearances such as CTAS descent advisories.For these situations, the modeled wind is added to the clearance airspeed to predict ground speed.If the pilot flies the airspeed precisely, wind model errors directly affect the predicted ground speed.These errors not only affect the time to fly, but they may also have a substantial impact on the TOD location.For constrained CTAS descents, the TOD location error will affect the thrust and/or drag needed to meet the BOD constraint and, therefore, the flyability and efficiency of the CTAS descent profile.For unconstrained descents, wind errors will also introduce errors in the altitude profile as well as TAS errors due to the altitude error.Wind errors have a Minimal impact on open-loop cruise segments that are based on track ground speed.For these segments, the wind model is used to estimate the airspeed at the initial position.The ground speed profile is then predicted based on the airspeed estimate and the winds along the path.If a constant airspeed profile is assumed, then the only variation in ground speed is caused by variations in wind and temperature along the path.During open-loop cruise segments, the ground speed error is primarily caused by the trackerinduced error with an atmospheric influence due to variations in the wind-temperature model error along the path. +Wind gradient along path.The main effect of wind gradient error is on the prediction of descent and ascent rate with a Minimal impact on time along the path for constrained descents.Sustained gradients observed during the test ranged from 1 to 3 knots/ 1000 ft altitude with substantially larger gradients occurring during peak jet stream conditions.As noted earlier, a gradient of 1 knot/1000 ft contributes approximately 3.5 percent to the descent rate of a 737.For a 20000-ft descent and a typical descent ratio of 3 n.mi./1000 ft, each knot of gradient error leads to a difference of 2 n.mi. in the optimum TOD.If the segment is flown with vertical constraints (i.e., TOD and BOD), then the error mainly affects the thrust or drag needed to meet the constraints and, therefore, the flyability and efficiency of the descent profile.If the segment is flown without vertical profile constraints, an unmodeled wind gradient leads to an error in the altitude profile which in turn may introduce a small error in the TAS profile for a constant Mach/CAS segment and an error in estimating the transition in airspeed at the BOD.Ascent rates may be more or less sensitive to wind gradient depending on the calm-wind ascent rate, which varies significantly with altitude and weight.An unmodeled wind gradient is expected to develop error in the predicted altitude profile and TOC.These altitude profile errors may lead to significant errors in ground speed caused by errors in the TAS and in wind speed caused by the uncertainty in altitude as well as an error in estimating the TOC transition from climb to cruise airspeed. +Temperature.The main impact of temperature (and pressure) is on the prediction of geometric (absolute) altitude rate with a Minimal impact on time along the path for constrained descents.For example, each 5°C error in temperature profile leads to approximately an error of 500 ft in the altitude to descend or ascend between FL350 and FL100.Like wind gradient, the main impact of temperature is on the time and distance to descend.For constrained descents, temperature errors primarily affect the thrust or drag required to meet the constraints.Although temperature errors also affect airspeed estimation during constant Mach/ CAS segments (approximately a 1-percent error in TAS for each 5°C error in temperature), the relatively small errors observed during the flight test had a negligible effect on the accuracy of the descent predictions.If the segment is flown without vertical profile constraints, a temperature error may contribute to an error in the altitude profile which in turn may introduce a small error in the TAS profile for a constant Mach/ CAS segment as well as an error in estimating the transition in airspeed at the BOD.For ascents, temperature not only affects the geometric altitude, it also affects the climb thrust of the airplane, both of which contribute to errors in predicting the altitude profile, TOC, and ground speed profile. +Airplane Performance ModelingErrors in the performance model affect trajectory prediction accuracy in a similar fashion to wind gradient.For constrained descents, the impact on time is Minimal with the main influence on the flyability and efficiency of the profile.Although the net thrust (and weight) has a direct effect on the time to accelerate or decelerate, these transitions tend to be short and have little effect on the trajectory prediction.For unconstrained descents, performance modeling errors may contribute to errors in the altitude profile which in turn may introduce a small error in the TAS profile for a constant Mach/CAS segment and an error in estimating the transition in airspeed at the BOD.The Phase I sensitivity analysis presented earlier indicated that the 5-percent error in the CTAS performance model for the TSRV led to a time error of 5 sec over a descent of 18000 ft.Earlier analysis of weight errors indicated that descent rate error varies with speed and is relatively insensitive to weight over a large portion of the speed envelope centered about the speed for maximum lift-to-drag ratio (ref.6).For ascents, performance model errors have a Primary impact on the accuracy of time and distance to climb with significant sensitivity to weight and speed profile.In addition, performance modeling errors may affect the accuracy of determining advisory limits such as the high-speed boundary or service ceiling in cruise.For future applications such as trajectory negotiation, precision between ATM and user (airborne or ground based) performance models might be important in order to accurately probe for conflicts as well as minimize deviations from user preferences. +Pilot Conformance6.4.4.1.Navigation.Navigation errors, depending on airplane equipage and knowledge of pilot intent, may have a Primary impact on trajectory prediction accuracy.As seen for the non-FMS cases, turn errors may contribute a significant error in predicted distance flown.Although the non-FMS cases studied in this test emphasized the uncertainty in the pilot's turn overshoot, the lack of error for the LNAV cases underscores the importance of turn model geometry which may have a significant effect on the predicted distance flown for typical turns associated with the extended terminal area and vectoring.In addition to the distance flown, turn overshoot and lateral crosstrack errors associated with conventional airway navigation may result in cross-track errors of up to several miles even within legal navigational limits defined by instrument flight rules.6.4.4.2.Speed.The sensitivity of trajectory prediction accuracy to speed conformance is significant.A speed conformance error affects a closed-loop trajectory segment in the same way that a ground speed (track) estimate error affects an open-loop segment.Although speed conformance was good during these flight tests, the TSRV speed-tracking performances (both manual pilot and FMS/autopilot) were not representative of speed conformance expected of airline pilots and commercial FMS equipment.Operational procedures must highlight the need for adherence to the predicted speed schedule in order to achieve good arrival time results. +RecommendationsThis paper presents a sample of en route trajectory prediction error sources under real-world operational conditions.Although the data provide a good "orderof-magnitude" basis, the data are not a statistically significant set.The recommendation is that a comprehensive trajectory accuracy sensitivity study be performed to provide a method for the analysis of the conflictprobe accuracy under operational conditions.Conflict prediction accuracy is derived directly from the relative trajectory prediction accuracy for an airplane pair.Trajectory prediction accuracy depends on the airplane type, atmospheric prediction accuracy, trajectory segments and orientation, and time horizon.A comprehensive sensitivity study would require the development of several sets of statistically significant error source data.The first and most significant error source is atmospheric prediction, which has a complex effect on trajectory prediction accuracy.A comprehensive analysis of atmospheric prediction accuracy, as it pertains to trajectory prediction, would help determine the sensitivity and overall expected performance of conflictprobe automation tools under operational conditions.Such a study should be conducted over an extended period of time (e.g., 1 year) to measure the frequency of significant errors due to seasonal variations in weather phenomena.The study should also cover a moderate-size airspace (e.g., an en route ARTCC) to capture the positional and trajectory orientation effects and during the normal hours of flight operations to capture temporal effects such as variations in sensor data availability.Previous evaluations have focused on the gross accuracy averaged over time and position (ref.17).Because the performance of conflict-probe tools varies with time and trajectory characteristics, the study must be focused on trajectory applications (i.e., provide a realistic correlation between the atmosphere and trajectories).Such a study would also be useful for (1) determining cost beneficial methods for improving atmospheric prediction accuracy where it is needed most for trajectory prediction and (2) creation of a data set to support the development of tools to predict the accuracy of atmospheric forecasts at the time of the forecast to provide an efficient bound for conflict-probe error buffers.The second error source that should be studied further is airplane tracking.Although the steady state accuracy of the FAA Host tracker may be adequate, the large track velocity errors associated with transients (maneuvers) are unacceptable for effective conflict prediction.These maneuvers may not be common during en route cruise, but they do occur frequently in the extended terminal area.Methods for improving track velocity accuracy or mitigating the impact of such errors on trajectory prediction tools are needed.Aside from ADS, two additional solutions exist: the use of advanced track filters and the use of logic to inhibit calculations based on Host track data during transient periods.The third error source relates to the modeling of airplane performance.Although errors in CTAS performance models do not significantly affect time profile accuracy in descent, model errors do affect the flyability and efficiency of DA-based clearances for non-FMS airplanes and have a small effect on the accuracy of the altitude profile.Performance modeling errors, including weight estimation, are expected to have a much greater impact on climb profile predictions in terms of both time and distance to climb.Generally, performance varies not only as a function of type but also between individual airframes of identical type (because of age and modification).Developing a database that indicates the performance variation over the fleet of airplanes operating in the national airspace system would be useful.This database should use input of airplane operators and manufacturers.The fourth source of errors, pilot conformance, may be useful to determine the accuracy to which speed and course clearances are conformed under operational conditions.Such a study would complement the data within this report (pilot conformance errors were minimized to isolate the other error sources).More importantly, it is critical to understand when, and under what conditions, CTAS does not have accurate knowledge of the intended course, speed, and TOD.The present flight tests evaluated trajectory predictions under the assumption that CTAS had accurate knowledge of the appropriate clearances.The validity of this assumption should be evaluated by a study of actual track data to determine how often and why the CTAS heuristics and controller inputs would fail to reasonably represent the intended clearance.The data gleaned from such a study would provide insight that would lead to improvements in the CTAS routing heuristics as well as reductions in the need for controller inputs.Finally, there is clearly the need for additional work on operational procedures for constrained descents which minimize the trajectory errors.In particular, the procedures should emphasize the need to maintain the CTAS-expected speed schedule throughout the descent in order to minimize time errors.Studies which document the differences in current descent procedures between different airplane types and different operators of the same airplane type would be useful in defining new common procedures.Field tests using the actual airplane operators and air traffic controllers, such as those conducted in reference 22, are useful for final validation and user acceptance of the new procedures. +Concluding RemarksThe Transport Systems Research Vehicle (TSRV) Boeing 737 based at the Langley Research Center flew 57 arrival trajectories that included cruise and descent segments; at the same time, descent clearance advisories from the Center-TRACON Automation System (CTAS) were followed.These descents were conducted at Denver for two flight experiments (Phase I in October 1992 and Phase II in September 1994).The actual trajectories (recorded onboard the TSRV) were compared with predictions calculated by the CTAS trajectory synthesis algorithms and the TSRV Flight Management System (FMS).The CTAS Descent Advisor was found to provide a reasonable prediction of metering fix arrival times during these tests.Overall arrival time errors (Mean + Standard deviation) were measured to be approximately 24 sec during Phase I and 15 sec during Phase II.These results, although not statistically significant, were obtained under real-world operational conditions and are representative of the level of performance which should be expected from active CTAS descent clearance advisories.The major source of error during these tests was found to be the predicted winds aloft used by CTAS.Overall along-track mean wind errors of 10 to 15 knots with standard deviations of about 15 knots were experienced during the cruise segments of both Phases I and II.Mean wind error reduced to between 5 and 10 knots during descent; however, the standard deviation remained at 10 knots or more.The sensitivity analysis of Phase I idle descents revealed that about two thirds of the mean time error and nearly all the variation in time error were due to wind errors.Analysis of Phase II runs also revealed wind errors to be the overriding factor in the arrival time errors measured during that test as well.Airplane position and velocity estimates provided to CTAS by the Air Traffic Control (ATC) Host radar tracker were found to be a relatively insignificant error source during these tests.Position errors were predominantly along track, with the tracker lagging the actual airplane position by an average of 6.3 sec with a standard deviation of 3.4 sec throughout Phase II.If all airplane positions are provided by the same radar tracking system, the mean along-track error tends to cancel when two trajectories are compared by CTAS for conflict probing.The cross-track component of radar tracking error was found to be relatively small, with an overall error of approximately 0.22 n.mi.standard deviation measured during Phase II.Ground speed errors during the stabilized initial condition locations for the test runs were also minimal, with a mean plus standard deviation error of less than 10 knots.Measurements of radar tracking performance at other flight conditions revealed significant ground speed errors when the airplane was turning.Ground speed errors of 100 knots or more (Mean + Standard deviation) recorded during turns rendered the radar tracking unusable as a source for airplane ground speed.These ground speed errors were found to persist for 1 to 3 min following a turn.Airplane performance modeling errors within CTAS were found to not significantly affect arrival time errors when the constrained descent procedures were used during these tests.The TSRV airplane performance differed from the CTAS Boeing 737-100 model data, in terms of lower net thrust minus drag (TMD), by approximately 5 percent over the descent.The principal effect of these modeling errors was on the calculated versus desired top of descent (TOD) for an efficient idle descent.Although the impact of these modeling errors on the time profile for descents was small, they are expected to have a significant impact on the predictions of ascent segments.The most significant effect related to the flight guidance used by the TSRV was observed to be the lateral path errors recorded when conventional VOR (very high frequency omnidirectional radio range) guidance was used during the non-FMS cases of Phase II.The Phase II runs involved a 60° turn during descent.Cross-track errors of 24000 ft (Mean plus Standard deviation) occurred following the turn during these cases, which contributed to an average 1.3 n.mi.longer range flown.This translated directly into approximately 13 sec of mean arrival time error for the non-FMS test cases.The use of FMS lateral navigation (LNAV) eliminated this error.Vertical trajectory errors, resulting from wind and airplane performance modeling errors, were also dependent on the method of flight guidance.Flight procedures which utilized the FMS-generated path for vertical guidance exhibited the largest vertical errors during the initial portion of the descent, whereas procedures using CTAS guidance (TOD and speed schedule) tended to build up errors during descent with the maximum occurring closer to the bottom of descent.The altitude errors recorded during these tests peaked at about 2000 ft (Mean plus Standard deviation) for both the non-FMS and FMS reference conditions, with the airplane being below predicted altitude for the non-FMS reference and above predicted altitude for the FMS reference conditions.The contribution of these altitude errors to the overall arrival time was determined to be insignificant.Overall, the constrained pilot procedures assisted by LNAV and VNAV (vertical navigation) guidance served to mitigate the impact of modeling errors on the accuracy of the altitude profile prediction.V ˙a g T D - ( ) W ----------------------gγ V ˙w - - = h ˙V a γ = D T W - = V ˙a V ˙w + g -------------------- h V a ------ +       h ˙h ṗ T k T k , s ---------- = C D D qS ref ------------ = q 1481δ am M 2 = ∆C D C D = C D,m - C D,mmodified by adding a constant 0.003 to for the revised TSRV drag model. +A.2. Idle ThrustUpdate of the idle thrust model required a careful review of the baseline TSRV thrust model.The analysis conducted in reference 24 provided the basis of the current TSRV engine model.As described in that report, idle thrust is a function of Mach number with an adjustment if the engine is operating at the minimum fuel flow limit.With this technique, a baseline idle thrust model was created for the TSRV airplane by using the manufacturer's performance data for the Boeing 737-100 airplane with Pratt and Whitney JT8D-7 engines.A function of engine pressure ratio (EPR) versus Mach number was generated which produced the idle thrust values presented in the manufacturer's data for idle fuel flows above the minimum limit (540 lb/hr).The generalized fuel flow model was then extended to include EPR values in the idle range.The resulting model provided a good match to the idle thrust and fuel values provided in the manual using the generalized fuel flow and thrust versus EPR functions.The process of updating the TSRV idle thrust model involved modifying this baseline idle EPR versus Mach relationship and determining an appropriate value for minimum fuel flow.The five idle descent runs of flight 679, which encompassed the flight envelope of the airplane utilized for this experiment, were analyzed for this purpose.Figure A2 shows the measured EPR at idle for all runs versus Mach number for both engines.As predicted by the engine model, a definite minimum EPR boundary is evident.A shift of 0.045 in the EPR from the baseline engine model resulted in a good match between the flight and model EPR limit.EPR values above the limit shown in figure A2 occur when the engine is operating at the minimum fuel flow limit.The original minimum fuel flow of 540 lb/hr was adjusted until a reasonable match to the average measured minimum fuel flow and correspond-ing EPR value was achieved.Figure A3 presents an example of minimum fuel flow for one of the flight 679 runs with the original and revised minimum fuel flow illustrated.A final check on the validity of the idle thrust model was done by comparing the predicted model values of idle thrust with the computed values based on measured EPRs for all the idle thrust descent runs.Figure A4 presents the composite of the mean and standard deviation of thrust error at discrete altitudes during the descents.The original model had mean errors of between 200 and 500 lb with maximum standard deviations of approximately 250 lb.The revised model reduces the mean errors to less than 100 lb with standard deviations of 200 lb or less.The largest values of standard deviation are a direct result of idle surge bleed operation in the altitude region of 20000 to 30000 ft.This unavoidable situation is discussed in greater detail in reference 24. +A.3. Descent Performance ModelIn order to determine the overall performance modeling error for descent calculations, the combination of idle thrust and drag errors must be considered.The stabilized descent points from the idle descent test runs were further analyzed to determine the error in the original model of thrust minus drag (TMD) compared with the measured flight results.Actual thrust was approximated by using the measured EPR and state conditions.Drag was computed by using the techniques described in the previous drag error analysis.Model values of thrust and drag came from the original models based on the state conditions and flight idle throttle setting.The TMD modeling errors were computed as a percentage of the baseline model values and plotted versus altitude in figure A5.As seen in the figure, the actual TMD varied from 2 percent greater (more negative) at 17000 ft to about 10 percent greater than the model TMD at 35000 ft.This compares with the constant drag error of approximately 11 percent.Figures . . .Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . +Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . +1 .1Figure12shows the flight card for each situation. +There were four basic trajectory prediction error sources encountered during these tests: Radar tracking errors Airplane performance model errors Atmospheric modeling errors Pilot conformance +Figures 15 and 1616Figures 15 and 16 present a summary of measured winds (resolved into components in the true north and east directions) for each flight within Phases I and II.The data are presented in terms of the mean and standard deviation of the wind, at common altitudes, over each descent run of a particular flight.The cruise altitude data are presented slightly differently for each Phase.For Phase I, a single data point (mean and standard deviation) is presented at cruise altitude based on the mean wind over the cruise segment of each run.The average length of the Phase I cruise segments was 9.8 n.mi.with a standard deviation of 6.5 n.mi.For Phase II, the cruise winds are presented at three positions corresponding to the analysis gates introduced in section 6.2.These data points include +Figure A1 presents drag coefficient error versusMach number.The data reveal a fair amount of scatter in the data; however, a constant offset of approximately 0.003 in C D (30 drag counts) is evident.The baseline Boeing 737-100 drag model was therefore +Figure A1 .A1Figure A1.Drag coefficient error from idle descent test runs of Phase I. +Figure A2 .A2Figure A2.Measured EPR at idle for descents of flight 679 with baseline and revised minimum EPR models shown. +Figure A3 .FuelA3Figure A3.Minimum fuel flow flight 679, run 3. Fuel flow, lb/hr 300 400 500 600 700 800 900 +Figure A4 .Figure A5 .A4A5Figure A4.Composite idle thrust error for all idle descent test runs. +Figure 2 .2Figure 2. Flight test area for Phase II. +Figure 3 .Figure 4 .Figure 5 .345Figure 3. TSRV Boeing 737-100 test airplane. +Figure 6 .Figure 7 .67Figure 6.TSRV control display unit (CDU). +Figure 8 .8Figure 8. TSRV mode control panel (MCP). +Figure 9 .9Figure 9. Experimental setup at Denver Center. +Figure 10 .10Figure 10.Vertical profile procedures as function of speed. +Figure 11 .11Figure 11.Test cards for Phase II descent using conventional FMS. +RFDProcedure: VNAV Using FMS TOD s Level at cruise altitude, on path, at test condition cruise Mach. +(a) Test conditions 1c and 3c; early descent.(b) Test conditions 1c and 3c; late descent. +Figure 12 .12Figure 12.Test cards for Phase II descent using FMS with CTAS top of descent. +RFDProcedure: VNAV Using CTAS TOD s Level at cruise altitude, on path, at test condition cruise Mach.Retract speed brake when vertical path is captured. +sSelect VCSS prior to 18000 ft.s Smoothly capture crossing conditions at DRAKO.(c) Test conditions 2c; early descent.(d) Test conditions 2c; late descent. +Figure 12 .s12Figure 12.Concluded. +Figure 13 .13Figure 13.Test cards for Phase II descent using range-altitude arc. +sSelect VCSS prior to 18000 ft.s Smoothly capture crossing conditions at DRAKO.(c) Test conditions 2d; early descent.(d) Test conditions 2d; late descent. +Figure 13 .s13Figure 13.Concluded. +Figure 14 .14Figure 14.Air temperature measurements and modeling errors. +Figure 15 .15Figure 15.Measured winds from Phase I test. +Figure 16 .16Figure 16.Measured winds from Phase II test. +Figure 17 .17Figure 17.CTAS wind model errors from Phase I. +Figure 21 .FL190Figure 22 .2122Figure 21.Analysis gates for trajectory comparisons. +Figure 23 .23Figure 23.Lateral paths flown during flight 729 using VOR guidance. +FMSFigure 24 .24Figure 24.Distance flown error relative to FMS path. +Figure 25 .25Figure 25.Altitude error summary from Phase I. +FMSFigure 26 .26Figure 26.Altitude error relative to CTAS path from Phase II flight test. +Figure 27 .27Figure 27.Altitude error relative to FMS path from Phase II. +Figure 28 .28Figure 28.CTAS speed errors from Phase I. +FigureFigure 28.Concluded. +Figure 29 .29Figure 29.CTAS speed errors from Phase II. +FigureFigure 29.Concluded. +Figure 30 .30Figure 30.FMS speed errors from Phase II. +FigureFigure 31 .31Figure 30.Concluded. +Figure 32 .Figure 34 .3234Figure 32.Time error relative to CTAS path from Phase II. + + +. . . .Abbreviations and Symbols MF metering fixADS N mag NASAmagnetic north automatic dependent surveillance National Aeronautics and Space AdministrationARTCC, Center Air Route Traffic Control Center ND navigation displayATC accel NOAA BOD PD BODG PFD C D PGA4D PGUI qAir Traffic Control National Oceanographic and Atmospheric Administration acceleration pilot discretion bottom-of-descent primary flight display bottom-of-descent gate drag coefficient, profile generation algorithm, 4D Drag ------------planview graphical user interface qS ref free-stream dynamic pressure, lb/ft 2C D,m CAS RFD CDI RTA CDU rms CRT S ref CTAS std. dev. D Tperformance model drag coefficient research flight deck calibrated airspeed required time of arrival course deviation indicator root-mean-square control and display unit reference wing area, ft 2 cathode ray tube standard deviation Center-TRACON Automation System airplane drag, lb airplane net thrust, lbDA T kDescent Advisor atmospheric temperature, KDME T k,sdistance measuring equipment standard day atmospheric temperature, Kdecel TASdeceleration true airspeedEPR TMA FAA TMDTraffic Management Advisor engine pressure ratio Federal Aviation Administration airplane net thrust minus drag, T -D, lbFAST TMUFinal Approach Spacing Tool Traffic Management UnitFFD TOCforward flight deck top of climbFL TODflight level top of descentFMS TODGFlight Management System top-of-descent gateGPS TRACONGlobal Positioning System Terminal Radar Approach Controlg TRKacceleration of gravity, 32.17 ft/sec 2 trackHA TSRVhigh altitude Transport Systems Research Vehicleh UTCtrue altitude, ft universal time coordinatedh p V apressure altitude, ft true airspeed, ft/secIC V winitial condition wind speed, ft/secIP VCSSinitial position for a test run velocity control stick steeringJ VNAVjet route vertical navigationKCAS VORknots calibrated airspeed very high frequency omnidirectional radio rangeLA Wlow altitude weight, lbLNAV M δ am MAG γlateral navigation atmospheric ambient pressure ratio Mach number magnetic air-mass flight path angle, radMAPSMesoscale Analysis and Prediction SystemMCPmode control panelv vi +Table 1 .1Test Conditions for Phase IThrust -DragDrag onlyModel error, Flight -Model, percent15 00020 00025 00030 00035 000Altitude, ft +Table 2 .2Test Conditions for Phase IITestSpeedAutomation levelLateralVerticalFlightconditionschedulepilot procedureguidanceguidancedeck1a0.72/0.72/280ConventionalVOR/DME Airspeed with CTASFFDnon-FMSTOD2a0.76/0.76/2403a0.76/0.76/3201b0.72/0.72/280Conventional FMSLNAVFMS withRFDVNAV TOD2b0.76/0.76/2403b0.76/0.76/3201c0.72/0.72/280FMS withLNAVFMS withRFDCTAS TODCTAS TOD2c0.76/0.76/2403c0.76/0.76/3201d0.72/0.72/280Range-altitude arcLNAVRange-altitude arcRFDwith CTAS TOD2d0.76/0.76/2403d0.76/0.76/320 +Table 6 .6Radar Track Position Error Statistics for Phase II FlightsElapsedAlong-track errorCross-track errorFlightflight time, hr:min:secMean, n.mi. Std. dev., n.mi.Mean, secStd. dev, sec Mean, n.mi. Std. dev., n.mi.7282:14:000.6840.3965.93.4-0.0240.199729a/b5:05:360.7770.3986.83.50.0080.248730a/b3:28:000.6880.3996.13.50.0060.2777312:19:360.7310.3906.33.2-0.0440.2077322:07:240.7190.3846.33.3-0.0370.1977331:34:360.7030.3826.23.4-0.0280.207Total . . . .16:49:120.7170.3926.33.4-0.0200.223 +Table 7 .7Radar Track Ground Speed and Track Angle Errors for Phase IIFlight SegmentsSegmentGround speed error, knots Mean Std. dev.Track angle error, deg Mean Std. dev.Level flight2.312.30.14.6Altitude change-2.312.90.75.1Turn37.058.94.927.8Postturn56.455.85.012.9 +Table 8 .8Mean and Standard Deviation Errors in Pilot Adherence to CTAS Descent Speed SchedulePhase IPhase IISpeedFFDRFDFFDRFDMean Std. dev. Mean Std. dev. Mean Std. dev. Mean Std. dev.Cruise Mach0.0050.0090.0010.0030.0100.0070.0010.004Descent Mach Descent CAS, knots0.008 -0.90.007 3.40.001 -0.20.009 3.10.009 1.50.008 5.50.004 0.30.008 4.8 +Table 9 .9Top of Descent Errors From Phase IIProcedureTOD error, n.mi. Mean Std. dev.Non-FMS1.21.0FMS TOD2.52.8CTAS TOD1.00.9ND arc0.50.4All runs1.41.7All procedures using CTAS TOD*0.90.8*Includes non-FMS, CTAS TOD, and ND arc. +Table 10 .10Arrival Time Errors (Actual -Predicted) at Metering Fix for Phase IProcedureArrival time error, sec Mean Std. dev.Idle descent16.69.9RFD constrained9.96.4FFD constrained16.414.8All runs14.79.6Table 11. Arrival Time Errors (Actual -CTAS predicted)at Metering Fix for Phase IIProcedureArrival time error, sec Mean Std. dev.Non-FMS FMS TOD CTAS TOD1.9 -4.6 -9.98.7 13.9 10.2ND arc All runs2.3 -2.713.8 12.3 +Table 13 .13Combinations of Airplane Performance and Weather Models Used in Sensitivity Analysis of Phase I Idle DescentsSet namePerformance modelWeather modelBaselineBoeing 737-100CTAS MAPSRevised performanceTSRVCTAS MAPSRevised weatherBoeing 737-100Flight measuredRevised bothTSRVFlight measuredTable 14. Arrival Time Error Resulting From ModelingErrors in Phase I Idle DescentsModel parameterTime error, sec Mean Std. dev.Performance5.01.5Weather12.18.8Both16.89.6 + (a) Mean along-track wind error component.(b) Standard deviation in along-track wind error. + + + +Appendix TSRV Performance Model UpdateThe stabilized cruise and descent conditions flown in Phase I were used to refine the performance model of the airplane to reflect the actual performance measured during the test.Data tables were then generated by this revised performance model.The following sections describe the methods used in updating the airplane performance model and present the resulting modifications made to the thrust and drag models. +A.1. DragThe first step in updating the airplane drag model was to compute the error in drag coefficient based on flight-extracted drag.The TSRV airplane was not instrumented to accurately extract drag information during unstable and maneuvering flight conditions.Calibrated angle of attack, sideslip, and longitudinal and lateral accelerations were not available in the recorded data.The benign cruise and descent trajectories, however, allowed the use of classical performance equations for computations of approximate airplane drag.This technique was deemed adequate for the purposes of this experiment.The standard point mass equations of motion in a vertical plane were used to extract drag from the measured flight data.These equations areCombining equations (A1) and (A2), and solving for drag give (A3)Because the altitude and altitude rate measurement were based on pressure altitudes, the following correc-tion was applied to correct for nonstandard temperatures and obtain true altitude rate: Application of these equations to the flight data was accomplished by first defining criteria for identifying stable flight segments for analysis.The following criteria were used based on the available recorded data:1. Normal acceleration between 31.0 and 33.0 ft/sec/sec 2. Roll attitude less than 5°3. Criteria 1 and 2 valid for at least 10 secThe stable flight segments consisted of a minimum of 10 sec and maximum of 30 sec while the criteria were valid.The parameters required for equations (A3), (A4), and (A5) were averaged over the segment to provide a single value of drag coefficient error for the segment.This technique was applied to the 13 trajectories flown with the idle thrust descent procedure.Actual trajectories of the airplane were compared with the trajectories predicted by the CTAS trajectory synthesis algorithms and airplane Flight Management System (FMS).Trajectory prediction accuracy was evaluated over several levels of cockpit automation that ranged from a conventional cockpit to performance-based FMS vertical navigation (VNAV).Error sources and their magnitudes were identified and measured from the flight data.The major source of error during these tests was found to be the predicted winds aloft used by CTAS.The most significant effect related to flight guidance was the cross-track and turn-overshoot errors associated with conventional VOR guidance.FMS lateral navigation (LNAV) guidance significantly reduced both the cross-track and turn-overshoot error.Pilot procedures and VNAV guidance were found to significantly reduce the vertical profile errors associated with atmospheric and airplane performance model errors. + + + + + + + Design of Center-TRACON Automation System + + Heinz;Erzberger + + + ThomasJDavis + + + StevenGreen + + + + Machine Intelligence in Air Traffic Management, Andre Benoit + + Oct. 1993 + + + + AGARD CP-538 + Erzberger, Heinz; Davis, Thomas J.; and Green, Steven: Design of Center-TRACON Automation System. Machine Intelligence in Air Traffic Management, Andre Benoit, ed., AGARD CP-538, Oct. 1993, pp. 11-1-11-12. + + + + + + DavidHWilliams + + + StevenMGreen + + Airborne Four-Dimensional Flight Management in a Time-Based Air Traffic Control Environment. NASA TM-4249 + + 1991 + + + Williams, David H.; and Green, Steven M.: Airborne Four-Dimensional Flight Management in a Time-Based Air Traffic Control Environment. NASA TM-4249, 1991. + + + + + Design and Evaluation of an Advanced Air-Ground Data-Link System for Air Traffic Control + + DenBraven + + + William + + NASA TM-103899 + + 1992 + + + Den Braven, William: Design and Evaluation of an Advanced Air-Ground Data-Link System for Air Traffic Control. NASA TM-103899, 1992. + + + + + Piloted Simulation of an Air-Ground Profile Negotiation Process in a Time-Based Air Traffic Control Environment + + DavidHWilliams + + + StevenMGreen + + + + NASA TM-107748 + + 1993 + + + Williams, David H.; and Green, Steven M.: Piloted Sim- ulation of an Air-Ground Profile Negotiation Process in a Time-Based Air Traffic Control Environment. NASA TM-107748, 1993. + + + + + Development and Evaluation of a Profile Negotiation Process for Integrating Aircraft and Air Traffic Control Automation + + StevenMGreen + + + William;Den Braven + + + DavidHWilliams + + NASA TM-4360 + + 1993 + + + Green, Steven M.; Den Braven, William; and Williams, David H.: Development and Evaluation of a Profile Negotiation Process for Integrating Aircraft and Air Traffic Control Automation. NASA TM-4360, 1993. + + + + + A piloted simulator evaluation of a ground-based 4D descent advisor algorithm + + StevenMGreen + + + ThomasJDavis + + + HeinzErzberger + + 10.2514/6.1987-2522 + + + Guidance, Navigation and Control Conference + + American Institute of Aeronautics and Astronautics + Aug. 1987 + 2 + + + + Available as AIAA-87-2522. + Green, Steven M.; Davis, Thomas J.; and Erzberger, Heinz: A Piloted Simulator Evaluation of a Ground- Based 4D Descent Advisor Algorithm. AIAA Guidance, Navigation and Control Conference-Technical Papers, Volume 2, AIAA, Aug. 1987, pp. 1173-1180. (Avail- able as AIAA-87-2522.) + + + + + Piloted simulation of a ground-based time-control concept for air traffic control + + ThomasJDavis + + + StevenMGreen + + 10.2514/6.1989-3625 + NASA TM-101085 + + + Guidance, Navigation and Control Conference + + American Institute of Aeronautics and Astronautics + 1989 + + + Davis, Thomas J.; and Green, Steven M.: Piloted Simu- lation of a Ground-Based Time-Control Concept for Air Traffic Control. NASA TM-101085, 1989. + + + + + + George;Hunter + + + Tara;Weidner + + + George;Couluris + + + John;Sorensen + + + RichardBortins + + CTAS Error Sensitivity, Fuel Efficiency, and Throughput Benefits Analysis + + Crown Commun., Inc.), Seagull Technol., Inc + July 1996 + + + + Subcontract under FAA Contract DTAFA01-94-Y-01046 to + Hunter, George; Weidner, Tara; Couluris, George; Sorensen, John; and Bortins, Richard: CTAS Error Sensitivity, Fuel Efficiency, and Throughput Benefits Analysis. 96150-02 (Subcontract under FAA Contract DTAFA01-94-Y-01046 to Crown Commun., Inc.), Seagull Technol., Inc., July 1996. + + + + + + GJCouluris + + + TWeidner + + + JASorensen + + Initial Air Traffic Management (ATM) Enhancement Potential Benefits Analysis + + Crown Commun., Inc.), Seagull Technol., Inc + Sept. 1996 + + + + Subcontract under FAA Contract DTFA01-96-Y-01009 to + Couluris, G. J.; Weidner, T.; and Sorensen, J. A.: Initial Air Traffic Management (ATM) Enhancement Potential Benefits Analysis. 96151-01 (Subcontract under FAA Contract DTFA01-96-Y-01009 to Crown Commun., Inc.), Seagull Technol., Inc., Sept. 1996. + + + + + Design Principles and Algorithms for Automated Air Traffic Management. Knowledge-Based Functions in Aerospace Systems + + HeinzErzberger + + AGARD-LS-200 + + 1995 + + + Erzberger, Heinz: Design Principles and Algorithms for Automated Air Traffic Management. Knowledge-Based Functions in Aerospace Systems, AGARD-LS-200, 1995. + + + + + RTCA, Inc. - Establishing the framework for a systems approach to aviation + + DavidWatrous + + + DavidWatrous + + 10.2514/6.1997-5640 + + + 1997 World Aviation Congress + + American Institute of Aeronautics and Astronautics + 1995 + + + Reprinted Feb. 1996. + Final Report of RTCA Task Force 3-Free Flight Imple- mentation. RTCA, Inc., 1995. (Reprinted Feb. 1996.) + + + + + Atmospheric data acquisition and interpolation for enhanced trajectory-prediction accuracy in the Center-TRACON Automation System + + MRJardin + + + HErzberger + + 10.2514/6.1996-271 + AIAA-96-0271 + + + 34th Aerospace Sciences Meeting and Exhibit + + American Institute of Aeronautics and Astronautics + Jan. 1996 + + + Jardin, M. R.; and Erzberger, H.: Atmospheric Data Acquisition and Interpolation for Enhanced Trajectory- Prediction Accuracy in the Center-TRACON Automa- tion System. AIAA-96-0271, Jan. 1996. + + + + + Terminal area trajectory synthesis for air traffic control automation + + Rhonda;Slattery + + + SteveGreen + + 10.1109/acc.1995.520941 + + + Proceedings of 1995 American Control Conference - ACC'95 + 1995 American Control Conference - ACC'95 + + American Autom Control Council + 1994 + + + Slattery, Rhonda; and Green, Steve: Conflict-Free Tra- jectory Planning for Air Traffic Control Automation. NASA TM-108790, 1994. + + + + + En-route descent trajectory synthesis for air traffic control automation + + RhondaASlattery + + + YiyuanZhao + + 10.1109/acc.1995.532248 + + + Proceedings of 1995 American Control Conference - ACC'95 + 1995 American Control Conference - ACC'95 + + American Autom Control Council + June 1996 + + + + Slattery, Rhonda A.; and Zhao, Yiyuan: En-Route Descent Trajectory Synthesis for Air Traffic Control Automation. Proceedings of the American Control Con- ference, June 1996, pp. 3430-3434. + + + + + Enabling user preferences through data exchange + + StevenMGreen + + + Tsuyoshi;Goka + + + DavidHWilliams + + + StevenMGreen + + + Tsuyoshi;Goka + + + DavidHWilliams + + 10.2514/6.1997-3682 + AIAA-97-3682 + + + Guidance, Navigation, and Control Conference + + American Institute of Aeronautics and Astronautics + 1997 + + + Green, Steven M.; Goka, Tsuyoshi; and Williams, David H.: Enabling User Preferences Through Data Exchange. AIAA-97-3682, 1997. + + + + + Mesoscale Weather Prediction with the RUC Hybrid Isentropic–Terrain-Following Coordinate Model + + StanleyGBenjamin + + + GeorgAGrell + + + JohnMBrown + + + TatianaGSmirnova + + + Rainer;Bleck + + 10.1175/1520-0493(2004)132<0473:mwpwtr>2.0.co;2 + + + Monthly Weather Review + Mon. Wea. Rev. + 0027-0644 + 1520-0493 + + 132 + 2 + + Aug. 1993 + American Meteorological Society + + + Benjamin, Stanley G.; Bleck, Rainer; Grell, Georg; Pan, Zaitao; Smith, Tracy Lorraine; Brown, John M.; Ramer, James E.; Miller, Patricia A.; and Brundage; Kevin A.: Aviation Forecasts From the Hybrid-B Version of MAPS-Effects of New Vertical Coordinate and Improved Model Physics. Fifth Conference on Aviation Weather Systems, American Meterol. Soc., Aug. 1993, pp. J-5-J-9. + + + + + An Hourly Assimilation–Forecast Cycle: The RUC + + StanleyGBenjamin + + + DezsöDévényi + + + StephenSWeygandt + + + KevinJBrundage + + + JohnMBrown + + + GeorgAGrell + + + Dongsoo;Kim + + + BarryESchwartz + + + TatianaGSmirnova + + + TracyLorraineSmith + + + GeoffreySManikin + + 10.1175/1520-0493(2004)132<0495:ahactr>2.0.co;2 + + + Monthly Weather Review + Mon. Wea. Rev. + 0027-0644 + 1520-0493 + + 132 + 2 + + July 1994 + American Meteorological Society + + + Benjamin, Stanley G.; Brundage, Kevin J.; Miller, Patricia A.; Smith, Tracy Lorraine; Grell, Georg A.; Kim, Dongsoo; Brown, John M.; and Schlatter, Thomas W.: The Rapid Update Cycle at NMC. Tenth Conference on Numerical Weather Prediction, Ameri- can Meterol. Soc., July 1994, pp. 566-568. + + + + + The NASA Langley Research Center 8-ft High Temperature Tunnel + 10.2514/5.9781600866678.0405.0425 + + + Advanced Hypersonic Test Facilities + + American Institute of Aeronautics and Astronautics + 1980 + + + + Staff of NASA Langley Research Center and Boeing Commercial Airplane Co.: Terminal Configured Vehicle Program-Test Facilities Guide. NASA SP-435, 1980. + + + + + + CharlesEKnox + + Description of the Primary Flight Display and Flight Guidance System Logic in the NASA B-737 Transport Systems Research Vehicle. NASA TM-102710 + + 1990 + + + Knox, Charles E.: Description of the Primary Flight Display and Flight Guidance System Logic in the NASA B-737 Transport Systems Research Vehicle. NASA TM-102710, 1990. + + + + + Design of Automated System for Management of Arrival Traffic. NASA TM-102201 + + Heinz;Erzberger + + + WilliamNedell + + + 1989 + + + Erzberger, Heinz; and Nedell, William: Design of Auto- mated System for Management of Arrival Traffic. NASA TM-102201, 1989. + + + + + Derivation of parallel algorithms from functional specifications to CSP processes + + AliEAbdallah + + 10.1007/3-540-60117-1_6 + + + Lecture Notes in Computer Science + + Springer Berlin Heidelberg + Aug. 1995 + + + + Computer Program Functional Specifications-Multiple Radar Data Processing. Model A4e1.3, NAS-MD-320, FAA, Aug. 1995. + + + + + Development and Initial Field Evaluation of Flight Deck Procedures for Flying CTAS Descent Clearances + + Patricia;Cashion + + + MFeary + + + NSmith + + + TGoka + + + HGraham + + + EPalmer + + + + Proceedings of the Eighth International Symposium on Aviation Psychology + + RichardSJensen + + + LoriARakovan + + the Eighth International Symposium on Aviation Psychology + + Assoc. of Aviation Psychologists + 1995 + + + + Cashion, Patricia; Feary, M.; Smith, N.; Goka, T.; Graham, H.; and Palmer, E.: Development and Initial Field Evaluation of Flight Deck Procedures for Flying CTAS Descent Clearances. Proceedings of the Eighth International Symposium on Aviation Psychology, Vol- ume 1, Richard S. Jensen and Lori A. Rakovan, eds., Assoc. of Aviation Psychologists, 1995, pp. 438-443. + + + + + Atmospheric Data Error Analysis for the 1994 CTAS Descent Advisor Preliminary Field Test + + MRJardin + + + SMGreen + + NASA/TM-1998-112228 + + 1998 + + + Jardin, M. R.; and Green, S. M.: Atmospheric Data Error Analysis for the 1994 CTAS Descent Advisor Pre- liminary Field Test. NASA/TM-1998-112228, 1998. + + + + + Impact of Mismodeled Idle Engine Performance on Calculation and Tracking of Optimal 4-D Descent Trajectories + + DavidHWilliams + + 10.23919/acc.1986.4789025 + + + 1986 American Control Conference + + IEEE + 1986 + 2 + + + + Williams, D. H.: Impact of Mismodeled Idle Engine Performance on Calculation and Tracking of Optimal 4-D Descent Trajectories. Fifth American Control Con- ference, Volume 2, IEEE, 1986, pp. 681-686. + + + + + Land use and land cover and associated maps for Seattle, Washington (covers western 50% of the Seattle map sheet only; east 50% will be covered by 1:100,000-scale map sheet) + 10.3133/ofr7713 + + + AGENCY USE ONLY (Leave blank) 2. REPORT DATE 3. REPORT TYPE AND DATES COVERED + + US Geological Survey + + + + AGENCY USE ONLY (Leave blank) 2. REPORT DATE 3. REPORT TYPE AND DATES COVERED + + + + + Report subtitle + + JonStark + + 10.5555/testdoi12344 + + + Report + + Stark Publishing + + + + + TITLE AND SUBTITLE 5. FUNDING NUMBERS + + + + + Author! Author!, Author! Author! Image 1 + + Author(s) + + 10.3998/mpub.11373292.cmp.305 + + null + University of Michigan Library + + + AUTHOR(S) + + + + + + + Organization Name + + S) AND ADDRESS + + + + ES + + + PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) + + + + + Spain (‘.es’) + + AlbertAgustinoyGuilayn + + 10.1093/oso/9780199663163.003.0035 + + + Domain Name Law And Practice + + Oxford University Press + + + + SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES) + + + + + The peak performing organization: An overview + 10.4324/9780203971611-11 + + + The Peak Performing Organization + + Routledge + + + + + SUPPLEMENTARY NOTES 8. PERFORMING ORGANIZATION REPORT NUMBER + + + + + Data monitoring and follow-up + 10.18356/dbfda0e4-en + + + Report of the Inter-Agency Task Force on Financing for Development + + United Nations + + + + + SPONSORING/MONITORING AGENCY REPORT NUMBER + + + + + SUMMARY STATEMENT OF FINDINGS RELATED TO THE DISTRIBUTION, CHARACTERISTICS, AND BIOLOGICAL AVAILABILITY OF FALLOUT DEBRIS ORIGINATING FROM TESTING PROGRAMS AT THE NEVADA TEST SITE + + KHLarson + + + JWNeel + + 10.2172/4124261 + + + DISTRIBUTION/AVAILABILITY STATEMENT 12b. DISTRIBUTION CODE + + Office of Scientific and Technical Information (OSTI) + + + + DISTRIBUTION/AVAILABILITY STATEMENT 12b. DISTRIBUTION CODE + + + + + + + Abstract + + + + Maximum 200 words + ABSTRACT (Maximum 200 words) + + + + + + + Subject Terms + + 10.7287/peerj.preprints.717/supp-2 + + null + PeerJ + + + SUBJECT TERMS + + + + + security classification + + Security + + + Of + + 10.1007/springerreference_24206 + + + LIMITATION OF ABSTRACT + + 18 + null + Springer-Verlag + + + REPORT + SECURITY CLASSIFICATION OF REPORT 18. SECURITY CLASSIFICATION OF THIS PAGE 19. SECURITY CLASSIFICATION OF ABSTRACT 20. LIMITATION OF ABSTRACT + + + + + + diff --git a/file782.txt b/file782.txt new file mode 100644 index 0000000000000000000000000000000000000000..8e1eeba08957433d434e99e2bdf544beac7ef0a3 --- /dev/null +++ b/file782.txt @@ -0,0 +1,487 @@ + + + + +Nomenclature +I. IntroductionRAFFIC on the surface of busy airports in the United States accounts for a large part of national flight delays, noise, and emissions. 2Generally, flights push back from their gates and taxi to a queue of flights waiting to use the runway.While the flights are queued, they burn fuel, release emissions, and produce noise.Several methods for reducing this wait time are described in the literature.One method is to keep the queue as short as possible without drying out operations on the runway by holding flights at their gates or spots.Another method is to maximize the usage of the runway by optimally sequencing operations on the runway.4] field tested a concept for metering the number of flights allowed to push back from the gate onto the airport taxiway.Results showed that the concept was able to save 3,900-4,900 US gallons of fuel during four eight-hour tests by decreasing the taxi time, with a increase in gate departure delay of 4.3 minutes per flight on average.Brinton et.al. 5 developed a Collaborative Departure Queue Management concept.This concept uses the ration-by-schedule algorithm that is used in Ground Delay Programs to assign flight-by-flight departure slots.The concept was field tested at Memphis International Airport.Initial results showed that the concept was able to provide slot times that controlled the departure queue length.Although these concepts produced benefits, their departure sequences were not guaranteed to be minimum-delay sequences, so other methods may be able to produce even greater benefits.An algorithm for generating minimum-delay and maximum-throughput departure sequences has been developed using the basic principles of mixed integer linear programming [6][7][8][9][10] and of dynamic programming. 11The dynamic programming version produced in less execution time similar results to the mixed integer linear programming version.In a standalone analysis, departure sequences produced by the dynamic programming version had 43% less delay and 1.3% more throughput than those produced by another algorithm that modeled the human controller. 12his algorithm was integrated into a concept for managing surface traffic named the Spot and Runway Departure Advisor (SARDA), which is being developed at NASA Ames Research Center. 13In 2010, high-fidelity human-inthe-loop (HITL) simulations of the concept managing traffic on the east side of Dallas/Fort Worth International airport were conducted.Initial results for the heavy scenario studied showed that the concept reduced departure delays by 64% and fuel consumption on the ground by 38%, while imposing little impact on perceived controller workload.1&14 There were little changes in the performance metrics for the normal scenario that was studied.The HITL simulations analyzed two traffic scenarios at a single airport.In addition, the simulations analyzed a single version of the SARDA concept.To expand research of the SARDA concept to include more traffic scenarios and airports and to investigate how changes to the concept would affect benefits, NASA developed a fast-time simulation of airport surface traffic named the Surface Operations Simulator and Scheduler (SOSS). 15The purpose of SOSS is to efficiently test and analyze concepts for managing traffic on airport surfaces.In this paper, SOSS was used to conduct a fast-time simulation of a delay-optimal scheduler managing traffic on the east side of Dallas/Fort Worth International Airport.This simulation was an initial step toward using SOSS to analysis the benefits of SARDA.Although the delay-optimal scheduler was the same as the one used in the 2010 HITL simulation of SARDA, the way the scheduler was applied to managing traffic was different.In the 2010 HITL simulation, flights crossing the runway were included in the optimal schedule, while in the fast-time simulation crossing flights were not included in the optimal schedule.In addition in the 2010 HITL simulation, flights were cleared from the spot, while in the fast-time simulation, flights were cleared from the gate.A summary of all the differences is provided in the next section.The differences were due to time limitations in adapting the scheduler from the HITL simulation to SOSS and lack of capabilities of SOSS.Performance in terms of delay, number of stops, and taxi time of the delay-optimal simulation was compared with the performance of a simulation of a firstscheduled-first-served scheduler managing traffic.In addition, equity in terms of which simulation had the most delayed flight was studied.This paper is organized as follows.First, the 2010 HITL SARDA simulation is described and compared with the fast-time simulation.Then, the details of the fast-time simulation tools and algorithms are explained, and the results are presented.The paper ends with conclusions. +II. Spot and Runway Departure AdvisorThis section describes the Spot and Runway Departure Advisor (SARDA) concept as it was simulated during the 2010 HITL simulation.First, airport surface locations and procedures that are used extensively in SARDA are explained.Then, SARDA is presented.Finally, a summary of the differences between the SARDA 2010 HITL simulations and the fast-time simulations is provided.At Dallas/Fort Worth International Airport, spots are locations on the surface where responsibility for control of flights passes from the ramp controller, who works for the airline, to the ground controller, who works for the FAA. +TTo pass through the spot and enter the taxiway, pilots must obtain clearance from the ground controller.The ground controller maintains control over flights in the taxiway until they enter the departure queue.The departure queue is the area just before the runway entrance where flights line up to takeoff.As flights enter the departure queue, the ground controller hands them off to the local controller.The local controller is responsible for enforcing safe operations on the runway.He clears departing, crossing, or arriving flights to enter and use the runway.Flights wait their turn in the departure queues or at the runway crossings until the local controller gives them clearance, at which time they may use the runway.The Spot and Runway Departure Advisor (SARDA) is a NASA developed concept for managing surface traffic.Its goal is to reduce taxi times and delay and, thereby, reduce fuel-burn, noise, and emissions.It accomplishes this by providing advisories to ramp, ground, and local controllers.Ground controllers use their advisory to clear flights from the spot, and local controllers use their advisory to sequence flight operations (departures and crossings) on the runway.The focus of SARDA to date as been on decision support tools for controllers.However, the long term vision for SARDA is that it also include a decision support tool for the airline ramp controller, which will be used to manage traffic in the ramp area.SARDA provides the ground controller with an advisory of when to clear flights at the spots to enter the taxiway.The advisories are designed to meter the number of flights entering the taxiway such that traffic on the taxiway is low and the departure queues are as small as possible without letting the runway operations dry out.For example, subject matter experts estimate the proper number of aircraft in the queue at Dallas/Fort Worth International Airport to be about five flights distributed across three queues that feed runway 17R.This prevents excessive flight taxi times, emissions, and noise.In addition to metering flights entering the taxiway, the spot clearances are also designed to satisfy a minimumdelay sequence of operations on the runway, where delay is defined as the difference between the flight's scheduled runway use time and the time it would have used the runway if there was no traffic.The sequence includes flight departures, arrivals, and runway crossings and is calculated by the SARDA tools.Although the flights are released from the spot in the delay-optimal sequence to use the runway, unexpected events during taxi (such as conflicts with other flights) can change the sequence.A goal of SARDA research, not addressed in this paper, is to understand how these events degrade the benefits of a delay-optimal schedule.SARDA provides the local controller with an advisory of the runway operations sequence.The local controller can use this sequence to decide which flight to allow to use the runway next.Depending on the runway, the local controller may have several options.In the south flow airport configuration at Dallas/Fort Worth International Airport for example, runway 17R has three primary departure queues and ten crossing locations.Flights may be waiting in any of these areas to use the runway.The advisory helps the controller identify the next aircraft to use the runway that will minimize delays.The SARDA concept used in the 2010 human-in-the-loop simulation calculated ground and local controller advisories with two programs: the runway scheduler and the spot release planner.The runway scheduler (RS) generated the minimum-delay sequences of operations on the runway, and the spot release planner (SRP) generated the spot clearance times. 13Both the RS and SRP had their own implementation of a delay-optimal scheduler.The delay-optimal scheduler calculated, using a dynamic programming approach, minimum-delay departure schedules for the runway.The RS sequence advisory and SRP spot clearance advisory were generated from independent calculations of minimum-delay departure schedules.The delay-optimal scheduler was adapted from the RS and the SRP to work with SOSS.Due to time constraints to complete the adaptation and limitations of SOSS, the scheduler used in this paper was not configured exactly as the SARDA concept in the 2010 HITL.The scheduler produced flight clearances at the gate, instead of at the spot, which is the case for the 2010 HITL.Since clearances were produced at the gate, the scheduler is hereafter referred to as the gate release planner (GRP).Another difference between the GRP fast-time simulations and the HITL was that in the HITL the local controller was asked to follow an advisory given by the RS that specified the sequence that flights should use the runway.In the GRP, an optimal sequence was not enforced at the runway entrances.Flights were allowed to use the runway on a first-available-first-served basis.There were several additional differences between the GRP simulation and the HITL simulations.In the HITL simulations, all three runway entrance queues were used and multiple runway crossings were set up and used extensively by the local controllers.In the GRP simulation, only one of the three runway entrance queues, the full queue, was used.This was due to limitations of the modeling capabilities of the Surface Operations Simulator and Scheduler (SOSS), which is currently being enhanced.Using the full queue only did not greatly affect the results because its length never extended to the point that it was blocking a taxiway intersection.Furthermore in the GRP simulation, all of the runway crossings ended up being single crossings because there was no logic simulating the actions local controls use to set up multiple crossings. +III. Fast-time SimulationThis section describes the fast-time simulation tools and setup.It illustrates how the main parts of the system, Surface Operations Simulator and Scheduler (SOSS) and the scheduler, fit together.It also introduces the details of the airport and flight dynamics models.SOSS connects with schedulers via NASA's common algorithm interface, which is the same interface used by all of NASA's airport simulation tools.This feature makes SOSS able to easily connect with any of the schedulers that NASA has developed because they also use the common algorithm interface.The common algorithm interface passes for each flight within the current time horizon Ptimes or wheels-on times, routes, and ETAs to the scheduler.The scheduler uses this information to create a new schedule, then passes the schedule back to SOSS in the form of a set of clearance times.The user controls how SOSS connects to the scheduler.Through SOSS's graphical user interface (GUI), the user can set the frequency of calls to the scheduler and the planning horizon.In this research, the scheduler was called every five minutes and the planning horizon was set to ten minutes.Several simulations were conducted with different call frequency and planning horizons, and these values produced adequate results with good execution speed.SOSS accepts a traffic scenario file that contains the filed flight schedule.This schedule includes a filed pushback time (Ptime), a taxi route, and a takeoff runway for departure flights and a landing runway, a wheels-on time, a taxi route, and gate for arrival flights.In the simulations, the Ptime is updated by the gate release time calculated by the gate release planner, and the wheels-on times are treated as absolutes that may not be modified.In practice, wheels-on times can be slightly modified by TRACON controllers who are controlling the final approach phase of the flight.In addition to the traffic scenario file, SOSS also reads in airport adaptation data that describes the node/link network model of the airport and the routes that flights will use to move from the gates to the runways and visa versa.An aircraft database is also read by SOSS.It contains parameters describing important characteristics of each type of aircraft.For executing a simulation, SOSS contains three models: an airport model, a flight dynamics model, and a tactical flight separation model.These are described in more detail in the following sections. +Airport ModelSOSS models the ramps, taxiways, and runways at airports with nodes and links.Any airport can be characterized by applying nodes and links to its specific ramp/taxi/runway layout.Nodes represent points on the airport surface, and links represent paths between nodes.Since links are straight lines, curved paths must be approximated with multiple nodes and links.Intersections where multiple paths meet are represented by nodes.Flight routes are paths through the predefined airport node/link network.They are defined by a set of airport nodes.Flight clearance times at nodes in the routes can be set by the scheduler.When a clearance time is set at a node in a flight's route, SOSS does not allow the flight to exit the node until after the clearance time has elapsed.Although in general a clearance time can be set at any node in a flight's route, in the present research flight clearances for each flight are set only at the gate.The simulation primarily focused on operations on the east side of Dallas/Fort Worth International Airport (DFW).However, it included a few flights that crossed sides.East-and west-side operations at DFW are almost independent of each other because they are controlled by separate towers and their areas of control are divided down the center of the airport.The two sides connect only by the north and south bridges, which flights use to cross the airport.The airport was in the south flow configuration, which is used about 70% of the time.Jet arrivals approached from the north and landed on runways 17C, 17L, and 18R ( 18R is on the west side of the airport, not shown in the Fig. 2) and jet departures took off from runways 17R and 18L (18L is also on the west side of the airport) towards the south.Flights departed from and arrived at gates on the west and east sides.The south flow configuration and focus on the east side of DFW is consistent with the SARDA 2010 HITL simulation.Figure 2 shows the node/link model with the runways labeled and outlined in black and the main east north/south taxiways highlighted in purple.The blowup pictures provide greater detail of the nodes and links.In clockwise direction the blowups are of the runway 17R departure queue, the intersection leading to runway 13L, and Terminal A ramp.Table 2 shows what each of the node colors denotes.Runway 13L is used primarily for turboprops, which where not included in the traffic scenario used in this paper.where s is path length, v is aircraft speed, and a is aircraft acceleration.Acceleration is defined as one of three values depending on the current and target speeds of the aircraft.If the current speed is equal to the target speed, a=0.If the current speed is less than the target speed, a=a max , and, if the current speed is greater than the target speed, a=a min , which is a negative value.An aircraft has three different target speeds: one for movement in the ramp area (v ramp ), one for movement in the taxiways (v taxi ), and one for movement in the runway queues (v queue ).The accelerations and the target speeds are defined in the aircraft parameters database.This database contains these and other parameters for 458 different types of aircraft.Table 3 gives the parameters for a Boeing 737 jet.These values are typical for the majority of the jets in the simulations.The target speed of the aircraft is determined by the aircraft speed control model.SOSS has two speed control models: open-loop and closed-loop.Open loop speed control simulates flight procedures used in the current system, while closed loop speed control simulates flight procedures in a future system.In open-loop control, the aircraft's target speed is determined from the aircraft parameters database.This target speed represents the preferred taxi speed that a pilot would use.As described above, this speed is constant for aircraft of the same type and in the same area: departure queue, taxiway, or ramp.In the real system, different pilots taxi at different speeds.The target speed used in SOSS represents an average of the distribution of speeds used in real life.In the present simulations, target speeds were treated as deterministic.Closed-loop control overrides the target speed from the database and adjusts it within bounds given by the database such that the flights rendezvous with their target nodes at just the scheduled times.For this to occur in practice, aircraft would require new flight deck equipment.The simulations in the current research were meant to model existing operations.Therefore, the open loop control option was used. +Tactical Flight Separation ModelThe tactical flight separation model simulates tactical actions that pilots and controllers take to maintain safe separation between aircraft.Separation is handled differently for flights using the runway than for flights taxiing through the ramp, taxi, and queuing areas.Tactical separations at the runway are determined by navigational safety and wake-vortex spacing constraints.In practice, controllers enforce a separation using a distance-based rule.SOSS enforces a separation by holding a flight at the entrance of the runway until a specified amount of time from the previous operation has elapsed.The time is calculated to achieve the correct distance between operations.Table 4 shows the times required between consecutive departures.The weight classes of the aircraft are denoted by small (S), large (L), heavy (H), and B757 (B75).The B757 has its own weight class because its weight is in the large class, but it creates more turbulence than other large aircraft.Rows specify the weight class of the leader, and columns specify the weight class of the follower.The separation time between a departure and a crossing, and vice versa, was 40 seconds.In SOSS, separation at the departure fix was also constrained.SOSS actually enforces the separation constraint at the runway take-off point because it currently does not model movement of aircraft in the air.In these simulations, consecutive flights headed for the same departure fix were spaced by 80 seconds.A spacing of 80 seconds at the departure fix translates into a five miles-in-trail restriction into the enroute airspace.This was not consistent with the SARDA 2010 HITL simulation.That simulation did not have constraints at the fixes.The method used by SOSS for maintaining separation in the ramp, taxi and queuing areas consists of a conflictdetection algorithm and a conflict-resolution algorithm.In contrast to the tactical separation model, the gate release planner did not attempt to give clearance times at the gate that would prevent conflicts.It was assumed that conflicts would be taken care of by the tactical separation algorithm.The conflict-detection algorithm uses the aircraft route to project the locations of all aircraft at future times, which are configurable by the user.In this research, 35.7, 75, 112.5, and 150 seconds in the future were selected because they allowed the detection algorithm to find most conflicts and did not noticeably slow SOSS execution speed.The projected locations of each aircraft are checked against the projected locations of the other aircraft to determine if there are any pairs of aircraft that will get too close to each other.If a pair is identified, the situation is called a conflict and actions are taken to prevent the two aircraft from getting too close.Conflicts are classified into one of several categories: in-trail, head-on, and merging.An in-trail conflict occurs when a trailing aircraft overtakes its lead traveling in the same direction on the same link, and a head-on conflict occurs when two aircraft traveling in different directions on the same link violate separation.A merging conflict occurs when two aircraft cross paths on an intersection at the same time.The resolution algorithm generates actions that prevent the conflict from turning into an actual collision.The resolution algorithm handles in-trail, merging, and head-on conflicts differently.For in-trail conflicts, the algorithm slows down the trailing aircraft.For merging conflicts, the algorithm determines which aircraft will arrive at the intersection first.Then, it slows the other aircraft so that the first can safely pass through the intersection.Finally, for head-on conflicts, the algorithm determines the set of links that make up the common path between the aircraft.Then, it identifies which aircraft will arrive first at its nearest link in the set.The aircraft that reaches its nearest link last is slowed or stopped outside the common path until the other aircraft safely passes. +B. SchedulerThe GRP produced gate-release times for flights using a two-step process.First using a delay-optimal scheduler, it built a runway take-off schedule.The delay-optimal scheduler was the same as that used in the core of the RS and SRP in the 2010 SARDA HITL, and it was based on a dynamic programming approach.The details of the dynamic programming approach are published in Refs.11 and 12.The schedule was constrained so that it satisfied wake vortex, navigation, and miles-in-trail separation requirements.In the second step, the gate-release time was calculated by subtracting the unimpeded taxi-time from the flight departure time.This clearance time did not account for traffic that the flight would encounter on its way to the runway.The tactical flight separation model was used to resolve conflicts that occured. +IV. ResultsTwo fast-time simulations of normal traffic at Dallas/Fort Worth International Airport were conducted to show the benefits of the gate release planner (GRP).These simulations illustrate how SOSS can be used to analyze the benefits of a concept for managing flights on the airport surface.The simulations are not an analysis of SARDA itself due to the differences listed in Section II.The GRP was used to control surface traffic in one simulation, and a first-come-first-served (FCFS) scheduler was used to control traffic in the other simulation.Performance of the GRP scheduler was compared with that of the FCFS scheduler.The FCFS scheduler was used for comparison because it was simple to model.It represents the case where operations are conducted in the same sequence that they are scheduled.Although it is similar to real operations, it is not the same because typically a controller will do positionconstrained re-sequencing to increase throughput.The departure spacing restrictions listed in Table 4 were enforced by SOSS.Similarly, 40 seconds were maintained between departures and crossings.Finally, an 80 second constraint was applied to consecutive departures using the same departure fix.In the GRP simulation, minimum-delay sequences of departures on runway 17R were generated by the optimal scheduler.Arrivals from 17C and 17L that needed to cross 17R were entered into the sequence on an availability basis.When there was a crossing that was ready to use 17R before the leading departure, it was cleared to cross.The schedules were updated every five minutes, starting at the initial simulation time of zero.Generally a higher update rate would be desired to capture as soon as possible any events that were not predicted.Five minutes was used here because the only unpredicted events were conflicts, and there were not enough conflicts to adversely affect the performance of the scheduler.The time horizon of the schedules was ten minutes.Time horizons that are too large can cause the scheduler to execute slowly because there are too many flights in the schedule.A ten-minute horizon caused the scheduler to execute quickly, less than a second, while not degrading the benefits of the simulation.Operations on runways other than 17R were sequenced using first-come-first-served.As discussed in the Section II, flights in the GRP simulation were controlled by strategically releasing them from their gates.No other strategic clearances were given to the flights as they moved through the ramp, taxi, and queue areas.However, as described in the tactical separation section of the paper, the flights did use tactical maneuvers to avoid separations violations.This caused the flights to actually reach the runway entrance at uncertain times that could differ from the runway-scheduled times.No other uncertainties were modeled in the simulation. +A. Traffic ScenarioThe traffic scenario was created by recording live surface traffic at Dallas/Fort Worth International for a peak hour in 2008.All flights with routes that did not intersect with runway 17R or the main inboard north/south taxiways K and L (shown in purple on Fig. 2) were filtered out.This produced full traffic levels in these areas, while minimizing the number of flights that were simulated.The traffic scenario consisted of 80 flights scheduled in a 42 minute period.Forty flights were arrivals and forty flights were departures.All of the flights were in the large weight class.Figure 3 illustrates the 17R operation departure rates.There are three curves.The demand curve represents the operation rate that would occur if the schedule were exactly followed and there were no separation constraints at the runway or along the taxiways.The FCFS curve illustrates the actual operations rates in the FCFS simulation, and the GRP curve illustrates the actual operations in the GRP simulation.The FCFS is drawn mostly on top of the GRP curve, except for a small portion from 700 to 1300 seconds.The operations rates were calculated by counting the number of departures and crossings occurring within a 5 minute sliding bin.Each point in the curve is created by shifting the bin one minute.The runway capacity is the operation rate where the runway cannot accept more operations without violating a separation constraint.It is shown on Fig. 3 where the FCFS and GRP curves flatten out.The capacity for 17R is about 7.5 operations per 5 minutes, depending on the mix † of operations.As shown in Fig. 3, demand exceeds capacity from approximately 500 to 900 seconds and from 1250 to 1750 seconds in the traffic scenarios.In Fig. 3, the FCFS and GRP curves are higher than the demand curves from approximately 1000 to 1250 seconds and 1750 to 2500 seconds.These intervals represent times when 17R departures have been delayed, and, even though demand has dropped, the actual operation rate is still at capacity so that the waiting flights can depart.Figure 4 shows the cumulative operations verses simulation time.Cumulative operations is like operations per bin except that a bin size does not need to be selected to create the plot.Again, the GRP curve is on top of the FCFS curve.At time 1300 the GRP and FCFS curves touch the demand curve, indicating that the backlog of flights waiting to use 17R has been served.After 1300 seconds the demand increases beyond the capacity of 17R and a backlog accumulates again.That backlog is finished be served at 2400 seconds.The maximum capacity of 17R is shown on this plot by the maximum slop of the line.There is very little difference between the GRP and FCFS curves.This indicates that in the GRP simulation the optimal scheduler was not able to find a sequence of runway operations that was able to increase the runway throughput above the level achieved by first-come-first-served.This is the result of only one simulation, more simulations would be needed to establish this result in general.The optimal scheduler may have had more opportunity to find better sequences if crossings had been included in the schedule and the inner and outer queues had been used.Also, all of the flights in the traffic scenario were in the large weight class.If there had been a more diverse set of weight classes, the scheduler would have had more opportunities to find better sequences. +B. MetricsIn this section, three metrics are explored: delay, number of stops, and taxi times.Delays were calculated by subtracting nominal unimpeded times from actual times.For example, to calculate total departure delay for a flight, the nominal time at which the flight was initially planning to takeoff is subtracted from the actual time the flight took off.These times can be taken at any node along the route, not just the take-off node.The number of stops is calculated by counting the number of times the speed of the aircraft drops below a specified threshold value.In this research, 0.1 knot was used as the threshold value because it is 100 times smaller than the target speed in the queue area, which was 10 knots.Taxi times were calculated by subtracting the release time at the gate from the take-off time at the runway.Table 5 gives averages, standard deviations, minimums and maximums of each metric for departures in the two simulations.Table 6 does the same for arrivals.Average delays are very similar for departures in the two simulations.This means the GRP was not able to reduce the total departure delay for flights.However, the average taxi time for departures in the GRP simulation was 45 seconds less than that for the FCFS simulation.This is because in the GRP simulation departures were held longer at their gates.Since engines are not turned on while at the gates, this lower taxi time directly translates into less fuel burned, emissions, and noise.Similarly, departures in the GRP simulation had fewer average stops.This is because by holding departures at the gates, the GRP reduces the According to Table 6, the GRP arrivals had a little less average delay.The slight difference is because holding flights at the gates produces less traffic for the arrivals to navigate through to get to their gates.Similarly, the GRP arrivals had fewer stops than the FCFS arrivals.In addition, the standard deviations are large.Thus, the less delay for GRP arrivals is not consistent. +V. Future WorkThe GRP simulation did not measurably increase the throughput rate of runway 17R or decrease average departure delay despite other research [6][7][8][9][10][11][12] that shows in standalone analysis that optimal departure scheduling can decrease delay.Future research could investigate other traffic scenarios, including runway crossings in the optimal scheduling process, and better prediction and control methods to see if optimal scheduling can increase the runway throughput.The GRP simulation was different from the SARDA 2010 HITL.To better confirm the results of SARDA 2010, the fast-time simulations could be made more like SARDA by including runway crossings in the delay-optimal scheduler, routing departures through the full, inner, and outer queues which feed runway 17R, and enforcing an optimal sequence at runway 17R.In addition, the GRP simulation used flight clearances at the gates.The 2010 SARDA simulation used flight clearances at the spots.A study could be conducted of how clearing flights at the spot or the gate would change the benefits.Aside from conflict detection and resolution, this study did not include uncertainty.Future research could include uncertainty in gate pushback times and uncertainty in human response times.This study focused on a single runway at Dallas/Fort Worth International Airport.A future study could apply the optimal scheduling to more than one departure runway, including runways with mixed operations or constraints with neighboring runways.In this study only a single traffic scenario was simulated.A future study could investigate many traffic scenarios. +VI. ConclusionA fast-time simulation of airport surfaces, the Surface Operations Simulator and Scheduler (SOSS), was able to analyze the benefits of a delay-optimal scheduler controlling surface traffic on the east side of Dallas/Fort Worth International airport.The delay-optimal scheduler used in the fast-time simulation was adapted from the 2010 Spot and Runway Departure Advisor (SARDA) human-in-the-loop (HITL) simulation.Although the fast-time simulation was meant to measure the benefits of the SARDA concept, differences between the fast-time and HITL simulations made it difficult to exactly compare the results.The differences were due to time and SOSS modeling limitations that are being resolved.It is desirable to fix these issues in SOSS so that it can be used to analyze SARDA against more traffic scenarios and changes than would be possible with HITL simulation alone.The delay optimal fast-time simulation had 8% less taxi time and 40% fewer stops relative to a simulation of a first-come-first-served scheduler managing traffic.The general trend of reduced taxi times and number of stops was also observed in the heavy traffic scenario of the 2010 SARDA HITL simulation.This is despite the differences between the HITL and fast-time simulations.In general, the reduced taxi times and fewer stops are due to the delayoptimal scheduler holding flights at the gate or spot longer.This reduces traffic in the ramp and taxiways and shortens the length of the departure queue.The average delay of the delay-optimal simulation was not less than that of the first-come-first-served simulation.This is in contrast to stand-alone analysis of delay-optimal schedulers, [6][7][8][9][10][11][12] which showed that they reduce delays and increase throughput relative to simple algorithmic models of human controllers.Equity of flight delays was also studied.It is not equitable for a few flights to be excessively delayed so that the other flights can enjoy less or no delays.The largest flight delay in the delay-optimal simulation was one and a half minute longer than the largest flight delay in the first-come-first-served simulation.The results of this paper need to be strengthened by more simulations with a variety of traffic scenarios.In addition, changes to the fast-time simulation such as including runway crossings in the delay-optimal departure schedule could reduce the average flight delay.SARDA = Spot and Runway Departure Advisor SOSS = Surface Operations Simulator and Scheduler SRP = Spot Release Planner TFS = Tactical Flight Separation +Figure 11illustrates how the pieces of the fast-time simulation fit together.SOSS simulates aircraft movement on the airport surface.For each flight in the schedule, it sends the flight's scheduled pushback time (PTime) or wheelson time, estimated unimpeded time of arrival (ETA) at the runway threshold or gate, and taxi route to the scheduler.Although in the 2010 SARDA concept the scheduler contained two programs: the runway scheduler (RS) and the spot release planner (SRP), in this research the scheduler only contained one program, a gate release planner (GRP).The GRP calculates gate release clearances for each flight based on its departure clearance time.The tactical flight separation (TFS) model simulates actions that pilots and controllers take to keep aircraft safely separated.It is embedded within the SOSS simulation. +Figure 1 .1Figure 1.System Diagram A. Surface Operations Simulator and Scheduler SOSS is a fast-time simulation of aircraft movement on the airport surface.It models flight departures and arrivals moving along airport runways, taxiways, and ramps.SOSS connects with schedulers via NASA's common algorithm interface, which is the same interface used by all of NASA's airport simulation tools.This feature makes SOSS able to easily connect with any of the schedulers that NASA has developed because they also use the common algorithm interface.The common algorithm interface passes for each flight within the current time horizon Ptimes or wheels-on times, routes, and ETAs to the scheduler.The scheduler uses this information to create a new schedule, then passes the schedule back to SOSS in the form of a set of clearance times. +Figure 2 .2Figure 2. Node/Link Model of East Side of Dallas/Fort Worth International Airport2.Aircraft Dynamic ModelThe aircraft dynamic model determines the movement of the aircraft through the node/link model.A mathematical model defines the aircraft dynamics and a control model defines how each aircraft attempts to achieve its target speed.The aircraft dynamic model uses one-degree-of-freedom kinematic equations.The equations are +Figure 3 .Figure 4 .34Figure 3. Runway 17R Operation Rates in Number of Departures and Crossings per Five-Minute Bin +waiting at the runway entrance.Departures that wait in line for less time have to stop and go less often. +Figure 5 .5Figure 5. Histogram of Departure Delays for FCFS and GRP Simulations +Table 1 . Comparison Between 2010 SARDA HITL and GRP1Table 1 compares the 2010 HITL simulation and the GRP simulation.Simulation CharacteristicDifference/Similarity2010 SARDA HITLGRPairportsimilarityDFW east-side south flowDFW east-side south flowtraffic scenariosimilarity2008 normal (also had2008 normal onlyheavy)delay-optimal schedulersimilaritydelay-optimal based on DPdelay-optimal based on DP(used in both RS and SRP)(used in GRP)frequency of scheduler calldifference12 seconds5 minutesclearancesdifferencespotgatesequencing at runwaydifferencedelay-optimalfirst-available-first-servedrunway queuesdifferencefull, inner, outerfullcrossingsdifferencemultiple crossingssingle crossingsweight classesdifferencesmall, large, heavyall largedeparture fix constraintdifferencenone80 seconds +Table 2 . Node Color Key2ColorMeaninglight bluedeparture queuedark bluetaxiway, gate, or holdredrunway crossinggreenrampyellowspotmagentarunway entranceorangerunway exit +Table 3 . Aircraft Dynamic Model Parameters for a Boeing 737 Jet3ParameterValuea max a min1.85 ft/sec 2 -1.85 ft/sec 2v queue10 ktsv taxi15 ktsv ramp10 kts +Table 4 . Separation in seconds between consecutive departures4SL H B75S45 68 82 82L45 45 68 68H45 45 82 68B75 45 45 82 68 +Table 5 . Departure Delays, Stops, and Taxi Times5Delay (sec)FCFS / GRP + + + + +AcknowledgmentsThe author would like to acknowledge Zhifan Zhu for his dedicated assistance with debugging and executing SOSS and Justin Montoya for explaining how the runway scheduler works. + + + +Note that in Tables 5 and6, flights are not necessarily consistent across the maximum and minimum rows.For example in Table 6, the flight with 36 stops was not the same as the flight with a taxi time of 870 seconds.Since SOSS categorizes the airport nodes and links into gate, ramp, spot, taxi, queue, and departure, the metrics can be broken down into these categories.Table 7 shows the departure metrics broken out by airport area.Taxi time is not applicable for the gate because it is defined as time from gate release to take-off.In the FCFS simulation, there is little delay in the gate, ramp, and spot sections of the airport.In the GRP simulation however, there is more delay at the gate because flights were strategically held at the gate by the scheduler.These delays at the gate lead to less delay in the taxi, queue, and departure areas.This shows that the queue of flights waiting to use 17R was smaller in the GRP simulation.A similar trend is seen in the number of stops data.Finally, the FCFS taxi times were higher in the queue area than the GRP taxi times.This again was caused in part by the GRP queue length being shorter.Although the length of the departure queue was not calculated in the simulations, it was visible in the simulation display.By visual inspection, the peak queue length in the GRP simulation was about five flights and in the FCFS simulation was about seven flights.In addition to efficiency, equity is also important.Greater throughput should not be achieved by excessively delaying a few flights.The equity of each simulation is illustrated by a histogram of the flight delays.If the tail of the histogram is thin and long, a small number of flights were excessively delayed.Figure 5 shows a delay histogram for the simulations.The GRP simulation has a longer tail.According to Table 5, the worst delay in the GRP simulation was 428 seconds, and the worst delay in the FCFS simulation was 343 seconds.The difference between the two is about one and half minutes. + + + + + + + Performance Evaluation of SARDA: An Individual Aircraft-based Advisory Concept for Surface Management + + YoonJung + + + TyHoang + + + MiwaHayashi + + + WaqarMalik + + + LeonardTobias + + + GautamGupta + + 10.2514/atcq.22.3.195 + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 22 + 3 + + 2011 + American Institute of Aeronautics and Astronautics (AIAA) + Berlin, Germany + + + Ninth USA/Europe Air Traffic Management Research and Development Seminar + + + Jung, Y., Hoang, T., Montoya, J., Gupta, G., Malik, W., Tobias, L., and Wang, H., "Performance Evaluation of a Surface Traffic Management Tool for Dallas/Fort Worth International Airport," Ninth USA/Europe Air Traffic Management Research and Development Seminar, 92, Berlin, Germany, 2011. + + + + + An Airspace Concept Evaluation System Characterization of National Airspace System Delay + + ShannonZelinski + + + TomRomer + + 10.2514/6.2004-6200 + AIAA 2004-6200 + + + AIAA 4th Aviation Technology, Integration and Operations (ATIO) Forum + Chicago, IL + + American Institute of Aeronautics and Astronautics + 2004 + + + Zelinski, S., and Romer, T., "An Airspace Concept Evaluation System Characterization of National Airspace System Delay," 4 th AIAA Aviation, Technology, Integration, and Operations Conference, AIAA 2004-6200, Chicago, IL, 2004. + + + + + Demonstration of reduced airport congestion through pushback rate control + + IoannisSimaiakis + + + HarshadKhadilkar + + + HamsaBalakrishnan + + + TomGReynolds + + + RJohnHansman + + 10.1016/j.tra.2014.05.014 + + + Transportation Research Part A: Policy and Practice + Transportation Research Part A: Policy and Practice + 0965-8564 + + 66 + + 2011 + Elsevier BV + Berlin, Germany + + + Ninth USA/Europe Air Traffic Management Research and Development Seminar, + + + Simaiakis, I., Khadilkar, H., Balakrishnan, H., Reynolds, T., Hansman, J., Reilly, B., and Urlass, S., "Demonstration of Reduced Airport Congestion Through Pushback Rate Control," Ninth USA/Europe Air Traffic Management Research and Development Seminar, , Berlin, Germany, 2011. + + + + + Analysis of Airport Performance using Surface Surveillance Data: A Case Study of BOS + + HarshadKhadilkar + + + HamsaBalakrishnan + + + BrendanReilly + + 10.2514/6.2011-6986 + AIAA 2011-6986 + + + 11th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Virginia Beach, VA + + American Institute of Aeronautics and Astronautics + 2011 + + + Khadilkar, H., and Balakrishnan, H., "Analysis of Airport Performance using Surface Surveillance Data: A Case Study of BOS," 11 th AIAA Aviation, Technology, Integration, and Operations Conference, AIAA 2011-6986, Virginia Beach, VA, 2011. + + + + + Ration by Schedule for airport arrival and departure planning and scheduling + + ChrisBrinton + + + StephenAtkins + + + LaraCook + + + StevenLent + + + TomPrevost + + 10.1109/icnsurv.2010.5503239 + + + 2010 Integrated Communications, Navigation, and Surveillance Conference Proceedings + Herndon, VA + + IEEE + 2010 + + + Brinton, C., Atkins, S., Cook, L., Lent, S., and Prevost, T., "Ration by Schedule for Airport Arrival and Departure Planning and Scheduling," 2010 Integrated Communications Navigation and Surveillance Conference, Herndon, VA, 2010. + + + + + A Mixed Integer Linear Program for Airport Departure Scheduling + + GautamGupta + + + WaqarMalik + + + YoonJung + + 10.2514/6.2009-6933 + AIAA 2010-7692 + + + 9th AIAA Aviation Technology, Integration, and Operations Conference (ATIO) + Hilton Head, SC + + American Institute of Aeronautics and Astronautics + 2009 + + + Gupta, G., Malik, W., and Jung, Y., "A Mixed Linear Program for the Airport Departure Scheduling Problem, "AIAA Aviation, Technology, Integration, and Operations Conference, AIAA 2010-7692, Hilton Head, SC, 2009. + + + + + Incorporating Active Runway Crossings in Airport Departure Scheduling + + GautamGupta + + + WaqarMalik + + + YoonJung + + 10.2514/6.2010-7695 + AIAA 2010-7695 + + + AIAA Guidance, Navigation, and Control Conference + Toronto, Canada + + American Institute of Aeronautics and Astronautics + 2010 + + + Gupta, G., Malik, W., and Jung Y., "Incorporating Active Runway Crossings in Airport Departure Scheduling," AIAA Guidance, Navigation, and Control Conference, AIAA 2010-7695, Toronto, Canada, 2010. + + + + + Managing departure aircraft release for efficient airport surface operations + + WaqarMalik + + + GautamGupta + + + YoonJung + + 10.2514/6.2010-7696 + AIAA 2010-7696 + + + AIAA Guidance, Navigation, and Control Conference + Toronto, Canada + + American Institute of Aeronautics and Astronautics + 2010 + + + Malik, W., Gupta, G., and Jung, Y., "Managing Departure Aircraft Release for Efficient Airport Surface Operations," AIAA Guidance, Navigation, and Control Conference, AIAA 2010-7696, Toronto, Canada, 2010. + + + + + A Mixed Integer Linear Program for Solving a Multiple Route Taxi Scheduling Problem + + JustinMontoya + + + ZacharyWood + + + SivakumarRathinam + + + WaqarMalik + + 10.2514/6.2010-7692 + AIAA 2010-7692 + + + AIAA Guidance, Navigation, and Control Conference + Toronto, Canada + + American Institute of Aeronautics and Astronautics + 2010 + + + Montoya, J., Zachary, W., Rathinam, S., and Malik, W., "A Mixed Integer Linear Program for Solving a Multiple Route Taxi Scheduling Problem," AIAA Guidance, Navigation, and Control Conference, AIAA 2010-7692, Toronto, Canada, 2010. + + + + + Effect of Uncertainty on Deterministic Runway Scheduling + + GautamGupta + + + WaqarMalik + + + YoonJung + + 10.2514/6.2011-6924 + AIAA 2011-6924 + + + 11th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Virginia Beach, VA + + American Institute of Aeronautics and Astronautics + 2011 + + + Gupta, G., Malik, W., and Jung, Y., "Effect of Uncertainty on Deterministic Runway Scheduling," AIAA Aviation Technology, Integration, and Operations Conference, AIAA 2011-6924, Virginia Beach, VA, 2011. + + + + + A Generalized Dynamic Programming Approach for a Departure Scheduling Problem + + SivakumarRathinam + + + ZacharyWood + + + BanavarSridhar + + + YoonJung + + 10.2514/6.2009-6250 + AIAA 2009-6250 + + + AIAA Guidance, Navigation, and Control Conference + Chicago, IL + + American Institute of Aeronautics and Astronautics + 2009 + 12 + + + Rathinam, S., Wood, Z., Sridhar, B., and Jung, Y., "A Generalized Dynamic Programming Approach for a Departure Scheduling Program," AIAA Guidance, Navigation, and Control Conference, AIAA 2009-6250 Chicago, IL, 2009. 12 + + + + + Runway Scheduling Using Generalized Dynamic Programming + + JustinMontoya + + + ZacharyWood + + + SivakumarRathinam + + 10.2514/6.2011-6380 + AIAA 2011-6380 + + + AIAA Guidance, Navigation, and Control Conference + Portland, OR + + American Institute of Aeronautics and Astronautics + 2011 + + + Montoya, J., "Runway Scheduling Using Generalized Dynamic Programming," AIAA Guidance, Navigation, and Control Conference, AIAA 2011-6380, Portland, OR, 2011. + + + + + A Concept and Implementation of Optimized Operations of Airport Surface Traffic + + YoonJung + + + TyHoang + + + JustinMontoya + + + GautamGupta + + + WaqarMalik + + + LeonardTobias + + 10.2514/6.2010-9213 + AIAA 2010- 9213 + + + 10th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Fort Worth, TX + + American Institute of Aeronautics and Astronautics + 2010 + + + Jung, Y., Hoang, T., Montoya, J., Gupta, G., Malik, W., and Tobias, L., "A Concept and Implementation of Optimized Operations of Airport Surface Traffic," 10 th AIAA Aviation, Technology, Integration, and Operations Conference, AIAA 2010- 9213 Fort Worth, TX, 2010. + + + + + Tower Controllers' Assessment of the Spot and Runway Departure Advisor (SARDA) Concept + + THoang + + + YJung + + + JHolbrook + + + WMalik + + + + Europe Air Traffic Management Research and Development Seminar + + 34 + 2011 + Berlin, Germany + + + Ninth USA + Hoang, T., Jung, Y., Holbrook, J., and Malik, W., "Tower Controllers' Assessment of the Spot and Runway Departure Advisor (SARDA) Concept," Ninth USA/Europe Air Traffic Management Research and Development Seminar, 34, Berlin, Germany, 2011. + + + + + A Simulator for Modeling Aircraft Surface Operations at Airports + + ZacharyWood + + + SivakumarRathinam + + + YoonJung + + + MatthewKistler + + 10.2514/6.2009-5912 + AIAA 2009-5912 + + + AIAA Modeling and Simulation Technologies Conference + Chicago IL + + American Institute of Aeronautics and Astronautics + 2009 + + + Wood, Z., Kistler, M., Rathinam, S., and Jung, Y., "A Simulator for Modeling Aircraft Surface Operations at Airports," AIAA Modeling and Simulation Technologies Conference, AIAA 2009-5912, Chicago IL, 2009. + + + + + + diff --git a/file783.txt b/file783.txt new file mode 100644 index 0000000000000000000000000000000000000000..d42fd2945c2c6c0c376b0e64d2b3bf248a9516ae --- /dev/null +++ b/file783.txt @@ -0,0 +1,599 @@ + + + + +I. IntroductionASA is researching and developing automation that provides advisories to ramp, ground, and local controllers, who manage air traffic on the airport surface. 1 The advisories tell the controller when to clear or hold flights at key airport surface locations, such as gates, spots, and runway crossings and thresholds.The advisories are designed to minimize taxi time and sequence departure and runway crossing operations to maximize throughput.The purpose of the automation is to improve the efficiency of surface operations without compromising safety and to possibly reduce controller workload.The scheduler is a primary part of the airport surface advisory automation.It is the mathematical algorithm that calculates the advisories that will be displayed to the controller.Many different formulations for schedulers are being studied.][4][5][6][7][8][9][10] Any scheduler that is selected for actual deployment should demonstrate in simulation that the advisories it provides meet the goals of the automation, i.e., improved traffic efficiency and no increase in controller workload.NASA tests schedulers and surface advisory automation in both fast-and real-time (human-in-the-loop) simulations.Relative to real-time simulation, fast-time simulation can process many scenarios and uncertainty cases for a modest cost of software development and staffing.Real-time simulation includes hands on interactions between human operator and automation and provides broad and crucial data about those interactions. 11,124][15] The purpose of SOSS is to develop and test schedulers.Schedulers that show promise are selected for further testing in real-time simulation.The development and testing process in SOSS must accurately predict the benefits that the scheduler and accompanying automation will produce in real operations.To this end, surface operations simulated in SOSS need to be realistic.Validation is the process of measuring the differences between simulation and reality. 16If the differences are too large, the benefits predicted by the simulation lack credibility.Typically, when the differences are large a simulation is tuned.This is the process of reducing the average differences by adjusting parameters of the simulation.Another option, when differences are large, is to forgo attempting to calculate absolute benefits the system would produce in the real world, and only calculate relative benefits between two simulations.This paper presents a validation of a SOSS simulation of airport surface traffic at Charlotte Douglas International Airport (CLT) during peak traffic hours.The traffic scenario is derived by extracting real flight schedules from recorded surveillance and FAA operations and performance data.A simple first-come-first-served scheduler loosely models control actions taken by controllers and pilots.SOSS simulates the traffic scenario to produce several validation metrics, taxi times and runway throughputs.The simulated metric values are compared with metric values calculated from the recorded field data to measure how well SOSS was able to simulate the operations that actually occurred.Two SOSS parameters, taxi speed and pushback duration times, are used to tune the simulation so that the simulated average taxi times match those calculated from the recorded field data.This paper begins by introducing the Surface Operations Simulator and Scheduler.Then, it describes the validation process, the data used to drive the process, and the metrics used to do the comparison.Finally, results of the validation are presented. +II. Introduction to the Surface Operations Simulator and SchedulerThe Surface Operations Simulator and Scheduler (SOSS) models operations on the airport surface.These operations include flight readiness, pushback procedure, taxi, takeoff, and landing.First the airport operations are described and then the SOSS models are presented. +A. Airport OperationsFlight readiness is the process of getting flights ready for pushback.Typically flights are not ready to pushback exactly at their scheduled departure time because the times required to load passengers, baggage, food, and fuel and for the crew to complete their safety check lists are uncertain.Results from Ref. 17 showed that from May to June 2012 at Dallas Fort Worth International Airport large airline flights pushed back on average 52 seconds early with a standard deviation of 148 seconds.Although SOSS has the ability to model the timing uncertainty of this process, it was not used in this study.The uncertainty in the actual pushback times did not have to be modeled because the actual pushback times for each flight in the scenario were available in the field data.The pushback procedure starts when a flight begins to pushback from the gate and ends when the pilot is given clearance from ramp control to begin taxiing.During this procedure several things are happening.The aircraft is being moved by the tug from the gate to the drop off point (which is usually close to the ramp taxiway centerline), and the pilot is completing his pushback procedure which includes spooling up one or more engines.It is uncertain how much time these procedures take.In Ref. 16 the pushback procedure duration averaged 202 seconds with a 189 second standard deviation.Moreover, there were outlier flights with pushback durations as little as 77 seconds and as large as 7 minutes.The taxi procedure moves flights from the area near the gates and terminals, called the ramp, out to the runways.While the flight is taxing, it crosses a location on the airport surface called the spot.The spot is the dividing point between the non-movement area and the aircraft movement area.The non-movement area includes the ramp and is controlled by the airline ramp control tower, while the aircraft movement area includes the taxiways and runways and is controlled by the FAA Air Traffic Control Tower (ATCT).Table 1 and Fig. 1 illustrate the areas covered by the non-movement and aircraft movement area and the spots.Control responsibility for flights is handed off between ramp controllers and air traffic controllers at the spot.The takeoff and landing procedures prevent runway incursions by ensuring that flights safely enter and exit runways and consecutive departure or arrival operations follow wake-vortex and departure fix separation constraints.If there is more than one crossing point on the runway, more than one flight may cross the runway at a time.However when flights are crossing the runway, departures and arrivals are not permitted.Generally, the runway limits the operations rate of an airport because only one departure or arrival or set of crossing flights can use the runway at a time. +B. SOSS ModelsThe SOSS airport surface model is a node/link network representing gates, ramps, spots, taxiways, crossings, and runways.Figure 1 illustrates the network for Charlotte Douglas International Airport (CLT).Table 1 contains a key that matches the node/link colors with the areas of the airport that they represent.The runway queue is at the entrance to the runway where departures waiting to takeoff line up.The runway names for south flow configuration and the direction of takeoffs and landings are denoted in Fig. 1.Arrivals use runways 18R, 18C, and 23, while departures use 18C and 18L.18C is used for arrivals and departures.Flight surface routes in SOSS are defined as ordered lists of nodes through the node/link network.Each departure possesses a route that takes it from its gate to its runway entrance, and each arrival has a route that takes it from its runway exit to its gate.Routes cross through one and only one spot.SOSS models the pushback procedure for departures along their first link.This link is between the gate node (first node) and first ramp node (second node) in the route.Generally, a flight's traversal time for a link is the length of the link divided by the speed of the flight, which is defined as the kinematic duration.In this study, the kinematic duration was not used as the traversal time for the first link for departures.Instead the user defined pushback time duration was used as the traversal time.The pushback time duration was one of the parameters used to tune SOSS as described later in the paper.A SOSS simulation is initialized with a traffic scenario.The traffic scenario contains a list of flights scheduled to depart or arrive during the simulation.Specific information about each flight in the list is required to build the traffic scenario.Call sign, aircraft type, gate number, runway number are needed for all flights.Additionally, scheduled This material is declared a work of the U.S. Government and is not subject to copyright protection in the United States.gate pushback times are needed for departures, and actual wheels-on times are needed for arrivals.The aircraft movement model uses kinematic equations of motion to move flights along links from one node to the next.Each type of aircraft has specified acceleration and deceleration values and set of target speeds.The model attempts to move the aircraft at the target speed.If the aircraft is stopped, the model accelerates it up to the target speed and then maintains its speed.If the aircraft speed is higher than the target speed the model decelerates the aircraft until it slows to the target speed.The target speed for each aircraft is selected based on the type of aircraft and the location of the aircraft on the airport surface.If the aircraft is located in the ramp area or runway queuing area the target speed is slower than if the aircraft is located on a taxiway.User parameters can be used to increase or decrease the target speed.The parameters for changing ramp speeds and taxi speeds were used to tune SOSS as described later in the paper.The process of accelerating or decelerating the aircraft to its target speed is interrupted when other traffic impedes the aircraft.SOSS takes conflict detection and resolution actions to keep aircraft properly separated on the node link network.This is accomplished by predicting when two flights are going to conflict and slowing down or stopping one of the flights.There are several type of conflicts that may occur.Tail-on conflicts occur when a trailing flight overtakes a flight traveling in the same direction on the same link.Intersection conflicts occur when multiple flights simultaneously arrive at intersection nodes, which are nodes connected to more than one link.Finally, headon conflicts occur when two flights are traveling on the same link in opposite directions.Head-on and intersection conflicts create the opportunity for gridlock which occurs when a pair or group of flights enter into a conflict and there is no way of resolving it.Flight separations at the runway are determined by navigational safety and wake-vortex spacing constraints.In practice, controllers enforce a separation using a distance-based rule.SOSS enforces a separation by holding a flight at the entrance of the runway until a specified amount of time from the previous operation has elapsed.The time is calculated to achieve the correct distance between operations.This material is declared a work of the U.S. Government and is not subject to copyright protection in the United States.Reference 15 describes the approach used to calculate the separation times for runway operations at CLT. Tables 23456show the separation times in seconds.The weight classes of the aircraft are denoted by small (S), large (L), heavy (H), and B757.The B757 has its own weight class because its weight is in the large class, but it creates more turbulence than other large aircraft.Columns specify the weight class of the leader, and rows specify the weight class of the follower.For example according to Table 2, the separation time between a leading H and a trailing L was 120 seconds.More than one table is needed to specify the separations because CLT has both mixed use runways and intersecting (crossing) runways.For example, 18C accommodates both arrivals and departures so Tables 3 and4 are required to specify the different separation requirements depending on whether an arrival proceeds or trails a departure.Similarly because 18L crosses 23, operations on 18L depend on operations on 23.Tables 5 and6 define the separation requirements for departures on18L depending on if they proceed or trail an arrival on 23.SOSS connects to schedulers through a socket and uses a protocol called the Common Algorithm Interface (CAI) to communicate with them.NASA's real-time simulation facilities also use the CAI to communicate with schedulers.Because both SOSS and the real-time simulation use the CAI, a scheduler that has passed development and testing using SOSS can be easily integrated into the real-time simulation environment.Figure 2 shows the SOSS system design.The arrows between the SOSS and Scheduler boxes depict scheduler calls that are made through the CAI.The user sets up the frequency and timing of the scheduler calls.For each call, SOSS sends to the Scheduler the current location, speed, route, and estimated times of arrival at key nodes in the airport node/link network of each flight operating on the surface.The Scheduler uses this information to calculate scheduled times of release for each flight at specific nodes in the airport node/link network.In a real-time simulation or in the field a scheduled time of release would be an advisory that would be displayed to a controller.The scheduled times of release are sent from the Scheduler to SOSS.SOSS controls each flight so that it does not leave a node before its scheduled time of release.If a flight arrives at a node before its scheduled time of release, SOSS holds the flight at the node until its release time.If a flight arrives at a node after its scheduled time of release, SOSS allows the flight to continue on its route without stopping.Not all nodes have scheduled times of release; only the nodes selected by the Scheduler have them.This is similar to actual surface operations where flights are only controlled at specific locations on the airport.In this study, SOSS used a very simple first-come-first-served scheduler. +III. ValidationThe validation process determines if the SOSS simulations are an accurate approximation of operations in the real world.It is accomplished by comparing metrics produced by SOSS simulations with those calculated from recorded surveillance and operational performance data.After the initial simulation, the simulated metrics did not compare well with actual metrics.SOSS was tuned to make the metrics compare better. +A. Comparison of Simulated Metrics to Actual MetricsFigure 3 depicts the process for measuring the difference between a SOSS simulation and real operations.The validation process begins with recorded surveillance and FAA operational performance data.These data are fed along two different paths depicted in the Fig. 3. Along the top path, a SOSS scenario file is created from the surveillance data.Then, the scenario file is used to initialize a SOSS simulation.Simulated flight tracks generated from the SOSS simulation are fed into the metric generator.The metric generator calculates simulated taxi times for each flight and simulated operation rate histories for each runway.Along the bottom path in Fig. 3, surveillance data and FAA operational performance data are fed into directly into a metrics generator.The metrics generator calculates actual taxi times for each flight and operation rate histories for each runway.The simulated metric values are compared with the actual metric values. +Model TuningAfter the initial simulation, the simulated taxi times did not compare well with the actual taxi times.The goal of the model tuning was to bring the simulated taxi times closer to the actual taxi times.The model tuning was accomplished by executing a series of SOSS simulations.For each simulation, SOSS input parameters were varied to move the simulated taxi times closer to the actual taxi times.The definition of the taxi time metric used in the model tuning is described in the metrics section.The SOSS input parameters used in the model tuning were the pushback duration time and the parameters that increase and decrease the ramp and taxi target speeds. +IV. DataNo single data set contained all of the information required to perform the validation.The three required data sources were Airport Surface Detection Equipment, Model X (ASDE-X), FlightStats, and the Airline Service Quality Performance (ASQP) system.This section describes each data source and the information that was used from it.Table 7 summarizes the data contained within the different sources.In addition, it lists the number of flights that were extracted from each source for the date and time period that was selected for the validation.Since three data sources were used, flights had to be cross referenced between sources.Call sign alone was not sufficient for cross referencing because call signs are not mutually exclusive between flights, i.e. two different flights may use the same call sign.In this study, cross referencing across data sources was accomplished by matching call sign, origin, and destination. +A. Airport Surface Detection Equipment, Model X (ASDE-X)Flights on the airport surface at CLT are tracked by multiple surveillance sources.Multiple sources are required for redundancy, accuracy, and coverage of the airport surface.The ASDE-X system integrates data from the multiple sources into a single stream of data that is used to drive tower and ramp controller displays and compute performance metrics for the FAA and airlines.Despite multiple sources, there are still coverage gaps in some terminal alleys and near some gates.In addition, tracks in the non-movement area are masked on tower air traffic controller displays because tower air traffic controllers are not responsible for flights in the that area.The ASDE-X data used in this study was obtained from ITT Excelis 18, 19 .ITT Excelis provides more data than is available in the raw surveillance data alone.For each flight, they also provide processed data such as the runway assignment, the aircraft type, and the wheels-on or wheels-off time.The ASDE-X data contained tracks for each flight that operated at CLT.In this study, airborne tracks were filtered out of the data so tracks for arrivals generally started with wheels-on and ended just inside the nonmovement area.Tracks for departures generally started around the edge of non-movement area and ended with wheels-off.Most flights had little track coverage in the non-movement area because that data was masked.Figure 4 illustrates tracks for a departure on runway 18C.As discussed later in the paper, the ASDE-X data were used to generate the traffic scenario.For most flights in the scenario, derived data from ITT Exelis, as opposed to track data, provided all of the necessary information.However some flights in the scenario were either missing data or their data was inaccurate.The data may have been the flight's runway assignment or gate-in, gate-out, wheels-on, or wheels-off time.For these flights, the track data would be used to visually estimate the missing data from a plot of the track data on a map of CLT similar to Fig. 4.ASDE-X contained 192 flights that operated during time period chosen for this study.Since ASDE-X provides surveillance data from the airport, it was assumed that 192 was the actual number of flights that operated during the time period.However, there is a possibility that there were some flights that operated in reality, but were filtered or dropped out of ASDE-X.The missing flights, if there were any, would have been military or other special flights. +B. FlightStatsThe FlightStats website 20 contains information about the on-time performance of flights.FlightStats gets its information from government, airline, airport, and reservation systems.In addition to on-time performance, FlightStats provides the departure and arrival gate that a flight used.The gate information from FlightStats was used to populate the gate assignment in the SOSS traffic scenario file.The FlightStats data set contained 181 flights that operated during the date and time period selected for this study. +C. Airline Service Quality Performance (ASQP)The Airline Service Quality Performance system 21 (ASQP) gives airline-reported on-time performance, flight delay, and cancellation statistics.It is driven by reports submitted by the airlines in accordance with Department of Transportation regulations.Airlines with one percent or more of total domestic scheduled passenger revenues must report on flights operating in any airport in the 48 contiguous states.In addition to those airlines that are required, some airlines voluntarily report.At the time of the writing of this paper, there were 14 airlines reporting.Included in the ASQP data is the so called Out-On-Off-In (OOOI) data, which refers to gate-out, wheels-off, wheels-on, and gate-in times.Typically, sensors onboard the airplanes capture the times that these events occurred and automatically send them via the Aircraft Communication Addressing and Reporting System (ACARS) to the airline database system.OOOI data was used to calculate actual taxi times as described in the Metrics Section.The ASQP data set contained 107 flights that operated during the date and time period selected for this study.The 85 flights that were in the ASDE-X data set, but not in the ASQP data set, were flights not operated by one of the 14 ASQP airlines. +V. Traffic ScenarioThe traffic scenario contains the list of flights that are scheduled to operate during the simulation.It is one of the input files that drives a SOSS simulation.A traffic scenario has a date and a time period which define when the simulation starts and ends.Only flights which operated between the start and end times on the date are included in the traffic scenario. +A. Building the Traffic ScenarioWhen building a traffic scenario for a validation study, a primary goal is to identify and include all of the flights that actually operated during the selected time period.If some flights are missed and not included in the traffic scenario, the traffic densities in the simulation will be lower than they were in reality, and this will distort the simulation metrics.A favorable feature of the ASDE-X data set was that it captured possibly (with the exception noted in the Data Section) all of the flights that operated at CLT during the selected time period.This data source was used to verify and identify the flights and include them in the traffic scenario.Multiple pieces of information were needed for each flight entry in the traffic scenario.These data were call sign, aircraft type, gate assignment, runway assignment, and start time.Start time for departures is the gate pushback time, and start time for arrivals is the wheels-on time.Since all of these were not included in a single data source, the data had to be extracted from both the ASDE-X and FlightStats data sets.Aircraft type, runway assignment, and wheels-on time were selected from ASDE-X, and gate assignment and gate-out time were selected from FlightStats.Cross referencing between ASDE-X and FlightStats was performed as described in the Data Section.ASDE-X contained 192 flights, and FlightStats contained 181 flights.All of the flights found in FlightStats were able to be cross referenced into the ASDE-X flight set.However, there were 11 flights that were found in ASDE-X but not in FlightStats.The flight tracks in ASDE-X were used to visually estimate the gate assignment and gate-out time for these flights.Fortunately these flights did not have a large affect on traffic in the ramp area around the main terminals because all of these flights were general aviation flights that were using gates in the general aviation terminal, which is on a separate side of the airport from the main airline terminals.These flights did not interact with the main traffic until they reached the runway. +B. Selecting the Date and Time PeriodThe date and time period for the validation were selected based on multiple factors, including weather, airport runway configuration, and traffic density.Because this study was an initial validation of the SOSS CLT model, it was desired to simulate conditions at CLT that occurred often and during good weather.The south flow configuration (see Section II.B) was selected because (i) it is used slightly more often than the north flow configuration, (ii) it uses all four runways (as opposed to the north flow configuration that uses only 3 runways), and (iii) it has slightly higher arrival and departure rates than the north flow configuration.It was desired to simulate current or at least recent traffic conditions, thus the date of the traffic scenario needed be close to the current date.This study started in March 2013.At the time, ASQP had not yet published data for March and February, so January was the closest month for which data was available from all three data sources.January 23 was a clear weather day at CLT.The airport was in south flow configuration for the majority of the day.The only period the airport was in the north flow configuration was during a few hours in the early morning.The time period was selected to capture a peak traffic density period during the day.Figure 5 shows the airport operations rates on January 23.The number above a peak denotes the number of operations (both arrivals and departures) that occurred in that peak.The blue box denotes the time period that was selected.This period was selected because it was during a late morning rush with the highest departure peak of the day, 26 departures per 15minutes.In addition, this period contained the peak with the second highest number of operations, 193 operations. +VI. MetricsTwo metrics were selected for doing the comparison between simulation and actual operations.These metrics were runway throughput and taxi time.This section describes how these metrics are calculated. +A. Runway ThroughputRunway throughput is the time history of the number of operations that occurred during a 15-minute sliding time bin.Two throughputs are calculated, arrival and departure.Arrival throughput is generated by counting the number of wheels-on times for arrivals within a 15-minute sliding time bin.Departure throughput is generated by counting the number of wheels-off times for departure within a 15-minute sliding time bin.The simulated and actual departure throughput histories should be as close as possible.When they are close, it verifies that the average separations in the separation matrices (Tables 23456) were a good model of the actual departure runway operations. +B. Taxi TimeTaxi time is the time required for a departure to transit from the gate to the wheels-off point or for an arrival to transit from the wheels-on point to the gate.The taxi time does not include time spent waiting at the gate.Reducing taxi time is desirable because during taxi engines are on, fuel is burned, and emissions are released.Taxi time for a departure is calculated as taxi time departure = wheels-off time -gate-out time,and taxi time for an arrival is calculated as taxi time arrival = gate-in time -wheels-on time.(For the validation study, the focus is on the difference between simulation and reality.Hence, the taxi time error for a individual flight is defined by the following equation: error = simulated taxi time -actual taxi time,where actual taxi time actual means the taxi time calculated from ASQP.To understand the size of the error relative to the size of the actual taxi time, the taxi time percent error is calculated as taxi time percent error = (error / actual taxi time)*100.The taxi time percent error measures the percent difference between the simulated taxi times and the actual taxi times.A taxi time error and percent error is calculated for each flight in ASQP, 107 flights out of a total of 192 flights.The taxi time errors and percent errors were averaged across the 107 flights and standard deviations were calculated.The taxi time percent error was the metric that was driven to zero for the model tuning.Taxi times for some of the 85 other flights may have been able to be derived from the FlightStats and ASDE-X data sets, but for this study that was not attempted.This is because the data for the non-ASQP airline flights were found to be more likely to have errors than the data for the ASQP airline flights.Identifying and fixing flights with inaccurate data was very time consuming because it required visually checking ASDE-X flight tracks. +VII. ResultsResults are presented for the model tuning, runway throughput metrics, and taxi time metrics. +A. Model TuningThe SOSS model was tuned to make the difference between simulated and actual taxi times zero.The metric that was selected for this process was the taxi time percent error as described in the Metrics Section.The taxi time percent error metric was averaged across both the arrivals and the departures.The tuning process drove both the This material is declared a work of the U.S. Government and is not subject to copyright protection in the United States.average arrival and departure taxi time percent errors to zero.The SOSS parameters used to achieve the model tuning were the pushback duration, the ramp target speed, and the taxi target speed.The initial and final values of the tuning parameters and the error metrics are listed in Table 8.For average departure and arrival errors, the first value is average error, Eq. ( 3), and the value in parenthesis is average percent error, Eq. ( 4).Inspection of the final values of these errors shows that even though the tuning process drove the average percent errors to zero, the final values of the average errors were not zero.In an exercise not reported in Table 8, the simulation was tuned to drive the average errors to zero.This produced very similar values of the tuning parameters.The negative initial value for average departure error in Table 8 meant that initially for departures actual taxi times were on average larger than simulated taxi times.The final pushback duration value adjusted for this by making the simulated pushback durations longer, 5 minutes 46 seconds versus 3 minutes 22 seconds.The positive initial value of the average arrival error meant that for arrivals simulated taxi times were on average larger than actual taxi times.The faster final ramp and taxi target speeds adjusted for this by shortening simulated arrival taxi times.All flights in the SOSS simulation had a pushback duration of 5 minutes and 46 seconds.Flights in actual operations had a wide variety of pushback durations.Similarly, in SOSS all flights in the ramp had a target speed of 13.7 knots and all flights on the taxiways had a target speed of 17.4 knots.However, flights in actual operations reached speeds higher or lower than these speeds.These difference illustrate the difficulty in predicting operations on the airport surface.It is not known on a flight-by-flight basis how long a pushback procedure will take, or what speeds an aircraft will attain in the absence of other traffic.This highlights the fact that although the model tuning drove the average departure and arrival taxi time percent errors to zero the standard deviations were not zero. +B. Arrival ThroughputsFigures 6 and7 show the simulated and actual arrival throughputs for runways 18R and 23, respectively.Both runways were used as arrival-only runways during the time period chosen for this study.The simulated and actual throughput curves lie directly on top of each other, which is why only a red curve shows in the figures.The blue curve is underneath.The fact that the curves lie on top of each other suggests that the arrival flights in the traffic scenario were correctly set up.This is to be expected since the wheels-on times for arrivals in the simulation were derived from the actual wheels-on times, as described in the Traffic Scenario Section.and9 show the simulated and actual departure throughputs for runways 18C and 18L, respectively.Here the simulated and actual curves do not lie exactly on top of each other.This is because departure throughputs are generated by counting the number of wheels-off times in a 15-minute bin and the simulated wheels-off times are not identical to the actual wheels-off times.It is desired that simulated wheels-off times be as close as possible to actual wheels-off times, but inaccuracies in the simulation produce differences.As denoted in Fig. 8 at about 3,400 seconds, the simulated departure throughput on runway 18C does not peak at the same level as the actual throughput.This indicates a loss of throughput in the SOSS simulation.The difference between the two peaks is 2 departures per 15 minutes which is 14% of the actual peak value.A similar loss in throughput shows at about 1,800 seconds for 18L in Fig. 9. However at the next peak at 3,300 seconds, the simulated and actual curves achieve the same maximum.The loss of throughput on 18C was investigated further by viewing playback visualizations of the simulation between 2,000 and 3,500 seconds.It was observed that the loss occurred because departure operations on 18C briefly dried out at about 2,800 seconds, which is the time when the loss first shows in Fig. 8. 18C dried out because two departures headed for 18C were caught in a traffic jam in the ramp.The traffic jam caused the two flights to arrive at the entrance to 18C late.Visual inspection of the ASDE-X tracks for these two flights revealed that in actual operations they were able to avoid the traffic jam.Figures 10 and11 show the cumulative departures on runways 18C and 18L, respectively.Because the simulated and actual curves do not differ from each other by more than several operations, it is seen that the loss in throughput noted in Figs. 8 and9 does not have a large affect on the cumulative departures.This material is declared a work of the U.S. Government and is not subject to copyright protection in the United States. +D. Taxi TimesFigure 12 shows histograms of the simulated and actual taxi times.Only the 107 ASQP flights were included in these histograms.The average simulated taxi time is 10 minutes and 29 seconds and the average actual taxi time is 11 minutes and 4 seconds.The average taxi times are not equal after model tuning because average taxi time percent errors were driven to zero, not the average taxi time errors.The simulated taxi time standard deviation is 5 minutes and the actual standard deviation is 5 minutes and 10 seconds.These results show that statistically SOSS is an accurate model of real operations.Figure 13 shows histograms of the arrival and departure taxi time percent errors.Only the 107 ASQP flights were included in these histograms.Included are the standard deviations in taxi time percent error, Eq. ( 4), and taxi time error in parenthesis, Eq. (3).The average taxi time percent errors were set to zero by the tuning process.The numbers in the x labels surrounded by parenthesis are negative.The arrival with 100% error in Fig. 13 is an outlier.Inspection of ASDE-X tracks showed that this flight in reality reached taxi speeds as high as 29 knots in the ramp, which is much larger than the 13.7 knots target speed that was used in simulation. +VIII. Future WorkThere are several next steps that this research could take.In this paper, only one traffic scenario was investigated.This was due to the work required to compile a traffic scenario from the field data.A future study could investigate more traffic scenarios, including scenarios with less traffic.The model tuning should be performed for each new scenario.The values of the tuned parameters for the different parameters would be compared.Ideally, the values of the tuned parameters should be approximately stable across scenarios if the pushback durations and ramp and taxi target speeds were similar for the different scenarios.Another future study would be to analyze the field data and measure the actual pushback procedure durations for each flight.The actual pushback durations could be used within SOSS to model exactly how long each flight took to pushback.This would eliminate the uncertainty in the model due to pushback procedure duration.With this uncertainty eliminated, the uncertainty due to taxi and ramp speeds, traffic interactions, and runway queue dynamics would be isolated and could be analyzed. +IX. ConclusionThis study showed that the Surface Operations Simulator and Scheduler (SOSS) was able to accurately on a statistical basis model real operations on the Charlotte Douglas International Airport surface.Using model tuning, the simulated and actual distribution of taxi times had the very close averages and standard deviations.In addition, the simulated and actual runway departure rates were approximate.The simulated runway departure rate peaked 14% less than the actual departure rate.However, the difference in peak departure rate was small enough and for a short enough period that it did not adversely affect the cumulative number of departures.Despite the success in matching the distributions of simulated and actual taxi times, on an individual flight basis SOSS did not predict well the exact actions of a particular flight.It was difficult to predict the duration of a specific flight's pushback procedure and what speed a pilot used to taxi across the airport surface.Any simulation of airport surface traffic must to contend with the difficulty of predicting these parameters.The pushback duration time and ramp and taxi speeds that were derived from the SOSS model tuning are good estimates of the average pushback durations and taxi and ramp speeds occurring in real operations.The tuned pushback duration of 5 minutes and 46 seconds is higher than the average duration measured in Ref. 17 at Dallas Fort Worth International airport.However, it compares well with unpublished estimates of pushback durations at Charlotte Douglas International Airport (CLT) that the authors have observed in actual operations.In addition, average ramp and taxi target times of 13.7 knots and 17.4 knots align well with observations of the flight speeds observed in the ASDE-X data.However, speeds observed in the ASDE-X data can be much higher, as high as 28 knots for an outlier flight.There were some limitations to this study.The study was accomplished for a specific airport surface and a specific traffic scenario.The results of this study may not hold for other airports or traffic scenarios.The model tuning process may need to be performed again for a different airport or a different traffic scenario.For example, observations of traffic at on the surface of CLT have shown that average taxi speeds change with the level of traffic.During periods of very little traffic, pilots tend to taxi faster, and during periods of heavy traffic, pilots tend to taxi slower.This observed effect would cause the tuned target speeds in SOSS to be different for a traffic scenario with little traffic.Because SOSS can on average realistically and accurately simulate operations on the airport surface, it is a good environment for developing and testing airport surface traffic schedulers.SOSS can be used to measure the anticipated benefits that a scheduler would produce if it were installed as part of an airport surface traffic management system and used at a real airport.ATCT = Air Traffic Control Tower CAI = Common Algorithm Interface CLT = airport code for Charlotte Douglas International Airport FAA = Federal Aviation Administration NASA = National Aeronautics and Space Administration SOSS = Surface Operations Simulator and Scheduler +NDownloaded by NASA AMES RESEARCH CENTER on August 14, 2013 | http://arc.aiaa.org| DOI: 10.2514/6.2013-4207Thismaterial is declared a work of the U.S. Government and is not subject to copyright protection in the United States. +Figure 1 .1Figure 1.Node/link network model of Charlotte Douglas International Airport +Figure 2 .Figure 3 .23Figure 2. Diagram of SOSS System Design +Figure 4 .4Figure 4. Flight Tracks for Departure on 18C in ASDE-X +Figure 5 .5Figure 5. Airport Operation Rates at CLT on January 23, 2013 +Downloaded by NASA AMES RESEARCH CENTER on August 14, 2013 | http://arc.aiaa.org| DOI: 10.2514/6.2013-4207 +Figure 6 .6Figure 6.Arrival Throughput on 18R Figure 7. Arrival Throughput on 23 +Figure 8 .Figure 9 .Figure 10 .8910Figure 8. Departure Throughput 18C Figure 9. Departure Throughput 18L +Figure 12 .Figure 13 .1213Figure 12.Histograms of Simulated and Actual Taxi Times + +Table 1 . Node/Link Color Key Node/Link Color Type of Movement Area Area of Airport blue1Downloaded by NASA AMES RESEARCH CENTER on August 14, 2013 | http://arc.aiaa.org| DOI: 10.2514/6.2013-4207non-movementgatebluenon-movementrampyellowbothspotgreenaircraft movementtaxiwaycyanaircraft movementrunway queuemagentaaircraft movementrunway takeoff pointorangeaircraft movementrunway exit pointredaircraft movementrunway crossing point +Table 2 . Separation Matrix for Consecutive Departures on all Runways (seconds)2SLH B757S60 60 120 120L60 60 120 120H60 60 9090B757 60 60 9090 +Table 3 . Separation Matrix for a Departure After an Arrival on 18C (seconds)3SL H B757S50 50 5050L60 60 6060H70 70 7070B757 60 60 6060 +Table 4 . Separation Matrix for a Departure Before an Arrival on 18C (seconds)4SL H B757S40 40 4040L28 28 2828H24 24 2424B757 28 28 2828 +Table 5 . Separation Matrix for a Departure on 18L After an Intersecting Arrival on 23 (seconds)5S L H B757S7 7 77L5 5 55H4 4 44B757 5 5 55 +Table 6 . Separation Matrix for a Departure on 18L Before an Intersecting Arrival on 23 (seconds)6SL H B757S40 40 4040L40 40 4040H40 40 4040B757 40 40 4040 +Table 7 . Summary of Data Used from Data Sources7Source# of Flights DataASDE-X192call sign, origin, destination, runway assignment, aircraft type, wheels-on timeFlightStats 181call sign, origin, destination, gate assignment, gate-outASPQ107call sign, origin, destination, gate-in, wheels-off time +Table 8 . Values of SOSS Tuning Parameters and Metrics Parameter Initial Value Final Value8Pushback duration3 min. and 22 seconds 17 5 minutes and 46 secondsRamp speed10 knots13.7 knotsTaxi speed15 knots17.4 knotsAverage departure error -2 min. + Downloaded by NASA AMES RESEARCH CENTER on August 14, 2013 | http://arc.aiaa.org| DOI: 10.2514/6.2013-4207Thismaterial is declared a work of the U.S. Government and is not subject to copyright protection in the United States. + + + + + + + + + A Concept and Implementation of Optimized Operations of Airport Surface Traffic + + YoonJung + + + TyHoang + + + JustinMontoya + + + GautamGupta + + + WaqarMalik + + + LeonardTobias + + 10.2514/6.2010-9213 + AIAA 2010- 9213 + + + 10th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Fort Worth, TX + + American Institute of Aeronautics and Astronautics + 2010 + + + Jung, Y., Hoang, T., Montoya, J., Gupta, G., Malik, W., and Tobias, L., "A Concept and Implementation of Optimized Operations of Airport Surface Traffic," 10 th AIAA Aviation, Technology, Integration, and Operations Conference, AIAA 2010- 9213 Fort Worth, TX, 2010. + + + + + A Generalized Dynamic Programming Approach for a Departure Scheduling Problem + + SivakumarRathinam + + + ZacharyWood + + + BanavarSridhar + + + YoonJung + + 10.2514/6.2009-6250 + AIAA 2009-6250 + + + AIAA Guidance, Navigation, and Control Conference + Chicago, IL + + American Institute of Aeronautics and Astronautics + 2009 + + + Rathinam, S., Wood, Z., Sridhar, B., and Jung, Y., "A Generalized Dynamic Programming Approach for a Departure Scheduling Program," AIAA Guidance, Navigation, and Control Conference, AIAA 2009-6250 Chicago, IL, 2009. + + + + + Runway Scheduling Using Generalized Dynamic Programming + + JustinMontoya + + + ZacharyWood + + + SivakumarRathinam + + 10.2514/6.2011-6380 + AIAA 2011-6380 + + + AIAA Guidance, Navigation, and Control Conference + Portland, OR + + American Institute of Aeronautics and Astronautics + 2011 + + + Montoya, J., "Runway Scheduling Using Generalized Dynamic Programming," AIAA Guidance, Navigation, and Control Conference, AIAA 2011-6380, Portland, OR, 2011. + + + + + A Mixed Integer Linear Program for Solving a Multiple Route Taxi Scheduling Problem + + JustinMontoya + + + ZacharyWood + + + SivakumarRathinam + + + WaqarMalik + + 10.2514/6.2010-7692 + AIAA 2010-7692 + + + AIAA Guidance, Navigation, and Control Conference + Toronto, Canada + + American Institute of Aeronautics and Astronautics + 2010 + + + Montoya, J., Wood, Z., Rathinam, S., and Malik, W., "A Mixed Integer Linear Program for Solving a Multiple Route Taxi Scheduling Problem," AIAA Guidance, Navigation, and Control Conference, AIAA 2010-7692, Toronto, Canada, 2010. + + + + + A Mixed Integer Linear Program for Airport Departure Scheduling + + GautamGupta + + + WaqarMalik + + + YoonJung + + 10.2514/6.2009-6933 + AIAA 2010-7692 + + + 9th AIAA Aviation Technology, Integration, and Operations Conference (ATIO) + Hilton Head, SC + + American Institute of Aeronautics and Astronautics + 2009 + + + Gupta, G., Malik, W., and Jung, Y., "A Mixed Linear Program for the Airport Departure Scheduling Problem," AIAA Aviation, Technology, Integration, and Operations Conference, AIAA 2010-7692, Hilton Head, SC, 2009. + + + + + Ration by Schedule for airport arrival and departure planning and scheduling + + ChrisBrinton + + + StephenAtkins + + + LaraCook + + + StevenLent + + + TomPrevost + + 10.1109/icnsurv.2010.5503239 + + + 2010 Integrated Communications, Navigation, and Surveillance Conference Proceedings + Herndon, VA + + IEEE + 2010 + + + Brinton, C., Atkins, S., Cook, L., Lent, S., and Prevost, T., "Ration by Schedule for Airport Arrival and Departure Planning and Scheduling," 2010 Integrated Communications Navigation and Surveillance Conference, Herndon, VA, 2010. + + + + + Managing departure aircraft release for efficient airport surface operations + + WaqarMalik + + + GautamGupta + + + YoonJung + + 10.2514/6.2010-7696 + AIAA 2010-7696 + + + AIAA Guidance, Navigation, and Control Conference + Toronto, Canada + + American Institute of Aeronautics and Astronautics + 2010 + + + Malik, W., Gupta, G., Jung, Y., "Managing Departure Aircraft Release For Efficient Airport Surface Operations," AIAA Guidance, Navigation, and Control Conference, AIAA 2010-7696, Toronto, Canada, 2010. + + + + + Analysis of a Surface Congestion Management Technique at New York JFK Airport + + AlexNakahara + + + TomReynolds + + + ThomasWhite + + + ChrisMaccarone + + + RonDunsky + + 10.2514/6.2011-6987 + AIAA 2011-6987 + + + 11th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Virginia Beach, VA + + American Institute of Aeronautics and Astronautics + 2011 + + + Nakahara, A. and Reynolds, T., "Analysis of a Surface Congestion Management Technique at New York JFK Airport," AIAA Aviation Technology, Integration, and Operations Conference, AIAA 2011-6987, Virginia Beach, VA, 2011. + + + + + Impact of Heavy Aircraft Operations on Airport Capacity at Newark Liberty International Airport + + IoannisSimaiakis + + + AlexanderDonaldson + + + HamsaBalakrishnan + + 10.2514/6.2011-6988 + AIAA 2011-6988 + + + 11th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Virginia Beach, VA + + American Institute of Aeronautics and Astronautics + 2011 + + + Simaiakis, I., Donaldson, A., and Balakrishnan, H., "Impact of Heavy Aircraft Operations on Airport Capacity at Newark Liberty International Airport," AIAA Aviation Technology, Integration, and Operations Conference, AIAA 2011-6988, Virginia Beach, VA, 2011. + + + + + Demonstration of reduced airport congestion through pushback rate control + + IoannisSimaiakis + + + HarshadKhadilkar + + + HamsaBalakrishnan + + + TomGReynolds + + + RJohnHansman + + 10.1016/j.tra.2014.05.014 + + + Transportation Research Part A: Policy and Practice + Transportation Research Part A: Policy and Practice + 0965-8564 + + 66 + + 2011 + Elsevier BV + Berlin, Germany + + + Ninth USA + Simaiakis, I., Khadilkar, H., Balakrishnan, H., Reynolds, T., Hansman R., Reilly, B., and Urlass S., "Demonstration of Reduced Airport Congestion Through Pushback Rate Control," Ninth USA/Europe Air Traffic Management Research and Development Seminar, 48, Berlin, Germany, 2011. + + + + + Tower Controllers' Assessment of the Spot and Runway Departure Advisor (SARDA) Concept + + THoang + + + YJung + + + JHolbrook + + + WMalik + + + + Europe Air Traffic Management Research and Development Seminar + + 34 + 2011 + Berlin, Germany + + + Ninth USA + Hoang, T., Jung, Y., Holbrook, J., and Malik, W., "Tower Controllers' Assessment of the Spot and Runway Departure Advisor (SARDA) Concept," Ninth USA/Europe Air Traffic Management Research and Development Seminar, 34, Berlin, Germany, 2011. + + + + + Performance Evaluation of SARDA: An Individual Aircraft-based Advisory Concept for Surface Management + + YoonJung + + + TyHoang + + + MiwaHayashi + + + WaqarMalik + + + LeonardTobias + + + GautamGupta + + 10.2514/atcq.22.3.195 + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 22 + 3 + + 2011 + American Institute of Aeronautics and Astronautics (AIAA) + Berlin, Germany + + + Ninth USA/Europe Air Traffic Management Research and Development Seminar + + + Jung, Y., Hoang, T., Montoya, J., Gupta, G., Malik, W., Tobias, L., and Wang, H., "Performance Evaluation of a Surface Traffic Management Tool for Dallas/Fort Worth International Airport," Ninth USA/Europe Air Traffic Management Research and Development Seminar, 92, Berlin, Germany, 2011. + + + + + A Simulator for Modeling Aircraft Surface Operations at Airports + + ZacharyWood + + + SivakumarRathinam + + + YoonJung + + + MatthewKistler + + 10.2514/6.2009-5912 + AIAA 2009-5912 + + + AIAA Modeling and Simulation Technologies Conference + Chicago IL + + American Institute of Aeronautics and Astronautics + 2009 + + + Wood, Z., Kistler, M., Rathinam, S., and Jung, Y., "A Simulator for Modeling Aircraft Surface Operations at Airports," AIAA Modeling and Simulation Technologies Conference, AIAA 2009-5912, Chicago IL, 2009. + + + + + Towards a Fast-time Simulation Analysis of Benefits of the Spot and Runway Departure Advisor + + RobertWindhorst + + 10.2514/6.2012-4975 + AIAA 2012-4975 + + + AIAA Guidance, Navigation, and Control Conference + Minneapolis, MN + + American Institute of Aeronautics and Astronautics + 2012 + 15 + + + Windhorst, R., "Towards a Fast-time Simulation Analysis of Benefits of the Spot and Runway Departure Advisor," AIAA Guidance, Navigation, and Control Conference, AIAA 2012-4975, Minneapolis, MN, 2012. 15 + + + + + Benefits Assessment of a Surface Traffic Management Concept at a Capacity-Constrained Airport + + KatyGriffin + + + AdityaSaraf + + + PeterYu + + + StevenStroiney + + + BenjaminLevy + + + GustafSolveling + + + John-PaulClarke + + + RobertWindhorst + + 10.2514/6.2012-5533 + AIAA 2012-5533 + + + 12th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference and 14th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference + Indianapolis, IN + + American Institute of Aeronautics and Astronautics + 2012 + + + Griffin, K., Saraf, A., Yu, P., Stroiney, S., Levy, B., Solveling, G., Clarke, J., and Windhorst, R., "Benefits Assessment of a Surface Traffic Management Concept at a Capacity-Constrained Airport," AIAA Aviation Technology, Integration, and Operations Conference, AIAA 2012-5533, Indianapolis, IN, 2012. + + + + + ACES terminal model enhancement + + GeorgeJCouluris + + + PaulCDavis + + + NathanCMittler + + + AdityaPSaraf + + + SebastianDTimar + + 10.1109/dasc.2009.5347530 + + + 2009 IEEE/AIAA 28th Digital Avionics Systems Conference + Orlando, FL + + IEEE + 2009 + + + Couluris, G., Davis, P., Mittler, N., Saraf, A., Timar, S., "ACES Terminal Modeling Enhancement," IEEE Digital Avionics Systems Conference, Orlando, FL, 2009. + + + + + Impact of Departure Prediction Uncertainty on Tactical Departure Scheduling System Performance + + AlanCapps + + + EdwardWalenciak + + + ShawnEngelland + + 10.2514/6.2012-5674 + AIAA 2012-5674 + + + + 12th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference and 14th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference + Indianapolis, IN + + American Institute of Aeronautics and Astronautics + 2012 + + + 18 ITT Exelis Symphony website + Capps, A., Day, K., Walenciak, E., and Engelland, S., "Impact of Departure Prediction Uncertainty on Tactical Departure Scheduling System Performance," AIAA Aviation Technology, Integration, and Operations Conference, AIAA 2012-5674, Indianapolis, IN, 2012. 18 ITT Exelis Symphony website, http://www.exelisinc.com/solutions/Symphony + + + + + The myClinicalOutcomes website: providing real-time, patient-level PROMs data + + DanWilliams + + 10.1308/147363512x13189526437991 + + + + The Bulletin of the Royal College of Surgeons of England + Bulletin + 1473-6357 + 1478-7075 + + 94 + 1 + + Feb. 24, 2012. 20 FlightStats website + Royal College of Surgeons of England + + + ITT Excelis + "Commercial Aviation Solutions VAS Gateway ICD -XML Format," ITT Excelis, 036-H101572, Feb. 24, 2012. 20 FlightStats website, http://www.flightstats.com/ 21 FAA Operations and Performance Data website, https://aspm.faa.gov/ + + + + + + diff --git a/file784.txt b/file784.txt new file mode 100644 index 0000000000000000000000000000000000000000..ccfacd8719f9137fbecce42a5b5d01eb792f910f --- /dev/null +++ b/file784.txt @@ -0,0 +1,413 @@ + + + + +IntroductionFor safety and passenger comfort, flights avoid airspace containing severe weather.One way of doing this is by rerouting flights around the airspace.For encounters predicted to occur in more than an hour, reroutes are currently selected from predefined sets, called playbooks and coded departure routes.Because the reroutes in these sets were pre-designed without knowledge of the specific weather events, they are usually not the most efficient.In practice, parts of playbook routes can be combined to refine a route, yet the parts are still pre-defined.An automated planning system could generate, extempore, reroutes that are tailored to the size and shape of the weather-impacted airspace and to the surrounding air traffic.In addition, the automated system could be integrated with other future automated systems designed to solve traffic congestion problems in all parts of the airspace [1 and 2].Several researchers have developed algorithms that generate reroutes around airspace with severe weather.In [2][3][4][5], grid-based search methods were used to find routes around severe weather.Reference [2] developed a sequential rerouting system and measured its performance as look-ahead-distance was varied.As the look-ahead-distance increased, rerouting delays decreased and the number of reroutes produced increased.Sridhar et.al. [6] proposed a geometric algorithm for generating a path stretch maneuver, consisting of auxiliary waypoints.It was used to route flights around congested sectors.The maneuver avoided the airspace by deviating the flight around the shorter side of the airspace.In [7], Love et.al. modified an automated conflict detection and resolution system to reroute flights around weather.The system not only prevents weather encounters, but also prevents conflicts with other aircraft.Although weather forecasts were used in these references, none analyzed the missed encounters due to the forecast uncertainty.In this paper, an automated planning system that routes airborne flights around airspace impacted by forecasted weather is developed.Since the weather forecasts are uncertain, i.e. they are not perfectly accurate, the system periodically re-plans based on the most recent weather forecast, flight tracks, and flight plan.Flights predicted to encounter hazardous airspaces within a predefined planning time horizon are included in the plan.Encounters predicted to occur before the beginning of the planning horizon are accounted for by another system, such as the one developed in [7].This is because that system builds reroutes that do not violate separation distances with other aircraft.Embedded within the planning system developed here is the algorithm from [6].It is used to build reroutes, sequentially, for each predicted encounter.When a flight has more than one predicted encounter, reroutes are generated in order of nearest time-to-encounter.The performance of the system is evaluated as the length of the planning horizon is varied between 20 and 100 minutes and the planning frequency is varied between 2 and 12 cycles per hour.Performance is measured by number of missed encounters, average delay per reroute, and number of reroutes per flight.This paper is organized as follows.First the weather forecast used in the experiment is presented.Next, the automated planning system is described.Results of a series of simulations designed to study the effect of the planning horizon and frequency on performance are illustrated.The paper concludes with a summary, conclusion, and future work. +Convective Weather ForecastsWeather, traffic congestion, military operations, and spacecraft launches are examples of events occurring within an airspace that may require flights to avoid it.The nature of an event determines the size, shape, and movement of the undesirable airspace.Although this paper is focused on airspace impacted by convective weather, the planning system can build reroutes around any hazardous airspace.There are various meteorological forecasts that predict the location, size, and severity of convective weather.These forecasts consist of multiple datasets describing various meteorological parameters, such as vertically integrated liquid and echo tops.The parameters need to be combined and interpreted to identify undesirable airspace.References [8][9][10] discuss a model for translating these forecasts into a set of dynamic airspaces that flights are likely to avoid.The model is named the Convective Weather Avoidance Model (CWAM).For the purposes of this paper, a CWAM forecast consists of a set of polygons that define the contours of airspaces containing convective weather.The set is divided into four subsets, each having a specified probability that flights will avoid the airspaces defined by the polygons.The probabilities are 40%, 60%, 80%, and 90%.In addition, the polygons are defined for specific altitudes, beginning at 25,000 ft. and ending at 45,000 ft. with 1,000 ft.increments.A set of CWAM forecasts is periodically produced every five minutes.A set includes 24 forecasts plus one nowcast, which represents the weather at the time that the set was produced.The forecasts range from five minutes to two hours with five-minute intervals.Green areas are 40% polygons, and blue areas are 60% polygons.Yellow areas are 80% polygons, and red areas are 90% polygons.Higher probability areas are drawn on top of lower probability areas.For areas representing the same convective weather cell, lower probability ones enclose higher probability ones.The shape and size of the polygons for all forecasts on 27 July 2006 were analyzed.Between 4 and 600 vertices defined each polygon, and in general, polygons were non-convex.Their cross sectional lengths ranged from 4 km. to 200 km., and their aspect ratios ranged from 1/1 to 1/10. +Automated Planning SystemThe automated planning system periodically generates reroutes for all flights predicted to encounter within a specified planning time horizon CWAM forecasted polygons.It consists of multiple functions, which are described next. +System ArchitectureFigure 2 shows a block diagram of the automated planning system.The system consists of three main blocks: the trajectory predictor, the encounter detector, and the resolver.Other blocks represent the national air traffic simulation and the CWAM forecasts.The two diamonds illustrate decision points.Planning occurs periodically based on a specified frequency, which was nominally 4 cycles per hour.In the experiment described in the results section, the frequency was varied.At the beginning of a planning cycle, current flight data, consisting of tracks and plans, are collected from the air traffic simulation, see the decision block in the upper half of Fig. 2.Flight tracks and plans are used in the trajectory predictor to build two-hour trajectories for each flight.The trajectories are discrete with one-minute resolutions.They are created using a four degree-offreedom flight dynamics model.Aircraft lift and drag coefficients, fuel burn rates, weights, and climb and descent speeds were selected from the Base of Aircraft Data [11].The encounter detector receives CWAM forecasts and the discrete trajectories.Each discrete point in a trajectory has a flight time.The polygons in the forecast for that time are checked to determine if the point is located inside one of them.An instance where a point lies in a polygon is called an encounter.Since the time resolution of the forecasts is five minutes and that of the trajectories is one minute, the forecasts are held valid until the next forecast.To minimize computations, the detector does not check parts of the trajectory between points for intersections with polygons.Since the trajectory points are one-minute apart, which translates to 7 nm for a flight traveling 420 knots, encounters where the flight does not penetrate the polygon by more than 7 nm could be missed.Although not done here, these encounters could be detected by adding buffer zones to the polygons, interpolating, or increasing the trajectory resolution.A list of encounters is passed from the detector to the resolver, which builds a reroute for each encounter.Since the resolver uses straight-line approximations for the trajectory, the reroutes may not successfully avoid all of the polygons.To check for failed reroutes, the resolver sends all of the reroutes, in the form of flight plan modifications, back to the trajectory predictor.New trajectories are predicted for flights with plan modifications, and the process is repeated.When no more encounters are detected, the process ends, and the flight plan modifications are sent to the flights in the air traffic simulation.The process will also end if it has iterated more than three times, even if there are unresolved encounters remaining.In the next planning cycle, both the weather forecasts and the flight trajectories will evolve.This evolution may allow a feasible path to be computed.Encounters that occur before the start of the planning horizon are not resolved.This is because reroutes for those encounters must also prevent the flight from violating separation criteria with other aircraft in the area.This problem is out of the scope of this research.The next two subsections detail the algorithm that creates the reroutes. +ResolverThe resolver uses the algorithm from [6] to build reroutes around polygons.This algorithm was selected because it is more computationally efficient than those in [2][3][4][5].Efficiency was important because the algorithm was used to plan reroutes for 53,000 flights operating in a 24-hour time period.The algorithm identifies the side with the shortest path around the polygon.Then, auxiliary waypoints on that side are inserted into the route.The steps of the algorithm are given in [6].They are repeated here for completeness. +Figure 3. Construction of a Reroute Around a PolygonLet P 1 through P 9 , in Fig 3, be the vertices of the polygon and P 0 and P f be the starting and ending points, respectively, of the reroute.Here, P 0 and P f were placed 10 nm before and after the polygon.1. Find the intersections of P 0 P f with the polygon and label them Q 1 and Q 2 .2. Find the midpoint of Q 1 Q 2 and label it Q m . +Divide the vertices by their location relative toQ 1 Q 2 .In Fig. 3, the right set is {P 6, P 7 , P 8 , P 9 }, and the left set is {P 1 , P 2 , P 3 , P 4 , P 5 }. +4.For each set, find the vertex (P l or P r ) furthest from Q m .In Fig. 3, P l is P 2 and, P r is P 6 .5. Select the vertex, P l or P r , closest to Q m and label the distance from Q m to it R.In Fig. 3, P 6 is closer to Q m .6. Construct a line of length R perpendicular to Q 1 Q 2 .It starts at Q m and points towards the side of the vertex chosen in step 5.In Fig. 3, this line points right.7. The end of the line drawn in step 6, point P n , is the new auxiliary waypoint in the flight plan.If any part of the new route intersects the polygon, define either P 0 P n or P n P f , whichever intersects the polygon, as the new P 0 P f and repeat, starting with step 1.On subsequent iterations in step 7, a new P n is generated.In reference [6], the old P n from the past iteration is discarded.Here, the old P n is inserted into the flight plan and kept.The cycle continues until the new route does not intersect the polygon, or it has iterated fifty times. +Flights with More than One EncounterThe algorithm described in the previous section only considers one encounter at a time.Some flights have more than one encounter within the planning horizon.The encounters for these flights are ordered by encounter time.The encounter with the earliest time is selected and solved, adding an auxiliary waypoint to the route.Auxiliary waypoints change the downstream path of the route and affect downstream encounters.For example, they may be eliminated, as the one occurring at 40 minutes in Fig. 4. Or, new encounters that were not there during the first detection may appear, see the encounter at 35 minutes in Fig. 4.Each time an auxiliary waypoint is added to the route, the downstream route must be rechecked for encounters.When there are downstream polygons that need to be avoided, the final reroute does not, in general, go around the shortest side of the group of polygons.In Fig. 4, it is clear that the final route would have been shorter had it gone around the left of the first polygon instead of the right. +ResultsThe automated planning system was implemented in a simulation of the air transportation system called the Airspace Concepts Evaluation System (ACES) [12 and 13].Simulations were conducted to assess the performance as the planning frequency and planning horizon were changed.In addition, reroutes for a single flight are tracked to illustrate how the planning system operates. +Simulation SetupAir traffic simulations were configured using CWAM forecasts generated from weather on Thursday, July 27, 2006.Convective weather that day was severe.It impacted the Midwest, especially affecting the Chicago area.Although the western United States was also impacted by severe convective weather, only weather in the northeastern quadrant of the United States was included in the simulations.This is because at the time the simulations were performed, CWAM forecasts were only available for that region.The CWAM probability of flight deviation was 60%, which was selected to be consistent with [2, 3, and 7].Flight plans were not selected from the same day that the weather was recorded because the traffic that day was routed around the weather.Instead, flight plans were taken from a clear weather day so that the routes would intersect the weather and test the planning system.They were extracted, using the Enhanced Traffic Management System, from the national airspace system on Tuesday, August 26, 2006.The traffic schedule contained 53,000 commercial flights that operated during a 30-hour time period.This represents a heavy traffic day.Winds were not modeled in the simulations.Airport arrival and departure rates and airspace capacities were unconstrained.Reroutes were only computed for flights in cruise.Multiple simulations varying three parameters of the system were executed.Table 1 shows the experiment matrix.The planning horizon was reduced in two different ways.In case 1, the far horizon boundary was moved backward, and in case 2, the near horizon boundary was moved forward.Figure 5 illustrates the relationship of the near and far boundaries with the full prediction horizon, which includes the near, planning, and far horizons.In case 3, the planning frequency was varied.The frequency is the inverse of planning interval, or time between successive plans. +Table 1. Experiment Matrix +CaseNear Horizon (min)Far +Performance MetricsThe performance of the system is measured using five metrics: number of successful reroutes, number of failed reroutes, number of popup encounters, number of non-popup encounters, and normalized delay per aircraft.Each planning cycle the resolver generates a reroute for each predicted encounter that is detected.A successful reroute solves the encounter and avoids the polygon.A failed reroute does not avoid the polygon.The sum of successful and failed reroutes is the total number of resolver attempts to solve predicted encounters.The success to failure ratio is a measure of the resolver performance.Ideally, the resolver should solve all of the predicted encounters, but for the reasons discussed in the resolver section and at end of the planning horizon results section, it does not.The resolver performance does not depend on the quality of the forecast.Predicted encounters occur when flights are predicted within the planning horizon to encounter forecast polygons.Another type of encounter, called an actual encounter, is identified after the simulation.Recorded flight tracks are checked against polygons from the CWAM nowcasts.Actual encounters are instances where flights actually encountered nowcast polygons.There are two types of actual encounters: popup and non-popup.Popup, also called missed, encounters are cases where the planning system did not have the opportunity to reroute the flight because the encounter was not detected within the planning horizon.This happens because the forecasts are poor.Non-popup encounters are cases where the system had an opportunity to create a reroute.Either the reroute was successful, but the forecast was poor, so the encounter reoccurred.Or, the reroute failed.The analysis here does not distinguish between the two.Figure 6 shows the taxonomy of an encounter. +Planning HorizonFigures 7 and8 show performance as the planning horizon is reduced.The planning interval was 15 minutes.The baseline case is where the planning system is turned off, i.e. the planning horizon is zero and the planning interval is infinite.For the other cases, the numbers on the horizontal axis are the near and far horizon boundaries.The green and pink bars denote the number of non-popup and popup encounters, and the blue and red bars denote the number of successful and failed reroutes.The percentage of successful reroutes was printed inside the blue bar.Finally, the light blue bars are the normalized average delays.The normalized delay was printed inside of the light blue bar for readability. +Figure 7. Performance with Varying Far HorizonBoundary (Case 1)Figure 7 shows performance for case 1 in Table 1.As the far boundary is moved backward, the planning horizon shrinks.Fewer encounters are detected, because ones predicted to occur later are excluded.The number of resolver attempts, denoted by the length of the red and blue bars, goes down.The number of actual encounters, denoted by the length of the pink and green bars, does not change much, and the normalized delay increases. +Figure 8. Performance with Varying Near HorizonBoundary (Case 2)Figure 8 shows performance for case 2 in Table 1.As the near boundary is moved forward, the planning horizon shrinks.Fewer encounters are detected, because ones predicted to occur sooner are excluded.The number of resolver attempts goes down.The number of actual encounters goes up, with most of the increase occurring in popup encounters.The normalized average delay remains approximately constant.As seen in both figures, the cases with the planning system reduce from the baseline case the number of actual encounters by a large amount.In the 20-120 case, the reduction is 79%, excluding popup encounters.As mentioned before, another system is required to plan reroutes for popup encounters.The resolver performs only reasonably well.There are still 25% -12% failures.Many of these are due to polygons with unusual geometries, to many polygons located in a small area, or to encounters occurring at the boundaries of the planning horizon.Using the convex hulls of the polygons could alleviate the problem with unusual geometry.A grid-based search, similar to the ones used in [2][3][4][5], may perform better, especially when there are many polygons located close together, which causes the method used here to generate reroutes that intersect the neighboring polygons.These performance data form a baseline to compare against future work. +Planning IntervalFigure 9 shows performance for case 3 in Table 1.The time horizon was 20-120 minutes.As the planning interval is increased, the number of resolver attempts goes down.However, the number of actual encounters slightly increases, with all of the increase in popup encounters.The normalized delays do not change.For the 10-minute interval, the number of non-popup encounters is reduced by 79% from the baseline. +Number of Reroutes per Hazardous AirspaceThe number of reroutes per flight needed to pass a hazardous airspace is an important performance parameter.It is a measure of the quality of the forecast and the reroute.Ideally, this number would be one.In that case, only one reroute would be sent to the flight to get around each hazardous airspace.In this section, this number will be investigated for a single flight and then for the whole simulation.The simulation used had a 20-120 minute planning horizon and 15 minute planning interval.Airspace 1 was located over the border of Kansas and Missouri.A predicted encounter between it and UPS2901 was detected at 13:20, see Fig. 10.There was 115 minutes-to-encounter, see Table 2.The planning system created a successful reroute.However, the 13:20 forecast for airspace 1 was too small.In subsequent forecasts, airspace 1 grew, and at 13:35, 15 minutes later, an encounter was detected again.The reroute produced by the planner failed.At 13:50, an encounter with airspace 1 was detected again.In addition, encounters with two new hazardous airspaces, 2 and 3 over Missouri and Illinois respectively, were detected.As illustrated in Fig. 11, the planner builds a successful reroute around all three hazardous airspaces.The forecast for airspace 1 continues to change, and at 14:05, UPS2901 is again predicted to encounter it.Figure 12 shows the successful reroute.Notice that a new large hazardous airspace over the junction of Indiana, Ohio, and Kentucky appears just past the far planning horizon boundary, marked as 120' in Fig. 12.This airspace is labeled 4 in Table 2.At 14:20, the system builds a successful reroute for airspaces 1 and 4. The reroute is presented in Fig. 13.One more reroute, generated at 14:50, was needed to pass UPS2901 around airspace 1. Table 2 and Figs.10-13 illustrated the number of reroutes per hazardous airspaces for one flight.For the whole simulation, Fig. 14 shows a histogram of the number of reroutes per flight needed to pass an encounter with a hazardous airspace.The majority of encounters required only one reroute.In the case of UPS2901, the encounter with airspace 1 required 6 reroutes, which is high.The other airspaces needed one to three, which is normal.The few encounters with 7 or higher reroutes were found to have anomalous reroutes. +ConclusionsAn automated planning system for rerouting flights around airspace containing convective weather was developed.The system detects encounters between flights and undesirable airspace.It builds reroutes for all encounters predicted within a specified planning horizon.To compensate for uncertainty in the weather forecast, it periodically re-plans at a specified planning frequency.The algorithm generates the reroutes using a simple geometric approach, which identifies a path around the shortest side of the airspace.When a flight has more than one encounter within the time horizon, the encounters are solved in order of earliest to latest.The planning system was implemented in a model of the national air traffic system.Simulations show that the planning system with a horizon of 20-120 minutes and an interval of 15 minutes is able to solve 79% of encounters that are detected.The reroutes delay flights on average about 3.3%.The algorithm that generates the reroutes is successful only 75% to 85% of the time.In most cases, one to three reroutes were required to route a flight around a hazardous airspace, but for the worst case, six reroutes were required.It is still unclear which planning horizon and frequency performs best.This is because the resolver needs to have a higher success rate.Based on the results shown here, extending the far horizon boundary past 40 minutes does not cost more missed encounters or excessive delays and reduces the number of reroutes.Moving the near horizon as far back as possible (10 minutes in this paper) reduces the number of actual encounters and increases the number of reroutes.However, the rerouting algorithm did not perform well for the 10-120 minute horizon (25% failures).Reducing the planning interval decreases the number of actual encounters and increases the number of reroutes.Future work will investigate several items.The performance of the algorithm that generates the reroutes needs to be improved.This could be accomplished by using a grid-based method and/or modifying the polygons so that they are convex.In addition, polygons located in close proximity could be combined into one.The reroutes could be optimized to minimize delays by generating alternate routes and selecting the best one.The reroutes need to be analyzed to understand their affect on traffic congestion and flight separation.Figure 11Figure 1 depicts a CWAM forecast covering an area slightly larger than the northeastern quadrant of the United States.The one-hour forecast was generated at 3:00 PM Eastern Daylight Time (EDT) on 27 July 2006 and was valid at 4:00 PM.In the figure, the polygons are filled in to make them more visible.Green areas are 40% polygons, and blue areas are 60% polygons.Yellow areas are 80% polygons, and red areas are 90% polygons.Higher probability areas are drawn on top of lower probability areas.For areas representing the same convective weather cell, lower probability ones enclose higher probability ones. +Figure 1 .1Figure 1.One-hour CWAM forecast for 4:00 PM EDT on 27 July 2006 +Figure 2 .2Figure 2. System Architecture +Figure 4 .4Figure 4. Application of Resolver Figure 4 illustrates the process for building reroutes for downstream encounters.The encounter predicted to occur at 25 minutes is considered first.Auxiliary waypoint A, which avoids the polygon at 25 minutes, is inserted into the route.The trial flight segment that connects auxiliary waypoint A to waypoint 2 is shown as a dashed line segment.This new flight segment is checked for possible intersections with other polygons on its path.It is found to intersect a polygon at 30 minutes.Auxiliary waypoint B is computed to avoid this polygon, and a new trial flight segment is computed, shown as the +Figure 5. Prediction Horizon +Figure 6 .6Figure 6.Taxonomy of an Encounter Normalized delay is the average delay per flight that was rerouted.It is normalized by the time for the flight to travel along its original flight plan from the departure to arrival airport. +Figure 9 .9Figure 9. Performance with Varying Planning Interval (Case 3) UPS2901 traveled from Ontario, California to Philadelphia, Pennsylvania.It departed at 12:00 PM EDT and cruised at 25,000 ft.On 27 July 2006, there were severe thunderstorms building and decaying from the Midwest to the East Coast.Six hazardous airspaces, caused by convective weather, were encountered by UPS2901.Table 2 summarizes the number of reroutes required to pass each hazardous airspace.The red numbers denote that the reroute at 13:35 failed.At 17:25 the flight had one popup encounter over Pennsylvania.It is not shown in Table 2 or Figs 10-13. +Figure 10 . 13 : 1 Figure 11 . 13 : 3 Figure 12 . 1 Figure 13 .101311113312113Figure 10.13:20 Reroute for Airspace 1 +Figure 14 .14Figure 14.Histogram of the Number of Reroutes per Flight Required to Pass an Encounter +Table 2 . Summary of UPS2901 Reroute Planning Cycles2HazardousPlanningPlanningTime-to-Airspace #Cycle #TimeEncounter(EDT)(minutes)1113:20115213:35100313:5085414:0570514:2050614:50202113:35115 + + + + +AcknowledgmentsAcknowledgement is due to Rich Pawlowicz, author of m_map, http://www.eos.ubc.ca/~rich/#M_Map.He created a map projection package for Matlab, which the authors used to render some of the figures. + + + + + + + + + Modeling and Optimization in Traffic Flow Management + + BanavarSridhar + + + ShonRGrabbe + + + AvijitMukherjee + + 10.1109/jproc.2008.2006141 + + + Proceedings of the IEEE + Proc. IEEE + 0018-9219 + 1558-2256 + + 96 + 12 + + December 2008 + Institute of Electrical and Electronics Engineers (IEEE) + + + Sridhar, Banavar, Shon Grabbe, and Avijit Mukherjee, December 2008, "Modeling and Optimization in Traffic Flow Management," Proceedings of the IEEE, Vol. 96, No. 12, pp. 2060- 2080. + + + + + Sequential Traffic Flow Optimization with Tactical Flight Control Heuristics + + ShonGrabbe + + + BanavarSridhar + + + AvijitMukherjee + + + + Journal of Guidance, Control, and Dynamics + + 32 + 3 + + May 2009 + + + Grabbe, Shon, Banavar Sridhar, and Avijit Mukherjee, May 2009, "Sequential Traffic Flow Optimization with Tactical Flight Control Heuristics," Journal of Guidance, Control, and Dynamics, Vol. 32, No. 3, pp. 810-820. + + + + + Design and Evaluation of a Dynamic Programming Flight Routing Algorithm Using Convective Weather Avoidance Model + + Hok + + + ShonNg + + + AvijitGrabbe + + + Mukherjee + + AIAA-2009- 5862 + + + AIAA Guidance, Navigation, and Control Conference + Chicago, Illinois + + August 2009 + + + Hok, Ng, Shon Grabbe, and Avijit Mukherjee, August 2009, "Design and Evaluation of a Dynamic Programming Flight Routing Algorithm Using Convective Weather Avoidance Model," AIAA-2009- 5862, Chicago, Illinois, AIAA Guidance, Navigation, and Control Conference. + + + + + Automated Route Generation for Avoiding Deterministic Weather in Transition Airspace + + JimmyKrozel + + + StevePenny + + + JosephPrete + + + JosephS BMitchell + + 10.2514/1.22970 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 30 + 1 + + 2007 + American Institute of Aeronautics and Astronautics (AIAA) + + + Krozel, Jimmy, Steve Penny, Joesph Prete, and Joeseph Mitchell, 2007, "Automated route generation for avoiding deterministic weather in transition airspace," Journal of Guidance, Control, and Dynamics, Vol. 30, No. 1, pp. 144-153. + + + + + Turn-Constrained Route Planning for Avoiding Hazardous Weather + + JimmyKrozel + + + ChangkilLee + + + JosephS BMitchell + + 10.2514/atcq.14.2.159 + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 14 + 2 + + 2006 + American Institute of Aeronautics and Astronautics (AIAA) + + + Krozel, Jimmy, Changkil Lee, and Joseph Mitchell, 2006, "Turn-Constrained Route Planning for Avoiding Hazardous Weather," Air Traffic Control Quarterly, Vol. 14, No. 2, pp. 159-182. + + + + + Integration of Traffic Flow Management Decisions + + BanvarSridhar + + + GanoChatterji + + + ShonGrabbe + + + KapilSheth + + AIAA-2002-5014 + + + AIAA Guidance, Navigation, and Control Conference + Monterey, CA + + August 2002 + + + Sridhar, Banvar, Gano Chatterji, Shon Grabbe, and Kapil Sheth, August 2002, "Integration of Traffic Flow Management Decisions," AIAA-2002-5014, Monterey, CA, AIAA Guidance, Navigation, and Control Conference. + + + + + Analysis of Automated Aircraft Conflict Resolution and Weather Avoidance + + JohnLove + + + WilliamChan + + + ChuLee + + 10.2514/6.2009-6995 + AIAA- 2009-6995 + + + 9th AIAA Aviation Technology, Integration, and Operations Conference (ATIO) + Hilton Head, South Carolina + + American Institute of Aeronautics and Astronautics + September 2009 + + + Love, John, William Chan, and Chu Han Lee, September 2009, "Analysis of Automated Aircraft Conflict Resolution and Weather Avoidance," AIAA- 2009-6995, Hilton Head, South Carolina. + + + + + An Approach to Verify a Model for Translating Convective Weather Information to Air Traffic Management Impact + + WilliamChan + + + MohamadRefai + + + RichDelaura + + 10.2514/6.2007-7761 + AIAA-2007-7761 + + + 7th AIAA ATIO Conf, 2nd CEIAT Int'l Conf on Innov and Integr in Aero Sciences,17th LTA Systems Tech Conf; followed by 2nd TEOS Forum + Belfast, Northern Ireland + + American Institute of Aeronautics and Astronautics + 2007 + + + Chan, William, Mohamad Refai, and Richard DeLaura, 2007, "An Approach to Verify a Model for Translating Convective Weather Information to Air Traffic Management Impact," AIAA-2007-7761, Belfast, Northern Ireland, AIAA Aviation Technology, Integration, and Operations Conference. + + + + + AMERICAN METEOROLOGICAL SOCIETY, JANUARY 5–10,1992 -- THE ATLANTA HILTON AND TOWERS, ATLANTA, GEORGIA + + RichDelaura + + + JamesEvans + + 10.1175/1520-0477-72.11.1811 + + + Bulletin of the American Meteorological Society + Bull. Amer. Meteor. Soc. + 0003-0007 + 1520-0477 + + 72 + 11 + + January 2006 + American Meteorological Society + Atlanta Georgia + + + DeLaura, Rich, and James Evans, January 2006, "An Exploratory Study of Modeling Enroute Pilot Convective Storm Flight Deviation Behavior," 12 th Conference on Aviation, Range and Aerospace Meteorology, American Meteorological Society, Atlanta Georgia. + + + + + Modeling Convective Weather Avoidance in Enroute Airspace + + RichDelaura + + + MikeRobinson + + + MargoPawlak + + + JimEvans + + + + 13 th Conference on Aviation, Range, and Aerospace Meteorology + New Orleans, LA + + American Meteorological Society + 2008 + + + DeLaura, Rich, Mike Robinson, Margo Pawlak, and Jim Evans, 2008, "Modeling Convective Weather Avoidance in Enroute Airspace," 13 th Conference on Aviation, Range, and Aerospace Meteorology, American Meteorological Society, New Orleans, LA. + + + + + Sensitivity Analysis of Aviation Environmental Impacts for the Base of Aircraft Data (BADA) Family 4 + + ANuic + + 10.2514/6.2021-0457.vid + + July 2003 + American Institute of Aeronautics and Astronautics (AIAA) + + + Revision 3.5," EEC Note No.11/03 + Nuic, A., July 2003, "User Manual for the Base of Aircraft Data (BADA), Revision 3.5," EEC Note No.11/03, Eurocontrol. + + + + + Build 4 of the Airspace Concept Evaluation System + + LarryMeyn + + + RobertWindhorst + + + KarlinRoth + + + DonaldVan Drei + + + GregKubat + + + VikramManikonda + + + SharleneRoney + + + GeorgeHunter + + + AlexHuang + + + GeorgeCouluris + + 10.2514/6.2006-6110 + AIAA-2006- 6610 + + + AIAA Modeling and Simulation Technologies Conference and Exhibit + Keystone, Colorado + + American Institute of Aeronautics and Astronautics + August 2006 + + + Meyn, Larry, Robert Windhorst, Karlin Roth, Donald Van Drei, Greg Kubat, Vikram Manikonda, Sharlene Roney George Hunter, Alex Huang, and George Couluris, August 2006, "Build 4 of the Airspace Concept Evaluation System," AIAA-2006- 6610, Keystone, Colorado, AIAA Modeling and Simulation Technologies Conference. + + + + + Validating the Airspace Concept Evaluation System for Different Weather Days + + ShannonZelinski + + + LarryMeyn + + 10.2514/6.2006-6115 + AIAA-2006-6115 + + + AIAA Modeling and Simulation Technologies Conference and Exhibit + Keystone, Colorado + + American Institute of Aeronautics and Astronautics + August 2006 + + + Zelinski, Shannon, and Larry Meyn, August 2006, "Validating the Airspace Concept Evaluation System for Different Weather Days," AIAA-2006-6115, Keystone, Colorado, AIAA Modeling and Simulation Technologies Conference. + + + + + + diff --git a/file785.txt b/file785.txt new file mode 100644 index 0000000000000000000000000000000000000000..a8a0fa1c94b7b9b6f5e1ca5936a599d1fa8db4f8 --- /dev/null +++ b/file785.txt @@ -0,0 +1,275 @@ + + + + +Nomenclature i= aircraft index R P min (i) = statistical minimum of path distance, nmi V TAS max (i) = statistical maximum of average true airspeed, knots S R (i) = minimum required leading/trailing wake vortex spacing, nmi R P (i) = path distance, nmi S new (i) = new leading/trailing wake vortex spacing, nmi S E (i) = excess spacing at runway threshold, nmi S B (i)= spacing buffer, nmi +S(i)= observed leading/trailing wake vortex spacing at runway threshold, nmi !T D (i) = delay time saved, sec !R P (i) = path distance saved, nmi T entry (i) = TRACON entry time, Unix time T runway (i) = runway threshold crossing time, Unix time T D (i) = estimated delay time, sec !T P (i) = delay savings due to path compression, sec !T S (i) = delay savings due speed compression, sec +I. Introductionrecent comprehensive study of aircraft arrivals at several of the busiest airports in the United States showed excess leading/trailing aircraft wake vortex spacing. 1 This excess spacing is an inefficiency in the terminal area that causes increased delay and fuel consumption and it degrades throughput.Due to the projection that air traffic is expected to grow 90% over the next twenty years as measured by revenue passenger miles, 2 this growth will certainly lead to more airport arrival operations that will amplify this excess spacing inefficiency.The Next Generation Air Transportation System (NextGen) is our Nation's solution for safely and efficiently handling this substantial increase in air traffic demand relative to current day operations. 3NextGen technologies are expected to reduce this excess spacing; therefore, reducing delay, fuel consumption and increasing throughput.Two approaches to assess the effectiveness of NextGen technologies are human-in-the-loop (HITL) simulations and statistical models.HITL simulations require supporting infrastructure (laboratory), participants, and a sufficient financial budget.Additionally, HITL simulations take significant time to organize and execute; however, they sometimes are the credible way to validate a new technology.In contrast, some assessments are more appropriately studied by statistical models.The technical tradeoff is the operational realism offered by a HITL simulation versus the faster, broader assessments enabled by statistical models.These two approaches can complement each other.For example, if a statistical model of a NextGen technology shows some potential benefits, the next step may be to access it with a HITL simulation.Data from HITL simulations can also be used to improve the statistical models.][6] Delay and throughput benefits have been determined based upon those results, which are specific to an airport, airport configuration, fleet mix, etc.Recently, two studies that make use of the statistical modeling approach have quantified the potential benefits from certain NextGen technologies such as precision scheduling, sequencing, and spacing and continuous descent operations (CDO). 1,7Both of these comprehensive studies made use of the same very large dataset of recorded traffic from 2010 (approximately 500,000 arrival flights).In Ref. 7, Robinson and Kamgarpour showed that the potential fuel savings from CDO during congested arrival periods is highly variable -day-to-day, aircraft to aircraft, etc.In Ref. 1, Zelinski analyzed the observed excess in-trail spacing for aircraft operating in instrument meteorological conditions (IMC) and visual meteorological conditions (VMC) to 29 runways in eight of the busiest Terminal Radar Approach Control Facilities (TRACONs).Zelinski assessed two major benefits of precision scheduling and spacing: (1) potential throughput increases and (2) potential flight time savings.The potential increase in throughput was determined using a separation buffer of 0.3 nmi.Performance-based navigation (PBN) such as Area Navigation (RNAV)/Required Navigation (RNP) 8 routes were modeled for each runway to assess delay savings.Flights were re-sequenced according to their new arrival times along these routes using the same speed profiles flown along the original routes.The delay savings for flights operating in VMC and IMC were estimated based on these original and new arrival times.The resource intensive procedure of trajectory reconstruction and subsequent examination to determine suitable RNAV routes was a key requirement of Zelinski's precision scheduling and spacing analysis. 9No alternative statistical method for assessing the potential benefits of precision scheduling and spacing automation tools on excess spacing can be found in the literature.In this study, we offer a statistical modeling method, like those discussed in Ref. 1 and Ref. 7., for assessing the potential benefits of terminal scheduling and spacing automation tools that does not require all of the complexities associated with HITL simulations.Moreover, the method presented in this paper does not require the complexity of reconstructing/generating trajectories to determine suitable RNAV routes.Our model estimates delay reduction when excess spacing is recovered by shortening each aircraft's path and (optionally) increasing its speed closer to its shortest and fastest reasonable limits, respectively.We refer to each model used to shorten the path and (optionally) increase the speed as the "traffic compression" model or "compression" model for short.This study is motivated by a desire to estimate flight time in the terminal area due to excess spacing at the runway threshold given a large arrival traffic sample (>1,000 flights) without requiring resource intensive trajectory reconstruction and examination procedures of each aircraft's route flown from the meter fix to the runway threshold.The study described in this paper quantifies potential delay savings, but is not motivated by a specific technology (e.g., RNP).NextGen technologies, however, are expected to result in more efficient terminal area operations-realizing at least some of the potential savings.Our statistical model can be viewed as a first order analysis that estimates delay savings due to reduced excess spacing.This work differs from Ref. 1 in four key ways.First, it does not require establishing RNAV/RNP routes through reconstruction of each aircraft's trajectory.In fact, the methodology used in this analysis uses trajectory states at just two key discrete points along the aircraft's arrival trajectory: entry into the TRACON and at the runway threshold.Second, the impact of speed variations, in addition to path variations, is explored.Third, the flights are segregated by meteorological condition before delay savings is estimated, whereas Ref This material is declared a work of the U.S. Government and is not subject to copyright protection in the United States.segregates the results after delay savings is estimated.Fourth, the minimum separation requirement of 2.5 nmi is used instead of 3 nmi. 10,11he methodology and traffic compression models used for the analyses are discussed in detail in Section II.Section III presents the results of the delay savings analysis.Lastly, we outline our conclusions regarding the statistical traffic compression model described in this paper in Section IV. +II. MethodologyThis analysis uses the recorded tracks of approximately 500,000 flights in eight different TRACONs.These flights occurred during January through May 2010.Figure 1 depicts a top-level flow diagram of the generation of the input data used in the compression models.The mathematical notation used in the last two blocks in the flow diagram are defined in the nomenclature section and will be described in detail later.The procedure begins with the recorded flight plans and tracks of the Center/TRACON Automation System (CTAS). 12A bundled set of post-processing scripts called Terminal Utilities performs various functions and calculations such as data integrity checks, required runway threshold separation, runway assignments, and determines various speeds and distances (e.g.ground, air, etc.).These calculations are described in more detail in Ref. 7. However, aircraft delay is not determined via the recorded radar tracks and subsequent Terminal Utilities post-processing because of prohibitively long execution times.In this paper, we estimate the aircraft delay as part of the compression models.It is not necessary that this delay estimate be precise because we are interested only in delay savings (i.e.relative delay values and not absolute).Now, following the execution of the Terminal Utilities, a database of numerous aircraft parameters is generated.Included in this database are the X and Y coordinates of the aircraft.For the purpose this analysis, a 40 nmi radius from the applicable runway threshold defines the notional TRACON boundary.Figure 2 shows the TRACON entry and runway threshold coordinates for aircraft landing at KATL 27L.The compression models make use of a subset of aircraft parameters captured upon entry into the TRACON and at the runway threshold (facilitated by the Terminal Utilities data processing-with the exception of meteorological conditions) and is listed in Table 1.The entry and runway subscribes correspond to parameters captured at the TRACON entry and runway threshold, respectively.The required spacing, S R , is a function of the leading and trailing aircraft weight classes (commonly referred to as "3/4/5 spacing" to denote the required minimum separation in nmi). 13The observed spacing, S , is the distance between the leading and trailing aircraft when the leading aircraft crosses the runway threshold.Path distance is the observed horizontal distance that the aircraft traverses.Ground distance is the estimated distance traveled by integrating the aircraft's observed ground speed with respect to time.Air distance is the estimated distance traveled by integrating the aircraft's estimated true airspeed (also with respect to time).Thus, the ground and air distances would be the same if there were no winds.Crossing times are captured at the TRACON entry, T entry , and runway threshold, T runway , in addition to the distances.Key flight attributes such as the aircraft identification (ID), arrival route (i.e., Standard Terminal Arrival Route (STAR)), and engine type (jet, turboprop, and piston) for each flight are also recorded.Airport meteorological conditions are retrieved from the FAA's Aviation System Performance Metrics (ASPM) quarter-hourly reports 14 and then fused into Table 1.The analysis uses the inputs provided in Table 1 to derive a reference delay time, T D (i) for each aircraft, i .This time establishes a reference time to facilitate relative delay savings calculations resulting from compressing the arrival traffic.The path distance flown for each aircraft, i , is calculated beginning with the first flight recorded on the first day of recordings ( i =1) and ending with the last flight recorded on the last day of recordings, R P (i) = P runway (i) !P entry (i).(1) This material is declared a work of the U.S. Government and is not subject to copyright protection in the United States.The air and ground distances are, respectively,R A (i) = A runway (i) ! A entry (i)(2)andR G (i) = G runway (i) ! G entry (i).(3)The transit time isT (i) = T runway (i) ! T entry (i).(4)The average true airspeed is determined using the air distance and transit time,V TAS (i) = R A (i) / T (i)(5)whereas the average ground speed uses of the ground distance and transit time,V GS (i) = R G (i) / T (i).(6)The average headwind for each aircraft is estimated as the difference between these average speeds:V W (i) ! V GS (i) "V TAS (i).(7)Arrival flights, for all days, are grouped by engine type, STAR, runway, and meteorological condition.This segregation results in a distribution of path distance, Eq. ( 1), and average true airspeed, Eq. ( 5) for each unique engine type/STAR/runway/meteorological condition grouping.Those aircraft that flew faster than the median are in a percentile greater than the 50 th , but less than the 100 th .Similarly, those aircraft that flew shorter distances than the median are in a percentile less than the 50 th , but greater than the 0 th .The strategic selection of a path distance and true airspeed percentile enables a delay time to be calculated.The extremes in the path distance and true airspeed distributions are avoided by limiting the minimum and maximum percentiles to the 10 th percentile (path) and 90 th percentile (speed).These outliers are included in the compression models, but do not receive any benefits.Hereafter, the minimum path distance that corresponds to a certain percentile is referred to as R P min (k) and the maximum (average) true airspeed that corresponds to a certain percentile is referred to as V TAS max (k) .The index k denotes that there are k unique engine type/STAR/runway/meteorological condition combinations.The minimum path distance for each aircraft, R P min (i), and maximum average true airspeed, V max TAS (i), correspond to the appropriate R P min (k) and V TAS max (k), respectively.The compression model does not allow path distances to increase.For those aircraft that have shorter than the statistically calculated minimum path distances (like the outliers mentioned above), the original path distance is used (i.e., if R P min (i) > R P (i) ), then R P min (i) = R P (i) ).A similar check on V max TAS (i) is not performed.Now, a new ground speed for each aircraft is determined from the maximum true airspeed and average winds,V GS new (i) = V TAS max (i) +V W (i).(8)The ratio of the minimum path distance and this new ground speed establishes a minimum transit time,T min (i) = R P min (i) / V GS new (i).Finally, the reference delay time for each arriving flight is established as the observed transit time less the minimum transit time,T D (i) = T (i) ! T min (i).(10)Delay time for each aircraft, T D (i) , in conjunction with the observed spacing, S from Table 1, establish a reference for comparison and are hereafter referred to as the baseline.Two types of traffic compression models are used to quantify the potential reduction of excess spacing at the runway threshold and the corresponding delay savings.The first model is referred to as the path compression model because the flight's path distance can be reduced, but its observed average true airspeed is retained.The second model is referred to as the path and speed compression model because the flight's path distance can be reduced and its average true airspeed can be increased in order to recover additional excess spacing.Both traffic compression models result in new path distances flown, S new (for comparison with S ), and new reference delay times, T D new (for comparison with T D (i) provided in Eq. ( 10).The path compression model is described first, followed by the path and speed compression model. +A. Path Compression ModelThe path compression model determines the excess spacing,S E (i) = S(i) ! S T (i),(11)which is the difference between the observed spacing and the target spacing, given below,S T (i) = S R (i) + S B (i),(12)where S R is the required in-trail separation, and S B is a spacing buffer.The spacing buffer is modeled as a normal random number with mean, µ , and standard deviation, ! ,S B (i) ~!(µ,! ).()13The mean spacing buffer is set to 0.5 nmi.Its standard deviation is prescribed as half of the mean (! = 0.25 nmi) based on the reasoning described in Ref. 1 indicating that 95% separation conformance is achieved within a buffer size at least twice that of the standard deviation.Eq. ( 13) models the in-trail spacing variations observed in actual operations.For each day of recorded aircraft traffic, Eqs.(14-16) determine the amount of path compression, the new path distance flown and the new excess spacing, respectively,!R P (i) = R P (i) " R new,P (i),(14)R new,P (i) = max R P min (i), R P (i) ! max(S E (i), 0) { } if i = 1 max R P min (i), R P (i) ! max(S E (i) + "R P (i !1), 0) { } if i > 1, # $ % & % (15) S E new (i) = S E (i) ! "R P (i) if i = 1 S E (i) ! ("R P (i) ! "R P (i !1)) if i > 1. # $ % & %(16)The minimum value for !R P (i) is zero occurring when R P min (i) = R P (i) .The new transit time after path compression is determined from the new path distance and original average ground speed T new,P (i) = R new,P (i) / V GS (i).This material is declared a work of the U.S. Government and is not subject to copyright protection in the United States. +The new runway threshold crossing time isT runway new,P (i) = T runway (i) !"T P (i),where !T P (i) = T (i) " T new,P (i).()19The new spacing and new delay time follow, respectively,S new (i) = S R (i) + S E new (i),(20)T D new (i) = T D (i) ! "T P (i).(21)The delay savings resulting from path compression and excess spacing reduction is found using the baseline delay provided in Eq. (10) and the new delay time from Eq. ( 21),!T D (i) = T D (i) " T D new (i). (22)Because the metric of interest is a time change, Eq. ( 22) is equivalent to Eq. ( 19).That is, the reduced delay is naturally the reduced transit time.The new excess spacing calculated using Eq. ( 16) is never negative, ensuring that the original arrival sequence is preserved. +B. Path and Speed CompressionThe path and speed compression model is an extension of the path compression model.The new transit time given in Eq. (17) above comes from flying a shorter path than the originally observed route while maintaining the same original average ground speed.Modeling speed changes in addition to route changes is another compression strategy described next.Here, the minimum transit time is calculated using the new path distance, Eq. (15), and new average ground speed, Eq. ( 8),T * min (i) = R new,P (i) / V GS new (i).(23)Next, the two components of the time change are determined.The first component is the difference from the values found in Eq. ( 17) and Eq. ( 23),!T 1 (i) = T new,P (i) " T * min (i).()24The second component is the ratio of the new excess spacing (given in Eq. ( 16)) to the new ground speed,!T 2 (i) = S E new (i) / V GS new (i). (25)Estimating the time change as T runway new,S (i) = T runway new,P (i) !"T S (i), This material is declared a work of the U.S. Government and is not subject to copyright protection in the United States.!T S (i) = min(!T 1 (i), !T 2 (i)),(26)T new,S (i) = T runway new,S (i) !T entry (i),T D new,S (i) = T D (i) !(T runway (i) !T runway new,S (i)).(29)Finally, the delay savings due to path and speed compression and excess spacing reduction is!T D (i) = T D (i) " T D new,S (i).(30)Algebraic manipulations (see Appendix) of Eqs. ( 18), ( 27), (29), and (30), reveal a more insightful expression for Eq. ( 30) showing that the delay savings is a linear combination of the delay savings of the path compression model, Eq. ( 19), and the speed compression model, Eq. ( 26),!T D (i) = !T P (i) + !T S (i).(31) +C. ScopeResults are shown for two busy runways in the United States.The two runways selected, KATL 27L and KDEN 35R, are chosen because they are independent runways dedicated to arrivals only.Zelinski identified several runway procedural constraints that could affect in-trail spacing in Ref. 1; the two runways chosen here are free from those constraints.Three types of results are shown: The sensitivity of the results to the compression model, the daily variation of the estimated delay savings and excess spacing reduction, and the excess spacing statistical distributions before and after path compression.Sensitivity analyses are performed on the R P min and V TAS max threshold percentiles in order to understand how they affect the estimated delay savings.As a reminder, these percentiles determine the reasonable "minimum" observed path distance and "maximum" average true airspeed, respectively.These percentiles are calculated on a per runway/STAR/engine type/meteorological condition basis.For the path compression model, the delay savings, Eq.(19), is unaffected by V TAS max , so two threshold percentiles are chosen for R P min to understand the range of !T D .The chosen lower bound is the 50 th percentile while the chosen upper bound is the 10 th percentile.The path and speed compression model uses these same two R P min threshold percentiles in combination with two V TAS max threshold percentiles (50 th and 90 th , the lower and upper limits of the true airspeed distribution, respectively).This results in a total of two discrete values of !T D for the path compression model and four discrete values for the path and speed compression model.The notation 10 R P min indicates the minimum path distance value, R P min , was assigned to the 10% percentile path distance value.Similar notation is used for the maximum average true airspeed, V TAS max .The two discrete values of !T D for the path compression model correspond to 50 R P min and 10 R P min .The four discrete values of !T D for the path and speed compression model correspond to (1) 50 R P min , 50 V TAS max , (2) 50 R P min , 90 V TAS max , (3) 10 R P min , 50 V TAS max , and (4) 10 R P min , 90 V TAS max .This same notation will be retained in the results section below.Now, varying the percentile that defines the statistical minimum, R P min , also affects the path distance saved, !R P , given in Eq. ( 14)and will be shown in the results section.Following the minimum path distance and maximum average true airspeed sensitivity analyses, the daily variations of delay savings, !T D , and excess spacing reduction, !R P , are examined for KATL 27L.The statistical minimum path distance is fixed at the 10 th threshold percentile ( 10 R P min ).Only results for the path compression model are presented.Lastly, cumulative distribution functions (CDFs) are used to compare the baseline (i.e., observed) in-trail spacing and the in-trail spacing resulting from the path compression model for the 2.5 nmi required minimum separation.The analysis compares the percentage of flights with excess in-trail spacing of 0.5 nmi or less.In these analyses, the delay time savings and path distance savings are provided in terms of the statistical median rather than the mean in order to lessen the affect of outliers in the data.The number of outliers is small (a fraction of a percent) and are the typical of data consistency errors encountered when analyzing large real air traffic data sets.).With this definition, a significant number of aircraft were not allowed to lengthen their path, which would have resulted in negative delay savings.Although the average delay savings was not zero, this paper references the median as the statistical metric.Flying shorter than average routes in the terminal area, 10 R P min , reducing the excess spacing, compresses the arrival traffic and saves almost 30 seconds per flight for KATL 27L (Fig. 3) and nearly 60 seconds for KDEN 35R (Fig. 4) during VMC periods.During IMC periods, about 40 seconds are saved for KATL 27L (Fig. 3) and about 108 seconds for KDEN 35R (Fig. 4).The path and speed compression model achieves the most delay savings at the lower and upper bounds on the minimum path distance and maximum true airspeed percentiles, respectively ( 10 R P min , 90 V TAS max ).In these circumstances, aircraft are flying faster on shorter routes (faster and shorter than the actual route/speeds flown).This material is declared a work of the U.S. Government and is not subject to copyright protection in the United States. +III. ResultsA couple of trends can be observed in Fig. 3 and Fig. 4. First, there is greater delay savings during IMC periods than VMC periods, and greater savings at KDEN 35R than KATL 27L.These results are consistent with more excess in-trail spacing during IMC periods and at KDEN 35R.Second, there is greater sensitivity to the R P min threshold percentile than the V TAS max threshold percentile.These results indicate that there is more variability in the path distances flown than the true airspeeds.This material is declared a work of the U.S. Government and is not subject to copyright protection in the United States.Figure 5 shows the effect of varying the "minimum" path distance threshold percentile on the median path distance saved, Eq. ( 14).Speed compression does not reduce the path distance flown; thus, only R P min threshold percentiles are examined.Open squares indicate the 50 th threshold percentile whereas asterisks indicate the 10 th threshold percentile.Consistent with the delay savings shown in Fig. 3 and Fig. 4, path reductions are achieved when flying shorter than average routes, IMC reductions are greater than VMC, and are greater at KDEN 35R.In other words, there is more recoverable excess spacing in IMC than in VMC and more at KDEN 35R than KATL 27L.This material is declared a work of the U.S. Government and is not subject to copyright protection in the United States. +B. Daily Variation of Results for KATL Runway 27LThe following three figures show data from analyses at KATL 27L. Figure 6 illustrates the daily variations of the delay savings, Fig. 7, the median path distance reduction, and Fig. 8, the number of flights during VMC and IMC periods.These analyses are presented for the path compression model with a fixed threshold value of 10 R P min .The delay savings shown in Fig. 6 varies from a few seconds per aircraft to roughly 115 seconds.Similar to the findings in Ref. 7, these results suggest that the day-to-day benefit pool of increased spacing precision will be quite variable.Similarly, Fig. 7 shows the daily variation of the path distance reduction.Values range from less than 1 nmi to about 5.5 nmi.Some of this daily variation is due to a low number of arrivals in the database for a given day -for example, the VMC results for Day 30.However, the cause(s) of much of the daily variation is not readily apparent.For example, Days 2, 3, 4, 15, and 27 each have about 500 flights and are exclusively VMC.Generally, these results are consistent with the findings reported in Ref.This material is declared a work of the U.S. Government and is not subject to copyright protection in the United States.Most of the days that show no delay savings are a result of missing arrivals in the database for that day, as seen in Fig. 8.The missing flights were caused by a traffic recording malfunction that has since been resolved.Most of the days that show no delay savings are a result of missing arrivals in the database for that day, as seen in Fig. 8. Days 5, 11, 19, 21, 22, 24, and 28 did not have any tracks recorded.In addition, Days 18, 30, 31, and 32 have few VMC arrivals recorded, and Days 7, 12, 26, and 31 have few IMC arrivals recorded.Therefore, some of the results for these days can be particularly sensitive to the sample size.For example, the Day 8 IMC results and Day 30 VMC results should be considered outliers due to their low sample sizes relative to the other traffic days.The missing flights were caused by a traffic recording malfunction that has since been resolved. +C. Cumulative Distribution FunctionsA cumulative distribution function (CDF) provides the probability of a random flight arriving with a leading/following aircraft wake vortex spacing less than or equal to x , where x is the spacing value of interest.This section provides the CDFs for (1) the observed in-trail spacing, S in Table 1 (referred to as the baseline) and(2) the in-trail spacing resulting from the path compression model, S new , from Eq. ( 20). S and S new are examined for those flights that require a minimum 2.5 nmi separation from the flight immediately ahead of it; most arrivals at KATL 27L and KDEN 35R are subject to that minimum required separation.Figure 9 shows that as the in-trail spacing increases so do the likelihood of more flights achieving that spacing.Ten percent of the flights have less than the required minimum separation for the baseline (actual operations) whereas roughly 6% of the flights have less than the required minimum separation after adjustments are made with the path compression model.In these cases, the leading aircraft of the pair was compressed forward in time and more in-trail spacing was achieved.A crossover point exists at about 2.8 nmi after which the path compression model results in more arrivals with less excess spacing relative to the baseline.These results show that 20% (below crossover point) of flights are subjected to a slight increase of in-trail spacing while the remaining 80% (above crossover point) achieve a moderate reduction in excess spacing -roughly 0.25 nmi.Another suitable point of comparison is 3 nmi as this corresponds to the mean buffer of 0.5 nmi prescribed in Eq. (13).Thirty percent of flights land with a 3 nmi separation or less for the baseline whereas the path compression model increases the probability by 10% (40% of the arrivals achieving a spacing buffer of 0.5 nmi or less).Figure 10 shows the CDFs for KATL 27L for aircraft operating in IMC periods; it is similar to the VMC CDFs in Fig. 9; however, there is no crossover point.For flights during IMC periods, there is a more substantial reduction in excess spacing that can be recovered -roughly 0.5 nmi.In other words, the observed excess in-trail separation present in the operations to KATL 27L is accompanied by enough corresponding excess path distance flown to allow its recovery.The percentage of arrivals landing with excess spacing of 0.5 nmi (3 nmi spacing) or less is 10% for the baseline and 22% for the path compression model; an increase of 12%, roughly the same as the arrivals under VMC periods.However, more arrivals during IMC periods benefit (i.e., land with less excess spacing in general) as a result of the path compression than during VMC.This observation is made when comparing the area between the baseline and path compression CDFs in Immediately apparent is the larger difference between the baseline and the path compression model relative to those shown for KATL 27L. Figure 11 (previous page) shows that ~5% of the baseline arrivals land with 3 nmi spacing (excess separation of 0.5 nmi).The path compression model achieves ~30%, a 25% increase over actual operations.Figure 12 presents the last CDF in this analysis for arrivals during IMC periods.Very few baseline arrivals, about 2%, land with excess spacing of 0.5 nmi compared to ~30% resulting from the path compression. +IV. ConclusionA statistical modeling method for accessing the potential benefits of terminal scheduling and spacing automation tools has been developed and described.The key to such a statistical model is a sufficiently large data set of recorded flights.The model is referred to as a traffic compression model because it estimates delay reduction when excess spacing is recovered by shortening each aircraft's path and (optionally) increasing its speed closer to its shortest and fastest reasonable limits, respectively.This model does not require the trajectory reconstruction/generation procedure of modeling RNAV routes.Instead, our compression models can quickly provide first order analysis of potential delay savings achieved by reducing excess leading/following aircraft wake vortex separation using aircraft parameters captured at just two discrete locations in the trajectory: TRACON entry and runway threshold crossing.Thousands of arrivals recorded over a five-month period in early 2010 enabled an analysis of observed wake vortex separation at KATL 27L and KDEN 35R (referred to as the baseline).A path compression model and its variant that models faster average airspeeds in the terminal area are utilized to examine the potential delay savings.Keys to the models are path and true airspeed distributions for aircraft flying on the same STAR to the same runway with the same engine type and operating in the same meteorological conditions.Selecting a percentile within the path and speed distribution produces two important parameters, the minimum path distance ( R P min ), and the maximum average true airspeed ( V TAS max ), respectively.Varying R P min and V TAS max results in a range of potential delay savings.From the results of the R P min and V TAS max sensitivity analysis (both path and path with speed compression models), we conclude the following:• Delay savings are more sensitive to R P min than V TAS max • More potential savings exist at KDEN 35R than KATL 27L• More potential savings in IMC than VMC This material is declared a work of the U.S. Government and is not subject to copyright protection in the United States.• At KATL 27L, median delay savings per flight varied from -5 to 33 seconds in VMC and 0 to 55 seconds in IMC • At KDEN 35R, median delay savings per flight varied from 0 to 80 seconds in VMC and 0 to 120 seconds in IMC Following the sensitivity analysis, the daily variations of delay and path savings were examined with the path compression model using 10 R P min .The reduction in excess spacing was also examined for those arrivals that required a 2.5 nmi separation using CDFs.This second half of the analysis found:• Uneven daily delay and path savings • At KATL 27L, 10% increase over the baseline in arrivals landing with an excess spacing of 0.5 nmi or less and 25% increase at KDEN 35RThis study found that potential benefits of scheduling and spacing automation tools can be estimated by a statistical traffic compression model that makes use of aircraft parameters captured at just two descrete trajectory locations.This method eliminates the need to reconstruct and examine the entire trajectory from the TRACON meter fix to the runway threshold for each aircraft.And, it can be used irrespective of how the excess leading/following aircraft wake-vortex separation is reduced.. 1 A1Downloaded by NASA AMES RESEARCH CENTER on August 14, 2013 | http://arc.aiaa.org| DOI: 10.2514/6.2013-4424 +Figure 1 .1Figure 1.Compression Model Input Generation Flow Diagram. +Figure 2 .Table 1 .21Figure 2. A80 TRACON Boundary Centered From KATL 27L. +( 17 )17Downloaded by NASA AMES RESEARCH CENTER on August 14, 2013 | http://arc.aiaa.org| DOI: 10.2514/6.2013-4424 +where the superscript S indicates speed compression model.This enables the calculation of a new runway threshold crossing time, new transit time and new delay time given below in Eqs.(27-29), respectively, +( 27 )27Downloaded by NASA AMES RESEARCH CENTER on August 14, 2013 | http://arc.aiaa.org| DOI: 10.2514/6.2013-4424 +Downloaded by NASA AMES RESEARCH CENTER on August 14, 2013 | http://arc.aiaa.org| DOI: 10.2514/6.2013-4424Thismaterial is declared a work of the U.S. Government and is not subject to copyright protection in the United States. +A.Figures 3 (above) and Fig. 4 (next page) show the effect of varying the "minimum" path distance and "maximum" average true airspeed threshold percentiles on the delay savings per flight at KATL 27L and KDEN 35R, respectively.Open circles denote the delay savings, Eq. (19), corresponding to the two R P min threshold values for the path compression model.Asterisks denote the delay savings, Eq. (31), corresponding to the four combinations of R P min and V TAS max threshold percentiles for the path and speed compression model.For the path compression model, no delay savings are achieved when defining the minimum path distance as the median path distance ( 50 R P min).With this definition, a significant number of aircraft were not allowed to lengthen their path, which would have resulted in negative delay savings.Although the average delay savings was not zero, this paper +Figure 3 .3Figure 3. Median Delay Time Saved, !T D , Per Flight In VMC And IMC at KATL 27L. +Figure 4 .4Figure 4. Median Delay Time Saved, !T D Per Flight In VMC And IMC at KDEN 35R. +Figure 5 .5Figure 5. Median Path Distance Saved, !R , Per Flight In VMC And IMC. +7. +Figure 6 .6Figure 6.Median Daily Variation of Delay Time Saved at KATL 27L in IMC and VMC. +Figure 7 .7Figure 7. Median Daily Variation of Path Distance Saved at KATL 27L in IMC and VMC. +Figure 8 .8Figure 8. Daily Variation of Number of Arrivals at KATL 27L in IMC and VMC. +Figure 9 .9Figure 9. Cumulative Distribution Function, KATL 27L, VMC. +Fig 9 and9Figure10shows the CDFs for KATL 27L for aircraft operating in IMC periods; it is similar to the VMC CDFs in Fig.9; however, there is no crossover point.For flights during IMC periods, there is a more substantial reduction in excess spacing that can be recovered -roughly 0.5 nmi.In other words, the observed excess in-trail separation present in the operations to KATL 27L is accompanied by enough corresponding excess path distance flown to allow its recovery.The percentage of arrivals landing with excess spacing of 0.5 nmi (3 nmi spacing) or less is 10% for the baseline and 22% for the path compression model; an increase of 12%, roughly the same as the arrivals under VMC periods.However, more arrivals during IMC periods benefit (i.e., land with less excess spacing in general) as a result of the path compression than during VMC.This observation is made when comparing the area between the baseline and path compression CDFs in Fig 9 and Fig 10.Now, after examining the CDFs for the 2.5 nmi reduced separation on final at KATL 27L, the CDFs for aircraft landing at KDEN 35R in VMC and IMC, are shown in Fig. 11 and Fig. 12, respectively. +Figure 10 .10Figure 10.Cumulative Distribution Function, KATL 27L, IMC. +Figure 11 .11Figure 11.Cumulative Distribution Function, KDEN 35R, VMC. +Figure 12 .12Figure 12.Cumulative Distribution Function, KDEN 35R, IMC. + This material is declared a work of the U.S. Government and is not subject to copyright protection in the United States. + (9) Downloaded by NASA AMES RESEARCH CENTER on August 14, 2013 | http://arc.aiaa.org| DOI: 10.2514/6.2013-4424Thismaterial is declared a work of the U.S. Government and is not subject to copyright protection in the United States. + Downloaded by NASA AMES RESEARCH CENTER on August 14, 2013 | http://arc.aiaa.org| DOI: 10.2514/6.2013-4424Thismaterial is declared a work of the U.S. Government and is not subject to copyright protection in the United States. + + + +AppendixSubstituting Eq. (18) into Eq.( 27) yieldsNow, substitute Eq. (A1) into Eq.( 29),T D new,S (i) = T D (i) !"T P (i) !"T S (i).(A2)Finally, substituting Eq. (A2) into Eq.(30),produces the expression given in Eq. (31). + + + + + + + Benefits of precision scheduling and spacing for arrival operations + + SZelinski + + 10.1109/dasc.2012.6382979 + + + + 2012 IEEE/AIAA 31st Digital Avionics Systems Conference (DASC) + Williamsburg, Virginia + + IEEE + October 14-18, 2012. 1 November 2012. June 2007 + 2 + + + Concept of Operations for the Next Generation Air Transportation System. Version 2.0 + Zelinski, S., "Benefits of Precision Scheduling and Spacing for Arrival Operations," 31 st Digital Avionics Systems Conference, Williamsburg, Virginia, October 14-18, 2012. 2 Federal Aviation Administration, Aerospace Forecast Fiscal Years 2012-2032 URL: www.faa.gov/about/office_org/headquarters_offices/apl/aviation_forecasts/aerospace_forecasts/2012-2032 [cited 1 November 2012]. 3 Joint Planning and Development Office, "Concept of Operations for the Next Generation Air Transportation System," Version 2.0, June 2007. + + + + + Design Considerations for a New Terminal Area Arrival Scheduler + + JaneThipphavong + + + DanielMulfinger + + + AlexanderSadovsky + + 10.2514/6.2010-9290 + + + 10th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Ft. Worth, Texas + + American Institute of Aeronautics and Astronautics + September 13-15, 2012 + + + Thipphavong, J., Mulfinger, D., "Design Considerations for a New Terminal Area Arrival Scheduler," 10 th AIAA Aviation Technology Integration and Operations Conference, Ft. Worth, Texas, September 13-15, 2012. + + + + + Design and Evaluation of the Terminal Precision Scheduling and Spacing System + + HNSwenson + + + June 14-17, 2011 + Berlin, Germany + + + USA/Europe Air Traffic Management Research and Development Seminar + + + Swenson, H. N., et al., "Design and Evaluation of the Terminal Precision Scheduling and Spacing System," 9 th USA/Europe Air Traffic Management Research and Development Seminar, Berlin, Germany, June 14-17, 2011. + + + + + Controller-managed spacing within mixed-equipage arrival operations involving flight-deck interval management + + JThipphavong + + 10.1201/b12321-14 + + + Advances in Human Aspects of Aviation + Chicago, Illinois + + CRC Press + June 10-13, 2013 + + + + 10 th USA/ + Thipphavong, J., et al., "Evaluation of the Controller-Managed Spacing Tools, Flight-deck Interval Management and Terminal Area Metering Capabilities for the ATM Technology Demonstration #1," 10 th USA/Europe Air Traffic Management Research and Development Seminar, Chicago, Illinois, June 10-13, 2013. + + + + + Benefits of Continuous Descent Operations in High-Density Terminal Airspace Considering Scheduling Constraints + + JohnRobinson Iii + + + MaryamKamgarpour + + 10.2514/6.2010-9115 + + + 10th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Fort Worth, Texas + + American Institute of Aeronautics and Astronautics + September 13-15, 2010 + + + Robinson III, J. E. and Kamgarpour, M., "Benefits of Continuous Descent Operations in High-Density Terminal Airspace Under Scheduling Constraints," 10 th AIAA Aviation Technology Integration and Operations Conference, Fort Worth, Texas, September 13-15, 2010. + + + + + The impact of advanced technology on Nex Generation Air Transportation: A case study of Required Navigation Performance (RNP) + + YoonjinYoon + + + MarkHansen + + 10.2514/6.2007-6697 + + + AIAA Modeling and Simulation Technologies Conference and Exhibit + Hilton Head, South Carolina + + American Institute of Aeronautics and Astronautics + August 20-23, 2007 + + + AIAA-2007-6697 + Yoon, Y. and Hansen, M., "The Impact of Advanced Technology on Next Generation Air Transportation: A Case Study of Required Navigation Performance (RNP)," AIAA-2007-6697, AIAA Modeling and Simulation Technologies Conference and Exhibit, Hilton Head, South Carolina, August 20-23, 2007. + + + + + A graph-based approach to nominal terminal routing + + SZelinski + + 10.1109/dasc.2012.6382333 + + + + 2012 IEEE/AIAA 31st Digital Avionics Systems Conference (DASC) + Williamsburg, Virginia + + IEEE + October 14-18, 2012. February 25. 2013. 1 February 2013 + 10 + + + Federal Aviation Administration. Order A80 7110.65E CHG2 + Zelinski, S., "A Graph-Based Approach to Defining Nominal Terminal Routing," 31 st Digital Avionics Systems Conference, Williamsburg, Virginia, October 14-18, 2012. 10 Federal Aviation Administration, "Standard Operating Procedures", Order A80 7110.65E CHG2, February 25, 2013 URL: http://www.faa.gov [cited 1 February 2013]. + + + + + Standard Operating Procedures (SOP) + 10.1002/9780471462422.eoct523 + + + Control Positions of Operation and Standard Operating Procedures (SOP) + + John Wiley & Sons, Inc. + + + + Order D01 + 11 Federal Aviation Administration, "Control Positions of Operation and Standard Operating Procedures (SOP)", Order D01 + + + + + + + H + + + + April 12. 2012. 1 February 2013 + + + 1H, April 12, 2012 URL: http://www.faa.gov [cited 1 February 2013]. + + + + + Design and Operational Evaluation of the Traffic Management Advisor at the Fort Worth Air Route Traffic Control Center + + HNSwenson + + JO 7110.65U CHG 2 + + + + Europe Air Traffic Management Research & Development Seminar + + June 16-19, 1997. March 7. 2013. 1 February 2013 + Sarclay, France + + + Air Traffic Organization Control + 12 Swenson, H. N., et al., "Design and Operational Evaluation of the Traffic Management Advisor at the Fort Worth Air Route Traffic Control Center," 1 st USA/Europe Air Traffic Management Research & Development Seminar, Sarclay, France, June 16- 19, 1997. 13 Federal Aviation Administration, "Air Traffic Organization Control", Order JO 7110.65U CHG 2, March 7, 2013 URL: http://www.faa.gov [cited 1 February 2013]. + + + + + U.S. Airport Infrastructure and Sustainable Aviation Fuel + + KristiMoriarty + + + AllisonKvien + + 10.2172/1768316 + + + + 14 Federal Aviation Administration, 2011-2015 National Plan of Integrated Airport Systems Report + + Office of Scientific and Technical Information (OSTI) + + + + cited 28 October 2012 + 14 Federal Aviation Administration, 2011-2015 National Plan of Integrated Airport Systems Report, URL: http://aspm.faa.gov [cited 28 October 2012]. + + + + + + diff --git a/file786.txt b/file786.txt new file mode 100644 index 0000000000000000000000000000000000000000..eb267a6e4e554801505bfb72232d02f554e2bb5d --- /dev/null +++ b/file786.txt @@ -0,0 +1,974 @@ + + + + +IntroductionFuture air traffic demand is expected to increase throughout the U.S. National Airspace System [1].To meet this expected increased demand, the FAA, in collaboration with NASA and industry partners, is transforming the airspace system with the Next Generation Air Transportation System (NextGen) [2].NextGen is expected to increase capacity, and improve efficiency and safety.Because NextGen addresses all phases in the air traffic system from the departure, through the climb, cruise, descent and arrival phases of flight, the capabilities are broad.They include, for example, airport surface management improvements [3], automatic dependent surveillance-broadcast (ADS-B) [4], performancebased navigation (PBN) [5], and modernizing ground-based automation systems [6,7]. +NASA's Air Traffic Management TechnologyDemonstration -1 (ATD-1) [8] is one such FAA/NASA/industry partner collaboration.ATD-1 technologies will be transferred to the FAA upon achieving the appropriate technology maturity.At NASA, the maturity of a technology is measured with technology readiness levels (TRLs) [9].Generally, NASA technology is transferable after achieving TRL 6.For the ATD-1 technologies, currently at TRL 5, this transfer represents the transition from laboratory test environments to an operational environment [10].To aid the FAA in its near-term investment decisions, a series of preliminary technology transfers began in October 2013 [11].A one-time, NASA-FAA simulation, called the Operational Integration Assessment (OIA), at the FAA's William J. Hughes Technical Center (WJHTC), is planned to start in 2015.ATD-1 integrates three separate NASA developed technologies that are expected to improve operations in the terminal airspace.These NASA technologies include: (1) an enhanced version of the FAA's Time-Based Flow Management (TBFM) [7,12,13], (2) a ground-based automation technology known as controller-managed spacing (CMS) [14][15][16], and (3) an advanced avionics airborne technology known as flight-deck intervalmanagement (FIM) [17][18][19].ATD-1 technologies have been extensively tested in large-scale HITL simulations [20][21][22][23][24][25][26][27] using general-purpose workstations, collectively referred to as the Multi-Aircraft Control System (MACS) [28,29], to study air transportation technologies.MACS performs multiple functions, including emulating the FAA's Standard Terminal Automation Replacement System (STARS) [30] radar display.In order to be operationally viable, ATD-1 technologies required extending the capabilities of the Raytheon-developed STARS platform to display certain ATD-1 technologies (exactly which technologies will be discussed later) to the terminal radar approach control (TRACON) controllers (hereafter referred to as terminal controllers) on terminal controller workstations (TCWs).Beginning in 2012, researchers at NASA Ames Research Center (NASA Ames) and Raytheon collaborated to augment the STARS platform by including CMS and FIM advisory tools to validate the feasibility of integrating these automation enhancements into the current FAA automation infrastructure.In the spring of 2013, NASA Ames acquired three STARS TCWs, and then integrated the ATD-1 technologies.CMS algorithms were added to the NASA enhanced version of the FAA's TBFM, and the advisories displayed on the STARS TCWs.The objective of this paper is to describe the validation of the ATD-1/STARS integration.HITL simulations were conducted in the summer of 2013 to evaluate the performance and acceptability of the integrated ATD-1 technologies within the STARS platform.These results were compared with the results obtained when the ATD-1 technologies were tested using MACS to emulate the STARS radar displays for the terminal controllers.The integration of the ATD-1 technologies within the STARS architecture represented a major technology maturation milestone and is a necessary and critical step prior to operational testing at WJHTC.This paper is organized as follows: an overview of the ATD-1 concept of operations and technologies is provided next, followed by an overview of the STARS platform.Then, the various components of the simulation are summarized, and the results discussed.The paper ends with concluding remarks and planned future work. +ATD-1 Integrated Arrival Scheduling and Spacing ConceptThe ATD-1 portfolio includes three distinct technologies that provide an integrated arrival concept [31] for scheduling, sequencing, and spacing.The first technology leverages the FAA's TBFM [7], the successor to the Traffic Management Advisor (TMA) [12,13] and extends it to include terminal metering (TMA-TM) [32] for conflict-free schedules to the runway and TRACON metering points.The second technology, CMS, provides a set of decision support tools for terminal controllers to better manage aircraft delay using speed control.Airborne relative spacing is achieved with the third ATD-1 technology, FIM.The integration of these three technologies enables an integrated arrival and spacing system with the following concept of operations.Beginning in air route traffic control center (ARTCC) airspace, prior to an aircraft's top-of-descent (TOD) and about 200 nautical miles (NM) from the runway, four dimensional (4D) trajectory predictions, so-called due to 3D position and time, determine the aircraft's arrival sequence and conflict-free scheduled times-ofarrival (STAs) at the TRACON boundary (usually the meter fix), meter points within the TRACON and the runway.The arrival sequence and STAs (referred to as the schedule) are frozen at about 130 NM from the meter fix and displayed to the ARTCC controllers (hereafter referred to as center controllers).The center controllers employ various tactical control strategies (e.g.speed and path assignments) to deliver the aircraft to the meter fix at or near its meter fix STA.Aircraft equipped with FIM avionics are issued a voice FIM clearance by the center controllers and begin automatically spacing (via speed control) behind a designated lead aircraft.Center controllers transfer responsibility of (hand-off) the aircraft to the terminal controllers prior to the meter fix.Terminal controllers make use of the CMS advisories and issue speed clearances to non-FIM equipped aircraft as required to adjust for any minor perturbations.TMA-TM and CMS are ground-based automation tools and when integrated together as a system have been evaluated in several HITL simulations and shown to have benefits in terms of increased throughput, decreased controller workload, and improved PBN operations [20][21][22][23][24][25][26][27].The TMA-TM/CMS system is commonly called Terminal Area Precision Scheduling and Spacing (TAPSS) [21][22][23][24][25] in NASA parlance, and Terminal Sequencing and Spacing (TSS) [26] within the FAA; however, TAPSS and TSS are essentially the same system.Figure 1 qualitatively shows the benefits of TSS by comparing the ground tracks for PHX arrivals in a west-flow configuration with and without TSS available to the terminal controllers.Immediately apparent is the reduction in path distance flown due to the reduction of radar vectors and shortened downwind legs enabled by TSS.Each of the ATD-1 technologies will now be described in detail. +Traffic Management Advisor with Terminal Metering (TMA-TM)TMA-TM is an arrival management automation system that leverages the FAA's TBFM, an improved version of the TMA, and extends it to include terminal metering for conflict-free schedules to the runway and metering points within the TRACON.TBFM is currently a scheduling tool used at ARTCCs whereas TMA-TM is an advanced prototype integrated into the FAA's TBFM release 3.12 (July 2011).For each aircraft, TMA-TM generates an estimated time-of-arrival (ETA) and a STA at all metering points, including the TRACON meter points.The ETA is the time that the aircraft would arrive at a certain location (e.g.meter fix) without considering separation requirements of other arrivals.The STA is the conflict-free arrival time at a certain location (e.g.meter fix).The time offset between ETA and STA is referred to as delay.TM enhancements include: (1) accurately representing area navigation (RNAV) and required navigation performance (RNP) routes by including additional custom waypoints and turn radius parameters (for RNP) that are tailored to the terminal airspace and arrival route structure, connecting the runway to the published standard terminal arrival route resulting in a single, continuous trajectory from ARTCC airspace to the runway threshold, (2) ensuring the 4D trajectory predictions make use of the published and standard operating procedure altitude and speed restrictions along the routes in the terminal area and adding operationally feasible altitude and speed constraints where required, and (3) a delay allocation strategy which, working backwards, first de-conflicts aircraft at the runway threshold, then TRACON meter points, followed by meter fixes using speedcontrol only.This delay allocation strategy assures that the STAs can be met with speed reductions, not speed increases.Depending on the airspace topology and aircraft type, aircraft can absorb up to about two minutes of delay within the terminal area using only speed-control; any remaining delay will not be taken as TRACON vectors, but rather will need to be absorbed by the ARTCC. +Controller-Managed Spacing (CMS) Advisory Tools +Flight-deck Interval Management (FIM)FIM is an advanced avionics capability that is designed to achieve a precise spacing goal by a certain point along the arrival route (usually the final approach fix (FAF)) between two aircraft landing on the same runway.FIM-equipped aircraft require the FIM avionics and the capability to receive ADS-B data from other aircraft within range.In accordance with the concept of operations, FIM-equipped aircraft are issued a FIM clearance by the center controller.This clearance information includes the required spacing goal, the target aircraft's planned route of flight and the target aircraft identifier.This clearance is provided by voice, and, in a laboratory environment, is entered by the pilot into Aircraft Simulation for Traffic Operation Research (ASTOR) glass cockpit display FIM emulators.Aircraft that are actively engaged in spacing should require little to no terminal controller intervention.The CMS advisory tools provide a mechanism for the terminal controllers to effectively monitor FIM operations. +Multi-Aircraft Control System (MACS)MACS is real-time air traffic control (ATC) simulation platform.Although it is not an ATD-1 technology itself, it is critical to the integration and evaluation of the ATD-1 technologies used in HITL simulations.MACS simulates long and short-range surveillance radar and provides realistic ARTCC and TRACON radar display emulators for ATC.For non-FIM aircraft, cockpit displays are emulated through desktop display monitors for the pilot simulation participants.Simulation manager user interfaces are provided for monitoring and control of the simulation.A MACS-based simulation begins with the generation of flight plans for each aircraft; these flight-plans use the aircraft's initial location and scenario start time as the coordination fix and time, respectively, as well as the appropriate flight routing, altitude, and speed. +Standard Terminal Automation Replacement System (STARS)Raytheon-developed STARS replaces older, outdated, hardware/software in the TRACON facilities and is currently being deployed through the FAA's TAMR program [6].Controllers interface with STARS through the TCWs.STARS offers several enhancements over the older systems: a key one being the improved capability of receiving and fusing tracks from multiple short-and long-range radars, and ADS-B, into a single, smooth one-second aircraft track update that is displayed to the controllers.STARS has two variants tailored to the size of the facility.A Local Integrated Terminal Equipment system (STARS-LITE) is intended for control towers without a TRACON.The second variant, Enhanced LITE (STARS-ELITE), is planned to be installed at small and medium size TRACON facilities.STARS-ELITE offers much of the same functionality and associated software as STARS, but with a smaller hardware footprint. +NASA Ames Research Center's Prototype Enhanced LITE STARS (STARS-ELITE)Beginning in 2012, engineers at NASA Ames and Raytheon collaborated on extending the STARS capabilities to display the CMS advisory tools on the TCWs.Early basic prototypes were designed and tested at Raytheon's Mt.Laurel, New Jersey facility.In the spring of 2013, the ATC laboratory at NASA Ames acquired a STARS-ELITE that consisted of three TCWs for terminal controllers and the software development environment (SDE) for STARS adaptation and software development.At the present time, the ATC lab is one of four sites with the SDE (the other three are Raytheon's Marlborough, MA and Mt.Laurel, NJ sites and the FAA's WJHTC) and one of three sites with the STARS-ELITE (the other two sites are McGuire Air Force Base, NJ and the FAA's WJHTC).To facilitate HITL simulations in the ATC lab, all the relevant systems were integrated: TMA-TM (scheduling and spacing), three ASTOR stations (FIM operations), CMS (advisory tools for terminal controller), three STARS-ELITE TCWs (displays CMS tools), and MACS (radar simulator, additional controller/pilot display emulators and traffic generator).Several system functional tests culminated in a week of HITL simulations.This HITL simulation used a prototype STARS-ELITE that had not yet implemented the assigned runways and sequence numbers of the CMS advisory toolset.Hereafter, for simplicity, we dispense with the distinction between STARS and STARS-ELITE, and just use the term STARS. +Human-in-the-loop (HITL) Simulation OverviewNineteen one-hour HITL simulations (hereafter referred to as simulation runs or just runs for short) were conducted over the course of one week.The objective was to validate the performance and acceptability of using an advanced prototype operational ATC system (STARS) to display CMS advisories.FIM clearances were not issued during the runs because the newest speed control law algorithm, Airborne Spacing for Terminal Arrival Routes (ASTAR12) [33], was still in development.All previous ATD-1 HITL simulations had used the MACS platform to emulate the STARS display scope, thus the approach used in the validation (described in more detail in the Validation Strategy section) was to compare and contrast performance of the newly integrated STARS with the MACS STARS emulation. +AirspaceAlbuquerque ARTCC (ZAB) and Phoenix TRACON (P50) are the primary facilities responsible for controlling aircraft arriving at PHX.The ATC laboratory at NASA Ames was configured to model simplified ZAB and P50 airspace.ZAB airspace was simplified by combining high and low altitude of the primary arrival sectors for four arrival directions.P50 airspace was modeled as a primary feeder and final sector: two sectors for the south and two for the north.Most aircraft entered P50 through one of the four arrival procedures: EAGUL5 and MAIER5 for the aircraft entering from the north and GEELA6 and KOOLY4 for aircraft entering from the south.PHX was configured for East Operations, with aircraft landing on runways 07R and 08.All simulation runs, with one exception (discussed later), conducted independent arrival runway operations with aircraft flying instrument flight rules (IFR) flight plans; thus, terminal controllers were responsible for separation and clearances to the runway.Due to the predominance of good weather in PHX, operations are primarily visual flight rules (VFR) for independent runway operations and the pilots are responsible for separation once they have been cleared for approach and have their lead aircraft in sight; however, this is not practical in a laboratory.Reference [34] provides the PHX airport arrival rate (AAR) guidelines under various operational conditions (i.e.airport weather condition and runway configuration).The AAR for an eastflow configuration using VFR for two runways is 74 aircraft per hour.In actual PHX operations, aircraft flying IFR landing on two (dependent) runways are required to use staggered instrument landing system (ILS) approaches, reducing the AAR to 48-52 aircraft per hour.The last simulation run conducted staggered ILS approaches.Large and heavy jets flew on the published standard RNAV routes; turboprops and small jets aircraft flew on non-RNAV arrival routes.Nominally, aircraft from the north are assigned to the northern most runway and aircraft from the south are assigned to the southern most runway.TMA-TM employs a runway-balancing algorithm to minimize system delay, potentially assigning some aircraft from the north to runway 07R (the southern runway) and arriving aircraft from the south to runway 08 (the northern runway).Most of these "crossover" aircraft will need to cross over the airport (hence the term "crossover") to land on their assigned runways; therefore, crossover routes were also designed with the assistance of P50 subject matter experts for RNAV and non-RNAV routes.The crossover routes were altitude separated (1,000 feet) to safely accommodate simultaneous north-to-south and southto-north crossovers.Figure 3 shows a schematic of the P50 airspace model just described. +Figure 3. Modeled P50 Airspace-East Arrival ConfigurationThe RNAV route from the north, known as the MAIER5, is connected to the final approach course.The RNAV route from the northeast, called the EAGUL5, ends at a fix called BASBL; thus, ATC issues radar vectors for those EAGUL5 aircraft traveling on the downwind leg to turn to base well before BASBL.The RNAV route in the southwest, called GEELA6, is also connected to the final approach course.Aircraft flying on the KOOLY4 RNAV arrival from the southeast, similar to the EAGUL5, also require radar vectors to turn the aircraft from downwind to base well before NEELE.Aircraft equipped with avionics that monitor RNAV performance, known as RNP, were not included in this HITL simulation to limit the experimental complexity, but TSS does support RNP procedures and capabilities.Black dashed lines represent the RNAV crossover routes with an arrow indicating the direction of travel.For clarity, non-RNAV routes for turboprop and non-RNAV equipped aircraft are not shown.Luke Air Force Base operates in the northwest region (to the west of BASBL) and its airspace is restricted, resulting in the irregular shape of the TRACON seen in the northwest of P50 between GEELA6 and MAIER5.Call-out symbols (e.g.Freeway, Quartz, etc.) denote the airspace sector name.The four meter fixes (GEELA, BRUSR, HOMRR, and SQUEZ) are also shown. +ScenariosTwo different traffic scenarios were simulated.The first scenario, referred to as east-flow #1 (EF1), was a snapshot of a 70-minute arrival rush period into PHX during 28 December 2011 simulating 50 arrivals including one turboprop aircraft.The second scenario was a higher demand scenario simulating 54 arrivals, six of which were turboprops, in a 60-minute period that is referred to as east-flow #2 (EF2).Four sets of winds were selected from a recent P50 wind analysis [35] that made use of nearly 8,700 one-hour, 40-km grid rapid update cycle (RUC) forecasts from 2011 [36].The analysis binned the wind patterns according to their effect on flight time changes for turbojets on the four standard terminal arrival/RNAV routes and the magnitude of the headwinds on final approach.For simplicity, these winds are referred to 'A', 'B', 'C', and 'D' winds.EF1 made use of the 'C' and 'D' winds, whereas EF2 alternated between the 'A' and 'B' winds.The result was two traffic scenarios each with two different wind topologies.The wind field that the aircraft simulated flying through (the environmental winds) was a variant of the RUC forecast winds used by the scheduler (i.e.TMA-TM) to model the realistic wind speed and direction differences (within expected operational limits) between the forecast and actual winds. +ParticipantsThirteen pilots and eight controllers staffed 21 positions.Three of the thirteen pilot positions were ASTOR pilot workstations, each simulating a B757 aircraft.The other ten pilots staffed workstations using MACS generic cockpit display emulators; each pilot was responsible for entering heading/speed commands, as instructed by ATC, for several aircraft.Four retired center controllers, with an average of 26 years of experience, staffed the four ZAB airspace sectors.Two of the center controllers were recently retired ZAB controllers.Four retired terminal controllers, averaging 30 years of experience, staffed the four TRACON positions that were configured as two feeder and two final positions.The QUARTZ/VERDE feeder/final pair controlled air traffic in the southern region of P50 and the APACHE/FREEWAY feeder/final pair was responsible for the northern half of P50 (see Fig. 3).All four terminal controllers who participated in the data collection HITL simulations were retired from the Southern California TRACON.Two of them were last-minute replacements for retired P50 controllers that were expected to, but could not, participate in the data collection HITL simulations.Three of the four terminal controllers had familiarity with CMS advisory tools, P50 airspace, and STARS from past HITL simulations.The fourth controller was new to CMS advisory tools, P50 airspace, and STARS; he received limited training during the first couple of simulation runs, and was considered reasonably trained by his peer controllers for this operation. +Controller and Pilot ProceduresControllers and pilots used headsets with a builtin microphone to communicate on designated frequencies.Center controllers' responsibilities included delivering the aircraft to the meter fix at or near its STA, keeping the aircraft safely separated, and issuing the expected runway assignments provided by TMA-TM.Center controllers issued optimized profile descent clearances (the phraseology is "DESCEND VIA") for the turbojet aircraft that use the standard terminal arrival/RNAV routes.For turboprops, the center controllers issued routing, speeds, and altitude.Prior to the aircraft crossing the TRACON boundary, the center controllers handedoff aircraft, via keyboard entries into the MACS ARTCC display system replacement desktop emulators, to the south or north terminal feeder controller.The CMS advisory tools were available for the terminal controllers to efficiently manage the traffic from center hand-off to the runway threshold.Additions made to the flight data-block (FDB) included displaying the CMS speed advisory or an early/late indicator in seconds.Feeder controllers made use of the timelines to strategically assess aircraft sequence information because the sequence number algorithm was not yet implemented in the STARS prototype.To avoid clutter, timelines were not displayed on the final controllers scope.The feeder controllers used the P50 general guidelines that aircraft arriving from the north would land on the north runway (08) and aircraft in the south would land on the south runway (07R).Aircraft that required crossing over the top of the airport (crossovers) were identified after the schedule was frozen and pointed out to the feeder controllers for proper coordination by a runway coordinator position.Feeder controllers handed-off aircraft to the final controllers prior to aircraft entering the final controller's designated airspace.The final controllers were responsible for safely merging the aircraft for the final approach to the runway.They issued heading clearances to the pilots to turn the aircraft from the downwind leg onto base and then from base to final.The pilot participants entered the ATC clearances into the graphical user interfaces on the desktop displays and verified the routing and assigned runway.Pilot participants at MACS stations were each responsible for multiple aircraft whereas pilots at ASTOR stations were each responsible for just a single aircraft. +Validation StrategyTo validate the performance and acceptability of CMS advisory tools displayed on STARS, roughly half of the simulation runs made use of the STARS and the other half used MACS for the TRACON radar display, enabling straightforward performance and acceptability comparisons.Terminal controller participants have used MACS in many past HITL simulations, so its performance and acceptability are known and documented [23][24][25][26][27][28][29][30].The three STARS TCWs were configured as QUARTZ (south feeder), VERDE (south final), and FREEWAY (north final).These TCWs were physically located side-by-side in the ATC laboratory in that order, from left to right.The fourth TRACON position, APACHE (north feeder), was a MACS station and was located to the right of FREEWAY.Located to the right of APACHE were three more MACS stations configured as (in this order): FREEWAY, VERDE, and QUARTZ.The terminal controller working APACHE sat at the same physical location for every simulation run.The other three terminal controllers each were responsible for the same separate airspace for each run, but sat at different physical locations depending on which TRACON radar scope was utilized (STARS or MACS).Prior to the first simulation run, information displayed to the terminal controllers on the MACS screens was customized to exactly match the displays on the STARS (e.g. the aircraft's FDB).Each station's display settings, including the CMS advisories, remained fixed throughout all of the simulation runs. +Test ConditionsTo keep the participants stimulated, traffic and winds were varied between simulation runs.For terminal controllers, the display scope (i.e. the station) was varied as part of the validation strategy.Each traffic scenario was paired with two different wind fields, resulting in four unique simulation runs on each TRACON display scope and repeated twice, resulting in eight simulation runs per station, as summarized in Table 1. +Data CollectionA total of 19 runs were conducted over five days.The test conditions were varied in such a way that ensured that no two successive simulation runs had the same test conditions, and, that over the course of five days, the test conditions remained balanced.The four scenarios per station shown in Table 1 were repeated accounting for the first 16 runs (4 scenarios x 2 stations x 2 runs).The last three runs used STARS exclusively for additional experimentation and were not critical to the validation of STARS displaying CMS advisory tools, but are included in the results.Two of the last three simulation runs performed advanced STARS prototyping with a declutter mode feature, so-called because it turned off the CMS advisories for the last five minutes of flight (about the last 10 NM).The last run simulated IFR using staggered ILS approaches.Output log files and voice recordings were captured for all 19 runs.Terminal controllers filled out questionnaires after each run and a separate end-of-simulation questionnaire at the end of week.As previously discussed in the HITL Simulation Overview section, FIM operations were not conducted; therefore, the results do not include FIM-specific metrics. +ResultsThe results compare performance and terminal controller feedback as a function of TRACON display scope (STARS versus MACS).Because one terminal controller was always stationed at a MACS position (APACHE), this participant's responses to the questionnaires are excluded in the controller feedback section. +MACS and STARS Performance ComparisonsData generated from simulation runs when three of the four terminal controllers used STARS stations are compared to data from when the same three controllers used MACS stations.Key metrics are compared for each simulation run and in the aggregate; the following plots display the aggregate results.The success rate for PBN measures the percentage of RNAV arrival procedures.A RNAV arrival procedure is considered uninterrupted if no radar heading vectors are given by ATC between the meter fix and the end of the RNAV procedure (turn to base for aircraft on EAGUL5 and KOOLY4 routes).The success rate ranged from 91 to 100%.The average PBN success rate based on terminal workstation (STARS or MACS) utilized during a particular run is shown in Fig. 4. +Figure 4. Average PBN Success RateBoth averages are high-above 95%.The average for the MACS stations is 2% greater than the STARS, and is minor.Whereas PBN success rate focuses on radar heading vectors issued to RNAV arrivals, ATC issues other types of clearances as needed such as speed and altitude.The average number of these three clearances (heading, speed, and altitude) for all flights (RNAV and non-RNAV) ranged from four to six clearances inside the terminal area.Figure 5 compares the average number of clearances for STARS and MACS and shows roughly the same average number of clearances (five) were required. +Figure 5. Average Number of Clearances Issued in TRACONDetermining the time flown below 10,000 feet is used as a way to measure operational efficiency (the less time spent flying below 10,000 feet-the more operationally efficient).This metric is an indirect measure of fuel efficiency.Direct fuel-usage calculations across multiple types of aircraft are not currently possible.The average time flown below 10,000 feet ranged from 630 to 715 seconds.The STARS runs results averaged to 654 seconds.This is 4% less (26 seconds) than the average time flown below 10,000 feet when using the MACS stations, as seen in Fig. 6.No obvious reason accounts for this 4% difference; it is considered acceptable. +Figure 6. Average Time Below 10,000 feetThe ground automation tools, collectively referred to as TSS, are expected to reduce excess spacing at the runway.The standard deviation of excess spacing is an indicator of the precision provided by the TSS system.Here, the standard deviation is measured at the final approach fix, instead of the runway threshold due to an earlier expected time-out of arriving flights when using STARS.This time-out prevented some flights from being logged as landed in the output data files.A change to STARS adaptation corrected this issue, but was made after the data collection simulation runs.The standard deviation ranged from 0.38 NM to 1.19 NM for the first 18 simulation runs with independent runway arrivals.Run 19 simulated staggered arrivals using STARS stations, which resulted in a standard deviation of 1.43 NM due to the additional stagger separation criteria for dependent runway approaches.Figure 7 shows the average standard deviation of spacing precision achieved when the terminal controllers worked at MACS stations and at STARS stations (including run 19).Because the spacing was measured at the FAF and not the runway threshold, as discussed above, the values shown in Fig. 7 include about an additional 0.5 NM of spacing due to compression that typically occurs between the FAF and the runway threshold.The average standard deviation is 7% less for STARS than MACS.The required spacing will be larger for staggered approaches; when excluding the run simulating staggered approaches, the standard deviation of spacing is 15% less for STARS than MACS.No cause for this difference is apparent, and is still under investigation. +Controller FeedbackOf the 19 runs in the study that tested the two station conditions, one run (the last run-run 19), was omitted from the data because it simulated different runway operations (staggered approaches) than the other 18 without tailoring the traffic scenario to account for the increased workload due to staggered approaches.There were 33 questions in the end-of-run survey asking about workload, acceptability, and utility of the stations and tools, resulting in a total of 72 sets of responses (4 x 18).Most questions asked for responses on a rating scale of 1 to 7 (usually very low to very high); some questions offered a choice of answers and some were free-response.Participants rated their level of workload using the NASA Task Load Index (TLX) [37], containing six subscales, each using a 1 to 7 scale ("very low" to "very high"-quotation marks indicate the choices available from the questionnaires).Participants did not use the full rating scale; workload was never rated a 7 ("very high").Standard deviations were between 1 and 1.5 for each subscale.The highest rating selected for the mental demand, effort, and frustration scales was 6 ("high load/activity"), which were selected once after a simulation run using STARS and once after a run using MACS.The highest rating selected for time pressure and physical demand was 5 ("reasonably high/activity").Because one of the sectors (APACHE) was not represented on a STARS station, its ratings were taken out into a separate category so those participants that worked on STARS and MACS stations were directly compared.Figure 9 shows the mean ratings for the six NASA TLX subscales.The performance subscale rating is reversed (rev) so that a lower rating is better than higher rating. +Figure 9. NASA Task Load Index (TLX)There is a significant difference between the participants' ratings of their mental demand (Z=-1.967,n=27, p=0.049) 1 , indicating that the three controllers who worked on MACS and STARS stations thought that using the MACS station demanded more mental activity than the STARS station.The other subscales are not significantly different when compared.Participants rated how acceptable they thought their flight operations were through the Controller Acceptance Rating Scale (CARS) [38], where they were guided through a set of yes/no questions to a 3point rating scale that was mapped to a larger scale.Overall they judged the operations on their safety and performance level.Combining the separate 3-point ratings leads to a single overall rating scaled from 1 to 10 for each participant.Controllers completed the CARS four times (after the last run of each day for the first four days).Controllers' mean acceptability ratings when using the MACS and STARS station are identical (mean=8.83)as shown in Figure 10.This indicates that they felt the system was acceptable and required little intervention from them to make it work. +Figure 10. Controller Acceptance Rating Scale (CARS)The lowest single rating offered was a 7 ("moderate intervention to maintain adequate system performance"), so all participants rated operations using both stations as "safe" and "controllable", and 93% of the time rated it as "adequate".Participants were asked to indicate when they had used the CMS advisory tools in a run.They were asked to note their usage of seven of the tools available to them (aircraft ground speed (GS) is available to controllers in current operations-it is not part of the CMS tools, but is included for completeness and comparison purposes).Only tool usage from participants who worked both the MACS and STARS stations is compared as shown in Fig. 11. +Figure 11. Tool UsageThe slot marker IAS was the most popular tool, used 98% of the time overall.The timeline was the least used tool as participants said they used it only 9% of the time.The low usage of the timeline, which displays strategic planning information temporally, is due to it requiring more experience to understand than was allotted for the replacement controllers.The slot marker IAS, aircraft IAS and speed advisory tools were reported to be used more often with the MACS station, whereas the slot markers, aircraft GS and timeline tools were used more often on the STARS station.However, these differences are slight-generally one usage rating different, except for the timeline, which was used four times more often in the STARS condition.Tool usage grouped into three categories results in: (1) those that were used often by everyone, (2) those that were used often by some participants, but not others, and (3) those that were used infrequently.The slot markers and slot marker IAS fall into the first category; everyone often used them.The aircraft GS and IAS were often used by two of the three controllers (but different pairs).Only one controller often used the early/late indicator and one the speed advisory.The timeline was not used by two of the controllers, and used infrequently by the third controller.The most experienced controller with the timeline was the one controlling the APACHE sector using a MACS station, but this controller's results were excluded because this controller did not control air traffic at a STARS station.Therefore, the tool usage patterns reflect different controller strategies and also different air traffic demands in the four TRACON sectors. +Concluding RemarksSeveral technologies expected to support ATC operations in the TRACON facilities prior to the end of this decade have been integrated and tested as a system.The Standard Terminal Automation Replacement System (STARS), developed by Raytheon, is the ground automation system that is replacing older technology in the TRACON facilities; it includes the terminal controller workstation (TCW).Engineers from Raytheon and NASA Ames collaborated to develop an enhanced prototype STARS that displays controller-managed spacing (CMS) advisory tools.The CMS algorithms were implemented in an advanced version of the FAA's TBFM release 3.12 that extends metering inside the terminal area.This advanced TBFM release 3.12 with terminal metering is called TMA-TM.The CMS advisory tools and TMA-TM are NASA ATD-1 products, and when integrated are known as the Terminal Sequencing and Spacing (TSS) system.The Operational Integration Assessment (OIA) at the FAA's WJHTC is expected to begin in 2015 to support the operational deployment of TBFM work package 3, potentially including TSS, in the 2015-2019 timeframe.In order to mature the technologies in preparation for testing at the WJHTC, retired controllers and pseudo-pilots participated in simulation runs over a one-week period at NASA Ames Air Traffic Control (ATC) laboratory, equipped with three STARS TCWs.To evaluate the performance and acceptability of the TSS system using STARS TCWs, three terminal controllers, each responsible for a sector of Phoenix TRACON (P50) airspace, performed their ATC duties using the Multi-Aircraft Control System (MACS) TRACON display emulators for half of the simulations and the STARS TCWs for the other half of the simulation runs.MACS has been used in past HITL simulations and its performance and acceptability by the participants has been established.Results compared controller performance and acceptability when performing ATC duties at STARS stations versus MACS stations.Analysis of the numerical data (i.e.non-subjective), showed results with similar or with acceptably minor differences for such metrics as RNAV conformance rate, average number of clearances issued in the TRACON, standard deviation of spacing at the final approach fix, time below 10,000 feet (an indirect measure of fuel efficiency), and maximum throughput when the STARS stations were utilized.Feedback from controller questionnaires showed a trend that the participants preferred the STARS stations.Although all aspects of workload, measured by NASA Task Load Indicator (TLX), were "low" using MACS and STARS stations, participants consistently reported lower workload when they were using the STARS station.Controllers' mean acceptability ratings, using the Controller Acceptance Rating Scale (CARS), were identical on STARS and MACS stations ("minimal controller compensation required to reach desired performance").The results from this HITL simulation validate the ATD-1/STARS integration and are an important milestone in transitioning the ATD-1 ground-based automation system, TSS, from TRL 5 to 6. +Future WorkSince the time of this research, HITL simulations testing the newest flight-deck interval management (FIM) algorithm, ASTAR12, have been conducted displaying the FIM advisories on a newer version of the prototype STARS.Runway assignments and sequence numbers have subsequently been implemented in the prototype STARS, and are displayed in the aircraft's FDB.A HITL simulation is planned for the fall of 2014 using a more recent version of TBFM (release 4.2) to demonstrate that the CMS algorithms will be compatible with the newer TBFM scheduling capabilities that were not present in release 3.12.After conducting the HITL simulation in the fall of 2014, the TSS technologies will be at TRL 6 and will be transferred to the FAA for further operational testing at the WJHTC beginning in early 2015.Figure 1 .1Figure 1.Ground Tracks Without TSS (Left) and With TSS (Right) [Star symbols indicate meter fixes] +Figure 2 .2Figure 2. Controller-Managed Spacing Advisory Tools indicators are provided.Timelines provide the controllers a way of determining the arrival sequence and overall demand.Figure 2 displays the CMS advisory tools. +Figure 22displays the CMS advisory tools. +Figure 7 .7Figure 7. Standard Deviation of Spacing at FAFReduced excess spacing potentially increases the AAR.The AAR is a dynamic parameter that captures the number of landings during any consecutive 60minute interval, and is referred to as throughput.The maximum throughput calculated for simulation runs 1 to 18 (two independent runway arrivals) varied from 83 to 108 aircraft per hour.Two dependent runway arrivals, simulated in run 19, achieved a maximum throughput of 71 aircraft per hour-36% greater than PHX AAR guidelines.Figure8compares the average maximum throughput achieved.When the terminal controllers worked at the STARS stations, the system averaged a maximum of 88 aircraft per hour, whereas the system averaged a maximum of 90 aircraft per hour when the MACS stations were utilized. +Figure 8 compares the average maximum throughput achieved.When the terminal controllers worked at the STARS stations, the system averaged a maximum of 88 aircraft per hour, whereas the system averaged a maximum of 90 aircraft per hour when the MACS stations were utilized. +Figure 8 .8Figure 8. Maximum Throughput +Table 1 . Test Matrix TRACON Scope1Number ofMinimumScenariosNumber(Traffic_Windof RunsType)STARS4 (EF1_C, EF1_D,8EF2_A, EF2_B)MACS4 (EF1_C, EF1_D,8EF2_A, EF2_B) + Z is the standardized test score, n is the number of samples, p is the null hypothesis result.The conventional interpretation here is that there is less than 1 chance in + (since p < 5%) that the difference is due to mere chance. + + + + +AcknowledgementsThe authors gratefully acknowledge Danny Vincent and the Human Solutions, Inc. team for developing the controller training materials and assistance during the simulations.We also thank P50 subject matter experts John Nolan and Glen Kanow for their invaluable guidance and suggestions regarding P50 operations.George Lawton, Joe Cisek, Matt Ma, and Tom Prevot for their hard work and dedication in establishing the interconnections in the ATC laboratory. +33rd Digital Avionics Systems ConferenceOctober 5 -9, 2014 + + + + + + + + + FAA aviation forecasts-fiscal years 1978–1989 + + Faa + + 10.1016/0160-7383(78)90349-3 + + + + Annals of Tourism Research + Annals of Tourism Research + 0160-7383 + + 5 + 4 + 460 + 2012-2032 + Elsevier BV + + + cited February 2014 + FAA, Aerospace Forecast Fiscal Years 2012- 2032, URL: http://www.faa.gov/about/office_org/headquarters_of fices/apl/aviation_forecasts/aerospace_forecasts/2012 -2032/ [cited February 2014]. + + + + + + + Faa + + + June 2013 + NextGen Implementation Plan + + + FAA, June 2013, NextGen Implementation Plan. + + + + + Validation of Simulations of Airport Surface Traffic with the Surface Operations Simulator and Scheduler + + RobertDWindhorst + + + JustinVMontoya + + + ZhifanZhu + + + SergeiGridnev + + + KatyGriffin + + + AdityaSaraf + + + SteveStroiney + + 10.2514/6.2013-4207 + + + 2013 Aviation Technology, Integration, and Operations Conference + Los Angeles, California + + American Institute of Aeronautics and Astronautics + August 2013. 2013. August + + + + Windhorst, Robert D., et al., August 2013, Validation of Simulations of Airport Surface Traffic with the Surface Operations Simulator and Scheduler, 2013 AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, 12-14 August, Los Angeles, California. + + + + + Aviation Safety: FAA Is Considering Changes to Aviation Medical Standards: Fact Sheet for Congressional Requesters + + Faa + + 10.1037/e433142008-001 + + + + Fact Sheet-Automatic Dependent Surveillance-Broadcast (ADS-B) + + Test accounts + 2010 + + + cited February 2014 + FAA, 2010, Fact Sheet-Automatic Dependent Surveillance-Broadcast (ADS-B), URL: http://www.faa.gov/news/fact_sheets/news_story.cfm ?newsid=7131 [cited February 2014]. + + + + + Roadmap for Performance-Based Navigation, Version + + JulyFaa + + + 2006 + + + FAA, July 2006, Roadmap for Performance- Based Navigation, Version 2.0. + + + + + + Faa + + + Terminal Automation Modernization and Replacement (TAMR) + + 2013 + + + cited February 2014 + FAA, 2013, Terminal Automation Modernization and Replacement (TAMR), URL: https://www.faa.gov/air_traffic/technology/tamr/ [cited February 2014]. + + + + + + BobTyo + + PM.pdf + + Time Based Flow Management Program (TBFM), FAA Industry Forum + + November 2012. February 2014 + + + Industry%20Forum%20Presentation%20TBFM + Tyo, Bob, November 2012, Time Based Flow Management Program (TBFM), FAA Industry Forum, URL: http://faaindustryforum.org/1%20- %20BOB%20TYO--- Industry%20Forum%20Presentation%20TBFM- PM.pdf [cited: February 2014]. + + + + + NASA's ATM Technology Demonstration-1: Transitioning Fuel Efficient, High Throughput Arrival Operations from Simulation to Reality + + ThomasPrevot + + + + Internal Conference on Human-Computer Interaction in Aerospace + Brussels; Belgium + + HCI-Aero + September 2012. 2012 + + + Prevot, Thomas, et al., September 2012, NASA's ATM Technology Demonstration-1: Transitioning Fuel Efficient, High Throughput Arrival Operations from Simulation to Reality, Internal Conference on Human-Computer Interaction in Aerospace (HCI- Aero 2012), Brussels, Belgium. + + + + + National Aeronautics and Space Administration (NASA) + + Nasa -Technology + + 10.4135/9781412939591.n797 + + + + Readiness Levels Demystified, 2010, NASA + + SAGE Publications, Inc. + January 2014 + + + cited + NASA -Technology Readiness Levels Demystified, 2010, NASA, URL: http://www.nasa.gov/topics/aeronautics/features/trl_d emystified.html [cited January 2014]. + + + + + NASA's ATM Technology Demonstration-1: Moving NextGen Arrival Concepts from the Laboratory to the Operational NAS + + Swenson + + + JohnEHarry + + + IiiRobinson + + + SteveWinter + + + + Journal of Air Traffic Control + + + July 2013 + + + Swenson, Harry, John E. Robinson III, Steve Winter, July 2013, NASA's ATM Technology Demonstration-1: Moving NextGen Arrival Concepts from the Laboratory to the Operational NAS, Journal of Air Traffic Control, pp. 22-44. + + + + + Air Traffic Management Technology Demonstration-1 (ATD-1) Tech Transfer Document Summary Version 1.0, NASA/TP-2013-216598 + + IiiRobinson + + + JohnE + + + September 2013 + Ames Research Center, Moffett Field, California + + + Robinson III, John E., September 2013, Air Traffic Management Technology Demonstration-1 (ATD-1) Tech Transfer Document Summary Version 1.0, NASA/TP-2013-216598, Ames Research Center, Moffett Field, California. + + + + + Design and Operational Evaluation of the Traffic Management Advisor at the Fort Worth Air Route Traffic Control Center, 1 st USA + + HarryNSwenson + + + + Europe Air Traffic Management Research and Development Seminar (ATM + + 1997. 1997 + Saclay, France + + + Swenson, Harry N., et al., 1997, Design and Operational Evaluation of the Traffic Management Advisor at the Fort Worth Air Route Traffic Control Center, 1 st USA/Europe Air Traffic Management Research and Development Seminar (ATM1997), Saclay, France. + + + + + The Dynamic Planner: The Sequencer, Scheduler, and Runway Allocator for Air Traffic Control Automation + + GregoryLWong + + NASA/TM-2000-209586 + + April 2000 + + + Wong, Gregory L., April 2000, The Dynamic Planner: The Sequencer, Scheduler, and Runway Allocator for Air Traffic Control Automation, NASA/TM-2000-209586. + + + + + Controller Advisory Tools for Efficient Arrivals in Dense Traffic Environments + + ToddJCallantine + + + EverettPalmer + + 10.2514/6.2009-6992 + + + 9th AIAA Aviation Technology, Integration, and Operations Conference (ATIO) + Hilton Head, South Carolina + + American Institute of Aeronautics and Astronautics + 2009 + + + Callantine, Todd J., Everett A. Palmer, 2009, Controller Advisory Tools for Efficient Arrivals in Dense Traffic Environments, 9 th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, Hilton Head, South Carolina. + + + + + How to compute a slot marker — Calculation of controller-managed spacing tools for efficient descents with precision scheduling + + ThomasPrevot + + 10.1109/dasc.2012.6382318 + + + 2012 IEEE/AIAA 31st Digital Avionics Systems Conference (DASC) + Williamsburg, Virginia + + IEEE + October 2012 + 31 + + + Prevot, Thomas, October 2012, How to Compute a Slot Marker-Calculation of Controller- Managed Spacing Tools for Efficient Descents with Precision Scheduling, 31 st Digital Avionics Systems Conference (DASC), Williamsburg, Virginia. + + + + + Controller-Managed Spacing - A Human-In-The-Loop Simulation of Terminal-Area Operations + + MichaelKupfer + + + ToddCallantine + + + JoeyMercer + + + LynneMartin + + + EverettPalmer + + 10.2514/6.2010-7545 + + + AIAA Guidance, Navigation, and Control Conference + Berlin, Germany + + American Institute of Aeronautics and Astronautics + 2011 + + + Kupfer, Michael, Todd Callantine, Lynne Martin, Joey Mercer, Everett Palmer, 2011, Controller Support Tools for Schedule-based Terminal Area Operations, 9 th USA/Europe Air Traffic Management Research and Development Seminar (ATM2011), Berlin, Germany. + + + + + Simulation Results for Airborne Precision Spacing along Continuous Descent Arrivals + + BryanBarmore + + + WilliamCapron + + + TerenceAbbott + + + BrianBaxley + + 10.2514/6.2008-8931 + + + The 26th Congress of ICAS and 8th AIAA ATIO + Yokohama, Japan + + American Institute of Aeronautics and Astronautics + 2004 + + + Barmore, Bryan, Terence Abbott, Karthik Krishnamurthy, 2004, Airborne-managed Spacing in Multiple Arrival Streams, 24 th International Congress of the Aeronautical Sciences (ICAS 2004), Yokohama, Japan. + + + + + Airborne Precision Spacing: A Trajectory-Based Aprroach to Improve Terminal Area Operations + + BryanBarmore + + 10.1109/dasc.2006.313683 + + + 2006 ieee/aiaa 25TH Digital Avionics Systems Conference + Portland, Oregon + + IEEE + 2006 + + + Barmore, Bryan, 2006, Airborne Precision Spacing: A Trajectory-based Approach to Improve Terminal Area Operations, 25 th Digital Avionics Systems Conference (DASC), Portland, Oregon. + + + + + Evaluation of an Airborne Spacing Concept to Support Continuous Descent Arrival Operations, 8 th USA + + JenniferLMurdoch + + + BrianEBarmore + + + BrianTBaxley + + + TerenceSAbbott + + + WilliamRCapron + + + + Europe Air Traffic Management Research and Development Seminar + + 2009. 2009 + ATM + Napa, California + + + Murdoch, Jennifer L., Brian E. Barmore, Brian T. Baxley, Terence S. Abbott, William R. Capron, 2009, Evaluation of an Airborne Spacing Concept to Support Continuous Descent Arrival Operations, 8 th USA/Europe Air Traffic Management Research and Development Seminar (ATM2009), Napa, California. + + + + + Human-in-the-loop Simulation of Trajectory-based Terminal Area Operations + + ToddJCallantine + + + EverettPalmer + + + MichaelKupfer + + + + 27 th International Congress of the Aeronautical Sciences (ICAS 2007) + Nice, France + + 2010 + + + Callantine, Todd J., Everett Palmer, Michael Kupfer, 2010, Human-in-the-loop Simulation of Trajectory-based Terminal Area Operations, 27 th International Congress of the Aeronautical Sciences (ICAS 2007), Nice, France. + + + + + Efficiency Benefits Using the Terminal Area Precision Scheduling and Spacing System + + JaneThipphavong + + + HarrySwenson + + + PaulLin + + + AnthonyYSeo + + + LeonardNBagasol + + 10.2514/6.2011-6971 + + + 11th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Virginia Beach, Virginia + + American Institute of Aeronautics and Astronautics + 2011 + + + Thipphavong, Jane, Harry Swenson, Paul Lin, Anthony Y. Seo, Leonard N. Bagasol, 2011, Efficiency Benefits Using the Terminal Precision Scheduling and Spacing System, 11 th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, Virginia Beach, Virginia. + + + + + Effects of scheduling and spacing tools on controllers' performance and perceptions of their workload + + LynneMartin + + + HarrySwenson + + + AlexanderSadovsky + + + JaneThipphavong + + + LiangChen + + + AnthonyYSeo + + 10.1109/dasc.2011.6096106 + + + 2011 IEEE/AIAA 30th Digital Avionics Systems Conference + Seattle, Washington + + IEEE + October 2011 + + + Martin, Lynne, Harry Swenson, Alexander Sadovsky, Jane Thipphavong, Liang Chen, Anthony Y. Seo, October 2011, Effects of Scheduling and Spacing Tools on Controllers' Performance and Perceptions of Their Workload, 30 th Digital Avionics Systems Conference (DASC), Seattle, Washington. + + + + + Evaluation of the terminal area precision scheduling and spacing system for near-term NAS application + + JaneThipphavong + + + HarrySwenson + + + LynneMartin + + + PaulLin + + + JimmyNguyen + + 10.1201/b12321-12 + + + Advances in Human Aspects of Aviation + San Francisco, California + + CRC Press + 2012 + + + + Thipphavong, Jane, Harry Swenson, Lynne Martin, Paul Lin, Jimmy Nguyen, 2012, Evaluation of the Terminal Area Precision Scheduling and Spacing System for Near-Term NAS Application, 4 th International Conference on Applied Human Factors and Ergonomics, San Francisco, California. + + + + + Development and evaluation of the terminal precision scheduling and spacing system for off-nominal condition operations + + HarryNSwenson + + + Jaewoo Jung + + + JaneThipphavong + + + Liang Chen + + + LynneMartin + + + JimmyNguyen + + 10.1109/dasc.2012.6382303 + + + 2012 IEEE/AIAA 31st Digital Avionics Systems Conference (DASC) + Williamsburg, Virginia + + IEEE + October 2012 + 31 + + + Swenson, Harry, Jaewoo Jung, Jane Thipphavong, Liang Chen, Lynne Martin, Jimmy Nguyen, October 2012, Development and Evaluation of the Terminal Precision Scheduling and Spacing System for Off-nominal Condition Operations, 31 st Digital Avionics Systems Conference (DASC), Williamsburg, Virginia. + + + + + Evaluation of the Terminal Area Precision and Spacing System for Performance-based Navigation Arrivals + + JaewooJung + + + HarrySwenson + + + JaneThipphavong + + + LynneMartin + + + LiangChen + + + JimmyNguyen + + + + 32 nd Digital Avionics Systems Conference (DASC) + Syracuse, New York + + 2013 + + + Jung, Jaewoo, Harry Swenson, Jane Thipphavong, Lynne Martin, Liang Chen, Jimmy Nguyen, 2013, Evaluation of the Terminal Area Precision and Spacing System for Performance-based Navigation Arrivals, 32 nd Digital Avionics Systems Conference (DASC), Syracuse, New York. + + + + + Evaluation of the terminal sequencing and spacing system for performance-based navigation arrivals + + JaneThipphavong + + + Jaewoo Jung + + + HarryNSwenson + + + LynneMartin + + + MelodyILin + + + JimmyNguyen + + 10.1109/dasc.2013.6712655 + + + 2013 IEEE/AIAA 32nd Digital Avionics Systems Conference (DASC) + Syracuse, New York + + IEEE + October 2013 + + + Thipphavong, Jane, Jaewoo Jung, Harry Swenson, Lynne Martin, Melody Lin, Jimmy Nguyen, October 2013, Evaluation of the Terminal Sequencing and Spacing System for Performance- based Navigation Arrivals, 32 nd Digital Avionics Systems Conference (DASC), Syracuse, New York. + + + + + Controller-managed spacing within mixed-equipage arrival operations involving flight-deck interval management + + JaneThipphavong + + 10.1201/b12321-14 + + + Advances in Human Aspects of Aviation + Chicago, Illinois + + CRC Press + 2013. 2013 + + + + Thipphavong, Jane, et al., 2013, Evaluation of the Controller-Managed Spacing Tools, Flight-deck Interval Management and Terminal Area Metering Capabilities for the ATM Technology Demonstration #1, 10 th USA/Europe Air Traffic Management Research and Development Seminar (ATM2013), Chicago, Illinois. + + + + + Co-Operative Air Traffic Management: Concept and Transition + + ThomasPrevot + + + ToddCallantine + + + PaulLee + + + JoeyMercer + + + VernolBattiste + + + EverettPalmer + + + NancySmith + + 10.2514/6.2005-6045 + + + AIAA Guidance, Navigation, and Control Conference and Exhibit + Madeira, Portugal + + American Institute of Aeronautics and Astronautics + July 2004 + + + ICCES-04 + Prevot, Thomas, Todd Callantine, Paul Lee, Joey Mercer, Everett Palmer, N. Smith, July 2004, Rapid Prototyping and Exploration of Advanced Air Traffic Concepts, International Conference on Computational and Engineering Sciences (ICCES- 04), Madeira, Portugal. + + + + + MACS: A Simulation Platform for Today's and Tomorrow's Air Traffic Operations + + ThomasPrevot + + + JoeyMercer + + 10.2514/6.2007-6556 + + + AIAA Modeling and Simulation Technologies Conference and Exhibit + Hilton Head, South Carolina + + American Institute of Aeronautics and Astronautics + August 2007 + + + Prevot, Thomas, August 2007, MACS: A Simulation Platform for Today's and Tomorrow's Air Traffic Operations, AIAA Modeling and Simulation Technologies Conference and Exhibit, Hilton Head, South Carolina. + + + + + + Faa + + + Standard Terminal Automation Replacement System (STARS) + + May 2012 + + + cited February 2014 + FAA, May 2012, Standard Terminal Automation Replacement System (STARS), URL: http://hf.tc.faa.gov/projects/stars.htm [cited February 2014]. + + + + + NASA's ATM technology demonstration 1 (ATD-1): Integrated concept of arrival operations + + BrianTBaxley + + + HSwenson + + + TPrevot + + + TCallantine + + 10.1109/dasc.2012.6382981 + NASA TM-2012-217585 + + + 2012 IEEE/AIAA 31st Digital Avionics Systems Conference (DASC) + + IEEE + July 2012 + + + Baxley, Brian T., et al., July 2012, Air Traffic Management Technology Demonstration-1 Concept of Operations, NASA TM-2012-217585. + + + + + Design and Evaluation of the Terminal Area Precision Scheduling and Spacing System, 9 th USA + + HarrySwenson + + + JaneThipphavong + + + AlexSadovsky + + + LiangChen + + + ChrisSullivan + + + LynneMartin + + + + Europe Air Traffic Management Research and Development Seminar + + 2011. 2011 + Berlin, Germany + + + Swenson, Harry, Jane Thipphavong, Alex Sadovsky, Liang Chen, Chris Sullivan, Lynne Martin, 2011, Design and Evaluation of the Terminal Area Precision Scheduling and Spacing System, 9 th USA/Europe Air Traffic Management Research and Development Seminar (ATM2011), Berlin, Germany. + + + + + An Overview of a Trajectory-Based Solution for En Route and Terminal Area Self-Spacing: Fourth Revision, NASA/CR-2013-TBD + + TerenceSAbbott + + + 2013 + Langley Research Center + Hampton, Virginia + + + Abbott, Terence S., 2013, An Overview of a Trajectory-Based Solution for En Route and Terminal Area Self-Spacing: Fourth Revision, NASA/CR-2013-TBD, Langley Research Center, Hampton, Virginia. + + + + + + Faa + + Order P50 TRACON 7210.4B ZAB 7210 + Phoenix Airport Arrival Rate (AAR) + + 2007 + 126 + + + FAA, 2007, Phoenix Airport Arrival Rate (AAR), Order P50 TRACON 7210.4B ZAB 7210.126B. + + + + + Major, John, (born 16 June 1945), consultant surveyor, 2002–07; General Manager, Borde Hill Estate, 2001–06 + + IiiRobinson + + + JohnE + + 10.1093/ww/9780199540884.013.u26367 + + June 2013. 20130607. 2013-06-07 + Oxford University Press + 2 + NASA Ames, California + + + Robinson III, John E., June 2013, 20130607 ATD-1 Wind Scenarios Description.docx, Version 2 (2013-06-07), NASA Ames, California. + + + + + Mesoscale Weather Prediction with the RUC Hybrid Isentropic–Terrain-Following Coordinate Model + + StanleyGBenjamin + + + GeorgAGrell + + + JohnMBrown + + + TatianaGSmirnova + + + RainerBleck + + 10.1175/1520-0493(2004)132<0473:mwpwtr>2.0.co;2 + + + Monthly Weather Review + Mon. Wea. Rev. + 0027-0644 + 1520-0493 + + 132 + 2 + + February 2004 + American Meteorological Society + + + Benjamin, Stanley G., George A. Grell, John M. Brown, Tatiana G. Smirnova, February 2004, Mesoscale Weather Prediction with the RUC Hybrid Isentropic-Terrain-Following Coordinate Model, Monthly Weather Review, Volume 132, pp. 473-494. + + + + + Development of NASA-TLX (Task Load Index): Results of Empirical and Theoretical Research + + SandraGHart + + + LowellEStaveland + + 10.1016/s0166-4115(08)62386-9 + + + Advances in Psychology + + PAHancock + + + Meshkati + + Amsterdam + + Elsevier + 1988 + + + + Hart, Sandra G., Lowell E. Staveland, 1988, Development of NASA-TLX (Task Load Index): Results of Empirical and Theoretical Research, P. A. Hancock and N Meshkati (Eds.) Human Mental Workload, Amsterdam, North Holland Press. + + + + + + KatharineKLee + + + KarolKerns + + + RandallBone + + + MonicarolNickelson + + Development and Validation of the Controller Acceptance Rating Scale (CARS): Results of Empirical Research, 4 th USA/Europe Air Traffic Management Research and Development Seminar (ATM2001) + Santa Fe, New Mexico + + December 2001 + + + Lee, Katharine K., Karol Kerns, Randall Bone, Monicarol Nickelson, December 2001, Development and Validation of the Controller Acceptance Rating Scale (CARS): Results of Empirical Research, 4 th USA/Europe Air Traffic Management Research and Development Seminar (ATM2001), Santa Fe, New Mexico. + + + + + + diff --git a/file787.txt b/file787.txt new file mode 100644 index 0000000000000000000000000000000000000000..5b6f2bf3b0a4d73e82ab4e75219c98cd7a992a93 --- /dev/null +++ b/file787.txt @@ -0,0 +1,720 @@ + + + + +I. IntroductionA ir traffic controllers work to maintain safe operations by directing aircraft departing from and arriving at airports.However, degradation of operating conditions or an increase in traffic (or both) at an airport can create difficult situations for the controllers.In order to prevent this from occurring, traffic flow managers monitor the system and issue traffic management initiatives to keep traffic at manageable levels.Ground delay programs are among the most aggressive traffic management initiatives, but if implemented at the wrong time or with the wrong parameters, may also decrease the efficiency of the National Airspace System.A ground delay program assigns arrival slots to a set of traffic destined for a particular airport, often resulting in a variable amount of pre-departure delay for each affected flight.Though pre-departure delay (holding on the ground) is generally preferred to airborne delay (holding in the air), it may result in unnecessary delay should conditions improve at the destination airport sooner than expected.Today's traffic flow managers rely on personal experience to create traffic management plans, including the issuance of ground delay programs.This has some potential drawbacks.First, two equally experienced managers may create equally effective but different plans for the same situation.Such unpredictability in response creates uncertainty for the airspace users, who as a result may fail to take proactive actions because of this variability.Second, less experienced traffic flow managers might develop less optimal plans: either less effective, too severe, or both inadequate and overly restrictive in some aspect.Finally, since no traffic flow manager has been on hand for all prior conditions, an unfamiliar situation may arise that would be challenging to effectively and efficiently manage without some decision support.We propose to address these issues by developing a decision support system that provides access to historical traffic flow management initiatives.These initiatives will be indexed by the relevant traffic and weather conditions that led to the actions taken by the traffic flow manager.Such a system would enable a traffic flow manager to expand upon personal knowledge by including the perspective of other traffic flow managers through a review of their past actions.At the core of this system is the ability to search for actions under a set of stated weather and traffic conditions, which may either be the current conditions or a "what-if" scenario.For example, the traffic flow manager could provide a traffic pattern and weather phenomenon to the decision support tool and inspect the actions taken in ten similar circumstances.This would suggest what actions could be taken, as well as which ones were taken more often, and should lead to more predictable, efficient, and timely decision making.Crucial to the success of such a decision support system is its search algorithm, in particular how well it ranks historical situations in order to their similarity to the scenario query.What makes a set of conditions similar?In this paper, we evaluate several techniques for rating the similarity of set of operating conditions.The remainder of the paper is organized as follows.A brief review of related work is provided in Section II.Sections III and IV describe the data used and metrics used in our study.Next, Section V outlines the models used to rank situations according to their estimated similarity to the scenario query.The results of our experiment is detailed in Section VI.Finally, we conclude with a summary and future work in Section VII. +II. Related WorkResearchers have previously explored search-based system for other weather-impacted fields, as well as other forms of traffic flow management decision support and prediction.Ji, Yuan and Yue used a case-based reasoning approach to retrieve representative weather cases from a database of weather events. 23They used similarity measure on several weather attributes that is equivalent to our weighted sum form with a log transform.Jain, Srinivas and Rauta developed a fuzzy logic system to forecast power (energy) loads, based on weather conditions. 24They used a similarity measure based on Euclidean distance, defined on several weather variables, to find the set of similar weather conditions to make the forecast.Juell and Paulson used weighted sum algorithm, as well as a neural net, to predict the dew point from other weather variables in a another case-based system. 25Reinforcement learning was used to tune the values of the weights.Elmore and Richman also use the Euclidean distance to find modes in weather data. 26They relate their choice of similarity metric to the more general Minkowski distance, and suggest other forms (such as the weighted sum) also be considered.Finally, Klein used weather-impacted traffic indices, as we do, to characterize weather in the National Airspace System. 27He used the cosine model to find similar days in terms of conditions at selected airports, but as he also factored in the overall magnitude, the approach is more comparable to the Euclidean distance model.To our knowledge, the proposed ability to search for past traffic management initiatives by the similarity of operating conditions is unique.The original concept of the decision support tool was suggested by Rios. 28The most analogous decision support tools, such as the Enhanced Traffic Management System, 29 provide situational awareness but do not provide historical context.Smith, Sherry and Donahue investigated the possibility of predicting GDPs. 30They used a support vector machine, trained on historical weather forecasts and airport arrival rates, to predict an arrival rate and GDP status given a new weather forecast.Nonetheless, the envisioned usage is different from our proposed decision support tool, as they recommend a particular action, whereas we propose to provide historical context.Our study fills the gap between decision support tools that only provide information on the current situation (and not access to past decisions), and prediction tools which utilize historical data but do not support the user in making decisions from that historical perspective. +III. DatasetA search system that meets our specifications would need to include both the situational parameters considered by the decision makers, as well their ultimate decisions.We fused two datasets, described below, to create our representation of the situation and corresponding action.These datasets provided hourly data from 2008 and 2009 for the FAA's OEP-35 (with the exception of Honolulu), which consists of the busiest U.S. airports.From this set, we eliminated any airport that did not have a minimum of fifty ground delay programs (GDPs) issued in 2009, leaving the airports listed in table 1.We use a form of WITI that is broken down into seven separate WITIs that capture different impacts at a given airport.These are:1. En-route convective weather, which captures the impact of convective weather on incoming and outgoing traffic, up to five hundred nautical miles away.2. Local convective weather, which captures the impact of convective weather within one hundred nautical miles of the airport.3. Wind, with or without accompanying precipitation.4. Snow, which captures all sorts of cold hazards.5. Instrument Meteorological Conditions (IMC), which captures the impact of poor visibility on traffic.6. Traffic volume, which captures volume impacts caused by weather disruptions elsewhere in the system.7. Other, capturing miscellaneous weather-related impacts that do not fall into the categories above.As we have the current conditions in addition to three forecast periods for each of the seven WITIs, we have 28 WITIs to characterize the conditions at each hour at any of the previously listed airports.Each WITI is a nonnegative real number.Fig. 1 shows an example of WITI scores where delays due to en-route convection and wind are predicted to increase, according to the forecasts.Traffic flow management decisions are made at various centers across the U.S. Fortunately, the difficult task of recording, standardizing and centralizing these decisions has already been done.Originally designed to provide situational awareness, the National Traffic Management Log (NTML) 7 also contains an archive of traffic flow management decisions in a database.This database contains both initial decisions as well as any subsequent revisions.An earlier study shows that the unified NTML has provided numerous benefits over the prior, loosely controlled logging system. 8We use the historical archive of the NTML to provide an hourly snapshot for the GDP status of the airports in our study.In addition to the NTML, we also scraped pages from the FAA's online advisory database. 9We matched these records to those in the NTML, cross-checking the data and also adding information about when the advisory was publicized. +C. ProcessingWe combined our WITI and NTML/advisory data to give us 8688 instances in 2008 and 8544 instances in 2009 of current and forecast conditions (from the WITI data source) and corresponding actions (from the NTML/advisory data sources), as given in figure 2. The 28 WITIs are as described in Section A (though the current conditions and three forecasts are not separately represented in figure 2).We treated ground stops as a type of GDP.The GDP cause is the reason given for the GDP, or "no GDP" if no GDP is active.However, over half of the GDP cause types in our database had no obvious relationship to weather (e.g., "Air Show").We treated all such non-weather GDPs as "no GDP" in our dataset.Likewise, some GDP causes were very similar from the WITI perspective (e.g., "Fog" and "Low Visibility"), and were mapped into a joint cause.In all, this left us with four GDP causes: in order of prevalence, they are "wind", "visibility", "convection", and "snow", and "no GDP".The scheduled start and end time reflects the proposal at that particular time and not the actual start and end times (the times a GDP is planned to occur are not always the same as when the GDP actually occurs); furthermore, these times are not defined when the cause is "no GDP".For this study, we did not track any other parameters of the GDP, such as scope and rate.WITI has been found to have a high correlation to overall national delays 10 and delays at specific airports. 11However, this does not necessarily mean that WITI values are strongly indicitive of ground delay programs.We ran a simple experiment to evaluate the relationship between our WITI scores and ground delay program causes (including "no GDP", as described above).We cast this as a classification problem to determine if a ground delay program is in effect, further divided into five separate classification problems, one for each of the five causes.For each WITI type, we found the rule with the highest classification accuracy by retrospectively picking the best WITI cutoff threshold (with different thresholds for different causes and airports).Each rule would make its prediction by comparing the actual WITI value with this cutoff threshold.For instance, a rule could be "predict a GDP due to snow in Atlanta if the observed snow WITI score exceeds 200."As the observed (non-forecast) WITIs are presumably more accurate than the forecast WITIs, we only evaluated the accuracy of these WITIs against the current GDP status (instead of planned GDPs, for the same reason).For further context, we compare these results to a baseline constant rule dubbed "zeror" (meaning zero rule), which always gives the same prediction without using any data.Figure 3 shows our evaluation on two measures, classification accuracy and normalized information gain over all airports, for the 2008 data.The leftmost chart shows classification accuracy, which is the fraction of times a prediction based on the WITI would yield the correct answer.All seven WITIs yield highly accurate predictions, but so does the constant "zeror" rule.This is because GDPs are somewhat uncommon, and always guessing that there is not a GDP produces highly accurate results.It is difficult to discern how much is gained by using the individual WITIs when evaluating classification accuracy, so a different evaluation measure is needed.The rightmost chart shows how much information can be gained by considering the individual WITI, according to information theory. 12We normalized this information gain to be on a 0-1 scale, so that 0 is no information and 1 is complete information.Since the "zeror" rule makes the same prediction in all cases, it has no information gain, and therefore cannot be seen on the chart.Unfortunately, the individual WITIs do not fare much better, with none providing even a fairly modest 20% gain in information.This means that searching for similiar events by WITI will not necessarily lead to events with the same control action, and that producing good search results will be challenging. +IV. Evaluation MetricsCore to the evaluation of information retrieval results is the notion of relevancy.In a typical information retrieval setting, the items in the search collection (known as a corpus) are evaluated for relevancy against several queries by either the users themselves or trained assessors.Typically, relevancy is seen as a binary measure (relevant/non-relevant), though levels of relevancy are sometimes used.This study needed to overcome the lack of available relevance data from actual traffic flow managers.However, since we know both the task (to make a decision regarding a possible GDP) and what decision was ultimately made, we can estimate what should be relevant.We cast the problem as that of taskbased information retrieval, 13,14 and use our understanding of the task and decision to simulate queries and relevance judgments.Specifically, as we regard the task as making a control action decision, a reasonable query is the observed and forecast conditions for that time.The query consists of only the WITIs (see figure 2, above) and not the GDP status, as the latter is the decision to be made by the user of the system.The historical instances from the same airport are used as the corpus.Once again, we use the envisioned usage of the system to simulate the relevance judgments.Since the decision support system is designed to assist traffic flow managers in making decisions on GDPs, we use the GDP status to derive the relevancy judgments.We treat the GDP decision that was made by the traffic flow managers in the situation as the (only) correct choice.Presumably, a system that favors instances from the corpus with the same GDP decision as was chosen for the queried situation would have been helpful for making that decision.Therefore, we define relevancy as a match between the corresponding action of the retrieved instance and the one chosen in the queried situation.(Note that the GDP actions are used only to define relevance, and are not a part of the query or similarity calculations.)However, in addition to the cause, the GDP decisions also have scheduled start and end times, which present a challenge to map to binary relevance.We opted to reformulate the problem as several parallel problems, each of the form "Will we need a GDP Y hours from now?", where Y was specified in hour increments up to the 6-hour WITI forecast, for a total of seven parallel problems.In practice, some instances with a different GDP status would be seen as similar, while some with the same might not (as conditions can vary greatly among situations with the same GDP status), but we believe our approximation to be as reasonable as possible without human assessors and without biasing the results by using WITIs in the relevance definition.Given our simulated queries and relevance judgments, we can evaluate the ranking of instances (as described in Section V) provided by the search system.The metrics we use are based on a metric defined on unranked results called precision.Given set of results produced for a query, the precision is simply the fraction of the set that is relevant to the query.Formally, given a query q, a set of results S q and the set of all relevant items R q , the precision is given in Eq. (1).p(S q , R q ) = |S q ∩ R q | |S q |(1)To translate this to a ranked list of results, we define a series of overlapping sets by including only the top m ranked results.We choose m as the smallest number that includes a specified number of relevant results, so that S q,1 includes a single relevant result, S q,2 includes two relevant results, and so on up to S q,|Rq| .One metric is the precision of S q,1 , which is smallest top m-ranked search results that include a single relevant result.This metric is known as the reciprocal rank, as it is equivalent to the reciprocal of the rank of the highest ranked relevant instance.It corresponds to the number of search results a user would need to inspect in order to find a single relevant result, when evaluating the search results in ranked order.It is a good metric when the user's task can be satisfied by any relevant result.Since we have a set of queries, we use the mean reciprocal rank (MRR) 15 as one of the two metrics we use to evaluate our results.Given the set of queries Q and other quantities as defined for Eq. ( 1), the mean reciprocal rank is defined in Eq. 2.M RR(Q) = 1 |Q| q∈Q p(S q,1 , R q,1 ) = 1 |Q| q∈Q |S q,1 ∩ R q | |S q,1 | = 1 |Q| q∈Q 1 |S q,1 |(2)However, there are shortcomings of the MRR metric from the aspect of our evaluation, stemming from the reciprocal rank on which it is based.Reciprocal rank models a user need that is satisfied with a single relevant result.However, it is not obvious that this would be the case for our decision problem, in fact it seems more likely that the traffic flow manager would need to review several comparable situations to arrive at a decision.Also, because the reciprocal rank is based on the rank of a single item, it tends to be a metric with high variance, though this is reduced by the averaging over queries in that occurs in the MRR metric.Instead of reciprocal rank, the average precision can be used, which is precision averaged over all sets S q,n (where reciprocal rank used only S q,1 ).Once again, this can also be averaged over a set of queries, leading to the mean average precision (MAP), 16 given in Eq. 3.M AP (Q) = 1 |Q| q∈Q 1 |R q | |Rq| i=1 p(S q,i , R q,i ) = 1 |Q| q∈Q 1 |R q | |Rq| i=1 i |S q,i |(3)The MAP metric does not assume a certain number of relevant results are needed for the user's task, instead averaging over all possible levels.This better fits our uncertainty of the traffic flow managers' needs, and is more commonly used in information retrieval evaluations. 17inally, as we have different categories of results (both from different airports and from queries of different GDP causes), we follow the example of Sebastiani 18 and use the microaverage to combine averages from different categories.Given an evaluation metric f () and k sets S 1 , . . ., S k , the microaverage over f () is given in Eq. 4.f micro (S 1 , . . . , S k ) = k i=1 |S i | k j=1 |S j | f (S i )(4)In our use, f () is either MRR or MAP, and since these are both averages themselves, the microaverage simplifies to the corresponding metric over all queries without regard to the category. +V. ModelsWe experimented with standard models from the information retrieval, machine learning and operations research communities to produce our rankings.We use a utility-based approach for ranking, which means that each instance is given an ordinal score (or utility, in this case a similarity measure), and the ranking is defined by the corresponding order of the scores.In the following, the query Q and instance X consist of the 28 WITIs (see Section C and fig.2), with q i and x i denoting the i th WITI value of the query and instance, respectively.Perhaps the most widely used similarity measure in information retrieval is the durable cosine, 19 defined in our domain as in Eq. 5.cos(Q, X) = 28 i=1 q i x i Q X (5)Since the WITIs are all nonnegative, the cosine will range from 0 to 1, with higher scores indicating more similarity and thus higher ranked.An important feature of the cosine similarity is that it is independent of the overall magnitude of both the query and instance vectors, a feature that makes sense for document retrieval but does not appear to be a good match for our domain.Rather, the cosine measure is sensitive only to the relative distribution among the vectors of the query and instance.Likewise, the common Euclidean distance is often used in machine learning algorithms, 20 defined on an multidimensional space.Since we have 28 WITIs, we have 28 dimensions, and so the Euclidean distance is defined as in Eq. 6.distance(Q, X) = 28 i=1 |q i -x i | 2 (6)Since Eq. 6 is a distance measure, larger values indicate less similarity and thus lower ranked.A feature of the Euclidean distance is the possibility of a "shortcut", i.e., that dissimilarity is lessened when the differences are distributed across different dimensions (compared to an equal total magnitude in the same dimension).It is unclear if this applies to our domain.Finally, we employ a simple weighted sum from operations research, 21 in an even simpler form as all our weights are 1.With equal weighting used, the weighted sum we used in our experiments is given in Eq. 7.wsum(Q, X) = 28 i=1 |q i -x i |(7)Like the Euclidean distance, the value of Eq. 7 grows as the differences increase, and we rank in decreasing order of scores.Implied by its formula, the weighted sum model assumes that the same differential change on a particular dimension will always result in the same change in the score, regardless of the other dimensions.Though it would stand to reason that any increase in WITIs should correspond to worse conditions, this independence property may not hold in our domain.Though the three models described have different properties and produce different rankings, they are in fact all equivalent to the parameterized Minkowski distance measure, 22 with an additional modification for the cosine model.The Minkowski distance takes an additional parameter p and is defined in our domain as in Eq. 8.M inkowski(Q, X, p) = 28 i=1 |q i -x i | p 1/p (8)The relationship to the Euclidean distance and the (evenly) weighted sum should be obvious by comparing the formulas.As it turns out, the ranking produced by the cosine model is equivalent to that of the Euclidean distance model if the vectors are first normalized to unit vectors.This may be easier to see when considering the geometry of the problem.As an alternative to the WITI values as given, we also experimented with a transformation of the WITI scores.Each WITI appeared to be approximately distributed by an exponential distribution, with many WITI values falling in the low range, and fewer and fewer spread out on the long right tail.We created transformed WITIs by mapping these values onto an approximation of their empirical cumulative distribution function.This has some potential advantages.First, the cumulative distribution function ranges from 0 to 1, so all transformed WITIs are on the same scale.Second, differences in the long tail are minimized, matching the intuition that large WITI values may capture qualitatively the same impact (at least to the fidelity in which we model the GDP).On the other hand, small differences among the more frequent low WITI values will be seen as more meaningful.The three original similarity models, combined with the same models using transformed WITIs, gives us six models to evaluate in experimentation.In addition to this, we added a random model to serve as a baseline.The random model scores instances randomly and produces a random ranking.Unlike our other models, a model that incorporates randomness can produce different results for different runs on the same data.To avoid this variability, we simply calculated what the theoretical average value of our metrics for the random model rather than report metrics on a limited number of runs. +VI. ResultsWe used all the instances from 2009 as queries, to be matched against instances from 2008 corpus for the same airport.This simulates a decision support tool built on 2008 data that went "live" in 2009.For each query, all instances from the same airport in the corpus were scored by each of the models in Section V and ranked in a list.Our metrics were recorded separately for each GDP cause, as well as the microaveraged results, for each airport.The qualitative results vary from airport to airport and are too numerous to report.Instead, we use the microaverage (as given in Eq. 4) over GDP causes to combine the results from different airports in the following results.In addition to breaking down the results by GDP type, we also evaluate the performance on queries that we consider "hard".We define the query as "hard" if for the given time period, the decision was to change the plan from what was in place previously.This could be changing the start or end times for a GDP (including cancellations), or changing the GDP cause.Figure 4 shows the data flow in our experiment.The WITI data (including the forecasted WITI, as in Fig. 1) is used as our representation of the operational situation.We divide this into two sets: the 2008 WITIs are used as the corpus (the body of data we will retrieve from), and the 2009 WITIs are used as our query set.Each query is evaluated by the retrieval models (described above in section V), with each producing a ranked list.The quality of these ranked lists are evaluated by comparing the corresponding actions of the query (from 2009) and each element in the ranked list.The final result is the MRR and MAP metrics, as defined in section IV and detailed below.Figure 5 shows MRR results microaveraged over airport and separated by the GDP cause of the query.Except for the cosine model on queries with no GDPs, all models easily outperform the random model on the MRR metric.Retrieving an instance without a GDP should be easy, since most of the time no GDP is in place.Therefore, even though the evaluation measure is high, the fact that the cosine model only slightly outperforms the random model when there is no GDP is unsatisfactory.For the queries that do have a GDP, the performance of the non-random models is largely comparable, though the cosine and transformed WITI cosine underperform for certain GDP types.Among the remaining models, it is nearly impossible to have a preference.Figure 6 shows MAP results microaveraged over airport and separated by the GDP cause of the query.The differences between models is somewhat more pronounced, though it is still difficult to pick a best or worst model.Again, except for cosine, all models clearly outperform the random model.The cosine remains the puzzling member of the bunch; it's performance is as poor as random guessing when no GDP is in place, as with the MRR metric, and it has a mediocre showing on the convection queries.On the other hand, it clearly outperforms the other models on the visibility, wind, and snow queries.For the most part, the other models give roughly similar performance.Initially, the performance of the cosine model, particularly when evaluated on mean average precision, was perplexing.The only difference between the cosine and Euclidean distance in terms of ranking is that cosine is not sensitive to the overall "length" (i.e., Euclidean norm) of the Weather Impacted Traffic Index scores, and the two give the same ranking when all instances have the same length.This would appear to be a disadvantage of the cosine model, and yet it has the best mean average precision performance for three of the four ground delay program causes.As it turns out, this Weather Impacted Traffic Index length varies considerably among ground delay programs with the same cause.This means that some instances with a given ground delay program action will have WITI scores much closer to instances without a ground delay program action than those with large lengths with the same ground delay program cause.The only ground delay program cause that did not have such a large length variance were the convection ground delay programs, and not surprisingly the cosine model underperformed for convection.By the same token, the transformed Weather Impacted Traffic Index (see Section V) reduces the Weather Impacted Traffic Index length variance, which explains why the weighted sum and Euclidean distance performed better on retrieving ground delay programs with the transform in place.Overall, if we were to pick a single model from this set, we would use either the weighted sum or Euclidean distance with the Weather Impacted Traffic Index transform for this reason. +VII. Conclusions and Future WorkIn this paper, we have proposed a decision support tool to assist traffic flow managers in deciding if a ground delay program is needed.The tool would assist not by making a recommendation, but by allowing the traffic flow manager to inspect past decisions by other traffic flow managers in similar conditions.We proposed six models to rank instances by similarity to given conditions and evaluated these on a database comprised of Weather Impacted Traffic Indices and corresponding ground delay program decisions for ten busy U.S. airports.All six methods evaluated outperformed a random ranking model on the metrics of mean reciprocal rank and mean average precision, both validating the choice of Weather Impacted Traffic Index to represent conditions and showing that the proposed decision support tool could help guide users towards events with comparable control actions.What is not clear is which of the models would be the best choice.This is difficult to assess as the models have different relative performance on different query types, as particular models are apparently better at detecting some events and worse at detecting others.The query types were not evenly represented in our database, nor do we feel that each query type is equally important to traffic flow managers, though we do not how important each type might be.In reality, none of the models we evaluated are likely to be the best for the task.One interesting possibility would be to evaluate additional parameter values for the Minkowski distance, or even derive it directly from the data.Another approach would be to combine the models we experimented with in some way.In particular, is there some way to get the advantages of the cosine models without incurring the cost?An alternative approach would be to use an automatic algorithm, such as boosting, 31 to combine the different models in a machine learning context.Indeed, there are quite a few possibilities for learning from the data to improve performance.In this paper, we have already used the data to form the empirical exponential distribution for the Weather Impacted Traffic Index transformation.Other transformations could be explored, allowing for a more powerful fit of the data.Another possibility for learning would be to learn weights or linear rescalings of the Weather Impacted Traffic Indices.There are precedents for such with each approach, and an uneven weighting should be expected for our domain as Weather Impacted Traffic Indices are not equally accurate.Finally, feedback from traffic flow managers would greatly improve the potential usefulness of the proposed decision support system.Indications of which instances are relevant to a query would give more accurate evaluations and allow us to better tune our models.Prioritization of the various ground delay program causes would allow us to create a more informed cost function to evaluate models.Lastly, additional input would refine our understanding of the needs of the traffic flow managers and better enable us to design a decision support tool to assist them.Figure 1 .1Figure 1.Example of WITI representation (from JFK airport).en-route local convection convection wind snow IMC volume other Now 12.2 0.0 0.0 0.0 0.0 0.0 0.2 2-hour forecast 64.7 0.0 9.2 0.0 0.0 0.0 0.0 4-hour forecast 56.3 0.0 111.4 0.0 0.0 0.0 0.0 6-hour forecast 168.8 0.0 70.8 0.0 0.0 0.0 0.0 +Figure 2 .Figure 3 .23Figure 2. Instance representation for weather conditions and corresponding GDP status en-route local GDP scheduled scheduled convection convection wind snow IMC volume other cause start time end time +Figure 4 .4Figure 4. Flow of data into evaluation results in our experiment. +Table 1 .1Weather-related GDP Statistics for airports in our datasetIATA CodeAirport GDPs in 2009 Total hours of GDP in 2009ATLHartsfield-Jackson Atlanta International213824BOSBoston Logan International91523CLTCharlotte Douglas International77191DENDenver International67162EWRNewark Liberty International2271959JFK New York John F. Kennedy International152852LGANew York LaGuardia1461406ORDChicago O'Hare International90503PHLPhiladelphia International142857SFOSan Francisco International1951006A. Weather Impacted Traffic IndexBoth weather and traffic must be captured in our representation of the relevant conditions. Both are richand complex, and their effective representation is an area of research in itself. Fortunately, we can leverageprevious research that lead to the development of the Weather Impacted Traffic Index (WITI) 1 to representboth weather and traffic. WITI is a model which seeks to capture specifically the effect weather has ontraffic. We use a version of WITI that includes a forecast component that had been previously developedby other researchers. 2 This forecast is important to consider in our search system, as GDPs are plannednot only based on the current conditions, but also on what is expected in the future. The weather forecastswere obtained for two hours, four hours, and six hours in the future. The traffic forecast was generated fromthe scheduled traffic. The weather forecasts came from the Collaborative Convective Forecast Product 3 andTerminal Aerodrome Forecasts, 4 whereas current weather conditions came from the National ConvectiveWeather Diagnostic 5 and Aviation Routine Weather Reports (often referred to by the French acronymMETARs). 6 +MAP results per GDP cause category, microaveraged over airport.WS is weighted sum, ED is Euclidean distance, CS is cosine, Ran is random, and '+' version indicate WITI transformation.No GDP QueriesWind QueriesVisibility Queries0.96 0.980.12 0.140.15MAP0.94MAP0.10MAP0.100.90 0.920.08 0.060.05060120180240300360060120180240300360060120180240300360PeriodPeriodPeriodConvection QueriesSnow QueriesHard QueriesMAP0.02 0.04 0.06 0.08 0.10MAP0.00 0.01 0.02 0.03 0.04 0.05 0.06MAP0.10 0.15 0.20 0.25 0.30 0.35060120180240300360060120180240300360060120180240300360PeriodPeriodPeriodRanWSWS+EDED+CSCS+Figure 6. + + + + +AcknowledgmentsThe authors would like to thank Sasha Klein for providing the WITI data, Rich Jehlen for providing access to the NTML and formulation of the proposed decision support tool, and Shon Grabbe for input and support for this project. + + + + + + + + + Assessing NAS Performance: Normalizing for the Effects of Weather + + MBCallaham + + + JSDearmon + + + ACooper + + + JHGoodfriend + + + DMoch-Mooney + + + GSolomos + + + + 4th USA/Europe Air Traffic Management R&D Symposium + + 2001 + + + Callaham, M. B., DeArmon, J. S., Cooper, A., Goodfriend, J. H., Moch-Mooney, D., and Solomos, G., "Assessing NAS Performance: Normalizing for the Effects of Weather," 4th USA/Europe Air Traffic Management R&D Symposium, 2001. + + + + + Using A Convective Weather Forecast Product to Predict Weather Impact on Air Traffic: Methodology and Comparison with Actual Data + + AlexanderKlein + + + SadeghKavoussi + + + DavidHickman + + + DavidSimenauer + + + MarkPhaneuf + + + ThomasMacphail + + 10.1109/icnsurv.2007.384153 + + + + 2007 Integrated Communications, Navigation and Surveillance Conference + + IEEE + + + + Collaborative Convective Forecast Product Product Description Document + Klein, A., Kavoussi, S., Hickman, D., Simenauer, D., Phaneuf, M., and MacPhail, T., "Predicting Weather Impact on Air Traffic," Integrated Communications Navigation and Surveillance (ICNS) Conference, 2007. 3 National Weather Service, "Collaborative Convective Forecast Product Product Description Document," http:// aviationweather.gov/products/ccfp/docs/pdd-ccfp.pdf. + + + + + Virtual Reality for Hazard Mitigation and Community Resilience: An Interdisciplinary Collaboration with Community Engagement to Enhance Risk Awareness + + NancyJStone + + + GuirongYan + + + Fiona Fui-HoonNah + + + ChamanSabharwal + + + KelseyAngle + + + FredE (gene)Hatch + + + SteveRunnels + + + VankitaBrown + + + GregorySchoor + + + ChristopherEngelbrecht + + 10.17705/1thci.00145 + + + AIS Transactions on Human-Computer Interaction + THCI + 1944-3900 + + 13 + 2 + + 1997 + Association for Information Systems + + + National Weather Service + + + National Weather Service, Weather Service Operations Manual, chap. D-31, National Oceanic and Atmospheric Admin- istration, 1997. + + + + + National Convective Weather Forecast Product, Version 2 (NCWF-2) + + NationalWeather + + + Service + + + + + National Weather Service, "National Convective Weather Forecast Product, Version 2 (NCWF-2)," http://products. weather.gov/PDD/ncwf-2_pdd.pdf. + + + + + National Oceanic and Atmospheric Administration + 10.4135/9781544377230.n58 + + + Federal Regulatory Guide + + CQ Press + 2005 + + + + Office of the Federal Coordinator for Meterological Services and Supporting Research, Federal Meteorological Handbook No. 1 , chap. 2, National Oceanic and Atmospheric Administration, 2005. + + + + + Federal Aviation Administration + 10.4135/9781483346526.n204 + + 2010 + SAGE Publications, Inc. + 17 + + + Federal Aviation Administration, Facility Operation and Administration, chap. 17-5, U.S. Department of Transportation, 2010. + + + + + Improving the Usability of an Automated Tool for the Recording, Coordination, and Communication of Traffic Management Initiatives + + BartBrickman + + + TanyaYuditsky + + 10.1177/154193120404800111 + + + Proceedings of the Human Factors and Ergonomics Society Annual Meeting + Proceedings of the Human Factors and Ergonomics Society Annual Meeting + 2169-5067 + 1071-1813 + + 48 + 1 + + 2004 + SAGE Publications + + + Brickman, B. and Yuditsky, T., "Improving the Usability of an Automated Tool for the Recording, Coordination, and Communication of Traffic Managment Initiatives," Proceedings of the Human Factors and Ergonomics Society 48th Annual Meeting, 2004, pp. 46-50. + + + + + Health hazard evaluation report: HHE-81-042-832, Federal Aviation Administration, New York Air Route Traffic Control Center, Ronkonkoma, New York. + 10.26616/nioshhhe81042832 + + + + Air Traffic Control System Command Center Advisories Database + + U.S. Department of Health and Human Services, Public Health Service, Centers for Disease Control, National Institute for Occupational Safety and Health + + + + 9 Federal Aviation Administration, "Air Traffic Control System Command Center Advisories Database," http://www.fly. faa.gov/adv/advAdvisoryForm.jsp. + + + + + National Airspace System Delay Estimation Using Weather Weighted Traffic Counts + + GanoChatterji + + + BanavarSridhar + + 10.2514/6.2005-6278 + + + AIAA Guidance, Navigation, and Control Conference and Exhibit + + American Institute of Aeronautics and Astronautics + 2005 + + + Chatterji, G. B. and Sridhar, B., "National Airspace System Delay Estimation Using Weather Weighted Traffic Counts," AIAA Guidance, Navigation, and Control Conference and Exhibit, 2005. + + + + + Airport delay prediction using weather-impacted traffic index (WITI) model + + AlexanderKlein + + + CCraun + + + RSLee + + 10.1109/dasc.2010.5655493 + + + 29th Digital Avionics Systems Conference + + IEEE + 2010 + + + Klein, A., Craun, C., and Lee, R. S., "Airport delay prediction using weather-impacted traffic index (WITI) model," 2010 IEEE/AIAA 29th Digital Avionics Systems Conference (DASC), 2010. + + + + + Elements of Information Theory + + ThomasMCover + + + JoyAThomas + + 10.1002/0471200611 + + + Elements of Information Theory + + John Wiley & Sons, Inc. + 2006 + + + + Cover, T. M. and Thomas, J. A., Elements of Information Theory, chap. 2, Wiley, 2nd ed., 2006, pp. 13-52. + + + + + A task-oriented approach to information retrieval evaluation + + WilliamHersh + + + JeffreyPentecost + + + DavidHickam + + 10.1002/(sici)1097-4571(199601)47:1<50::aid-asi5>3.0.co;2-1 + + + Journal of the American Society for Information Science + J. Am. Soc. Inf. Sci. + 0002-8231 + 1097-4571 + + 47 + 1 + + 1996 + Wiley + + + Hersh, W., Pentecost, J., and Hickam, D., "A task-oriented approach to information retrieval evaluation," Journal of the American Society for Information Science, Vol. 47, 1996, pp. 50-56. + + + + + Task-based information searching + + PerttiVakkari + + 10.1002/aris.1440370110 + + + Annual Review of Information Science and Technology + Ann. Rev. Info. Sci. Tech. + 0066-4200 + + 37 + 1 + + 2003 + Wiley + + + Vakkari, P., "Task-based information searching," Annual Review of Information Science and Technology, Vol. 37, No. 1, 2003, pp. 413-464. + + + + + The seventh text REtrieval conference (TREC-7) + + EMVoorhees + + + DKHarman + + 10.6028/nist.sp.500-242 + + + Proceedings of the 8th Text REtrieval Conference + the 8th Text REtrieval Conference + + National Institute of Standards and Technology + 1999 + + + Voorhees, E. M., "TREC-8 Question Answering Track Report," Proceedings of the 8th Text REtrieval Conference, 1999. + + + + + + EMVoorhees + + + DHarman + + TREC: Experiment and Evaluation in Information Retrieval + + MIT Press + 2005 + + + Voorhees, E. M. and Harman, D., editors, TREC: Experiment and Evaluation in Information Retrieval, MIT Press, 2005. + + + + + + CDManning + + + PRaghavan + + + HSchütze + + Introduction to Information Retrieval + + Cambridge University Press + 2008 + 18 + + + Manning, C. D., Raghavan, P., and Schütze, H., Introduction to Information Retrieval, Cambridge University Press, 2008. 18 + + + + + Machine learning in automated text categorization + + FabrizioSebastiani + + 10.1145/505282.505283 + + + ACM Computing Surveys + ACM Comput. Surv. + 0360-0300 + 1557-7341 + + 34 + 1 + + 2002 + Association for Computing Machinery (ACM) + + + Sebastiani, F., "Machine learning in automated text categorization," ACM Computing Surveys, Vol. 34, No. 1, 2002. + + + + + Automatic text processing: the transformation, analysis, and retrieval of information by computer + + GSalton + + 10.5860/choice.27-0351 + + + Choice Reviews Online + Choice Reviews Online + 0009-4978 + 1523-8253 + + 27 + 01 + 27-0351-27-0351 + 1989 + American Library Association + Reading, MA + + + Salton, G., Automatic Text Processing: The Transformation, Analysis, and Retrieval of Information by Computer , Addison Wesley, Reading, MA, 1989. + + + + + Instance-based learning algorithms + + DavidWAha + + + DennisKibler + + + MarcKAlbert + + 10.1007/bf00153759 + + + Machine Learning + Mach Learn + 0885-6125 + 1573-0565 + + 6 + 1 + + 1991 + Springer Science and Business Media LLC + + + Aha, D. W., Kibler, D., and Albert, M. K., "Instance-Based Learning Algorithms," Machine Learning, Vol. 6, No. 1, 1991, pp. 37-66. + + + + + Multi-Criteria Decision Making Methods + + EvangelosTriantaphyllou + + 10.1007/978-1-4757-3157-6_2 + + + Applied Optimization + + Springer US + 2000 + + + + Triantaphyllou, E., Multi-Criteria Decision Making Methods: A Comparative Study, chap. 2, Kluwer Academic Publish- ers, 2000. + + + + + + + TFCox + + + MA ACox + + + Scaling + + + 2000 + 63 + + + Cox, T. F. and Cox, M. A. A., Multidimensional Scaling, chap. 3, Chapman and Hall/CRC, 2nd ed., 2000, p. 63. + + + + + A Method of Weather Cases Generation Based on Similarity Rough Set + + SaiJi + + + ShenfangYuan + + + JianYue + + 10.1109/icmss.2009.5302086 + + + 2009 International Conference on Management and Service Science + + IEEE + 2009 + + + + Ji, S., Yuan, S., and Yue, J., "A Method of Weather Cases Generation Based on Similarity Rough Set," Proceedings of the International Conference on Management and Service Science (MASS '09), IEEE, 2009, pp. 1-4. + + + + + Short term load forecasting using fuzzy adaptive inference and similarity + + AmitJain + + + ESrinivas + + + RasmimayeeRauta + + 10.1109/nabic.2009.5393627 + + + 2009 World Congress on Nature & Biologically Inspired Computing (NaBIC) + + IEEE + NaBIC 2009. 2009 + + + + Jain, A., Srinivas, E., and Rauta, R., "Short term load forecasting using fuzzy adaptive inference and similarity," Proceedings of the World Congress on Nature & Biologically Inspired Computing (NaBIC 2009), IEEE, 2009, pp. 1743 -1748. + + + + + Case-based systems + + PJuell + + + PPaulson + + 10.1109/mis.2003.1217629 + + + IEEE Intelligent Systems + IEEE Intell. Syst. + 1541-1672 + + 18 + 4 + + 2003 + Institute of Electrical and Electronics Engineers (IEEE) + + + Juell, P. and Paulson, P., "Using Reinforcement Learning for Similarity Assessment in Case-Based Systems," IEEE Intelligent Systems, Vol. 18, No. 4, 2003, pp. 60 -67. + + + + + Euclidean Distance as a Similarity Metric for Principal Component Analysis + + KimberlyLElmore + + + MichaelBRichman + + 10.1175/1520-0493(2001)129<0540:edaasm>2.0.co;2 + + + Monthly Weather Review + Mon. Wea. Rev. + 0027-0644 + 1520-0493 + + 129 + 3 + + 2001 + American Meteorological Society + + + Elmore, K. L. and Richman, M. B., "Euclidean Distance as a Similarity Metric for Principal Component Analysis," Monthly Weather Review , Vol. 129, No. 3, 2001, pp. 540 -549. + + + + + Day's Weather in the NAS: Visualization of Impact, Quantification, and Comparative Analysis + + AKlein + + + + Integrated Communications Navigation and Surveillance (ICNS) Conference + + 2006 + + + Klein, A., "Day's Weather in the NAS: Visualization of Impact, Quantification, and Comparative Analysis," Integrated Communications Navigation and Surveillance (ICNS) Conference, 2006. + + + + + Aggregate Statistics of National Traffic Management Initiatives + + JosephLRios + + 10.2514/6.2010-9382 + + + + 10th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + + American Institute of Aeronautics and Astronautics + 2010 + 29 + + + Enhanced Traffic Management System + Rios, J. L., "Aggregate Statistics of National Traffic Management Initiatives," Proceedings of the 10th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, 2010. 29 Federal Aviation Administration, "Enhanced Traffic Management System," http://hf.tc.faa.gov/projects/etms.htm. + + + + + Decision support tool for predicting aircraft arrival rates from weather forecasts + + DavidASmith + + + LanceSherry + + 10.1109/icnsurv.2008.4559186 + + + 2008 Integrated Communications, Navigation and Surveillance Conference + + IEEE + 2008 + + + ICRAT-2008 + Smith, D. A., Sherry, L., and Donohue, G., "Decision Support Tool for Predicting Aircraft Arrival Rates, Ground Delay Programs, and Airport Delays from Weather Forecasts," Proceedings International Conference on Research in Air Transportation (ICRAT-2008), 2008. + + + + + Improved boosting algorithms using confidence-rated predictions + + RobertESchapire + + + YoramSinger + + 10.1145/279943.279960 + + + Proceedings of the eleventh annual conference on Computational learning theory + the eleventh annual conference on Computational learning theory + + ACM + 1999 + 37 + + + Schapire, R. E. and Singer, Y., "Improved Boosting Algorithms Using Confidence-Rated Predictors," Machine Learning, Vol. 37, No. 3, 1999. + + + + + + diff --git a/file788.txt b/file788.txt new file mode 100644 index 0000000000000000000000000000000000000000..9c3718d3fd14992bec900032048e3807171fe218 --- /dev/null +++ b/file788.txt @@ -0,0 +1,518 @@ + + + + +IntroductionAirspace is divided into sectors to distribute air traffic controller workload.The goal is to balance the workload amongst all controllers, while not overwhelming any single controller.Overloading an air traffic controller is unsafe, and aircraft rerouting and delays are used to prevent overloading.Optimally partitioning the airspace, hereafter known as sectorization, requires a metric that quantifies controller workload.Previous research identified such a quantification, known as Dynamic Density, that includes a wide variety of air traffic and airspace metrics [1][2][3][4][5][6].This has been narrowed to a subset of key metrics known as the Simplified Dynamic Density [7,8].Geosect is a tool that partitions the airspace into sectors, while optimizing a cost function representing the magnitude of controller workload for each sector [9][10][11][12].Geosect accomplishes this using computational geometric techniques and constraints to produce sector shapes that are acceptable to air traffic controllers.In its current implementation, Geosect uses a cost function that is a combination of the number of aircraft predicted to be in a given sector (aircraft count) and the average amount of time each aircraft is predicted to spend in each sector (dwell time).While aircraft count and dwell time are both components of the Simplified Dynamic Density metric, they do not account for all factors that contribute to air traffic controller workload.For example, having ten aircraft evenly spaced and flying in the same direction in a sector causes less workload than having five aircraft all converging on the same point from different directions.The focus of this study is to evaluate the benefits of using Simplified Dynamic Density as a cost function in Geosect to partition airspace.Benefits are measured in terms of sector capacity and average system delay.Benefits are also measured using the Simplified Dynamic Density metric as a gauge of controller workload.This sort of benefit analysis is similar to that of Zelinski [13].At the time the Zelinski study was conducted, Geosect had not reached the necessary level of maturity to be included in the study.This paper gives a brief overview of the Geosect Airspace Partitioning Tool, followed by an explanation of the cost functions implemented in Geosect.The method by which the cost functions were tested is explained, followed by an analysis of the results.This is then followed by a discussion of future work.Finally, conclusions are drawn from the analysis. +2.D.1-2 +Overview of GeosectGeosect partitions or sectorizes airspace using computational geometric techniques.The particular version of Geosect examined in this study uses a top down approach with binary partitions.Given a region of airspace, Geosect partitions that airspace into two sectors.Each of these sectors is then evaluated and, if necessary, partitioned into two smaller sectors.Geosect continues partitioning until the desired number of sectors (an input parameter) has been created.Various geometric constraints ensure that the resulting sector shapes are acceptable to controllers in the presence of the anticipated traffic flow.Geosect's current implementation balances controller workload through the use of a cost function based on aircraft count and dwell times.Geosect begins with a single airspace region (usually the boundaries of an Air Route Traffic Control Center or "Center"), predicted aircraft track hits within the region, polylines representing the dominant air traffic flows through the region (as determined by the user), and the location of major airports and special use airspaces (SUAs) in the region (see Figure 1).In order to partition the region into sectors that are geometrically acceptable to air traffic controllers, Geosect sets up exclusion areas and search nodes. +Figure 1. Geosect Input ComponentsExclusion areas (blue circles in Figure 2) are created around major airports, SUAs, and dominant flow intersections.Geosect does not create partitions that encroach on these exclusion areas.This ensures that critical points such as airports, SUA corners, and dominant flow intersections are not too close to a sector boundary.Moreover, as partitions are tested and accepted by Geosect, exclusion zones are set up around the intersections of the partitions.This prevents the creation of points where four or more sectors meet.Such a point is undesirable because it creates hand-off ambiguities for air traffic controllers. +Figure 2. Exclusion Areas in Cleveland CenterGeosect constructs a series of external and internal search nodes (black squares in Figure 3).The external search nodes are equally spaced on the perimeter of the region to be partitioned.The internal search nodes are placed midway between the dominant flows.The segments joining these search nodes make up the search space for segments of each candidate partition.This ensures two properties of the partition.First, the external search nodes ensure that the partition extends from a point on the perimeter to another point on the perimeter.Second, the internal search nodes constrain bends in the partition to occur as far away from the dominant flows as possible. +Figure 3. Initial Search Nodes in Cleveland CenterGeosect further constrains candidate partitions to those whose orientation is orthogonal to the long axis of the region.Thus, if the region to be partitioned is long and skinny, Geosect will not partition the region 2.D.1-3 into two skinnier sectors.In addition, Geosect excludes candidate partitions that cross dominant flows at small angles (i.e.nearly parallel).Allowing small angles can create hand-off ambiguities for air traffic controllers.Given these geometric constraints, Geosect examines a series of partitions that divide the region into two smaller sectors.The cost function is applied to each of the sub-regions, and the highest cost of the two is associated with the partition.From the set of possible of partitions for a given region, Geosect selects the partition with the lowest cost associated with it.The resulting smaller sectors are then placed on a priority queue, which is ordered by a priority function, for further partitioning.The process of partitioning the highest priority sector in the queue and then adding the resulting smaller sectors back on the queue is repeated until the user-defined number of sectors has been created. +Cost FunctionsThis section describes the cost functions implemented in Geosect for this study.The inputs to the cost functions are aircraft radar track hits.Each track hit includes information on the aircraft's location, altitude, ground speed and heading.The track hits are one minute apart. +Aircraft Count/Dwell Time HybridThe first cost function examined by this study is a hybrid of aircraft dwell time and average aircraft count.The aircraft dwell time is the total time all aircraft spend in a given sector.It is desirable to maximize dwell time to reduce the number of handoffs required between sectors and thus the amount of controller coordination required.Therefore, the dwell time function is implemented as an inverse cost function where lower dwell times translate to higher cost.The aircraft dwell time function tends to create sectors that conform to the dominant flow of traffic.The average aircraft count function is the number of aircraft present in a sector averaged over the period the sectorization will take place.As described in the previous section, each candidate partition is evaluated based on the largest cost of the resulting two sectors.If the number of sectors created is less than half the desired number of sectors, then the aircraft dwell time function is used.After half the desired number of sectors has been created, then the average aircraft count cost function is used to evaluate the candidate partitions.The goal of this hybrid cost function is to create large super-sectors that align with the dominant flows.These super-sectors are then divided into smaller sectors to reduce and balance each controller's workload.The sectorization based on a hybrid of aircraft count and dwell time uses maximum aircraft count as its priority function.Maximum aircraft count is the greatest number of aircraft predicted to be within a sector at any given moment.The sector in the priority queue with the greatest maximum aircraft count is selected as the next sector to partition. +Occupancy Count ComponentThe Simplified Dynamic Density (SDD) metric is a weighted combination of seven components.The first SDD component, occupancy count, is the number of aircraft track hits within a sector averaged over a 15-minute period.More aircraft present in a sector at the same time implies higher controller workload.Occupancy count, x1 s,k , for sector s and 15minute period k is given byx1 s,k = n s,k 15 ,(1)where n s,k is the number of aircraft track hits in s during k.x1 s,k is a component of the SDD cost function.It, along with the other components, appears as a term in Eq. ( 4), shown later in this paper. +Composite Proximity Level ComponentProximity level is a quantification of how close the track hits of two aircraft are to each other in space and time.Aircraft that are predicted to be near each other in time and space could potentially come into conflict and increase the air traffic controller's workload.Even if they do not conflict, aircraft pairs with high proximity levels require monitoring by the air traffic controller.The track hits of each pair of aircraft are compared and assigned a proximity level.Table 1 lists the proximity levels from more severe to less severe.For example, if two aircraft are separated by between 5 and 7.5 nautical miles horizontally and less than 1000 feet vertically, and their track hits' timestamps were less than 10 seconds apart, then the aircraft pair are assigned a proximity level of 2. +2.D.1-4The composite proximity level, x2 s,k , for sector s and 15-minute period k is a composite of the identified Proximity Levels according tox2 s,k = 4 p 1,s,k + 2 p 2,s,k + p 3,s,k + p 4,s,k 4 ,(2)where p 1,x,k , p 2,x,k , p 3,x,k , and p 4,x,k represent the number of level 1, 2, 3, and 4 proximities, respectively. +Altitude Transition Count ComponentAltitude transition count, x3 s,k , is the number of aircraft track hits where the aircraft's absolute altitude change rate is greater than 500 feet per minute.Aircraft that are changing flight levels can increase an air traffic controller's workload as they merge or cross through streams of other aircraft. +Sector Transfer Count ComponentSector transfer count, x4 s,k , is the number of aircraft that enter and exit sector s during 15-minute period k.Note that an aircraft leaving sector r and entering sector s is counted in the sector transfer count for both s and r.Sector transfers require communication and coordination between sector controllers and pilots, and this adds to controller workload. +Sector Density ComponentSector density, x5 s,k , is the number of aircraft track hits in sector s during 15-minute period k (same as occupancy count x1 s,k ) divided by the volume of the sector (in km 3 ).Higher densities imply higher workload because the air traffic controllers have less airspace available to resolve conflicts. +Heading Variance ComponentHeading variance, x6 s,k , is a quantification of the variation in aircraft headings within a sector.The heading variance is given byx6 s,k = 1 N s,k (h i,s,k i=1 N s ,k ∑ -m s,k ) 2 ,(3)where m s,k is the mean of all N s,k headings, h i,s,k , in sector s during 15-minute period k.The motivation behind this metric is that it is easier to control streams of aircraft that are flying in the same direction than aircraft with a variety of headings and potentially crossing trajectories. +Speed Variance ComponentSpeed variance, x7 s,k , is based on the aircraft ground speeds in a manner similar to heading variance. +Simplified Dynamic DensityThe seven components of the SDD are combined in a weighted sum, ⎟ s,k , for sector s and 15-minute period k according toχ s,k = 2.2(x1 s,k ) + 1.2(x2 s,k ) + 0.2(x3 s,k ) + 0.4(x4 s,k ) + 3000(x5 s,k ) + 0.0005(x6 s,k ) + 0.0005(x7 s,k )(4)The SDD cost function, X s , is computed as the mean of ⎟ s,k over all of the 15-minute periods. +Method for Evaluating Cost FunctionsGeosect generated sectors in two altitude strata for Cleveland Center.Cleveland Center was selected for this study because its en route airspace includes a variety of crossing, climbing, and descending air traffic.High altitude sectors covered the first strata from 24,000 to 34,900 feet.Super high sectors 2.D.1-5 covered the second strata from 35,000 to 60,000 feet.This is the same stratification used by most of Cleveland's present day sectorization.The locations of the airports in Detroit, Cleveland, and Pittsburgh were input to Geosect to provide partitioning exclusion zones.Aircraft track hit data were from a day in 2005 that was not impacted by weather and spanned approximately 32 hours.The current sectorization includes 29 total sectors, covering the high and super high altitude strata.Since this averages to 15 sectors per stratum, Geosect was configured to generate 15 sectors for each altitude stratum using the aircraft count/dwell time cost function and the SDD cost function described above.Each Geosect generated sector was evaluated using SDD as the workload measure.Each sector of the current sectorization was similarly evaluated using SDD as the workload measure.The maximum, mean, and standard deviation over all of the sectors were computed for each sectorization.In addition, for each sectorization, the high and super high sectors were input into a simulation using the Airspace Concept Evaluation System (ACES) [14].First, ACES simulated current (2005) levels of air traffic unconstrained by sector capacities.This unconstrained traffic data was then used to compute the Monitor Alert Parameter (MAP) values for each sector.The Federal Aviation Administration (FAA) uses MAP values as an indication of sector capacity [15].MAP values range from 5 (lowest capacity) to 18 (highest capacity).Next, ACES simulated the same traffic but imposed delays so that each sector's capacity, as given by its MAP value, was not exceeded.The resulting average delay over the entire system was recorded. +Results and AnalysisThe SDD function is used as the workload measure in the analysis of each Geosect generated sectorization as well as the current day Cleveland sectorization.The maximum is a sectorization's worst-case workload, and the mean is a sectorization's average workload over the entire period under consideration.The standard deviation is interpreted as a sectorization's workload balance among its individual sectors.The smaller the standard deviation, the better the balance.In addition, the average MAP value and average system delay for each sectorization are derived from ACES simulations and compared. +WorkloadFigure 4 shows the maximum SDD measure for each of the sectorizations.(Note that the sectorization based on the aircraft count/dwell time hybrid cost function is abbreviated as AC/DTH in the following figures.)Using SDD as a measure of workload, both Geosect sectorizations had smaller worst-cases than the current sectorization.The sectorization generated by Geosect using SDD as its cost function had the least worst-case. +Capacity and DelayMean MAP values are shown in Figure 7. Interpreting MAP value as an indication of capacity, neither of the Geosect sectorizations had an average sector capacity as high as the current sectorization.However, as shown in Figure 8, the capacity is better balanced between the sectors created by Geosect using the aircraft count/dwell time hybrid than the current day sectors.Given that dwell time is closely related to the computation of MAP values, it makes sense that the aircraft count/dwell time hybrid sectorization scored better than the SDD sectorization.ACES also computed the average system delay for each sectorization, and the results are shown in Figure 9.The current day sectorization has less average delay than either Geosect sectorizations.Among the Geosect generated sectorizations, the aircraft count/dwell time hybrid sectorization had less delay than the SDD sectorization.Again, this may be due to the fact that dwell time is closely related to the MAP value calculations used by ACES. +Summary of ResultsTable 2 summarizes which sectorizations did the best under the various evaluations used in this study.Using the aircraft count/dwell time cost function, Geosect does a good job of balancing workload and capacity.However, current day sectorization outperforms Geosect in the areas of average workload, sector capacity and system delay.Design of the current sectorization is an art form requiring the efforts of an experienced airspace designer who designs the sectors in conjunction with the design of the air traffic routes.As a result, the current sectorization scored well in this study since the traffic used to evaluate the sectorizations followed these routes.Given a different set of traffic that do not follow the routes, either due to weather or some sort of wind-optimized routing, the evaluations may have yielded different results.The current sectorization has also benefited from years of evolution while the design of automated sectorization tools is fairly new.The ACES simulations show that the current sectorization design strives to optimize airspace capacity while reducing system delay.While Geosect does a good job of balancing workload and capacity, further development is required to bring it up to the level of the current sectorization, and this is covered in the next section. +Future WorkWithin each altitude stratum (high and super high), the current day sectorization included sectors that occupied only portions of the stratum.In those cases, sectors overlapped each other with different sectors covering different portions of the altitude stratum.The version of Geosect used in this study was limited by its two-dimensional nature.Each sector created by Geosect was expected to cover the full altitude stratum, and the stratum is specified by the user.A future study should look at using Geosect to generate three-dimensional sectors, where Geosect would pick its own altitude strata.The resulting sectors would overlap would not necessarily all have the same floors and ceilings.At the time of this study, the dominant flows used by Geosect were created by hand.In future, Geosect will use automatically generated dominant flows.This will make dominant flows more consistent between different Centers, and will make Geosect easier to use with different altitude strata.Finally, at the time of this study, the user specified how many sectors for Geosect to generate.A future enhancement would allow Geosect to create an undetermined number of sectors until some criterion was met.This would alleviate the problem 2.D.1-8 of small sectors being generated merely because Geosect had not reached its user-specified limit. +ConclusionTwo sets of sectorizations of Cleveland Center were generated by Geosect using different cost functions.These sectorizations and the current day sectorization were compared by applying the Simplified Dynamic Density measure (representing workload) to each sector.In addition, the Airspace Concept Evaluation System was used to simulate current air traffic.From these simulations, sector capacity and average system delay were computed for each sectorization.Geosect generated sectorizations using the aircraft count/dwell time hybrid cost function gave a better workload balance and sector capacity balance than the current sectorization.The sectorization resulting from the Simplified Dynamic Density cost function had a lower maximum workload measure than the other sectorizations.However, Geosect's sectorizations incurred greater delay and did not generate as much sector capacity as the current sectorization.Figure 5 Figure 6 Figure 6 .566Figure 4. Maximum SDDFigure5shows the mean SDD measure for each of the sectorizations.The current sectorization has a significantly smaller average workload than either of the Geosect sectorizations.Geosect's average workload was slightly lower using the SDD cost function versus the sectorization based on the aircraft count/dwell time hybrid. +Figure 8 .8Figure 7. Mean MAP Value +Figure 9 .9Figure 9. Average System Delay +Table 1 . Proximity Level Criteria1VerticalHorizontalTimeProx.SeparationSeparationSeparationLevel(ft.)(nm)(seconds)1< 1000< 5< 102< 10005 to 7.5< 10< 5< 203< 10007.5 to 10< 105 to 7.5< 20< 5< 304any< 5< 10 +Table 2 . Summary of the Best Sectorizations2ComparisonBest Sectorizationmaximum workloadSDDaverage workloadcurrentaircraft count/workload balancedwell time hybridAverage Sector Capacity currentaircraft count/Sector Capacity BalanceDwell time hybridAverage System Delaycurrent + + + + +AcknowledgementsThe author thanks Girishkumar Sabhnani at State University of New York (SUNY) at Stony Brook (now at Metron Aviation) for his assistance and expertise with Geosect, and Charlene Cayabyab for her ACES support at NASA Ames Research Center. + + + +Email AddressesGregory L. Wong: Gregory.L.Wong@nasa.gov28th Digital Avionics Systems Conference October [25][26][27][28][29]2009 + + + + + + + Dynamic Density: An Air Traffic Management Metric + + IVLaudeman + + + SGShelden + + + RBranstrom + + + CLBrasil + + NASA- TM-1998-112226 + + April 1998 + Moffett Field, California + + + Laudeman, I. V., S. G. Shelden, R. Branstrom, C. L. Brasil, April 1998, Dynamic Density: An Air Traffic Management Metric, NASA- TM-1998-112226, Moffett Field, California. + + + + + Integration of Traffic Flow Management Decisions + + BanavarSridhar + + + GanoChatterji + + + ShonGrabbe + + + KapilSheth + + 10.2514/6.2002-5014 + + + AIAA Guidance, Navigation, and Control Conference and Exhibit + + American Institute of Aeronautics and Astronautics + + + + + Sridhar, B., K. S. Sheth, S. Grabbe, 1-4 + + + + + Relationship between Air Traffic Demand, Safety and Complexity in High-Density Airspace in Europe + + TamaraPejovic + + + FedjaNetjasov + + + DusanCrnogorac + + 10.5772/intechopen.88801 + + + Risk Assessment in Air Traffic Management + Orlando, Florida + + IntechOpen + December 1998 + + + December 1998, Airspace Complexity and its Application in Air Traffic Management, 2nd USA/Europe Air Traffic Management R&D Seminar, Orlando, Florida. + + + + + Measures for air traffic controller workload prediction + + GanoChatterji + + + BanavarSridhar + + 10.2514/6.2001-5242 + AIAA-2001-5242 + + + 1st AIAA, Aircraft, Technology Integration, and Operations Forum + Los Angeles, California + + American Institute of Aeronautics and Astronautics + 16-18 October 2001 + + + Measures for Air Traffic Controller Workload Prediction + Chatterji, G. B., B. Sridhar, 16-18 October 2001, Measures for Air Traffic Controller Workload Prediction, AIAA-2001-5242, Proceedings of the First AIAA Aircraft Technology, Integration, and Operations Forum, Los Angeles, California. + + + + + Dynamic density: measuring and predicting sector complexity [ATC] + + PKopardekar + + + SMagyarits + + 10.1109/dasc.2002.1067920 + + + Proceedings. The 21st Digital Avionics Systems Conference + The 21st Digital Avionics Systems ConferenceBudapest, Hungary + + IEEE + June 2003 + + + Kopardekar, P., S. Magyarits,, June 2003, Measurement and Prediction of Dynamic Density, 5th USA/Europe Air Traffic Management R&D Seminar, Budapest, Hungary. + + + + + Dynamic Density and Complexity Metrics for Realtime Traffic Flow Management, 5th USA + + AMasalonis + + + MCallaham + + + CWanke + + + + Europe Air Traffic Management R&D Seminar + + June 2003 + Budapest, Hungary + + + Masalonis, A., M. Callaham, C. Wanke, June 2003, Dynamic Density and Complexity Metrics for Realtime Traffic Flow Management, 5th USA/Europe Air Traffic Management R&D Seminar, Budapest, Hungary. + + + + + + + PKopardekar + + + ASchwartz + + + SMagyarits + + + JRhodes + + + + Airspace Complexity Measurement: an Air Traffic Control Simulation Analysis, 7th USA + + 2-5 July 2007 + Barcelona, Spain + + + Europe Air Traffic Management R&D Seminar + Kopardekar, P., A. Schwartz, S. Magyarits, J. Rhodes, 2-5 July 2007, Airspace Complexity Measurement: an Air Traffic Control Simulation Analysis, 7th USA/Europe Air Traffic Management R&D Seminar, Barcelona, Spain. + + + + + Dynamic Airspace Configuration Analysis, Deliverables for Subtasks 13 & 14 under Contract GS-35F-0308K; Order No. A65616D + + SKlein + + + MRodgers + + + HKaing + + + PLucic + + + KLeiden + + + + Final Report Part + + 3 + 3 December 2008 + NASA Ames Research Center + Moffett Field, California + + + DAG CE-6 Golden Nuggets Fast-Time Modeling and Simulation Studies + Klein, S., M. Rodgers, H. Kaing, P. Lucic, K. Leiden, 3 December 2008, DAG CE-6 Golden Nuggets Fast-Time Modeling and Simulation Studies, Final Report Part 3: Dynamic Airspace Configuration Analysis, Deliverables for Subtasks 13 & 14 under Contract GS-35F-0308K; Order No. A65616D, NASA Ames Research Center, Moffett Field, California. + + + + + Simplified dynamic density: A metric for dynamic airspace configuration and NextGen analysis + + AlexanderKlein + + + MarkDRodgers + + + KennethLeiden + + 10.1109/dasc.2009.5347539 + + + 2009 IEEE/AIAA 28th Digital Avionics Systems Conference + Orlando, Florida + + IEEE + + + + Klein, A., M. Rodgers, K. Leiden, Simplified Dynamic Density: a Metric for Dynamic Airspace Configuration and NEXTGEN Analysis, 28 th Digital Avionics Systems Conference, Orlando, Florida. + + + + + Dynamic Airspace Configuration Management Based on Computational Geometry Techniques + + JoeMitchell + + + GirishkumarSabhnani + + + RobertHoffman + + + JimmyKrozel + + + ArashYousefi + + 10.2514/6.2008-7225 + + + AIAA Guidance, Navigation and Control Conference and Exhibit + + American Institute of Aeronautics and Astronautics + + + + Mitchell, J. S. B., G. Sabhnani, J. Krozel, R. + + + + + + AHoffman + + + Yousefi + + Dynamic Airspace Configuration Management based on Computational Geometry Techniques, AIAA 2008-7225, AIAA Guidance, Navigation, and Control Conference + Honolulu, Hawaii + + 18-21 August 2008 + + + Hoffman, A. Yousefi, 18-21 August 2008, Dynamic Airspace Configuration Management based on Computational Geometry Techniques, AIAA 2008- 7225, AIAA Guidance, Navigation, and Control Conference, Honolulu, Hawaii. + + + + + Geometric Algorithms for Optimal Airspace Design and Air Traffic Controller Workload Balancing + + AmitabhBasu + + + JosephS BMitchell + + + GirishkumarSabhnani + + 10.1137/1.9781611972887.8 + + + 2008 Proceedings of the Tenth Workshop on Algorithm Engineering and Experiments (ALENEX) + SIAM Proceedings in Applied Mathematics + + Society for Industrial and Applied Mathematics + 2008. 2008 + + + + Basu, A., J. S. B. Mitchell, G. Sabhnani, 2008, Geometric Algorithms for Optimal Airspace Design and Air Traffic Controller Workload Balancing, Proc. Tenth Workshop on Algorithm Engineering and Experiments, ALENEX'08, SIAM Proceedings in Applied Mathematics, (Journal version submitted to ACM Journal of Experimental Algorithmics, 2008). + + + + + Geometric Algorithms for Optimal Airspace Design and Air Traffic Controller Workload Balancing + + AmitabhBasu + + + JosephS BMitchell + + + GirishkumarSabhnani + + 10.1137/1.9781611972887.8 + + + 2008 Proceedings of the Tenth Workshop on Algorithm Engineering and Experiments (ALENEX) + Smith College, Northampton, Massachusetts + + Society for Industrial and Applied Mathematics + 10-11 November 2006 + + + + Basu, A., J. S. B. Mitchell, G. Sabhnani, 10- 11 November 2006, Geometric Algorithms for Optimal Airspace Design and Air Traffic Controller Workload Balancing, Proc. 16th Fall Workshop on Computational and Combinatorial Geometry, Smith College, Northampton, Massachusetts. + + + + + Geometric Algorithms for Dynamic Airspace Sectorization + + GRSabhnani + + + 2009 + Stony Brook, New York + + + PhD Dissertation, Stony Brook University + + + Sabhnani, G. R., 2009, Geometric Algorithms for Dynamic Airspace Sectorization. PhD Dissertation, Stony Brook University, Stony Brook, New York. + + + + + A Comparison of Algorithm Generated Sectorizations + + ShannonZelinski + + 10.2514/atcq.18.3.279 + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 18 + 3 + + 29 June -2 July, 2009 + American Institute of Aeronautics and Astronautics (AIAA) + Napa, California + + + Zelinski, S., 29 June -2 July, 2009, A Comparison of Algorithm Generated Sectorizations, Eighth USA/Europe Air Traffic Management 2.D.1-9 Research and Development Seminar (ATM2009), Napa, California. + + + + + Build 4 of the Airspace Concept Evaluation System + + LarryMeyn + + + RobertWindhorst + + + KarlinRoth + + + DonaldVan Drei + + + GregKubat + + + VikramManikonda + + + SharleneRoney + + + GeorgeHunter + + + AlexHuang + + + GeorgeCouluris + + 10.2514/6.2006-6110 + + + AIAA Modeling and Simulation Technologies Conference and Exhibit + + American Institute of Aeronautics and Astronautics + + + + Meyn, L., R. Windhorst, K. Roth, D. Van + + + + + + GDrei + + + VKubat + + + SManikonda + + + GRoney + + + AHunter + + + GHuang + + + Couluris + + Build 4 of the Airspace Concept Evaluation System, AIAA 2006-6110, AIAA Modeling and Simulation Technologies Conference + Keystone, Colorado + + August 2006 + + + + Federal Aviation Administration. Order JO + Drei, G. Kubat, V. Manikonda, S. Roney, G. Hunter, A. Huang, G. Couluris, 21-24 August 2006, Build 4 of the Airspace Concept Evaluation System, AIAA 2006-6110, AIAA Modeling and Simulation Technologies Conference, Keystone, Colorado. [15] Federal Aviation Administration, Order JO + + + + + Examining Change + 10.4135/9781483329680.n4 + + + + Schools That Change: Evidence-Based Improvement and Effective Change Leadership + + Corwin Press + 2008. March 12, 2009 + 8 + + + + Monitor Alert Parameter + 3V, 2008, Section 8. Monitor Alert Parameter, [online publication] URL: http://www.faa.gov/air_traffic/publications/atp ubs/FAC/Ch17/s1708.html, effective date: February 14, 2008, includes change effective July 31, 2008 and change 2 effective March 12, 2009. + + + + + + diff --git a/file789.txt b/file789.txt new file mode 100644 index 0000000000000000000000000000000000000000..a2d4a0b4f4e89280ae42b2fb66b531c2b3ed7278 --- /dev/null +++ b/file789.txt @@ -0,0 +1,1962 @@ + + + + +I. INTRODUCTION A. OverviewThe Center-TRACON Automation System (CTAS) is a set of automation tools that assist air traffic controllers and traffic management coordinators (TMCs) in the Air Route Traffic Control Center (ARTCC, also known as Center) and the Terminal Radar Approach Control (TRACON) with the planning and control of arrival air traffic [1][2] [3].CTAS is designed to generate advisories that assist air traffic controllers to improve airport capacity and reduce delays.CTAS has shown benefits such as reducing controller workload and increasing situational awareness without decreasing safety.Currently, there are three primary tools, in various stages of development, that make up CTAS.The Traffic Management Advisor (TMA) assists Center controllers and TMCs with the scheduling of arrival aircraft within the Center's airspace [4][5] [6].The Center's airspace extends from approximately 35 to 200 nautical miles from the airport (see figure 1).The En Route/Descent Advisor (E/DA) works in conjunction with the TMA and assists Center controllers in meeting the schedules set by the TMA [7].The Final Approach Spacing Tool (FAST) provides advisories to TRACON air traffic controllers to balance runways and sequence arrival aircraft within 40 nautical miles of and less than 10,000 feet altitude above the airport (see figure 1) [8][9] [10].CTAS also includes a capability referred to as Conflict Probe which assures that CTAS advisories are conflict free [11] [12].The Dynamic Planner (DP) is the main computational engine of the TMA and computes the sequences and schedules of arrival aircraft in the Center.Before the 1970s, the United States air traffic control system concerned itself primarily with maintaining proper separation between aircraft.The TMCs dealt mainly with special situations such as weather and equipment failures.Under normal situations, the traffic was allowed to run freely until it became necessary to hold the traffic in the terminal area.However, holding traffic at low altitudes is not fuel efficient.The energy crisis of the 1970s resulted in an increasing emphasis on fuel efficiency.This, coupled with the air traffic controller strike of 1981, made it clear that managing the traffic flow was necessary to increase fuel efficiency, and to smooth out the workload on controllers while maintaining safety [13].Thus, TMCs were given the added responsibility for traffic flow management.The TMA assists the Center TMCs and controllers in a number of ways.First, the TMA helps TMCs anticipate the future traffic flow by predicting where an aircraft will be at any point in the future.The TMA also uses these predictions to compute the desired sequences and arrival times to various reference points to improve the overall flow of traffic.To help optimize the arrival times, the TMA computes runway assignments for each aircraft.In addition, the TMA assists with traffic analysis by generating statistics and reports about the traffic flow.The TMA is to be deployed nationally as part of the Federal Aviation Administration's (FAA) Free Flight Phase I program.The DP's role in the TMA is to compute the orderly sequence, arrival times, and runway assignments to ensure a smooth flow of traffic into the terminal area.Normally, the DP sequences the aircraft so that they arrive in a first-come-first-served (FCFS) order, but TMCs can override this order by inputting sequence constraints.In addition, the TMC may input scheduling constraints which restrict the traffic flow or affect the required separation between aircraft.These scheduling constraints may be necessary due to current runway capacity, traffic density, aircraft type distribution, airport configuration, and weather conditions.This paper describes the details of the DP's modules and algorithms.There has been considerable effort in the development of schedulers for air traffic.Most of these include some variation of FCFS order [13] [14].There have also been attempts at designing a scheduler that specifically optimizes a cost function [15].Some of these schedulers lack sufficient accuracy, while others are incompatible with the number of constraints that are required for use by TMCs in busy air traffic control facilities such as Ft.Worth Center.The DP is conceptually similar to many of these schedulers previously developed, but it has been designed to:1. explicitly account for a variety of specified traffic constraints, 2. provide schedule updates that respond to the CTAS trajectory predictions, 3. optimally assign delay between the TRACON and Center, and 4. allow for the inclusion of advanced optimization features at a later date. +B. Paper OrganizationThis paper describes the details of the DP modules and algorithms.This paper begins with a view of the DP from outside of the DP.The inputs to the DP and the outputs from the DP are summarized.This is followed by descriptions of the other CTAS processes required to run concurrently with the DP.Next is a list of terms and definitions.The view switches to the internals of the DP with a list and explanation of the indicators and flags that the DP uses to identify special situations.This is followed by a description of the DP's principal modules.These are the main body, sequencing, scheduling, runway allocation, and miles-in-trail advisor modules.The important data structures and algorithms for each module are explained.Next, this paper describes the ETA Hovering mechanism used to compensate for errors in each aircraft's coordination fix time.This is followed by a discussion on the broadcast of the schedules computed by the DP.Next this paper describes the DP's object-oriented design methodology.Finally, this paper describes the DP's current status and future direction. +II. DESCRIPTION A. General DescriptionThe flight paths of some typical arrival aircraft are shown in figure 2. As each arrival aircraft flies through the Center's airspace, it crosses the outer meter arc.The outer meter arc is approximately 60 nm outside its Figure 1.Air Traffic Control Airspaces associated meter fix.The aircraft then crosses the meter fix which lies at the boundary between the airspaces of the Center and TRACON.For some Centers, the DP groups meter fixes together, and these meter fix groups are called gates.From the meter fix, the arrival aircraft flies through the TRACON's airspace, crosses the final approach fix (FAF), and touches down on the runway.The DP schedules an arrival time at the outer meter arc, meter fix, FAF, and runway threshold for each aircraft.The outer meter arc 1 , meter fix, FAF, and runway threshold are collectively known as Reference Points, and each of these computed arrival times is called a Scheduled Time of Arrival (STA).To account for various traffic, weather, and airport conditions, the TMC (the primary user of the DP) can control the schedule by inputting scheduling constraints such as separation distances and acceptance rates.The DP will obey these scheduling constraints when computing the STA for each arrival aircraft.In addition, the DP sequences the aircraft to arrive at the meter fix in FCFS order.The TMC may alter this sequence by inputting specific sequence constraints.Furthermore, the DP, through a process known as Runway Allocation, will assign aircraft to runways that reduce the overall delay.The TMC may override runway assignments and manually assign an aircraft to a particular runway.All sequencing, scheduling, and runway allocation take place while the aircraft is in the Center's airspace (approximately 25 to 300 miles from the arrival airport).Moreover, scheduling of some aircraft takes place before the aircraft have even entered Center's airspace.The DP only requires an aircraft's flight plan for scheduling.This can occur as early as 90 minutes before the aircraft enters the Center's airspace.The DP updates these sequences, schedules, and runway assignments constantly to adapt to changes in the traffic situation, changes in the environment, or in response to inputs by TMCs. 1 All of the points that make up the outer meter arc are collectively treated as a single point by the TMA.Thus, an outer meter arc arrival time is the time that the aircraft will cross any point along the outer meter arc.The inputs to the DP are summarized below. +Flight PlansFlight plans contain fundamental information on each aircraft that is in, or due to enter, the Center's airspace.The flight plans are submitted by the airlines and processed by the Center's Host computer.Flight plans may also be amended by air traffic controllers while the aircraft is under their command.The DP receives a flight plan for each aircraft being processed by CTAS.The DP uses the following flight plan information for each aircraft:• Aircraft's identification • Aircraft's type and characteristics • Aircraft's planned route of flight • Anticipated time when the aircraft will enter the Center's airspace• Flight plan status (see below)The following are the possible flight plan status indicators which DP receives from the Center's Host computer:Estimated Flight Plan.The aircraft is entering the Center's airspace from an adjacent Center.Proposed Flight Plan.The aircraft is anticipated to depart from an airport within the Center's airspace.Departed Flight Plan.The aircraft has departed from an airport within the Center's airspace. +Track UpdatesCTAS receives radar tracks, from the Center's Host computer, on aircraft that are in the Center's airspace.Those aircraft which have radar tracks are known as active aircraft.The DP maintains flight plans for both active and inactive aircraft. +Estimated Times of Arrival (ETAs)The TMA's Route Analyzer (RA) and Trajectory Synthesizer (TS) programs work together to provide the DP with Estimated Times of Arrival (ETAs) to each of the reference points shown in figure 2. The RA computes the horizontal route of the arriving aircraft as well as the various speed restrictions at various points along the route.The TS then takes this information, along with highly accurate aircraft and weather models, and computes a complete 4-dimensional trajectory from the aircraft's current location to touchdown at the runway threshold.From this trajectory, the RA can extract the ETAs to the reference points and send these ETAs to the DP.The ETAs are recomputed with each radar update.From the DP's perspective, the ETAs are a prediction of when each aircraft will cross each reference point if there were no other aircraft in the airspace.Hence, the DP treats the ETAs as the earliest allowed STA.When building a schedule involving all aircraft, each aircraft's STA may have to be delayed in order to avoid a conflict with other aircraft.Thus, barring any manual interaction by the controller or TMC, each aircraft's STA will be equal to or later than the aircraft's ETA. +Overcrossing TimesThe RA sends to the DP the time that an aircraft enters the TRACON's airspace from the Center's airspace.Within the DP, this time is referred to as the overcrossing time.Usually, this is the time that the aircraft crosses the meter fix.However, some aircraft do not actually cross a meter fix when entering the TRACON's airspace, and the RA has special logic to account for this.In either case, the DP interprets the overcrossing time as the time that the aircraft has crossed the meter fix. +Scheduling ConstraintsWhen the DP computes the STAs, it obeys the scheduling constraints.These scheduling constraints are entered by the TMC to reflect the actual current and future airport capacity, mix of traffic, weather conditions, staffing level, runway topology, and air traffic control procedures.When there are no scheduling constraints, the computed STA for each aircraft will be equal to its ETA.However, when the traffic is heavy enough or the scheduling constraints are restrictive enough, the DP will begin to delay aircraft to accommodate the scheduling constraints.As a result, aircraft STAs will be later than their ETAs.The scheduling constraints are listed below and grouped by similarity in software implementation. +• Separation DistanceMiles-in-trail.This is the minimum horizontal distance allowed between aircraft as they cross a particular meter fix.Wake Vortex Separation.This is the minimum distance allowed between aircraft as they land at a particular runway. +• Occupancy TimeRunway Occupancy Time.This is the minimum amount of time allowed between landings at a particular runway.The TMC may enter this scheduling constraint to account for runway stopping conditions or extra time required to clear the runway. +• Blocked IntervalMeter Fix Blocked Interval.This is a period of time during which no aircraft are allowed to cross a particular meter fix. +Runway BlockedInterval.This is a period of time during which no aircraft are allowed to land on a particular runway. +Sequence ConstraintsThe DP normally schedules aircraft to arrive at the meter fix in FCFS order based on their ETAs to the meter fix.However, the TMC can enter sequence constraints which force certain aircraft to be scheduled before or after other aircraft.The DP must take all such constraints into account when generating a sequence at the meter fix. +C. Outputs from the Dynamic PlannerThe outputs from the DP are summarized below.• STAs• Runway Assignments +Scheduled Times of Arrival (STAs)The DP's calculations result in a set of STAs for each aircraft at the following reference points (see figure 2).• Outer meter arc +Runway AssignmentsIn a process known as Runway Allocation, the DP will assign each aircraft to a runway.These runway assignments are designed to reduce the total delay of all aircraft and thus optimize the schedule.Additionally, the TMC can override the DP's computed runway assignment and manually assign an aircraft to a particular runway. +D. Other CTAS Components RequiredThe DP is a significant software component of the TMA, computing the schedule, sequences, and runway assignments based on the current air traffic situation.The DP cannot work alone.It relies on the TMA's other software components (see figure 3).The most important of these are described below. +Weather Data Acquisition Daemon (WDAD) and Weather Data Processing Daemon (WDPD)The Weather Data Acquisition Daemon (WDAD) and Weather Data Processing Daemon (WDPD) acquire and process the atmospheric data that CTAS receives from an external host computer.Currently, atmospheric data is provided by the National Centers for Environmental Prediction (NCEP). +Host Data Acquisition and Routing (HDAR) and Input Source Manager (ISM)The Host Data Acquisition and Routing (HDAR) process and the Input Source Manager (ISM) serve as the interface between the Center's Host computer and CTAS.This is a two-way interface through which flight plan information and radar tracks are sent from the Host to CTAS while schedules and flight plan amendments are sent from CTAS to the Host. +Communications Manager (CM)The Communications Manager (CM) controls the communication between each CTAS software process and maintains a central database to keep all the processes in synch with one another.All of the DP's input and output data (such as schedules and flight plan amendments) are received from, or sent to, the CM.It is the CM's responsibility to distribute the data to the proper CTAS processes.In addition, CM maintains a centralized database that includes all of the scheduling constraints entered by the user.Every time a CTAS process is started and connects to the CM, the database is transferred to the newly connecting process to synchronize it with the rest of CTAS. +Trajectory Synthesizer (TS)The Trajectory Synthesizer (TS) is the computational engine of CTAS [16][17].The thoroughness and accuracy of the TS is a significant improvement in generating ETAs over that of previous engineering efforts.The DP is a beneficiary of this improvement and can create realizable schedules given the highly accurate ETAs provided by the TS. +Route Analyzer (RA)The Route Analyzer (RA) generates all possible realistic horizontal routes for an aircraft.These routes extend from the aircraft's current position to its end point based on a set of site-adaptable analysis rules.These horizontal routes are computed for each aircraft every time a radar update is received.The horizontal routes, along with the aircraft's initial condition, desired end conditions, and intermediate altitude and speed restrictions, determined by the RA, are passed along to the TS.From the resulting TS output, the RA extracts the fast, nominal, and slow ETAs to every eligible runway threshold, FAF, assigned meter fix and outer meter arc.Currently, the DP only uses the nominal ETAs.The fast and slow ETAs are reserved for future scheduling research.Additionally, the RA extracts the ground speeds at these points.The RA then sends these ETAs and ground speeds to the DP.Finally, when the RA detects that an aircraft has crossed its meter fix or has otherwise entered the TRACON's airspace from the Center's airspace, it sends the DP the overcrossing time of that aircraft. +Timeline and Planview Graphical User Interfaces (TGUI and PGUI)The Timeline Graphical User Interface (TGUI) and the Planview Graphical User Interface (PGUI) provide the principal means by which the user may interact with the DP.All scheduling and sequencing constraints and any other manual inputs to the DP are done through the Figure 3. TMA Software Components various panels and displays provided by the GUIs.The DP's output, such as schedules and aircraft status, is displayed to the user on the GUIs.The GUIs also provide other information which serve to enhance the TMA user's situational awareness. +E. TerminologyThe following terms and concepts are used to describe the DP.Some terms may be specific to the DP and are not part of the air traffic control (ATC) vernacular. +Blocked SlotBlocked slots can best be thought of as phantom aircraft.A blocked slot may be used to hold open a space where a real but untracked aircraft is anticipated to be.The TMC may create a blocked slot relative to a meter fix or a runway.For a meter fix blocked slot, the user specifies the blocked slot type, assigned meter fix, arrival airport, and ETA to the meter fix.For a runway blocked slot, the TMC specifies the blocked slot type, assigned runway, and ETA to the runway.The blocked slot type is selected by the TMC from a set of engine type and weight class combinations.A flight plan is created by the CM using an aircraft model representative of the engine type and weight class the user has selected (see table 1).This flight plan is then distributed throughout CTAS as an inactive aircraft (since there are no tracks for it).For a meter fix blocked slot, the TMC specifies the meter fix ETA of the blocked slot.The ground speeds at the meter fix, FAF, and runway threshold, and the ETAs to the FAF and runway threshold are computed by the RA based on nominal trajectory information for the representative aircraft model.For a runway blocked slot, the TMC specifies the runway ETA.The ground speed at the runway is computed by the RA based on nominal trajectory information for the representative aircraft model. +Schedulable Objects (SOs)The DP computes STAs for Schedulable Objects (SOs).Aircraft and blocked slots are collectively known as SOs. +Acceptance Rate IntervalAlthough the acceptance rate is expressed as the number of SOs per hour, the actual time period used by the DP is a fraction of an hour.This time period is known as the Acceptance Rate Interval and varies from site to site. +Stream ClassesCenter arrival traffic is categorized into stream classes.Each stream class contains SOs with similar scheduling characteristics.Currently, this categorization is based on engine type, destination airport, and assigned meter fix. +Super Stream ClassesDuring runtime, stream classes with similar scheduling characteristics may be grouped into super stream classes.Every stream class will be included in one super stream class though several stream classes may be placed in the same super stream class. +Reference PointThe runway threshold, FAF, meter fix, and outer meter arc are all reference points.For each SO, the DP computes STAs to various reference points, and the scheduling constraints are applied at these reference points. +Transition TimeThe transition time is the time an SO takes to fly from one reference point to another if there were no other SOs in the system.The transition time is computed by taking the ETA of the destination reference point and subtracting the ETA of the source reference point.For example, if the meter fix ETA is 1045Z and the FAF ETA is 1102Z for a particular SO, the transition time from the meter fix to the FAF is 17 minutes. +Flight RulesThe DP's computation of STAs to the runway depends on the flight rules in effect for the airport's configuration.The two flight rules handled by the DP are Instrument Flight Rules (IFR) and Visual Flight Rules (VFR).These flight rules, within the DP's context, are defined below. +Instrument Flight Rules (IFR)Under IFR conditions, the controllers guide the aircraft all the way to touchdown.Hence, the DP will compute the STA for each aircraft at the runway threshold to meet the runway and airport scheduling constraints.The FAF STA is computed by subtracting the transition time between the FAF and the threshold from the threshold STA. +Visual Flight Rules (VFR)Under VFR conditions, the controllers guide the aircraft to the FAF.Hence, the DP will compute the STA for each aircraft at the FAF to meet the runway and airport scheduling constraints.The threshold STA is computed by adding the transition time between FAF and the threshold to the FAF STA. +ConfigurationsThe TMC specifies an airport's current and future configurations and the time of the future configuration changes.The configuration defines which runways are active, which runways are dependent on each other, and which flight rules are in effect.When scheduling to dependent runways, the DP treats the runways within a dependent set as if they were a single runway.Associated with each configuration is a flow parameter set.One flow parameter set is selected for a configuration from a group of flow parameter sets available for that configuration. +Flow Parameter SetsFlow parameter sets are scheduling constraint macros associated with each airport configuration.A flow parameter set is designed to set a number of scheduling constraints to control the traffic flow rate to the airport. +STA FreezeAn SO will become STA-Frozen when its ETA at the meter fix is less than or equal to M minutes in the future.The value of M is known as the STA Freeze Horizon and varies from stream class to stream class and from site to site.A typical value is 19 minutes for jets.Inside the STA Freeze Horizon, an SO's STA will not change when the schedule is updated.Exception: STA-Frozen SOs will have their STAs recomputed and, possibly, changed in response to scheduling events that correspond to Scheduling Modes 1 through 5 (see table 2). +Sequence FreezeAn SO will become Sequence-Frozen when its ETA at the meter fix is less than or equal to N+M minutes in the future.The value of N is known as the Sequence Freeze Horizon and M is the STA Freeze Horizon.Both N and M vary from stream class to stream class and from site to site.A typical value for N is 5 minutes.Because both N and M are non-negative numbers, aircraft that are STA-Frozen are also Sequence-Frozen.Unlike STA-Frozen SOs, Sequence-Frozen SOs can have their STAs changed during the scheduling process.For Scheduling Modes 4, 5, 8, and 9 (see table 2), Sequence-Frozen aircraft will be scheduled such that they maintain their super stream class sequence.For example, if aircraft A and B are both Non-Sequence-Frozen aircraft in the same super stream class, and A's ETA is earlier than B's ETA, then A will be placed in the sequence ahead of B. Later, if A and B become Sequence-Frozen and A's ETA becomes later than B's ETA, then A will still be placed in the sequence ahead of B.Sequence Freeze only applies to meter fix sequences.Sequences at the runway threshold or FAF may change.Also, some events will result in the resequence of Sequence-Frozen SOs.These events correspond to Scheduling Modes 1, 2, 3, 6, and 7 (see table 2). +Scheduling ModesEach scheduling event is mapped to a scheduling mode.The scheduling mode influences which SOs are rescheduled.Some events affect all SOs, while some events affect only those SOs whose ETA is at, or later than, a certain point in time.Additionally, some scheduling events require that all or some of the SOs be resequenced as part of the scheduling process.For each possible freeze state (STA-Frozen, Sequence-Frozen, and Other), the scheduling mode indicates whether or not sequences from the previously computed schedule are to be maintained in the new schedule.The scheduling modes are listed in table 2. +F. Special SO IndicatorsThere are a number of indicators or flags associated with each SO.These flags may affect how the SO is treated during scheduling and runway allocation.Some of these flags may be set as a result of the user's interaction with the DP.Additionally, some flags are set by the DP automatically in response to changes in data or the receipt of various events. +Departed AircraftArrival aircraft that are to depart from airports within the Center's airspace are represented in the system by proposed flight plans.Proposed flight plans create uncertainty because their filed departure times are often inaccurate.Aircraft have been known to actually depart up to three hours earlier or later than the time indicated by the flight plan.These highly inaccurate departure times result in highly inaccurate ETAs.If the DP were to schedule the aircraft using such a faulty ETA, the resulting STA may be unrealizable.Therefore, these aircraft require special treatment.The DP will not schedule proposed flight plans unless the TMC has manually input the aircraft's departure time.When the TMC is informed of the accurate departure time of the aircraft, the TMC enters the departure time into the TGUI.The TGUI informs the DP, and the DP will set a status flag for that aircraft indicating that the aircraft has been "manually departed."Additionally, the TGUI will pass along the departure time as a flight plan amendment to the aircraft's coordination time.The DP will dispose of any previously computed ETA for the aircraft and will wait to receive a new ETA based on this coordination time.When the new ETA is received, DP will be able to compute an STA for the aircraft and thus hold a slot for that aircraft until the aircraft becomes airborne and radar tracks are received.Once track data is received, the aircraft is treated the same as any other tracked aircraft. +Expired AircraftNormally, after an aircraft lands or is otherwise removed from the Center Host computer's database, the Host computer instructs CTAS to delete the aircraft's flight plan.Occasionally, CTAS does not receive such a message, and the flight plan remains in CTAS.If this aircraft were left unchecked, the DP would continue to schedule the aircraft as though it were simply an inactive flight plan, and this aircraft would take up a slot in the schedule.To assure that such an aircraft will not take up a slot in the schedule, the DP will flag the aircraft as expired and exclude it from the scheduling and runway allocation processes.An aircraft is flagged as expired if all of the following criteria are true:• Aircraft is inactive.• Aircraft is not a proposed flight plan.• The aircraft's runway ETA is in the past.Only aircraft can be flagged as expired.It is not necessary for a blocked slot to expire because the CM will delete a blocked slot's flight plan after a certain amount of time. +Pop-up AircraftAn aircraft that is flagged as a pop-up is excluded from the scheduling and runway allocation processes.Note, only aircraft can be flagged as a pop-up.Blocked slots are never flagged as pop-ups.The DP's initial design and implementation included provisions for pop-up aircraft. 2However, the current DP implementation does not flag any aircraft as a pop-up.The pop-up functionality may be reactivated as part of a future enhancement. +Priority AircraftThe user may designate an aircraft as being a priority aircraft.This functionality is used in emergency situations where the aircraft must land as soon as possible.A priority aircraft will have a different ETA computed for it by the RA.This ETA is based on a quicker and more direct route to the runway threshold.Priority aircraft will normally be scheduled at their ETA unless this puts them in conflict with SOs not being rescheduled, manually scheduled SOs, or other priority aircraft.The position of a priority aircraft in the sequence is irrelevant when scheduling to the ETA. +Landed AircraftWhen an aircraft's track is within a certain distance and altitude of its arrival airport, the CM informs all CTAS processes that the aircraft has landed.This is necessary because the Center's Host computer does not inform CTAS when an aircraft has landed.Landed aircraft are not eligible for scheduling nor runway allocation.However, the number of aircraft that have landed in the past hour must be counted when applying an acceptance rate scheduling constraint. +Manually Scheduled SOsThe user may manually set the STA of an SO.The user may supply the STA to a reference point such as the meter fix or runway threshold, and the DP will compute the STA to the other reference points.The difference between the ETAs to the various reference points is added or subtracted from the user-entered STA to derive the STA to the other reference point.The DP will automatically freeze the STA of a manually scheduled SO.If the manually set STA places the SO in conflict with other STA-Frozen SOs, then the conflict will not be resolved.On the other hand, Non-STA-Frozen SOs will be scheduled in such a way as to avoid a conflict with the manually scheduled SO.The DP's philosophy regarding manually scheduled SOs is that the user has assigned the STA to an SO, and it is up to the DP to schedule the remaining SOs around the manually scheduled ones. +Suspended SOsThe user may suspend or unsuspend an SO.A suspended SO is not scheduled, and its slot is given up when the SO is first suspended. +Holding AircraftAircraft which have been placed in holding require special handling by the scheduler.Some holding aircraft are instructed, by the controller, to follow a holding pattern over a meter fix until it is taken out of holding.As a result, the aircraft's track alternates between the Center's airspace and the TRACON's airspace.The DP compensates for this by treating aircraft that are in holding as if they are in the Center's airspace even if the aircraft's track indicates that the aircraft is in the TRACON's airspace.However, currently, no CTAS process informs the DP that an aircraft is in holding.The detection of an aircraft in holding is an area of future research. +G. The Main BodyAll of the DP's activity is event driven, and the DP's main body controls the program flow based on these events.Most events come from other CTAS processes and are received via messages while other events are generated by the DP itself.When a message is received from another CTAS process, the DP will update its databases as appropriate and then execute one of the following three activities.Do nothing.This occurs when the message simply requires that the DP update its database.Place a rescheduling or runway allocation event on the list of pending events for processing later.Some events of the same event type occur in bunches.To improve computing performance, these events are collected in a pending list and are processed as a group only after an event of a contrasting event type is received.If the pending list is full 3 or if the current event differs from those in the pending list, then the appropriate rescheduling or allocation actions on one or more aircraft for the pending events are executed.The pending list is then flushed, and the current event is placed in the pending list. +Carry out an immediate reschedule or runway allocation.If there are events in the pending list, then rescheduling or runway allocation is executed on one or more aircraft for the pending events.The pending list is then flushed, and rescheduling or runway allocation is carried out on one or more aircraft for the current event.The DP will send messages to the other CTAS processes as a result of rescheduling or runway allocation.Runway allocation activity is always followed by rescheduling in the DP.If an aircraft's assigned runway is changed as a result of runway allocation, then a flight plan amendment indicating the runway change is sent by the DP to the other CTAS processes.Similarly, if rescheduling an aircraft shifts the aircraft's STA from a time before a configuration change to a time after the configuration change, then a flight plan amendment indicating that the aircraft is under the influence of a different airport configuration is sent by the DP to the other CTAS processes.After rescheduling is complete, a schedule message is sent to all of the other processes indicating the newly computed STAs.The schedule message will also include information about the changes in the aircraft status as a result of rescheduling.In addition to responding to messages sent by other CTAS processes, the DP generates its own internal rescheduling or runway allocation triggering events.If no rescheduling has been executed for six seconds, then the DP will carry out a reschedule known as a Periodic Reschedule.If an aircraft is about to have its STA frozen, then a runway allocation event for that aircraft is generated internally.Rescheduling may also be triggered internally when an aircraft's ETA is hovered.The mechanism of ETA hovering is described in a later section. +H. SequencingAn integral part of scheduling is sequencing.A sequence is the order in which aircraft are to arrive at a particular scheduling reference point.The DP handles sequencing through its Sequencer and Sequence Constraint modules.Aircraft to be scheduled are first sequenced in an FCFS order within each stream class based on their ETAs to the meter fix.However, depending on the scheduling mode, the sequence may be further restricted so that aircraft which are Sequence-Frozen maintain their sequence relative to other Sequence-Frozen aircraft.Furthermore, the user may input sequence constraints to the DP which force certain aircraft to be scheduled before or after other aircraft.The sequencer must take all of these restrictions into account when generating a sequence at the meter fix. +Sequence ConstraintsThe sequence may deviate from the FCFS rule as a consequence of sequence constraints.A sequence constraint is an instruction to the DP to force certain aircraft to follow behind other aircraft.Currently, sequence constraints are entered by the users via CTAS's graphical user interfaces, PGUI and TGUI.These sequence constraints are sent to CM, and CM forwards them to the DP.So, from the DP's point of view, the sequence constraints are always generated by other CTAS processes.Sequence constraints always involve two aircraft, have a time stamp and priority associated with them, and occur in two varieties.The first variety of sequence constraint is referred to as a "direct" constraint.If aircraft B is constrained directly behind aircraft A, then the DP will create a sequence that places aircraft B immediately following A even if this means delaying either aircraft so that B can follow behind A. A direct constraint is notated here with a double directed line pointing to the aircraft ahead.For example, the sequence constraint just described is shown in figure 4.The second type of constraint is an "indirect" sequence constraint.If aircraft B is constrained indirectly behind A, then the DP will create a sequence that places B somewhere after A. This constraint allows for other aircraft to be sequenced behind A and in front of B so long as B is somewhere behind A. An indirect constraint is notated here with a single directed line pointing to the 3 The capacity of the pending list is site-dependent.Currently, for all adapted sites, the list can hold 50 pending events.Since sequences are computed independently for each super stream class, any sequence constraints involving aircraft that are in different super stream classes are ignored.For example, suppose aircraft B, a turbo prop, is constrained directly behind aircraft A, a jet.If the TMC has not combined these two engine types into the same super stream class, this sequence constraint will be ignored. +Sequence Constraint DeconflictionThe most complicated function related to sequencing is sequence constraint deconfliction.If the DP has been issued sequence constraints that conflict with each other, the DP applies an order of precedence to determine which constraints are active and which are inactive at the time of the sequencing.Each sequence constraint received by the DP has a relative priority associated with it.In the current NASA proof-of-concept software, sequence constraints received from the TGUI have a priority of 10 and those received from the PGUI have a priority of zero 4 .If two sequence constraints are in conflict, then the sequence constraint with a higher priority number will have precedence.If the sequence constraints have the same priority, then the sequence constraint entered more recently is used.The deconfliction of sequence constraints is dynamic.That is, two sequence constraints might be in conflict during one scheduling cycle, but they may not be in conflict during a later scheduling cycle.Therefore, a sequence constraint is not removed from the DP unless the DP is explicitly told to remove it, or if one of the aircraft listed in the constraint has been removed from the system.Instead, sequence constraints that are in conflict with constraints that have a higher priority (or are newer if the priorities are equal) are made inactive for that scheduling cycle.Each node in the graph must obey certain fan-in and fanout limits.The allowable fan-in for a particular node (the number of directed edges pointing to a particular node) is either one direct constraint or zero or more indirect constraints.Similarly, the allowable fan-out for a particular node (the number of directed edges pointing away from a particular node) is either one direct constraint or zero or more indirect constraints.Finally, no edge may begin from and end at the same node.See figure 6 for an example of a valid Sequence Constraint Graph and figure 7 for three examples of invalid Sequence Constraint Graphs.The process of deconflicting sequence constraints consists of iterating over each constraint q i in order from highest to lowest priority.Within each priority, the order is from newest to oldest.Each sequence constraint q i is compared with each of the sequence constraints q j that has been examined during a previous iteration and has been marked as active.Each q i is compared against each q j using the following three checks.1. Individual Check 2. Chain Check 3. Graph Update 4 The sequence constraint priority values are contained within the sequence messages that the DP receives from other CTAS processes.These values may be changed to suit the operational environment without changing the DP.If q i fails any of these checks when compared against any of the previously examined active sequence constraints q j , then q i is marked as inactive.An inactive sequence constraint will not affect the sequence during the current scheduling cycle.On the other hand, if q i passes all of the preceding checks against each q j , then q i is marked as active.An active sequence constraint will influence the sequence during the current scheduling cycle, and the remaining sequence constraints to be checked will be examined for conflict with each of the active constraints.Each of these checks is described in greater detail below.Individual Check.Several simple quick checks are conducted between the sequence constraint q i and the sequence constraints that have already been accepted during the preceding iterations.These checks are broken down into two phases.During the first phase of individual checks, the constraint q i is compared on a oneto-one basis with each constraint q j which has already been accepted.During the second phase of individual checks, the constraint q i , if it is an indirect sequence constraint, is compared against the indirect constraints contained within the Sequence Constraint Graph.The Sequence Constraint Graph is the set of sequence constraints q j which have been deconflicted and are currently active.If the comparisons show that q i is redundant or contradicts a previously accepted constraint q j , then q i is made inactive.The specific cases which result in making q i inactive are as follows.If a previously accepted constraint q j is exactly like the constraint q i , then q i is made inactive.This is known as rejection by redundancy.For example, if q j and q i are those shown in figure 8, then q i is made inactive in favor of the higher priority or newer constraint which was previously accepted.If a previously accepted constraint q j has the same two SOs as q i but in reverse order, then q i is made inactive.For example, figure 9 shows q i which is the reverse of a previously accepted constraint q j .Due to this contradiction, the q i is made inactive.If the combination of the constraint q i and a previously accepted constraint q j would result in one SO being constrained directly ahead of or behind two different SOs, then q i is made inactive.For example, figure 10 shows a case where the combination of q i and a previously accepted constraint q j constrains two different SOs (B and C) directly behind the same SO (A).Also, figure 11 shows a case where the combination of q i and q j constrains the same SO directly behind two different SOs.In both cases, q i is made inactive.If a previously accepted constraint q j constrains one SO directly ahead of or behind a second SO while the constraint q i only constrains the first SO indirectly ahead of or behind the second SO, then q i is made inactive (rejection due to redundancy).In figure 12, the constraint q i does not add any more information because the previously accepted constraint is more restrictive.Therefore, q i is made inactive.The second phase of the individual checks consists of comparing the indirect constraint q i against the previously accepted indirect constraints that are Note that this second phase is only concerned with indirect constraints.If q i is a direct constraint, then it passes this second phase of individual checks.Furthermore, the construction of the Sequence Constraint Graph may alter some of the indirect sequence constraints that are added to the graph.The purpose of this second phase of individual checks is to compare the indirect constraint q i against those altered indirect constraints contained within the graph.Those constraints contained within the graph which are unaltered have already been compared against q i in the first phase of the individual checks described above.If the indirect constraint q i is the same as an indirect constraint within the graph, then q i is made inactive (rejection due to redundancy).An example of this case is shown in figure 13.If the indirect constraint q i is the reverse of an indirect constraint within the graph, then q i is made inactive.In the example shown in figure 14, q i is made inactive because it is the reverse of a constraint contained within the graph.Chain Check.Next, the constraint q i is checked against the Sequence Constraint Graph to see if q i conflicts with a chain of constraints contained within the graph.If either the preceding SO or the following SO (or both) in q i is not contained in the Sequence Constraint Graph, then q i passes the Chain Check.If both the ahead SO and the behind SO in the constraint q i are contained in the Sequence Constraint Graph, then more detailed comparisons between q i and the graph are conducted.For each SO contained in the Sequence Constraint Graph, a list containing all SOs constrained anywhere ahead of that SO is created.The constraint q i is then compared against these lists to determine if q i contradicts or duplicates the information contained in these lists.Cases where q i is made inactive are described below.If the constraint q i is the reverse of a constraint chain contained in the Sequence Constraint Graph, then q i is made inactive.An example is shown in figure 15.If the constraint q i is a duplicate of a constraint chain contained in the Sequence Constraint Graph, then the current constraint may be redundant.For example, in figure 16, q i is made inactive because it is redundant.An exception to the case described above is shown in figure 17.If q i is a direct constraint, then it replaces a less restrictive indirect constraint involving the same SOs when there are no intervening SOs.Graph Update.In order to remain within the fan-in/fanout limits, the Sequence Constraint Transitivity Rule (SCTR) is used to adjust an existing Sequence Constraint Graph when new sequence constraints are inserted.The SCTR is detailed in the sidebar at right.Building the directed Sequence Constraint Graph consists of examining each sequence constraint q i in order, beginning with the newest/highest priority sequence constraint and ending with the oldest/lowest priority sequence constraint.Each acceptable sequence constraint q i is added to the graph, and adjustments are made to the graph to comply with the fan-in/fan-out limitations.Each constraint falls into one of the cases described below.Case 1: The sequence constraint q i is already in the graph.The constraint q i is ignored in favor of the newer/higher priority constraint already in the graph (see figure 18).Case 2: Neither the SO ahead nor the SO behind in constraint q i appears in the graph.A new node is created for the SO ahead and the SO behind in the constraint q i , and the constraint itself is added.In this case, q i is disjoint from the Sequence Constraint Graph thus far (see figure 19). +The Sequence Constraint Transitivity Rule (SCTR)Let A, B, C, D, X, Y, and Z be distinct SOs.If B is constrained directly behind A, and C is constrained indirectly behind A, then C is constrained indirectly behind B (see figure 20).Similarly, if Y is constrained directly ahead of X, and Z is constrained indirectly ahead of X, then Z is constrained indirectly ahead of Y (see figure 21).This may be extended to include chains of direct sequence constraints (see figure 22) A new node is created for the SO in the constraint q i that is not already in the graph.An edge is added that links the new node with the SO that is in q i and already in the graph (see figure 23).Simply inserting q i could exceed the fan-in/fan-out limit because of direct constraints in the graph.In this case, the SCTR is applied, and the new edge is added at the end of the chain of direct constraints (see figure 24 and figure 25).Case 4: The constraint q i is a direct constraint, and either the SO ahead or the SO behind in the constraint (but not both) is already in the graph.A new node is created for the SO in the constraint q i that is not already in the graph.An edge is added that links the new node with the SO that is in q i and already in the graph (see figure 26).Simply inserting q i could exceed the fan-in/fan-out limit because of indirect constraints in the graph.In this case, the SCTR is applied, and the indirect constraints in the graph are moved to the end of the chain of direct constraints (see figure 27 and figure 28).No new nodes have to be created.The indirect constraint q i is simply added to the graph (see figure 29).Simply inserting q i could exceed the fan-in/fan-out limit because of direct constraints already in the graph.In this case, the SCTR is applied, and the indirect constraint q i is moved to the end of the chain of direct constraints (see figure 30, figure 31, and figure 32). Figure 27.SCTR Applied to a Direct Constraint No new nodes have to be created.In some cases, the constraint q i can simply be inserted into the graph (see figure 33).Simply inserting q i could exceed the fan-in/fan-out limit because of indirect constraints already in the graph.In these cases, the SCTR is applied after q i is inserted, and the indirect constraints are moved to the end of the chain of direct constraints (see figure 34 and figure 35). +Building the Combined Sequence to the Meter FixesAll SOs are sequenced to their respective meter fixes and placed into one combined sequence.The scheduler can then extract sequencing information from each individual super stream class as required.The sequencer begins by taking a list of SOs provided by the scheduler and places them into a sorted list that represents the previous sequence.The SOs are placed in this previous sequence, from earliest to latest, according to the meter fix STAs computed during the previous scheduling cycle.If two SOs have the same STA, then the tie is resolved using each SO's ETA to the meter fix.Some SOs don't have STAs because they were added to the system since the last scheduling cycle or changes have made their previous STA invalid.These SOs are placed at the end of the previous sequence after those SOs with the latest STAs.Step 2: Deconflict manual sequence constraints.The manual sequence constraints, those entered by the user, are deconflicted (see section II.H.2).The resulting Sequence Constraint Graph contains a consistent set of sequence constraints which will be applied in Step 7.Step 3: Create a preliminary version of the new sequence.The preliminary sequence represents the new sequence before any manually entered sequence constraints are applied.Each SO within the preliminary sequence has a flag which this document refers to as the "localsequence-frozen" flag.This ad hoc flag is only used while constructing sequences and should not be confused with the SO's Sequence-Frozen indicator used throughout the DP.At the beginning of the sequencing process, the local-sequence-frozen flag for each SO is set to FALSE.Step 4: Place Sequence-Frozen SOs into the preliminary sequence.Next, if the scheduling mode requires that the sequence of Sequence-Frozen SOs is maintained, then SOs that are Sequence-Frozen are copied from the previous sequence (created in Step 1) and placed in a preliminary sequence (created in Step 3) in the same order.These Sequence-Frozen SOs have their local-sequence-frozen flag set to TRUE.Step 5: Place Non-Sequence-Frozen SOs into the preliminary sequence.For each of the remaining SOs, S i , regardless of the scheduling mode, the following is carried out.S i 's meter fix ETA is compared against the meter fix ETA of each S j already in the sequence.Exception: If S j is manually scheduled, then S i 's meter fix ETA is compared against S j 's meter fix STA.S i is inserted into the sequence after the last SO with a time (ETA or STA depending on the circumstances just described) that is equal to or earlier than S i 's meter fix ETA.Step 6: Unfreeze SOs mentioned in a sequence constraint.After the preliminary sequence has been built, each SO mentioned in the deconflicted sequence constraints has its local-sequence-frozen flag set to FALSE.This allows manual sequence constraints, when they are applied, to change the position of Sequence-Frozen SOs.Thus, at this point, local-sequence-frozen SOs are those that have their positions constrained only by being Sequence-Frozen and are not constrained by manual sequence constraints.Step 7: Apply sequence constraints and create the final sequence.In the final phase of sequencing, the sequence is modified to comply with the non-conflicting manual sequence constraints.The SOs are copied from the preliminary sequence and placed in the final sequence.This processing is done in the order that they should appear in the final sequence.Thus, when an SO is constrained behind one or more SOs, then it is not added to the end of the final sequence until all of the SOs constrained ahead of it have been added to the final sequence.The details of this final step are as follows.First, the SOs in the preliminary sequence which meet the following criteria are examined.The SO is not already in the final sequence, andThe SO is not local-sequence-frozen, orThe SO is local-sequence-frozen, and there are no local-sequence-frozen SOs ahead in the preliminary sequence, orThe SO is local-sequence-frozen and the first localsequence-frozen SO ahead in the preliminary sequence is already in the final sequence.For each SO, S i , which satisfies the preceding criteria, the DP's sequencer examines S i to see if placing it at the end of the final sequence violates any sequence constraints.If S i is constrained behind an SO which has not yet been placed in the final sequence, then there is a sequence constraint violation.On the other hand, if S i is the preceding SO in a direct constraint, then the following SO in that constraint would have to be added to the end of the final sequence after S i .This creates a dependency between S i and the SO constrained directly behind.Thus, S i violates a sequence constraint if the behind SO violates a sequence constraint.If the behind SO is itself the ahead SO in a direct sequence constraint, then it violates a sequence constraint if the other SO violates a sequence constraint.This continues recursively along the chain of direct sequence constraints.Once it has been determined that adding S i to the end of the final sequence does not violate any sequence constraints, S i is added to the final sequence.If another SO is constrained directly behind S i , then that other SO is added to the final sequence as well. +I. SchedulingThe goal of scheduling is to prepare a plan that delivers SOs from the Center to the TRACON in a smooth manner with minimal delay.This plan consists of STAs to various reference points, the outer meter arc, meter fix, FAF, and runway threshold for each SO.When scheduling SOs, each SO's ETA to each of the reference points represents the time of arrival if no other SOs were in the system.Thus, the ETA is used as the initial STA for each SO.A number of scheduling constraints are applied which delay the SO.This results in an STA that is farther in the future than the ETA.These scheduling constraints are entered by the TMC to reflect current procedures and conditions at the airports, in the TRACON, and in the Center.All of the scheduling constraints are applied, though some constraints may be more restrictive than others from SO to SO.The resulting schedule will consist of STAs for each SO that is as close to the SO's ETAs as possible while complying with all scheduling constraints.In addition, the scheduling process sets various status flags for each SO.The STAs and flags are then sent to the other CTAS processes. +Scheduling EventsThe DP reschedules all or some of the SOs in response to various events.These events are listed in table 3 along with the scheduling modes (see section II.E.15) used when responding to these events. +Scheduling Event Time Reference PointThe scheduling events are further broken down into immediate events and pending events.Immediate events will trigger a reschedule at once while pending events are placed in a pending list for rescheduling later.This deferred scheduling actually takes place when one of the following occurs:1.An immediate scheduling event is received.2. A pending event is received that is not the same type as events stored in the pending list.Each event listed in table 3 is of a separate type.3. A pending event is received and the pending list is full.Since all of the events in the pending list are of the same type (see item 2 above), processing the entire list of pending scheduling events requires only a single schedule computation.Once the events in the pending list have been processed, the pending list is cleared. +Order of Consideration at the RunwayAfter the SOs have been scheduled to the meter fixes, they are scheduled to the runway.Because a sequence generated at the runway could contradict the sequence at the meter fix, sequences at the runway are not computed, and sequence constraints relative to the runway are disallowed.Instead, an Order of Consideration at the runway is computed by the Scheduler class.The Order of Consideration is the order in which SOs have their runway threshold and FAF STAs computed.This does not necessarily mean that the SOs will be scheduled to arrive at the runway in this order, but only that they are computed in this order.Thus, the scheduler will have greater latitude in computing the STAs for SOs which are earlier in the order versus SOs which are later in the order.The sequence at the meter fix will remain unchanged by the Order of Consideration algorithm.Finally, B1 and then B2 are scheduled to the runway.Because of the other aircraft already scheduled, B1 and B2 will be delayed to satisfy the 1.5 minute separation requirement at the runway. +Delay FeedbackDue to scheduling constraints at the airport and its runways, an SO may be delayed at the runway threshold.The measure of this delay is based on the sum of the meter fix STA (STA mf ) and the TRACON Transition Time (T mf->rwy ).The TRACON Transition Time is defined as the time required to fly from the meter fix to the runway threshold when there are no other SOs present.Thus, T mf->rwy = ETA rwy -ETA mf (1) where ETA rwy is the ETA to the runway threshold and ETA mf is the ETA to the meter fix.In addition, scheduling constraints may cause delays at the meter fix.Such delays make it impossible for an SO to meet the runway ETA.The earliest time that an SO can arrive at the runway threshold is its Proposed Time of Arrival (PTA rwy ) and is computed according to:PTA rwy = STA mf + T mf->rwy(2)Finally, it follows that the amount of delay at the runway threshold is:Delay rwy = STA rwy -PTA rwy(3)To optimize performance, this delay must be distributed between the Center and the TRACON.Adding delay to the meter fix STA as a result of excessive delay at the runway threshold is handled in the DP by a mechanism known as Delay Feedback.After a preliminary STA is computed to the meter fix and the runway threshold for a particular SO, the delay at the runway is examined.The maximum amount of delay that can be absorbed in the TRACON is a parameter of the DP.If the runway delay is within the amount that can be absorbed in the TRACON, then no adjustments to the SO's STAs are necessary.However, if the amount of delay is greater than the amount that can be absorbed in the TRACON, then the excess delay is fed back or added to the meter fix STA.This delays the SO at the meter fix.If delaying the SO at the meter fix results in a violation of a scheduling constraint, then the process of computing the SO's STA is repeated.The times that were just computed serve as the earliest allowed times of arrival during this next scheduling iteration.This is necessary because any meter fix delay beyond that required by the delay feedback could make the runway threshold STA unrealizable.A similar mechanism is used by the DP to distribute the delay between the low altitude arrival sector and the high altitude sector.The SO's meter fix STA is used by the low altitude arrival sector while the outer meter arc STA is used by the high altitude arrival sector.No scheduling constraints are applied at the outer meter arc.The Outer Meter Arc to Meter Fix Transition Time (T oma->mf ) is the time it takes an SO to fly from the outer meter arc to the meter fix when there are no other aircraft in the airspace.T oma->mf = ETA mf -ETA oma (4) Once the meter fix STA (STA mf ) is computed, the outer meter arc STA (STA oma ) is computed according to:STA oma = STA mf -T oma->mf -AMDT (5)where AMDT is the Amount of Delay Time.Without the AMDT, the high altitude arrival sector would be required to absorb all Center delays.By subtracting the AMDT, the low altitude arrival sector is forced to absorb up to the amount represented by the AMDT.The AMDT is a sitedependent parameter of the DP. +Scheduling ProcessThe scheduling events described in section II.I.1 trigger the scheduling process.In addition, the Runway Allocator will request the computation of schedules for each runway assignment it is considering for an SO.The Runway Allocator compares these schedules to determine which runway assignment is best.In either case, the scheduling process follows the steps summarized in figure 36.The numbers shown in figure 36 correspond to the step numbers given in this section's description. +Figure 36. Top Level Scheduling Flow ChartStep 1: Determine if the current scheduling event requires immediate rescheduling (see section II.I.1 and table 3).If this is an immediate rescheduling event, then go to Step 4.Step 2: The current scheduling event is a pending event.If the list of pending scheduling events is full or the events in the list are different from this rescheduling event, then go to Step 4.Step 3: Add this event to the list of pending scheduling events and quit.Reaching this step means that the current scheduling event is a pending event, there is room in the pending list, and this event is of the same type as those in the pending list.This event is added to the pending list, and the scheduling process is ended at this point.This event will be processed later, along with the other pending events in Step 4 in response to a different scheduling event.Step 4: Begin processing the events in the list of pending scheduling events (if any) by determining the earliest start time for the reschedule.All of the pending scheduling events in the list are of the same type.They use the same scheduling mode and method.Therefore, only one reschedule is necessary for the entire set of events.To ensure that all events are covered, each event is examined to find the earliest reschedule start time necessary.Step 5: Carry out the scheduling process for the pending event starting with Step 6, and remove the events from the list of pending scheduling events.Scheduling triggered by a pending event is handled in a manner similar to scheduling triggered by an immediate event.The start time of the reschedule is that time found in Step 4 above.Since a single rescheduling process is enough to handle all of the pending scheduling events in the list, the list can be cleared after the reschedule is complete.Step 6: Reset the STAs of all expired and popup SOs.Since the last schedule was computed, some SOs may have been flagged as expired (see section II.F.2) or as pop-ups (see section II.F.3).Resetting their STAs ensures that any previous STAs are not stored in the CM nor sent to the Center's Host computer.Additionally, the STAs of these SOs are not displayed on any of the GUIs.Step 7: Determine the point in time where rescheduling is to be executed.Not all SOs are affected by the events triggering rescheduling.In most cases, SOs due to arrive later are usually influenced by the SOs due to arrive sooner.In contrast, SOs due to arrive sooner are usually not influenced by the SOs due to arrive later in the absence of sequence constraints.However, sequence constraints can cause some SOs to influence the schedule of SOs that are due to arrive earlier.A rescheduling start time must be derived from the scheduling event.SOs whose ETAs or STAs are equal to or later than the rescheduling start time are rescheduled.The rescheduling start time is equal to the Scheduling Event Time minus the Zone of Influence.The Scheduling Event Time and the Zone of Influence are described below.The rescheduling start time may be relative to the meter fix, FAF, or runway.For example, adding a meter fix acceptance rate means that the rescheduling start time must be relative to the meter fix.The DP determines from the event which reference point to use when determining which SOs are to be rescheduled (see table 3).Scheduling Event Time.Some scheduling events have an explicit Scheduling Event Time associated with them.For example, the user may enter a scheduling constraint and the time when the scheduling constraint should be active.This activation time is the Scheduling Event Time.Other scheduling events have a Scheduling Event Time which is based on an SO's ETA.These are scheduling events that have a specific SO associated with them.For example, the user may request that rescheduling be executed for a particular SO and all SOs following it.In such a case, the Scheduling Event Time is the ETA of the SO. +Zone of Influence. It is not sufficient to set the Rescheduling Start Time equal to the Scheduling EventTime.The nature of some scheduling constraints makes it necessary to reschedule SOs that are due to arrive slightly earlier than the Scheduling Event Time.The Zone of Influence is the mechanism by which the DP accounts for the rescheduling of SOs due to arrive earlier than the Scheduling Event Time.The Zone of Influence is set based on the maximum of the Acceptance Rate Interval (see section II.E.3) and the sum of the largest occupancy time constraint and largest separation distance constraint (see section II.I.10) active at the Scheduling Event Time.Step 8: Get all SOs with ETAs that are earlier than the rescheduling start time (see Step 7) and insert them into the schedule based on their previously computed STAs.SOs earlier than the rescheduling time are unaffected by the current scheduling event, so their STAs are left unchanged.However, they must be included in the schedule since they may affect the STAs of those SOs that are being rescheduled.Step 9: Get all SOs that are to be rescheduled as a result of this scheduling event.For each SO, a series of conditions are checked to determine if the SO should be rescheduled.These conditions are summarized in a truth table (see table 5).If the SO is eligible for rescheduling, it is placed in a list that is used in the following steps.For example, the fifth column from the right in table 5 shows that an SO is eligible for rescheduling if the following conditions are true:• The SO has valid ETAs.• The SO's ETA is not earlier than the rescheduling start time.• The SO has not landed.• The SO has not been suspended.• The SO is not a pop-up.• The SO is not expired.• The SO is active.The "*" indicates that if the above conditions are true, then it does not matter if the SO is a proposed flight plan or not, and it does not matter if the SO has departed or not.Step 10: Update the sequence to the meter fix of the SOs that are being rescheduled.The SOs selected for rescheduling are sorted into a sequence to the meter fix.The sequence is based on an FCFS order based on meter fix ETA.The sequence is further refined so that SOs which are Sequence-Frozen maintain their sequence relative to each other as dictated by the scheduling mode (see section II.E.15).The sequence is also constructed so that any sequence constraints received by the DP are also observed.For a detailed description of the sequencing process, see section II.H.3.Step 11: The SOs are broken down into sets that have different scheduling priorities.The SOs are broken down into Scheduling Priority Sets.Each set has a scheduling method associated with it.The definitions of these sets and their corresponding scheduling methods vary depending on the scheduling mode (see table 6, table 7, and table 8).Each scheduling method is detailed in its own section later in this paper.The SOs with a higher priority are scheduled first and have the best chance of actually being scheduled at their ETAs.The SOs with a lower scheduling priority are scheduled around the SOs with a higher scheduling priority.Each table lists the Scheduling Priority Sets in order from highest priority to lowest.Step 12: Each set of SOs is processed one at a time from the highest priority to the lowest priority.Each SO within each set is inserted into the new schedule using one of three methods depending on the scheduling mode and the set's scheduling priority.Each method is summarized below and is described in its own section. +Insert withoutRescheduling.This method inserts the SO into the schedule based on the STA computed in a previous scheduling cycle (see section II.I.5). +Reschedule at ETA and Insert.This method is usually applied when scheduling priority aircraft.The SO is scheduled at its ETA and is only delayed to avoid a conflict with an SO that has already been scheduled such as another priority aircraft or an STA-Frozen SO.Sequence is not considered by this method (see section II.I.6). +Reschedule after Aircraft Ahead and Insert.The bulk of the SOs are scheduled using this method.The sequence computed in Step 10 is used when computing the STA of an SO under this method (see section II.I.7).Each method, except Insert without Rescheduling, executes the following steps:1. Compute STAs to the meter fix in sequence.2. Compute STAs to the runway in order of consideration.3. Adjust meter fix STAs to account for delay feedback.4. Compute outer meter arc STAs.In addition, all three methods create a Schedule Linked List data structure.This list is used to identify which SOs have already been scheduled and can affect the STA calculation of the SO currently being scheduled.Another data structure used during scheduling is the set of Acceptance Rate Bins.The Acceptance Rate Bins track the number of SOs which have been scheduled in a particular period of time.This data structure is used when applying an acceptance rate scheduling constraint to the SO currently being scheduled.For more details on Acceptance Rate Bins, see section II.I.11.Once the current SO has been scheduled, its STA determines where the SO should be inserted into each of these data structures.Step 13: If this scheduling activity is part of the processing of the list of pending scheduling events (see Step 5), then the processing of the pending scheduling events is complete.Begin processing the current scheduling event starting at Step 6.However, if this scheduling activity is part of the processing of the current scheduling event, then proceed to the next step.Step 14: Prepare the scheduling message containing the newly computed STAs and other SO status indicators, and send the message to the rest of CTAS.To minimize the amount of message traffic over the network, not all SOs are contained in the scheduling message.SOs which have a valid STA are included in the message.In addition, SOs which have had certain flags set by the DP are included in the message even if no STAs have been computed for them.These special flags are examined by the GUIs and affect the display of the applicable SOs.Step 15: Update each SO's associated airport configuration and runway.An SO may be delayed enough to cause it to be associated with a different airport configuration than it was prior to scheduling.In this case, the DP assigns the SO to the default runway of the new airport configuration.For each SO which has had its associated airport configuration changed or both its associated airport configuration and assigned runway changed, a flight plan amendment is sent notifying the rest of CTAS of the change. +Insert without ReschedulingThis method is used when an SO is not to have its STA changed by the scheduling process.The SO is simply inserted into the Schedule Linked List and Acceptance Rate Bins based on the STA computed during a previous scheduling update.Even though its STA is not changed, including the SO in the scheduling data structures ensures that its influence on the STAs of other SOs is taken into account. +Reschedule at ETA and InsertThis method is used for scheduling priority SOs.Although a sequence is computed, the sequence is used only as the order in which the SOs are processed.An attempt is made to schedule each SO at its ETA disregarding the sequence.If scheduling the SO at the ETA would cause it to violate a scheduling constraint, then the SO is delayed an amount sufficient to comply with all scheduling constraints.The steps executed under this method are summarized in figure 37.The numbers included in figure 37 correspond to the step numbers listed in the following description.Step 1: Determine the processing order of the SOs by determining the sequence at the meter fix (see section +II.H.3).This method does not actually obey the meter fix sequence when computing the STAs for the SOs in the set.The sequence is simply used as the order in which SOs are processed by this method.Step 2: For each SO to be scheduled using this method, Step 2A through Step 2D are executed.Step 2A: If an overcrossing time has been received for this SO or the SO is completely within the TRACON's airspace, then the meter fix STA is set to the overcrossing time.Go to Step 2D.Step 2B: If the conditions in the previous step are not true, then get the meter fix ETA.This becomes the Proposed Time of Arrival (PTA).The PTA represents a temporary STA until the final STA is computed.Since the meter fix ETA is the earliest possible time that an SO can be scheduled to arrive at the meter fix, the SO's PTA is initialized to the SO's ETA.As scheduling constraints are applied, the SO's PTA will be delayed.Step 2C: Modify the PTA to satisfy all Center scheduling constraints.Use the modified PTA as the meter fix STA.Each of the Center scheduling constraints is applied to the PTA possibly pushing the PTA later.The more restrictive scheduling constraints will have a greater impact on a particular SO, and which constraint is more restrictive will vary from SO to SO.The following is a list of scheduling constraints which affect the computation of the meter fix STA.Eventually, when this SO has had its runway STA computed, and its meter fix STA has been adjusted for delay feedback, its meter fix STA will be permanently assigned.A permanently assigned STA will impact the STAs of other SOs as they are being scheduled to the runway.Step 3: For each SO to be scheduled using this method, Step 3A through Step 3F are executed.The sequence computed in Step 1 is used as the processing order when computing the runway STAs.Step 3A: Get the temporary meter fix STA just computed in Step 2.Step 3B: If an overcrossing time has not been received for this SO and the SO is not completely within the TRACON's airspace, adjust the meter fix STA to account for SOs that have had some delay fed back to the Center as a result of executing Step 3 on those other SOs.If the current SO has an overcrossing time or is completely within the TRACON's airspace, then use the overcrossing time as the meter fix STA.No adjustment of this meter fix STA is necessary since the SO has already crossed the meter fix.If, however, this SO does not have an overcrossing time and is in the Center's airspace, then the meter fix STA may require adjustment so that it satisfies the scheduling constraints at the meter fix.The meter fix STA of the current SO is only compared against those SOs with permanently assigned meter fix STAs.In other words, the current SO is compared against those SOs which have already had their runway STAs computed and their delay fed back to the Center.If the meter fix STA is modified, then adjustments are made to the Schedule Linked List and the Acceptance Rate Bins to correspond to the modified meter fix STA.Step 3C: Set the runway PTA to the sum of the meter fix STA and the transition time from the meter fix to the FAF or runway threshold (see equation 1 and equation 2).If the SO does not have a meter fix STA (usually because the SO is completely within the TRACON's airspace and no overcrossing time was received), then the runway PTA is set to the runway ETA.Otherwise the value that equation 2 yields is used as the runway PTA.Whether to use the FAF or the threshold depends on the airport's configuration (see section II.E.8).Step 3D: Compute the SO's runway STA and apply any delay feedback to its meter fix STA.This step is described later.For details, see section II.I.8.Step At this point, the current SO has been successfully scheduled to the runway threshold, FAF, and meter fix.The current SO is permanently placed in the Schedule Link List and Acceptance Rate Bins.This means that the current SO can affect the schedules of all remaining SOs that need to be scheduled.Step 4: For each SO to be scheduled using this method, compute the outer meter arc STA using equation 5. +Reschedule after Aircraft Ahead and InsertThis method is used for most of the common cases of scheduling.The sequence computed in Step 10, section II.I.4 is followed when computing the meter fix STAs and maintained when computing the runway STAs.Note that SOs in different super stream classes are sequenced independently of each other.Thus, there is no restriction as to the order of SOs from different super stream classes relative to each other.Once the preliminary meter fix STAs have been computed, the runway STA for each SO is computed.There is no rigid sequence to the runway because such a sequence might contradict the sequence at the meter fix.Therefore, the DP constructs an Order of Consideration (see section II.I.2).This is the order in which SOs are considered for the available slots at the runway.Under most circumstances, the sequence at the runway will correspond to the Order of Consideration.However, the freedom exists to allow the sequences at the runways to differ from the Order of Consideration in order to maintain the sequence at the meter fix.The steps executed by this method are summarized in figure 38.The numbers included in figure 38 correspond to the step numbers listed in the following description.Step 1: For each SO, in the order of the meter fix sequence, Step 1A through Step 1D are executed.Step 1A: If an overcrossing time has been received for this SO or the SO is completely within the TRACON's airspace, then the meter fix STA is set to the overcrossing time.Go to Step 1D.This is the same as Step 2A of section II.I.6.Step 1B: If the conditions in the previous step are not true, then set the PTA to be the later of the nominal meter fix ETA and the meter fix STA of the SO ahead of the current SO in the sequence and in the same super stream class.The meter fix ETA is used as the earliest possible time that an SO can be scheduled to arrive at the meter fix if there are no other aircraft to consider.Additionally, in order to maintain the sequence at the meter fix, the current SO cannot be scheduled earlier than the SO just ahead of it in the same super stream class.Thus, the PTA is set to the later of either the nominal meter fix ETA or the STA of the SO ahead of the current SO which is in the same super stream class.Scheduling constraints and delay feedback may delay this aircraft so that its meter fix STA is even later than the time computed in this step.Step 1C: Adjust the PTA to satisfy all Center scheduling constraints.Use the adjusted PTA as the meter fix STA.This is the same as Step 2C of section II.I.6.Step 1D: If an STA was successfully computed in Step 1A or Step 1C, then insert the current SO into the Schedule Linked List and the appropriate Acceptance Rate Bins.Step 2: Schedule SOs to the runway in Order of Consideration (see section II.I.2).For each SO to be scheduled to the runway, Step 2A through Step 2I are executed.Step 2A: Of the SOs yet to be scheduled at the runway, find the one with the earliest meter fix STA.The meter fix STA of the earliest SO is used in the following step when determining the super stream class of each SO.The grouping of stream classes into super stream classes can vary over time if future super stream class scheduling constraints have been entered by the TMC.Therefore, the meter fix STA of the earliest SO is used for the sake of determining which SO should be considered next for runway scheduling.Step 2B: Determine the SO with the earliest meter fix STA within each super stream class.The time computed in the previous step is used to determine the super stream class of each SO.Step 2C: Of the earliest SOs within each super stream class, select the one with the earliest runway ETA.The selected SO will be the next SO to consider for runway scheduling.This mechanism generates a sequence at the runway that resembles FCFS sequencing while maintaining the sequence computed at the meter fix.By limiting the selection of the next SO to consider for runway scheduling to those with the earliest meter fix STAs, the sequence at the meter fix is maintained.Step 2D: With the SO just selected, get the temporary meter fix STA just computed in Step 1.Step 2E: If an overcrossing time has not been received for this SO and the SO is not within the TRACON's airspace, adjust the meter fix STA to account for delay feedback of the SO ahead in the sequence during a previous iteration of Step 2.If the current SO has an overcrossing time or is within the TRACON's airspace, then use the overcrossing time as the meter fix STA.No adjustment of this meter fix STA is necessary since the SO has already crossed the meter fix.If, however, this SO does not have an overcrossing time and is in the Center's airspace, then the meter fix STA may require adjustment so that it satisfies the scheduling constraints at the meter fix while remaining in its proper place in the meter fix sequence.To accomplish this, the meter fix STA of the SO ahead of the current SO in the meter fix sequence is used as a starting point.This temporary meter fix STA is then adjusted to satisfy the scheduling constraints at the meter fix.The meter fix STA of the current SO is only compared against those SOs with permanently assigned meter fix STAs.In other words, the current SO is compared against those SOs which have already had their runway STAs computed and their delay fed back to the Center.This temporary meter fix STA is compared against the meter fix STA originally computed in Step 1, and the later of the two becomes the new adjusted meter fix STA for the current SO.If the meter fix STA is modified, then adjustments are made to the Schedule Linked List and the Acceptance Rate Bins to correspond to the modified meter fix STA.Step 2F: Set the runway PTA to the sum of the meter fix STA and the transition time from the meter fix to the final approach fix or runway threshold (see equation 1).This is the same as Step 3C of section II.I.6.Step 2G: Compute the SO's runway STA and apply any delay feedback to its meter fix STA.This step is described later.For details, see section II.I.8.Step 2H: If delay feedback caused the current SO to violate a Center scheduling constraint, then recompute the meter fix and runway STAs by going back to Step 2A.As explained in section II.I.3, if delay feedback puts the current SO's meter fix STA in violation of a scheduling constraint, then the SO's meter fix STA must be recomputed.This is because it will be necessary to delay the meter fix STA further to avoid violating the scheduling constraint, and this new meter fix STA could make the runway STA computed in Step 2G impossible to meet.The meter fix STA plus delay feedback, after adjustment for scheduling constraints, is used as the meter fix STA when the process returns to Step 2A with the current SO.Step 2I: If delay feedback did not cause the current SO to violate any Center scheduling constraints, then permanently insert the SO into the Schedule Link List and Acceptance Rate Bins.At this point, the current SO has been successfully scheduled to the runway threshold, FAF, and meter fix.The current SO is permanently placed in the Schedule Link List and Acceptance Rate Bins.This means that the current SO can affect the schedules of all remaining SOs to be scheduled.Step 3: For each SO to be scheduled using this method, compute the outer meter arc STA using equation 5. +Schedule to RunwayThis section describes the scheduling of a single SO to the runway.The algorithm described here assumes that the meter fix STA, before delay feedback, has already been computed.It also assumes that the runway PTA has already been determined.Under VFR conditions, the scheduling constraints are applied to the FAF STAs since the aircraft are under the direction of controllers up to the FAF.Under IFR conditions, the constraints are applied to the runway threshold STAs since the controllers guide the aircraft all the way to touchdown.The process of scheduling a single SO to the runway is as follows.Step 1: Apply the runway and airport scheduling constraints to the PTA and possibly delay the PTA.The following runway and airport scheduling constraints are applied to the PTA.• Airport acceptance rate • Runway acceptance rate • Required wake vortex separation • Runway separation buffer • Runway occupancy timeThe resulting PTA becomes the new STA at the runway reference point (either FAF or threshold).Step 2: Compute the STA to the other runway reference point.If the reference point where the scheduling constraints are being applied is the FAF, then the threshold STA is computed in this step.The transition time between the FAF and the threshold is added to the STA computed in the preceding step.If the reference point where the scheduling constraints are being applied is the threshold, then the FAF STA is computed in this step.The transition time between the FAF and the threshold is subtracted from the STA computed in the preceding step. +Scheduling ConstraintsScheduling constraints allow the TMC to control the flow of traffic.If there are no scheduling constraints and there is no other traffic in the system, an SO will be scheduled at its nominal ETA.Realistically, however, the TMC will enter scheduling constraints that will delay an SO.That is, an SO's STA may be later than its ETA because of these scheduling constraints.The TMC will enter scheduling constraints to ensure proper spacing, favor one flow of traffic over another, compensate for adverse weather conditions, or just generally model normal air traffic control procedures.All of the scheduling constraints are accounted for by the DP when computing STAs.Some constraints will be more restrictive than others, and these more restrictive constraints will have a greater impact on STAs than others.However, which constraint is more restrictive is not always clear.At certain times, one constraint will be more restrictive than another for certain aircraft while a different constraint will be more restrictive for other aircraft at a different time.This makes it necessary for the DP to consider all scheduling constraints when computing STAs.The scheduling constraint's start time is the time when the constraint becomes active.SOs with ETAs that are equal to or later than the start time are affected by that constraint.Additionally, a scheduling constraint can override another scheduling constraint of the same type depending on their respective activation times.For example, suppose that a constraint limiting runway 18R's acceptance rate to 40 aircraft per hour were set to go active at time 1905Z.This is represented in figure 39 by the RWY_FC timeline tag at 05 past the hour.Further, suppose that a constraint limiting runway 18R's acceptance rate to 60 aircraft per hour were set to go active at time 1920Z.This is represented in figure 39 by the RWY_FC timeline tag at 20 past the hour.Next, suppose that UAL382 has an ETA of 1910Z while UAL365 has an ETA of 1922Z, and both aircraft are assigned to 18R.Initially, UAL382's STA will be affected by the 40 aircraft per hour constraint, while UAL365 will be affected by the 60 aircraft per hour constraint.However, if delays push UAL382's STA to 1920Z or beyond, then it will be affected by the 60 aircraft per hour constraint.Scheduling constraints can be divided into two airspace categories (see table 9).The Center scheduling constraints restrict the flow of traffic at the meter fixes and have a direct impact on meter fix STAs.They will also have an indirect impact on runway threshold and FAF STAs since these ultimately depend on the meter fix STAs.On the other hand, the TRACON constraints restrict the flow of traffic at the runway threshold or the FAF.Under VFR conditions, the TRACON constraints directly affect the FAF STAs.Under IFR conditions, the TRACON constraints directly affect the runway threshold STAs.Additionally, the TRACON scheduling constraints will have an effect on the meter fix STAs as a result of the feeding back of delay from the TRACON to the Center (see section II.I.3). +Figure 39. Runway Flow Change ExampleIn addition to conceptually categorizing the scheduling constraints into airspace categories, the object-oriented design of the DP divides the scheduling constraints into several classes (see table 9).Scheduling constraints are placed into a class because they share common algorithms for their application and common types of requisite data.These classes are explained in greater detail in the sections that follow.A separate list is maintained for each type of scheduling constraint.Each list is sorted by activation time from earliest (past) to latest (future).Whenever a new Acceptance Rate constraint is added, the list is purged of constraints whose activation time precedes the current time with the exception of the constraint with the most recent activation time.The constraint with the most recent activation time is retained since it is the constraint used for the present time.The new constraint is then inserted into the list according to its activation time.However, if the new constraint has the exact same activation time as an existing constraint, then the new constraint replaces the old one.Existing constraints can be deleted from the list by specifying their activation time. +Separation DistanceThe Separation Distance scheduling constraint restricts the horizontal distance between aircraft when they cross a reference point such as the meter fix, FAF, or runway threshold.The requisite data for this constraint are the time that this scheduling constraint is to become active and the minimum number of miles of separation required between SOs.Because the DP's schedules are time-based, these separation distances must be converted into units of time separation.For example, for jets crossing over a meter fix, a 5 mile separation might translate into a 60 second separation.Thus, the DP would schedule one jet at least 60 seconds behind another.However, translating the separation distance to time varies from aircraft to aircraft.Although the aircraft may be flying according to some fixed airspeed, the winds aloft can affect the ground speed.Thus, one jet might cover 5 miles in 60 seconds while another might require 80 seconds to cover 5 miles because of a headwind.Therefore, the DP uses the predicted ground speed of each aircraft at the reference point to derive a reasonable translation from separation distance to time according to equation 6.s sep is the required separation distance, v g is the predicted ground speed at the reference point, and t sep is the resulting time separation. +Miles-in-Trail or Super Stream ClassSeparation.This scheduling constraint defines the minimum allowed horizontal separation between SOs within the same super stream class and directly affects the meter fix STAs.The --------= separation distance, applied at the meter fix, for each super stream class is independent of the separation distance of every other super stream class.This constraint is sent to the DP along with a definition of which stream classes are placed into which super stream classes.Wake Vortex Separation.This scheduling constraint restricts the minimum horizontal distance between SOs destined to individual or dependent runways.The airport configuration defines which runways are dependent on each other.SOs that are assigned to different but dependent runways are separated from each other as if they were assigned to the same runway.The required separation can vary depending on the wake vortex category of the SO ahead and the SO behind.These wake vortex categories, based on weight class and engine type, are:1. small piston, 2. small turboprop, 3. large turboprop, 4. large jet, 5. heavy jet, and 6.Boeing 757.Note that the Boeing 757 is placed in a separate wake vortex category due to its unique wake characteristics.The minimum required separation is specified by the user through the Wake Vortex Separation Matrix, which contains the required separation for each pair of wake vortex categories.Thus, a small piston following a large turboprop may have one required separation while a large turboprop following a small piston may have a different required separation.The Wake Vortex Separation constraint is sent to the DP via a Runway Flow Change message.Included in this message is an optional separation buffer.The separation buffer distance is added to each value in the Wake Vortex Matrix and can be used to compensate for uncertainty in the data. +Acceptance RateThe goal of the Acceptance Rate algorithm is to schedule as many SOs as possible to fully utilize, but not exceed, the Acceptance Rate.For example, suppose the only active scheduling constraint is an Airport Acceptance Rate constraint.An SO will be scheduled at its ETA unless doing so will exceed the Acceptance Rate.If the Airport Acceptance Rate is exceeded, then the SO will be delayed to a point where its STA no longer exceeds the Airport Acceptance Rate.At the heart of the Acceptance Rate is a data structure containing Acceptance Rate Bins.Each bin represents 30 seconds of time and contains the number of SOs scheduled within that 30 second time interval.As each SO is scheduled, the count in the bin corresponding to the SO's STA is incremented.If an SO's STA is changed as a result of delay feedback, for example, then the count in the bin corresponding to the old STA is decremented while the count in the bin corresponding to the new STA is incremented.The algorithm begins with the current SO's PTA.This is the earliest possible time that the SO may be scheduled to cross the reference point.The 30 second bin corresponding to the PTA is determined, and a window the size of the Acceptance Rate Interval (see section II.E.3) is extended into the past from the PTA's bin.The number of SOs in each bin within this window is summed.If the total is less than the Acceptance Rate, then scheduling the current SO at the PTA would not violate the Acceptance Rate in that window.Subsequently, the window is advanced into the future by one bin.Again, the number of SOs already scheduled within the window is counted.If the number is less than the Acceptance Rate, then the window is advanced again by one bin into the future.This continues until the total count within a window equals or exceeds the acceptance rate, or the window has moved far enough into the future that it no longer contains the PTA's bin.If the total equals or exceeds the Acceptance Rate, then the PTA must be delayed.Since moving the PTA to a bin that is still within the current window would not improve the situation, the PTA is delayed to the first bin in the future just beyond the current window.Once the PTA has been delayed, the whole process of creating a sliding window and counting the number of SOs within that window is repeated.The result is a PTA that satisfies the Acceptance Rate constraint.At the conclusion of every complete scheduling cycle, the Acceptance Rate Bins are emptied of all SOs with the exception of landed aircraft (see section II.F.5).Landed aircraft will not be scheduled in any future scheduling cycle, and they will never be re-entered into the Acceptance Rate Bins.As a result, it is necessary to maintain their presence in the Acceptance Rate Bins between scheduling cycles.However, landed aircraft are not counted against the Acceptance Rate under scheduling modes 2 through 5. Thus the Acceptance Rate Bins are completely cleared of both landed and nonlanded aircraft before scheduling under these scheduling modes.The example in figure 40 shows how the Acceptance Rate algorithm is applied.In this example, the Acceptance Rate is 20 SOs per 10 minute period.Each column of numbers represents the bins and the number of SOs which have already been scheduled into each bin.Column (A) shows the PTA and the first window.The window is 10 minutes wide and corresponds to the 10 minute Acceptance Rate Interval.The number of SOs within the window is 15, so the window is slid up one bin at a time.Each time, the number of SOs within that window is counted.Column (B) shows where the window is positioned when it is found that the number of SOs equals the Acceptance Rate.This means that there is no more room to schedule the current SO at its PTA.Doing so would increase the number of SOs in the window to 21 which exceeds the Acceptance Rate.The PTA is delayed to PTA'.Delaying the PTA to any bin earlier than the bin corresponding to PTA' would not help since it would still exceed the Acceptance Rate in the current window.The process of creating a window that ends at the PTA, counting SOs, and sliding the window up by one bin is repeated. +Figure 40. Acceptance Rate ExampleColumn (C) shows a window where the number of contained SOs is again 20.PTA' is then delayed to PTA'' as a result.PTA'' turns out to be the earliest time that does not exceed the Acceptance Rate.Column (D) shows the last window that is checked.If there were no other scheduling constraints in the system, then the SO's STA would be set to PTA''.Currently, the algorithm schedules these remaining SOs at the beginning of each hour.Thus, in our example, 22 SOs would be scheduled to arrive during the first 10 minute period of the hour while only 19 SOs would be scheduled to arrive during each of the other 10 minute periods.As this paper is being written, feedback from the field indicates that scheduling the extra SOs to arrive during the first 10 minute period of each hour is unsatisfactory.Therefore, the handling of the remaining SOs is currently under examination, and there is currently a plan to modify the implementation so that the remaining SOs are more evenly distributed throughout the hour.TRACON Acceptance Rate.This constraint limits the number of SOs per hour that may cross any and all meter fixes.Note that the TRACON Acceptance Rate constrains all traffic crossing the meter fixes regardless of engine type, stream class, or destination airport.Gate Acceptance Rate.This constraint limits the number of SOs per hour that may cross any and all of the meter fixes contained within a single gate regardless of engine type, stream class, or destination airport.Meter Fix Acceptance Rate.This constraint limits the number of SOs per hour that may cross a particular meter fix.Note that the Meter Fix Acceptance Rate constrains all traffic crossing the indicated meter fix regardless of engine type.However, different meter fix acceptance rates may be entered for the same meter fix but different destination airports.For example, an acceptance rate of 24 aircraft per hour may be entered for meter fix BAMBE and traffic destined for DFW.At the same time, a different acceptance rate of 18 may be entered for meter fix BAMBE and traffic destined for Dallas Love Field airport (DAL).Airport Acceptance Rate.This constraint limits the number of SOs per hour that may land at any and all runways of a particular airport.Runway Acceptance Rate.This constraint limits the number of SOs per hour that may cross the threshold or FAF of a particular runway.This constraint is sent to the DP via the Runway Flow Change message. +Occupancy TimeThe Occupancy Time scheduling constraint adds additional time to the required time separation translated from the Separation Distance constraint.Runway Occupancy Time.This constraint adds additional time to the separation times translated from the Wake Vortex Separation.It is often used to account for the extra time that may be required to stop an aircraft on a slippery runway and clear it before the next aircraft lands.Alternatively, by setting all of the entries in the Wake Vortex Matrix to zero, the Runway Occupancy Time constraint can be used to separate landing SOs strictly by time.This constraint is sent to the DP via the Runway Flow Change message.The requisite data for this constraint are the time that this scheduling constraint is to become active, the runway affected, and the number of seconds of separation to be added to the time translated from the separation distance. +Blocked IntervalsThe Blocked Interval Scheduling Constraint prevents any SOs from being scheduled to cross a particular reference point during the specified interval of time.The Blocked Interval can be used, for example, to allow time for an airport configuration change or to avoid a severe weather cell.In addition to a start time, a Blocked Interval has an associated end time after which SOs may be scheduled.The Blocked Interval algorithm begins with the current SO's PTA.The PTA is compared against all Blocked Intervals for the SO's assigned meter fix and runway.If the PTA is found to be between the start and end times, inclusive, of a Blocked Interval, then the PTA is delayed to the coincide with the Blocked Interval's end time. +J. Runway AllocationWithout any additional optimization, the DP, as part of TMA, has been shown to be beneficial to controllers and TMCs.Taking the accurate ETAs generated by the RA, computing the STAs in the DP to meet the TMC's scheduling constraints, and displaying this information on the TGUI and PGUI allows TMCs to ensure a safe and smooth flow of traffic from the Center into the TRACON.It also gives the TMCs a look into the future to assist with staffing decisions.The schedules computed by the DP can be optimized by allocating runways to SOs to reduce delay.The goal is to assign SOs to runways which reduce the delay of all SOs in the system. +Runway Allocation EventsFrom the TMC's point of view, it is undesirable to have an SO switch runways constantly even if this would continually optimize the schedule.Therefore, only certain events trigger the runway allocation process.Some of these events will cause an immediate runway allocation while others are placed in a pending list for later runway allocation.This is the same pending list as the one used by the scheduling process (see section II.I.4).When processing the events in the pending list, the DP distinguishes the scheduling events from the allocation events and takes the appropriate action.The various runway allocation event types are listed in table 10.The runway allocation process assigns an SO to each allowable runway and generates a temporary schedule based on each runway assignment.These schedules are compared when determining which runway assignment is best.The Scheduling Modes listed in table 10 are the modes used when generating these temporary schedules, and they vary from event to event.Finally, table 10 indicates which runway allocation events are processed immediately and which are deferred for later processing.The runway allocation event processing is summarized in figure 41.The numbers shown in figure 41 correspond to the step numbers given in this section's description. +Step 1: Determine if the current runway allocation event requires immediate runway allocation (see table 10). If this is an immediate allocation event, then go toStep 4.Step 2: The current runway allocation event is a pending event.If the list of pending events is full or the events in the list are different from the current allocation event, then go to Step 4.Step 3: Add this event to the list of pending events and quit.Reaching this step means that the current allocation event is a pending event, there is room in the pending list, and this event is of the same type as those in the pending list.This event is added to the pending list, and the runway allocation event processing is ended at this point.This event will be processed later, along with the other pending events in Step 4 in response to a different runway allocation event.Step 4: Process the events in the list of pending allocation events (if any) and flush the list.The SOs requiring runway allocation are collected from each pending event.Since all the pending events are of the same type, the same allocation parameters can be used for the entire set of SOs just collected.The details of the runway allocation process are explained in the next section.Step 5: Process the current allocation event.The details of the runway allocation process are explained in the next section.Step 6: Perform a final reschedule.During the runway allocation process, many partial and temporary schedules are computed.At the end of the allocation process, one final reschedule is executed to make sure each of the affected SOs has an updated schedule to its newly assigned runway.Step 7: Prepare and send out a schedule message containing the newly computed STAs.The scheduling message contains the STAs of the SOs.The message also indicates which runway corresponds to each STA just computed.Step 8: Each SO's airport configuration is updated.This is the same as Step 15 in section II.I.4.Step 9: Flight plan amendments consisting of a change in configuration, runway, or both are sent to the other CTAS processes.An SO may be associated with a different airport configuration than it was prior to executing the runway allocation process.Additionally, an SO may be assigned to a new runway as result of the runway allocation process.For each of these SOs, a flight plan amendment is sent notifying the rest of CTAS of the change.then the event is treated as a rescheduling event instead (see table 3).The SO selection process begins by determining which runways have an increasing, decreasing, or steady acceptance rate as a result of the runway flow change event.Only SOs whose runway threshold ETAs or STAs are equal to or later than the time of the runway flow change are considered for selection.In addition, SOs whose runway assignments have been locked as a result of manual runway assignments are not selected.From the remaining SOs, the selection process determines which SOs require runway allocation depending on which of the following conditions is true.Condition 1: Exactly one runway has an increasing acceptance rate, and the acceptance rates of the other runways are steady or decreasing.SOs from runways with decreasing acceptance rates and Non-STA-Frozen SOs on steady acceptance rate runways are selected.Condition 2: More than one runway has an increasing acceptance rate.SOs from runways with decreasing or increasing acceptance rates and Non-STA-Frozen SOs on steady acceptance rate runways are selected.SOs from runways with decreasing acceptance rates are selected. +Selecting SOs for Runway Allocation Due to the Addition or Deletion of a Runway Blocked Interval.This method of selecting SOs for runway allocation is executed in response to the addition or deletion of a runway blocked interval (event 1 in table 10).Only SOs whose runway threshold ETAs or STAs are between the start and stop time of the blocked interval (inclusive) are considered for selection.From this group of SOs, this method selects SOs for runway allocation in a manner similar to that used when a runway flow change is added or deleted (see above).In this case, the addition of a blocked interval to a runway is similar to reducing the acceptance rate of that runway, and the deletion of a blocked interval is similar to increasing the acceptance rate. +Runway Allocation ProcessOnce a runway allocation event has been received and the SOs that require allocation have been selected, the actual runway allocation process is executed on the selected SOs.The SOs are processed in order of increasing runway threshold ETAs.Placing the SOs in this order is the responsibility of the SO selection methods described in the previous section.The runway allocation process is summarized in figure 42.The numbers shown in figure 42 correspond to the step numbers given in the this section's description.Step 1: Determine the Runway Allocation Mode.Many of the runway allocation events trigger similar runway allocation methods.Those events with similar allocation methods are grouped together into Runway Allocation Modes.There are currently three runway allocation modes, and the mapping from runway allocation events to runway allocation modes is shown in table 10.The Runway Allocation Mode derived from the runway allocation event is then used in the next step.Step 2: Get the Runway Category from the Runway Decision Tree.Decision trees are used throughout CTAS as a way of encoding site-dependent rules.Similar in functionality to nested case statements in C, a decision tree identifies a category based on a set of criteria.Because of its generic nature, decision trees can be applied in many different ways and can vary from site to site.The category that Figure 42.Runway Allocation Process results from the traversal of a decision tree can then be used to identify which set of rules is to be followed.Thus, decision trees provide a mechanism to execute a set of rules specifically designed for a certain set of criteria.Runway allocation is one of these CTAS processes that utilize a decision tree.The runway allocation decision tree is contained in a site-dependent data file called dp_runway_decision_tree.The decision tree is traversed for each SO to determine its runway category.The criteria used by the runway decision tree are listed below.assign_runway.This specifies which runway to assign to the SO if the criteria listed are met.The name of the runway will follow the keywords assign_runway, or the words PREVIOUSLY_ALLOCATED_RWY will appear in place of the runway name.In the latter case, the runway to which the SO had been previously assigned is the one that is examined.from_runway R.This criterion is true if the best runway chosen so far is R.For example, in the last assign_runway block in listing 2, in order to accept 17L as the best runway, the best runway chosen so far must be 17C plus the result of check_criteria which is explained below.check_criteria.This specifies a criterion to be met for this runway assignment to be the best runway chosen so far.It is followed by a rule and possibly a value contained in curly braces.These rules are as follows.DEFAULT_RUNWAY {}.This criterion is always true.When this criterion appears in an assign_runway block no other criterion can be contained within that block.It is used to specify the runway to assign to the aircraft if all of the other possible assignments are less desirable.The SO is temporarily assigned to this runway, and a temporary schedule is computed.The runway threshold STAs of all SOs are added together, and this sum is known as the System Schedule Time.For example, in listing 2, 13R is initially considered the best runway.This may change as the subsequent assign_runway blocks are processed. +USE_RUNWAY_IF_DELAY_REDUCED {D}.The SO is temporarily assigned to the runway specified on the immediately preceding assign_runway line.A temporary schedule is computed along with the associated System Schedule Time.If this System Schedule Time is less than the System Schedule Time of the best runway assignment seen so far by D minutes, then this criterion is true.If the other criteria within the assign_runway block hold up, then this runway becomes the best runway seen so far.For example, in listing 2, 13R is initially the best runway.In the second assign_runway block, the SO is assigned to 18R and a schedule is computed.If the System Schedule Time associated with assigning the SO to 18R is better than the System Schedule Time associated with assigning the SO to 13R by more than 0.0 minutes, then 18R becomes the best runway so far.The number of minutes D can be used to prevent an SO from jumping back and forth between runways just because the System Schedule Time is reduced by a couple of seconds. +RUNWAY_ACCEPTANCE_RATE_CHANGE {}.This criterion is true if either of the following conditions is true.• The SO's original runway has an increasing or steady acceptance rate and the runway specified in assign_runway has an increasing acceptance rate.• The SO's original runway has a decreasing acceptance rate and the runway specified in assign_runway has an increasing or steady acceptance rate.This criterion steers the runway selection to runways that have improving acceptance rates when compared with the SO's original runway.For example, if one runway has an increasing acceptance rate, then SOs on other runways need only compare their original runways against the runway with the increasing acceptance rate.As another example, if the SO is originally on a runway with a decreasing acceptance rate, then all runways with increasing or Once the acceptance rates have been apportioned to the Requested Super Stream Classes, the separation distance for each of the requested super stream classes is computed.For each Requested Super Stream Class, j, the average ground speed, s j , is used to convert the acceptance rate, AR j , to the separation distance d j , according to the following equation.(11) Finally, a response message is created.This message contains separation distances for each of the super stream classes.Note that if a Requested Super Stream Class has no qualifying aircraft, then the average ground speed and, hence, the separation distance cannot be computed for that Requested Super Stream Class.In this case, an exceptional value is stored in the message.The recipient recognizes this exceptional value as an indication that a separation distance was not successfully computed for a particular super stream class.The message also contains a flag indicating if MINTA was successful in computing the separation distances given the inputs and the data available.In addition, the message contains an indication of whether the desired TRACON acceptance rate, AR TRACON , is estimated to be exceeded if the recommended separation distances are followed.This is due to the provision in the algorithm that guarantees that a minimum of 10% of the desired TRACON acceptance rate (see equation 9 and equation 10) is apportioned to each super stream class. +L. ETA HoveringETA Hovering is a mechanism that the DP uses to compensate for any inaccuracies in the coordination fix time contained in a flight plan.Before CTAS receives active tracks for an aircraft, the RA will compute the aircraft's ETA based on information contained in the aircraft's flight plan.Sometimes, however, an aircraft will become active at a different time than indicated by the flight plan.If left alone, the ETA based on the inactive aircraft's flight plan will move inside the STA Freeze Horizon.The STA, which is based on this ETA, may be frozen at a time as early as the ETA.If the aircraft becomes active much later than indicated by the flight plan, then the new active aircraft ETA computed by RA will be much later than the ETA computed for the flight plan.This will make it impossible for the aircraft to meet the STA which was computed while the aircraft was inactive.The ETA hovering mechanism hovers the ETA of such an inactive aircraft outside the STA Freeze Horizon.Periodically, the DP checks all of the aircraft and hovers the ETAs of those aircraft which are eligible.If an eligible aircraft is found to be within the ETA Hover Horizon, then the ETA Hover Amount is added to the aircraft's ETA.The ETA Hover Horizon is currently set to be 180 seconds outside the STA Freeze Horizon, and ---------------= this value may be changed by the user as appropriate for the deployment site.The additions to the ETA are maintained in the DP.The other CTAS processes are not aware of the change in ETA except via the hovering's effect on the aircraft's STA.It's possible that an eligible aircraft may be hovered several times to keep it outside of the STA Freeze Horizon until it goes active.ETA hovering is applied only to aircraft.Blocked slots never have their ETAs hovered.Moreover, only aircraft which satisfy all of the following criteria are eligible for ETA hovering:• The aircraft must be inactive.That is, no tracks have been received, and the aircraft's ETA is computed using information contained in its flight plan.• The aircraft cannot be STA-Frozen.• The aircraft has not landed (see section II.F.5).• The aircraft is not a pop-up (see section II.F.3).• The aircraft has not been manually departed (see section II.F.1).• The aircraft is not a proposed flight plan.• The aircraft is not awaiting a new ETA as a result of a flight plan amendment.It is possible that the new ETA will make an aircraft ineligible for ETA hovering, so the DP waits until the ETA is received before hovering the aircraft.• The aircraft is not a departed flight plan.• The aircraft's flight time to the meter fix is not less than the ETA Hover Horizon. +M. BroadcastThe Communications Manager (CM), in response to certain events, will prevent the sending of STAs to the controllers' Planview Displays (PVDs) and PGUIs via a mechanism known as Broadcast Blocking.The events which trigger Broadcast Blocking are those which have a direct effect on the STAs of STA-Frozen SOs.Since only STAs of STA-Frozen aircraft are displayed on the controllers' PVDs, these are the type of changes that will be noticed by the sector controllers.Such changes are referred to as an aircraft list "ripple."Broadcast Blocking allows the TMC to make multiple adjustments to the traffic flow and make sure all of the sector controllers are ready for the inevitable "ripple."The DP is not even aware that schedules are blocked.The DP continues to generate schedules and these schedules are updated on the TMC's TGUI and PGUI.The responsibility of blocking the schedules is left to the CM through which all message traffic flows.To turn Broadcast Blocking off, the TMC issues a Broadcast All command from the TGUI.When the DP receives the request to Broadcast All, it prepares a schedule message containing the STAs of all aircraft and sends it to the CM.The CM will turn the Broadcast Blocking off and forward the DP's schedule message to all PGUIs, TGUIs, and the two-way interface.In addition to the Broadcast All command, the TMC may request that the schedule for a particular aircraft be broadcast using a Broadcast command.When the DP receives such a request, it prepares a schedule message containing the scheduling information for the requested aircraft.The CM will forward this schedule message to all PGUIs, TGUIs, and the two-way interface.However, CM will not turn the Broadcast Blocking off.Turning the Broadcast Blocking off requires that the TMC issue the Broadcast All command. +N. Design MethodologyA team of software engineers, aerospace researchers, and air traffic control experts was formed to establish the requirements for the DP.The requirements reflected the lessons learned from previous scheduler implementations, the experience gained in the field, and the possible direction of future research.Once the requirements were established, a team of software engineers applied the techniques of Object-Oriented Analysis (OOA) and Object-Oriented Design (OOD) as described by Rumbaugh et al. [18].OOA and OOD techniques were selected because of the ease with which the implementation task could be divided among the programming resources as well as the ease with which the design could be maintained.OOA and OOD also resulted in a design that was flexible enough to meet the dynamic requirements of the air traffic control researchers.The resulting design was implemented in ANSI C.Although C is not an Object-Oriented Programming (OOP) language, the software engineering team established a set of programming guidelines that resulted in C code that resembled an object-oriented implementation.The software engineering team as well as future software developers must maintain the discipline to follow these guidelines in order to maintain an object-oriented program since the C language offers very few tools to enforce object-oriented programming. +III. CONCLUDING REMARKSThe DP computes the aircraft sequence, STAs, and runway assignments to ensure an orderly, efficient, and conflict-free flow of traffic into the terminal area as part of the TMA tool of CTAS.The DP sequences the aircraft so that they arrive in an FCFS order at the meter fix unless the TMC overrides this order via manually entered sequence constraints.Also, the STAs computedFigure 2 .2Figure 2. Typical Arrival Flight Paths +Figure 4 .4Figure 4. Direct Constraint +Deconflicting sequence constraints requires building a directed graph called the Sequence Constraint Graph.Each node in the graph (represented by a circle in the accompanying figures) represents an SO involved in a constraint, and each SO can appear, at most, once in the graph.Each directed edge (represented by an arrow in the accompanying figures) extends from the SO behind in the sequence constraint to the SO ahead in the sequence constraint.Additionally, each edge indicates if the sequence constraint is a "direct" constraint (represented by double lines in the following diagrams) or an "indirect" constraint (represented by a single line in the following diagrams). +Figure 5 .Figure 6 .Figure 7 .567Figure 5. Indirect Constraint +Figure 8 .8Figure 8. Redundant Constraint +Figure 13 .13Figure 13.Redundant Constraint +Figure 18 .18Figure 18.Trying to Add an Existing Constraint +Figure 20 .20Figure 20.SCTR Example #1 +Figure 23 .23Figure 23.Indirect Constraint Added to the Graph +Figure 28 .28Figure 28.SCTR Applied to a Direct Constraint +Figure 31 .31Figure 31.SCTR Applied to an Indirect Constraint +Figure 32 .32Figure 32.SCTR Applied to an Indirect Constraint +Figure 35 .35Figure 35.SCTR Applied to a Direct Constraint +Figure 38 .38Figure 38.Reschedule after Aircraft Ahead and Insert Flow Chart +Condition 3 :3One or more runways has a decreasing acceptance rate, and no runways have an increasing acceptance rate. +aerodynamic and propulsion characteristics, and preferred speeds), airspace structure data, and near real-time weather information.In addition, the Route Analyzer (RA) (see section II.D.5) provides the TS with each aircraft's initial condition, horizontal route of flight, desired end conditions, and intermediate speed and altitude constraints.The output of the TS consists of nominal, slow, fast, and meet-time ETAs to various reference points, aircraft ground speeds at the reference points, and the nominal 4D trajectory for each aircraft.Of particular importance to the DP's calculations are the ETAs to, and the ground speeds at, each of the following reference points.It provides the other CTAS processes with accurate 4-dimensional trajectories and the associated ETAs at various reference points for each aircraft.Inputs to the TS include aircraft model data for all aircraft types (including +Table 1 .1Blocked Slot Aircraft TypesBlocked SlotRepresentativeFAATypeAircraft ModelDesignationsmall pistonCessna 172C172smallBeechcraft KingBE20turbopropAir 200largeEmbraer EMB-120 E120turboproplarge jetBoeing 727B727heavy jetBoeing DC-10DC10Boeing 757Boeing 757B757 +Table 2 .2Scheduling ModesScheduling Mode #Which SOs Included a +Table 3 .3Scheduling EventsScheduling Event TypeSchedulingImmediate orScheduling EventMode #PendingTime ReferenceEventPointFOR_BLOCKED_INTERVAL_METER_FIX5ImmediateMeter FixAddition or deletion of a meter fix blocked intervalFOR_FLOW_CHANGE_AIRPORT5ImmediateRunwayChange in the airport acceptance rateFOR_FLOW_CHANGE_RUNWAY5ImmediateRunwayChange in the occupancy time or required separationdistance at a runway (This event is usually treated as arunway allocation event. However, if the runway allocatordetects that a runway acceptance rate has not changed as aresult of this event, then this event is treated as a schedulingevent.)FOR_FLOW_CHANGE_GATE5ImmediateMeter FixChange in the gate acceptance rateFOR_FLOW_CHANGE_METER_FIX5ImmediateMeter FixChange in the meter fix acceptance rateFOR_FLOW_CHANGE_STREAM_CLASS5ImmediateMeter FixRedefinition of super stream classes and/or change in therequired separation distance at the meter fixFOR_FLOW_CHANGE_TRACON5ImmediateMeter FixChange in the TRACON acceptance rate +Table 3 .3Scheduling Events (Continued)Scheduling Event TypeSchedulingImmediate orMode #PendingEvent +Table 3 .3The Order of Consideration algorithm is executed when runway STAs are computed.It assumes that a preliminary meter fix STA has already been computed for each SO.The algorithm begins by determining the SO with the earliest meter fix STA within each super stream class.Among these SOs, the SO with the earliest runway ETA is selected as the next SO in the order of consideration.Next, this SO has its runway threshold and FAF STAs computed and any necessary delay is fed back to its meter fix STA.This algorithm is repeated until all SOs have been scheduled to the runway.Scheduling Events (Continued)Scheduling Event TypeSchedulingImmediate orScheduling EventMode #PendingTime ReferenceEventPointFOR_USER_REQUEST_AT_MFspecified inImmediateAll SOs requested:TMC requests a reschedule via a meter fix timeline menurequestRunwayOtherwise: MeterFixFOR_PERIODIC_RESCHEDULE8ImmediateRunwayAt least 6 seconds since a reschedule involving an even-numbered modeConsider the example in table 4. Aircraft A1 and A2 arein stream class A. B1 and B2 are in stream class B. C1and C2 are in stream class C. The table also shows thepreliminary meter fix STA and the runway ETA for eachof these aircraft. For this example, assume that aircraftmust maintain 1 minute separation at the meter fix fromother aircraft within its stream class. Also, assume thatthe required separation at the runway is 1.5 minutes. +Table 4 .4Order of Consideration Example B1, and C1 have the earliest meter fix STA for each stream class.Among these, C1 has the earliest runway ETA, so its runway STA is computed first.C1's runway STA is computed to be 12:10:00Z.ID StreamPrelim.RunwayComputedClassMF STAETARunway(Zulu)(Zulu)STA(Zulu)A1 A12:00:0012:12:00 12:12:00A2 A12:02:0012:14:00 12:15:00B1 B12:00:0012:15:00 12:16:30B2 B12:01:0012:16:00 12:18:00C1 C12:00:0012:10:00 12:10:00C2 C12:03:0012:13:00 12:13:30Since A2 has the earlier runway ETA, it is next in the Order of Consideration.The only slot available to A2 is 1.5 minutes behind C2.Thus, A2 is delayed at the runway by a minute and is scheduled to land at 12:15:00Z. +Table 5 .5Scheduling Eligibility Truth TableSO Statushas valid ETAs0******11111ETA is earlier than the rescheduling start time* a 1*****00000landed** 1****00000suspended***1***00000pop-up****1**00000expired*****1*00000proposed flight plan********0101departed********0011(User manually departed this SO.)active******010000(DP has received radar track data for this SO.)schedule flight plans******0*1111(The DP has been set to include inactive SOs inthe scheduling process. This is normallyswitched on.)Eligible for Scheduling000000011011a. * = any value +Table 6 .6Schedule All Including STA-Frozen SOs (modes 2, 3, 4, and 5)Scheduling Priority SetScheduling MethodSOs with ETAs earlier thanInsert withoutthe rescheduling start timeReschedulingManually scheduled SOsInsert withoutReschedulingPriority SOsReschedule at ETAand InsertOther SOsReschedule afterAircraft Ahead andInsert +Table 7 .7Schedule All Non-STA-Frozen SOs (modes 1, 6, 7, 8, and 9)Scheduling Priority SetScheduling MethodSOs with ETAs earlier thanInsert withoutthe rescheduling start timeReschedulingManually scheduled SOsInsert withoutReschedulingSTA-Frozen priority SOsInsert withoutReschedulingSTA-Frozen non-priorityInsert withoutSOsReschedulingNon-STA-Frozen priorityReschedule at ETASOsand InsertOther SOsReschedule afterAircraft Ahead andInsert +Table 8 .8Schedule All Non-Sequence-Frozen SOs (modes 10 and 11)Scheduling Priority SetScheduling MethodSOs with ETAs earlier thanInsert withoutthe rescheduling start timeReschedulingSequence-Frozen SOsInsert withoutReschedulingNon-Sequence-FrozenReschedule at ETApriority SOsand InsertOther SOsReschedule afterAircraft Ahead andInsert +Reschedule at ETA and Insert Flow Chart have any effect.Note that it is necessary to compute the meter fix STA for SOs that have already crossed over the meter fix because STAs in the past have an impact on present and future STAs via the acceptance rate constraints.For example, suppose that a particular meter fix has an acceptance rate of 24 aircraft per hour, and 20 aircraft have already crossed over the meter fix in the past 50 minutes.The scheduling constraint should prevent more than 4 aircraft from crossing the meter fix in the next 10 minutes.Only by computing the past meter fix STAs of those 20 aircraft using their overcrossing times can this limitation of 4 aircraft be properly enforced.The overcrossing time (see section II.B.4) is used as the meter fix STA and is not adjusted to comply with Center scheduling constraints since these constraints no longer Figure 37. +If an STA was successfully computed from Step 2A or Step 2C, then insert the current SO into the Schedule Linked List and the appropriate Acceptance Rate Bins.• TRACON acceptance rate• Meter fix acceptance rate• Gate acceptance rate• Super stream class separation Miles-in-Trail• Meter fix blocked intervalsStep 2D: The meter fix STA just computed is stored in theSchedule Linked List and Acceptance Rate Bins as atemporary meter fix STA. It is used to detect schedulingconflicts between this SO and any other SOs which aresubsequently processed in Step 2. However, since this is a temporary meter fix STA assignment, other SOs, which are being scheduled to the runway threshold and have their delay fed back to the Center, ignore this SO's meter fix STA when applying the scheduling constraints. +3E: If delay feedback caused the current SO to violate a Center scheduling constraint, then recompute the meter fix and runway STAs by going back to Step 3A.As explained in section II.I.3, if delay feedback puts thecurrent SO's meter fix STA in violation of a schedulingconstraint, then the SO's meter fix STA must berecomputed. This is because it will be necessary to delaythe meter fix STA further to avoid violating thescheduling constraint, and this new meter fix STA willmake the runway STA computed in Step 3D impossibleto meet. The meter fix STA plus delay feedback, afteradjustment for scheduling constraints, is used as themeter fix STA when the process returns to Step 3A withthe current SO.Step 3F: +If delay feedback did not cause the current SO to violate any Center scheduling constraints, then permanently insert the SO into the Schedule Link List and Acceptance Rate Bins. +Table 9 .9Scheduling ConstraintsScheduling ConstraintAirspace Category Scheduling Constraint ClassMeter Fix Acceptance RateCenterAcceptance RateGate Acceptance RateCenterAcceptance RateTRACON Acceptance RateCenterAcceptance RateSuper Stream Class Separation DistanceCenterSeparation Distance(Miles-in-Trail)Meter Fix Blocked IntervalCenterBlocked IntervalAirport Acceptance RateTRACONAcceptance RateRunway Acceptance RateTRACONAcceptance RateRunway Occupancy TimeTRACONOccupancy TimeWake Vortex SeparationTRACONSeparation DistanceRunway Blocked IntervalTRACONBlocked Interval +The algorithm translates the hourly acceptance rate, given as SOs per hour, into the number of SOs per Acceptance Rate Interval.For example, a rate of 120 SOs per hour at DFW would translate into 20 SOs per 10 minute period because the Acceptance Rate Interval is 10 minutes at DFW.In addition, the algorithm must handle the case in which the hourly acceptance rate is not evenly divisible by the number of Acceptance Rate Intervals per hour.Continuing our Dallas/Ft.Worth example, if the hourly acceptance rate were 117 SOs per hour, then this would be translated to 19 SOs per 10 minute period.However, this would result in a total acceptance rate of only 114 SOs per hour.The algorithm must try to schedule an additional 3 SOs within each hour. +Table 10 .10Runway Allocation EventsRunway Allocation Event TypeSchedulingImmediateRunway AllocationMode #or PendingModeEvent1. Add/delete runway blocked interval5ImmediateBlocked IntervalChange Allocation2. Add/delete runway flow change a5ImmediateRunway AcceptanceRate Change Allocation3. Change in the airport configuration of an SO5ImmediateAllocate WithinConfiguration4. TMC manually sets departure of a "proposed flightplan" aircraft in Center +Boeing 747 destined for DFW and assigned to the BAMBE meter fix.The first line in the file asks for which airport United 242 is destined.In this case, the destination is DFW.Next, the decision tree asks to which meter fix has United 242 been assigned.United 242 has been assigned to BAMBE, so the runway allocation mode must be determined.The event triggering runway allocation is an airport configuration change, and according to table 10, the runway allocation mode is ALLOCATE_WITHIN_CONFIGURATION.Next, the airport configuration is examined.Since SOUTH_4_VFR is one of the configurations listed, United 242's engine type is examined.Since United 242 is neither a turbo prop nor a piston aircraft, the default value is used.The resulting runway allocation category is DFW_BA_S_4_13R_DEF, and this is used in Step 3.• SO's destination airport • Airport configuration for the SO • SO's meter fix • SO's engine type An excerpt from the dp_runway_decision_tree file is shown in listing 1. Suppose that there has been a change to the SOUTH_4_VFR airport configuration, and United 242 requires runway allocation. Further, suppose that United 242 is a Step 3: For each SO, follow the site-specific allocation rules. Associated with each runway category is a set of rules for determining the best runway for a particular SO. These rules are contained in the site-dependent data file called dp_runway_category_definitions. If, after following these rules, it is determined that the best runway for a particular SO is different from the runway previously assigned, then a flight plan amendment is sent to the other CTAS processes informing them of the change in runway assignment.An excerpt from the dp_runway_category_definitionsfile is shown in listing 2. Continuing with the United 242example, this excerpt shows the runway category, calledDFW_BA_S_4_13R_DEF, determined from theprevious step. The rules in this category are processedfrom top to bottom. Each of these rules is explainedbelow.criteria WHICH_AIRPORT {value DFWcriteria WHICH_METER_FIX {value BAMBEcriteria WHICH_RWY_ALLOC_MODE {value ALLOCATE_WITHIN_CONFIGURATIONcriteria WHICH_CONFIGURATION {value SOUTH_4_VFRvalue SOUTH_4_IFRvalue SOUTH_4_VFR_BA_GG_REROUTEvalue SOUTH_4_IFR_BA_GG_REROUTEcriteria WHICH_ENGINE{value TURBO_PROPvalue PISTONcategory DFW_BA_S_4_13R_TURBO_DEFvalue DEFAULTcategory DFW_BA_S_4_13R_DEF}Listing 1. dp_runway_decision_tree Data File Excerpt +Let P be the number of Specified Super Stream Classes.The acceptance rates for the P Specified Super Stream Classes are subtracted from the desired TRACON acceptance rate, AR TRACON .The result, AR req , is the total TRACON acceptance rate to be apportioned to the super stream classes for which the TMC has not specified a separation distance.However, for practical reasons, the minimum that AR req can be is 10% of the desired TRACON acceptance rate AR TRACON (see equation 10).Next, the remaining super stream classes, which did not have a separation distance specified for them, are processed.These super stream classes are known as Requested Super Stream Classes.The remaining acceptance rate, AR req , is divided among the Requested Super Stream Classes in proportion to the number of aircraft in each super stream class.Because the minimum that AR req can be is 10% of the desired TRACON acceptance rate, AR TRACON , it is possible that the separation distances computed by MINTA may result in a TRACON Acceptance Rate that exceeds the desired TRACON Acceptance Rate, AR TRACON .The number of qualifying aircraft is counted for each super stream class.Simultaneously, the average ground speed for each super stream class is computed for use during a later step when the acceptance rate is converted to a separation distance.For example, suppose there are three super stream classes.Further suppose that the first super stream class has 10 aircraft within the specified time period, and the other two super stream classes have 5 aircraft each.Since the first super stream classes accounts for 50% of the total traffic, 50% of AR req is apportioned to the first super stream class.So, if AR req is 60 aircraft per hour, then the first super stream class is given an acceptance rate, AR j , of 30.The other two super stream classes are each given a rate of 15.acceptance rate, AR i , according to the followingequation.category DFW_BA_S_4_13R_DEFassign_runway13Rcheck_criteria DEFAULT_RUNWAY (7){}assign_runway18Rcheck_criteria USE_RUNWAY_IF_DELAY_REDUCED{0.0}assign_runway17Ccheck_criteria USE_RUNWAY_IF_DELAY_REDUCED{1.0}assign_runway17Lfrom_runway17Ccheck_criteria USE_RUNWAY_IF_DELAY_REDUCED{1.0}Listing 2. dp_runway_category_definitions Data File Excerpt(8)(9)(10) + In the DP's original design an aircraft was defined as a pop-up if the first ETA associated with the aircraft's first radar track occurred within the STA Freeze Horizon. + a.Only changes in the runway acceptance rates trigger runway allocation.Other changes, like occupancy time or wake vortex separation, only trigger rescheduling (see section II.I.1). + A ground speed of zero is an indication that the Route Analysis (RA) and Trajectory Synthesizer (TS) programs failed to compute a ground speed for that aircraft. + + + +Selecting SOs for Runway AllocationDepending on the event triggering the runway allocation process, different methods are used to determine which SOs require allocation.Common to these different methods is the restriction that the following SOs do not undergo runway allocation.• Inactive SOs with proposed flight plans • Expired SOs • Pop-up SOs • Suspended SOs Additionally, each of the selection methods described below sorts the selected SOs in order of their threshold ETAs.Selecting a Single SO for Runway Allocation.Most runway allocation events involve a single SO (see events 3 through 11, 15, and 16 in table 10).The SO that undergoes runway allocation is specified as part of the event.Selecting All SOs for Runway Allocation.Runway allocation event 13 in table 10 involves all SOs.In response to this event, all SOs are sent through the runway allocation process.Selecting All SOs Following a Particular SO for Runway Allocation.Runway allocation event 14 in table 10 occurs when the user specifies that a particular SO, along with all SOs that follow it, requires runway allocation.All SOs that are destined for the same airport as the selected SO are examined.Those SOs whose runway threshold ETAs or STAs are equal to or later than the runway threshold ETA of the selected SO are put through the runway allocation process. +Selecting SOs for Runway Allocation Due to a RunwayFlow Change.This method of selecting SOs for runway allocation is executed in response to a runway flow change event (event 2 in table 10) that changes the acceptance rate of one or more runways.If a runway flow change event does not change any acceptance rates, +ADD_OR_REMOVE_BLOCKED_INTERVAL {}.Under the current implementation, this criterion is treated the same as RUNWAY_ACCEPTANCE_RATE_CHANGE with the idea that adding a blocked interval is the same as decreasing a runway's acceptance rate, and deleting a blocked interval is the same as increasing a runway's acceptance rate.Continuing the example of United 242, the first line in listing 2 tells the runway allocator to first assign United 242 to runway 13R.The resulting schedule is computed, and let's say the associated System Schedule Time is 45000.0minutes.Next, United 242 is temporarily assigned to runway 18R.Another schedule is computed, and the associated System Schedule Time is 44996.0minutes.Since the parameter to the USE_RUNWAY_IF_DELAY_REDUCED criterion is 0.0, runway 18R will become the best runway examined so far if the new System Schedule Time is less than the System Schedule Time computed for runway 13R.Indeed, the System Schedule Time for 18R is less than that for 13R, so 18R becomes the best runway examined so far.Next, United 242 is temporarily assigned to runway 17C, and the resulting System Schedule Time is 44995.5 minutes.Although the new System Schedule Time is less than that computed for runway 18R, the difference does not exceed the USE_RUNWAY_IF_DELAY_REDUCED parameter of 1.0.Therefore, 18R remains the best runway examined so far.Finally, the last assign_runway block is processed.The criterion from_runway 17C means that 17L is only considered if the best runway so far is 17C.This is not the case, so the runway allocator does not need to consider runway 17L.The final result is that United 242 is assigned to runway 18R.If 18R is different from the runway that United 242 was assigned to before the execution of the runway allocation process, then a flight plan amendment is sent to the other CTAS processes informing them of the change in runway assignment.Step 4: Compute a schedule using the newly assigned runways.Once all of the SOs involved in runway allocation have been assigned to runways, a final schedule is computed (see section II.I.4).Subsequently, a schedule message containing the new STAs is sent to the rest of CTAS. +K. Miles-in-Trail AdvisorThe Miles-in-Trail Advisor (MINTA) functionality in the DP analyzes the future traffic flow and computes the super stream class separations (Miles-in-Trail) to meet a TRACON acceptance rate specified by the TMC.Less busy super stream classes will be given larger separations in order to relieve the pressure from the busier super stream classes.This functionality is still undergoing research and development, but the current functionality is described below.The TMC, through the TGUI, sends a Miles-in-Trail request which includes the following information. +Start Time and Stop Time.The start time and stop time specify the period for which the advisory is to be used.The MINTA functionality will analyze aircraft only from this period when formulating a response to the TMC's request. +Desired TRACON Acceptance Rate (AR TRACON ).The MINTA functionality will compute the separation distances for each super stream class so that the desired TRACON acceptance rate is fully utilized.Super Stream Class Definitions.These definitions specify which stream classes are grouped into which super stream classes for the time period specified above.This is similar to specifying the super stream class separation distance scheduling constraint (see section II.I.10). +Manual Separation Distances.The TMC may specify the desired separation distance for zero or more super stream classes.MINTA will compute the separation distances for the remaining super stream classes such that the TRACON acceptance rate is fully utilized.Once the parameters have been specified by the TMC, MINTA first processes the super stream classes for which the TMC has specified the separation distances.These super stream classes are known as Specified Super Stream Classes.The average ground speed of qualified aircraft (not blocked slots) in each of these Specified Super Stream Classes is computed.To be qualified, an aircraft must have a non-zero ground speed 5 at the meter fix.Additionally, if the aircraft is STA-Frozen, then its meter fix STA must be between the start and stop times specified by the TMC.If the aircraft is not STA-Frozen, then its nominal meter fix ETA must be between the start and stop times specified by the TMC.For each Specified Super Stream Class, i, the average ground speed, s i , is used to convert the specified separation distance, d i , to an by the DP meet all of the spatial and flow constraints entered by the TMC.These constraints reflect the current and anticipated runway capacities, traffic densities and distributions, airport configurations, and weather conditions.The DP assigns aircraft to the active runways to optimize the schedule while taking into account any operational runway assignment procedures.The above results are updated in response to changing events and TMC inputs at a rate comparable to the live radar update rate.OOA and OOD methods were used in the design of the DP.This has made it relatively easy to modify the DP in response to feedback from the field.The DP is the first component of CTAS to use object-oriented techniques in its design, and its success has led to a wider use of object-oriented techniques in the design of other CTAS processes.The DP continues to evolve as new research is conducted.Research is currently being conducted to improve the Miles-in-Trail Advisor [19], improve the fairness in the distribution of delay between meter fixes, and basing the STA Freeze Horizon on the location of the aircraft as opposed to its ETA.Future work includes making the modifications necessary to fully integrate TMA with FAST, and adding a method of detecting when aircraft have been placed into holding, and its effect on the STAs of other aircraft.The DP has been installed at both Denver Center [20] and Ft.Worth Center for testing and evaluation [4].Feedback from the field has been both positive and useful, and a number of changes have been implemented in response to requests from the field.Currently, the DP is in daily use as a flow-visualization tool at Atlanta, Denver, Los Angeles, and Miami Centers.In addition, the DP is in daily use at Ft. Worth Center as the primary arrival planning tool.In the near future, the DP will also be the primary arrival planning tool for Altanta, Denver, Los Angeles, and Miami Centers. + + + + + + + Flight Preparation, Flight Plans and Flight Itineraries + + AlexisJ DFafard + + Flight Plans . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + 10.1007/978-981-19-3599-2_19 + + + Canadian Air Law for Pilots + + Springer Nature Singapore + + + + + Flight Plans . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 + + + + + Exome Sequencing in early disease diagnosis: are we on the right track? + 10.15761/du.1000113 + + + Diabetes Updates + Diabetes Updates + 2631-5483 + + 5 + 1 + 4 + + Open Access Text Pvt, Ltd. + + + Track Updates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 + + + + + Seismic anisotropy estimated from P-wave arrival times in crosshole measurements + + YanghuaWang + + Estimated Times of Arrival (ETAs). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + 10.1111/j.1365-246x.2010.04907.x + + + Geophysical Journal International + 0956-540X + + 184 + 3 + + + Oxford University Press (OUP) + + + Estimated Times of Arrival (ETAs). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 + + + + + overcrossing + + Overcrossing Times . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + 10.1007/978-3-642-41714-6_151360 + + + Dictionary Geotechnical Engineering/Wörterbuch GeoTechnik + + Springer Berlin Heidelberg + + + + + Overcrossing Times . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 + + + + + SCHEDULING UNDER FINANCIAL CONSTRAINTS + + JCarlier + + Scheduling Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + 10.1016/b978-0-444-87358-3.50013-4 + + + Advances in Project Scheduling + + Elsevier + + + + + Scheduling Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 + + + + + Protein Sequence Constraints + + DanielThorLavelle + + Sequence Constraints. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + 10.18130/v3gr88 + + null + University of Virginia + + + Sequence Constraints. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + + + A queueing model in which arrival times are scheduled + + MakotoDoi + + C. Outputs from the Dynamic Planner . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 1. Scheduled Times of Arrival (STAs) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + + Yow-MowChen + + C. Outputs from the Dynamic Planner . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 1. Scheduled Times of Arrival (STAs) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + + HideoŌsawa + + C. Outputs from the Dynamic Planner . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 1. Scheduled Times of Arrival (STAs) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + 10.1016/s0167-6377(97)00049-7 + + + Operations Research Letters + Operations Research Letters + 0167-6377 + + 21 + 5 + + + Elsevier BV + + + C. Outputs from the Dynamic Planner . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 1. Scheduled Times of Arrival (STAs) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 + + + + + Predicting Arrival and Departure Runway Assignments with Machine Learning + + Runway Assignments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + 10.2514/6.2021-2400.vid + + + American Institute of Aeronautics and Astronautics (AIAA) + 5 + + + Runway Assignments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 + + + + + Hierarchical Reuse Daemon + + D. Other CTAS Components Required. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + 10.1007/0-306-47034-9_2 + . . . . . . . . 5 + + + The Fractal Structure of Data Reference + + Kluwer Academic Publishers + null + + + + D. Other CTAS Components Required. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 1. Weather Data Acquisition Daemon (WDAD) and Weather Data Processing Daemon (WDPD). . . . . . . . 5 + + + + + Face-ism and shape-ism at work + + ISM) . . . . . . . . . . . . . . . . . . + + 10.1057/9780230369764.0027 + + + The Talented Manager + + Palgrave Macmillan + null + 6 + + + Host Data Acquisition and Routing (HDAR) and Input Source Manager (ISM) . . . . . . . . . . . . . . . . . . . . 6 + + + + + CM law + + Communications Manager (CM) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + 10.1016/s1061-9259(96)80081-6 + + + The Case Manager + The Case Manager + 1061-9259 + + 7 + 3 + + + Elsevier BV + + + Communications Manager (CM) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 + + + + + CLUSTERGEN: a statistical parametric synthesizer using trajectory modeling + + AlanWBlack + + Trajectory Synthesizer (TS). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + 10.21437/interspeech.2006-488 + + + Interspeech 2006 + + ISCA + + 6 + + + Trajectory Synthesizer (TS). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 + + + + + A computer program integrating a multichannel analyzer with gamma analysis for the estimation of {sup 226} Ra concentration in soil samples + + JEWilson + + Route Analyzer (RA) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + 10.2172/10175123 + + + Office of Scientific and Technical Information (OSTI) + 6 + + + Route Analyzer (RA) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 + + + + + Graphical User Interfaces + + TGUI and PGUI) . . . . . . . . . . . . . . . . . . . . . . . . . . . + + 10.1007/springerreference_65011 + + null + Springer-Verlag + + + Timeline and Planview Graphical User Interfaces (TGUI and PGUI) . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + + + Battery Terminology + 10.4271/j1715/2_202108 + + null + SAE International + 7 + + + E. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 + + + + + Blind-blocked + + Blocked Slot. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + 10.1093/gao/9781884446054.article.t009222 + + + Oxford University Press + 7 + + + Blocked Slot. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 + + + + + Transforming PV installations toward dispatchable, schedulable energy solutions + + MesaScharf + + Schedulable Objects (SOs). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + 10.2172/1357461 + + + Office of Scientific and Technical Information (OSTI) + 7 + + + Schedulable Objects (SOs). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 + + + + + Acceptance rates into the main disability programme vary widely across countries + 10.1787/a73db096-en + + + Organisation for Economic Co-Operation and Development (OECD) + 7 + + + Acceptance Rate Interval . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 + + + + + Stream computing + + MikeHouston + + Stream Classes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + 10.1145/1401132.1401151 + + + ACM SIGGRAPH 2008 classes + + ACM + + 7 + + + Stream Classes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 + + + + + Super Hyper Dominating and Super Hyper Resolving on Neutrosophic Super Hyper Graphs and Their Directions in Game Theory and Neutrosophic Super Hyper Classes + + Super Stream Classes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + 10.33140/jmtcm.01.03.09 + + + Journal of Mathematical Techniques and Computational Mathematics + JMTCM + 2834-7706 + + 1 + 3 + 7 + + Opast Group LLC + + + Super Stream Classes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 + + + + + reference point + + Reference Point . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + 10.1007/springerreference_23214 + + null + Springer-Verlag + 7 + + + Reference Point . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 + + + + + Docking Time + + Yoon + + Transition Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + 10.2979/transition.115.116 + + + Transition + Transition + 0041-1191 + + 115 + 116 + + Indiana University Press + + + Transition Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 + + + + + Ground-simulation investigation of VTOL instrument flight rules airworthiness criteria + + JLebacqz + + Flight Rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + + BScott + + Flight Rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + 10.2514/6.1984-2105 + + + 11th Atmospheric Flight Mechanics Conference + + American Institute of Aeronautics and Astronautics + + 8 + + + Flight Rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 + + + + + Final Environmental Assessment (EA) for Modification of Airspace Units R-3008A/B/C from Visual Flight Rules (VFR) to VFR-Instrument Flight Rules (IFR) at Moody Air Force Base, Georgia + + U.S. Air Force Moody Afb United States + + Instrument Flight Rules (IFR) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + 10.21236/ad1008872 + + + Defense Technical Information Center + 8 + + + Instrument Flight Rules (IFR) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 + + + + + Final Environmental Assessment (EA) for Modification of Airspace Units R-3008A/B/C from Visual Flight Rules (VFR) to VFR-Instrument Flight Rules (IFR) at Moody Air Force Base, Georgia + + U.S. Air Force Moody Afb United States + + Visual Flight Rules (VFR). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + 10.21236/ad1008872 + + + Defense Technical Information Center + 8 + + + Visual Flight Rules (VFR). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 + + + + + Life Configurations + + CarlosRafaelRuta + + Configurations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + 10.1515/9783110338737.vii + + + Life Configurations + + DE GRUYTER + null + 8 + + + Configurations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 + + + + + Table 3: Proposed parameter sets. + + Flow Parameter Sets. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + 10.7717/peerj-cs.1391/table-3 + + null + PeerJ + 8 + + + Flow Parameter Sets. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 + + + + + Sta, Henri de + + STA Freeze . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + 10.1093/benz/9780199773787.article.b00174121 + + + Oxford University Press + 8 + + + STA Freeze . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 + + + + + The Process Sequence in Summary + + Sequence Freeze. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + 10.1039/9781847557704-00013 + + + Freeze-drying of Pharmaceuticals and Biopharmaceuticals + + The Royal Society of Chemistry + + 8 + + + + Sequence Freeze. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 + + + + + Project Scheduling with Multiple Activity Execution Modes + + Scheduling Modes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + 10.1007/0-306-48142-1_8 + + + International Series in Operations Research & Management Science + + Kluwer Academic Publishers + null + + + + Scheduling Modes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + + + What’s So Special about the Body? + + AnnePhillips + + F. Special SO Indicators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + 10.23943/princeton/9780691150864.003.0002 + + + Our Bodies, Whose Property? + + Princeton University Press + + 8 + + + F. Special SO Indicators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 + + + + + FIVE. He departed with thoughts of home, He departed with thoughts of home, He departed towards another place. -Honey-Ant Men's Song + + Departed Aircraft . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + 10.1515/9780822396123-006 + + + At Home in the World + + Duke University Press + + + + + Departed Aircraft . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 + + + + + Expired + + Expired Aircraft . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + 10.2307/j.ctt5vkfv2.29 + + + Blowout + + University of Pittsburgh Press + null + + + + Expired Aircraft . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 + + + + + Pop up Display | pop Up Trade Show Display |Pop up Booths and banners + + Pop-up Aircraft . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + 10.4016/18225.01 + + + SciVee, Inc + 10 + + + Pop-up Aircraft . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 + + + + + Small Business Innovation Research to Support Aging Aircraft + + Priority Aircraft . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + 10.17226/10092 + + + National Academies Press + 10 + + + Priority Aircraft . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 + + + + + The Landed Interest + + EricLJones + + Landed Aircraft . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + 10.1007/978-3-319-74869-6_1 + + + Landed Estates and Rural Inequality in English History + + Springer International Publishing + + + + + Landed Aircraft . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 + + + + + Papers Scheduled for Holiday Meetings + + Manually Scheduled SOs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + 10.1785/gssrl.10.3.10c + + + Seismological Research Letters + 0895-0695 + 1938-2057 + + 10 + 3 + + + Seismological Society of America (SSA) + + + Manually Scheduled SOs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 + + + + + SOS Ciudades Iquitos 2011: a cidade e as águas | SOS Ciudades Iquitos 2011: the city and water | SOS Ciudades Iquitos 2011: la ciudad y las aguas + + FábioMarizGonçalves + + Suspended SOs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + 10.24220/2318-0919v10n1a1924 + + + Oculum Ensaios + Oculum Ensaios + 1519-7727 + 2318-0919 + + 10 + 1 + 6 + + Cadernos de Fe e Cultura, Oculum Ensaios, Reflexao, Revista de Ciencias Medicas e Revista de Educacao da PUC-Campinas + + + Suspended SOs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 + + + + + Air Force's Combat Aircraft: A Future Holding into the Past + + GordonPGreaney + + Holding Aircraft. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + 10.21236/ada522950 + + + Defense Technical Information Center + 10 + + + Holding Aircraft. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 + + + + + Main body + + Mc + 0000-0001-9496-4852 + + G. The Main Body. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + 10.31219/osf.io/gn2jr + + + Center for Open Science + 10 + + + G. The Main Body. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 + + + + + Single-machine Sequencing + + H. Sequencing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + 10.1002/9781119262602.ch2 + + + Principles of Sequencing and Scheduling + + John Wiley & Sons, Inc. + + + + + H. Sequencing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 + + + + + Protein Sequence Constraints + + DanielThorLavelle + + Sequence Constraints. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + 10.18130/v3gr88 + + null + University of Virginia + 11 + + + Sequence Constraints. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 + + + + + Towards Sequence-to-Sequence Reinforcement Learning for Constraint Solving with Constraint-Based Local Search + + HelgeSpieker + + Sequence Constraint Deconfliction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 3. Building the Combined Sequence to the Meter Fixes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + 10.1609/aaai.v33i01.330110037 + + + Proceedings of the AAAI Conference on Artificial Intelligence + AAAI + 2159-5399 + 2374-3468 + + 33 + 01 + + + Association for the Advancement of Artificial Intelligence (AAAI) + + + Sequence Constraint Deconfliction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 3. Building the Combined Sequence to the Meter Fixes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 + + + + + Scheduling + + ShaharuddinSalleh + + I. Scheduling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + + AlbertYZomaya + + I. Scheduling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + 10.1007/978-1-4615-5065-5_1 + + + Scheduling in Parallel Computing Systems + + Springer US + + + + + I. Scheduling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 + + + + + JaCa-MM: A User-centric BDI Multiagent Communication Framework Applied for Negotiating and Scheduling Multi-participant Events - A Jason/Cartago Extension Framework for Diary Scheduling Events Permitting a Hybrid Combination of Multimodal Devices based on a Microservices Architecture + + JuanLuisLópez Herrera + + Scheduling Events . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + + HomeroVladimir RiosFigueroa + + Scheduling Events . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + 10.5220/0006751703180330 + + + Proceedings of the 10th International Conference on Agents and Artificial Intelligence + the 10th International Conference on Agents and Artificial Intelligence + + SCITEPRESS - Science and Technology Publications + + + + Scheduling Events . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + + + Task Order Number 5TS5702D035P: Testing Alternative Aircraft and Runway/Taxiway Deicers - Phase 2 + + SusanVan Scoyoc + + Order of Consideration at the Runway . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + 10.21236/ada432921 + + + Defense Technical Information Center + + + Order of Consideration at the Runway . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + + + Table 4. Activation and deactivation of the brain regions involved in feedback delay (delay – non-delay). + + Delay Feedback . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + 10.7554/elife.42265.013 + + null + eLife Sciences Publications, Ltd + + + Delay Feedback . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + + + PROCESS INDUSTRY SCHEDULING + + Scheduling Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + 10.1007/springerreference_7057 + + null + Springer-Verlag + + + Scheduling Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + + + Hardmetal insert tooling + + Insert without Rescheduling. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + 10.3403/00063475 + + null + BSI British Standards + + + Insert without Rescheduling. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + + + Hydrocodone: To reschedule, or not to reschedule? + + SharonKPark + + ETA and Insert . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + 10.1016/s1042-0991(15)31187-7 + + + Pharmacy Today + Pharmacy Today + 1042-0991 + + 19 + 9 + 54 + + Elsevier BV + + + Reschedule at ETA and Insert . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + + + Is it Possible to Treat Insert Dislocation With a Fixed Insert Tibial Component After Primary Oxford Phase 3 Mobile Unicompartmental Knee Arthroplasty? + + GökhanBSever + + Reschedule after Aircraft Ahead and Insert. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + 10.1097/bto.0000000000000449 + + + Techniques in Orthopaedics + 0885-9698 + + Publish Ahead of Print + + Ovid Technologies (Wolters Kluwer Health) + + + Reschedule after Aircraft Ahead and Insert. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + + + Runway schedule determination by simulation optimization + + TCHolden + + Schedule to Runway. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + + FWieland + + Schedule to Runway. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + 10.1109/wsc.2003.1261618 + + + Proceedings of the 2003 International Conference on Machine Learning and Cybernetics (IEEE Cat. No.03EX693) + the 2003 International Conference on Machine Learning and Cybernetics (IEEE Cat. No.03EX693) + + IEEE + null + + + Schedule to Runway. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + + + SCHEDULING UNDER FINANCIAL CONSTRAINTS + + JCarlier + + Scheduling Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + 10.1016/b978-0-444-87358-3.50013-4 + + + Advances in Project Scheduling + + Elsevier + + + + + Scheduling Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + + + Optimizing Propeller Separation Distance to Enhance Multi-rotor UAV Aerodynamics + + Separation Distance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + 10.2514/6.2023-4462.vid + + + American Institute of Aeronautics and Astronautics (AIAA) + + + Separation Distance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + + + Acceptance rates into the main disability programme vary widely across countries + 10.1787/a73db096-en + + + Organisation for Economic Co-Operation and Development (OECD) + + + Acceptance Rate. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + + + The Double Occupancy Problem + + NikkEffingham + + Occupancy Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + 10.1093/oso/9780198842507.003.0005 + + + Time Travel + + Oxford University Press + + + + + Occupancy Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + + + Figure 9: Tone accuracy and Pinyin accuracy of Picture Naming for LV (low variability), HV (high variability) and HVB (high variability blocked) training groups. Error bars show 95% confidence intervals. + + Blocked Intervals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + 10.7717/peerj.7191/fig-9 + + null + PeerJ + + + Blocked Intervals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + + + Merging Optimization Method with Runway Allocation Optimization maximizing Runway Capacity + + DaichiToratani + + J. Runway Allocation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + 10.2514/6.2019-0353 + + + AIAA Scitech 2019 Forum + + American Institute of Aeronautics and Astronautics + + + + J. Runway Allocation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + + + Merging Optimization Method with Runway Allocation Optimization maximizing Runway Capacity + + DaichiToratani + + Runway Allocation Events. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + 10.2514/6.2019-0353 + + + AIAA Scitech 2019 Forum + + American Institute of Aeronautics and Astronautics + + + + Runway Allocation Events. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + + + Merging Optimization Method with Runway Allocation Optimization maximizing Runway Capacity + + DaichiToratani + + Selecting SOs for Runway Allocation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + 10.2514/6.2019-0353 + + + AIAA Scitech 2019 Forum + + American Institute of Aeronautics and Astronautics + + + + Selecting SOs for Runway Allocation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + + + Merging Optimization Method with Runway Allocation Optimization maximizing Runway Capacity + + DaichiToratani + + Runway Allocation Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + 10.2514/6.2019-0353 + + + AIAA Scitech 2019 Forum + + American Institute of Aeronautics and Astronautics + + + + Runway Allocation Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + + + Determining the NAS Impact of Flights Affected by Miles-In-Trail (MIT) and other Traffic Manageme... + + KMiles + + -in-Trail Advisor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + 10.2514/6.2022-3907.vid + + + American Institute of Aeronautics and Astronautics (AIAA) + + + K. Miles-in-Trail Advisor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + + + Modal Decomposition Analysis of Hovering Rotor Wake Breakdown + + L. ETA Hovering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + 10.2514/6.2021-0737.vid + + + American Institute of Aeronautics and Astronautics (AIAA) + + + L. ETA Hovering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + + + Writing broadcast news scripts + + M. Broadcast . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + 10.4324/9780203342671_chapter_4 + + + Writing for Broadcast Journalists + + Taylor & Francis + null + 39 + + + M. Broadcast . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + + + Concluding Remarks - IV + + RenéHudec + + N. Design Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . III. CONCLUDING REMARKS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IV. REFERENCES. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + 10.22323/1.331.0092 + + + Proceedings of Frontier Research in Astrophysics – III — PoS(FRAPWS2018) + Frontier Research in Astrophysics – III — PoS(FRAPWS2018) + + Sissa Medialab + + + + N. Design Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . III. CONCLUDING REMARKS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IV. REFERENCES. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + + + + + + Iv + + + DGReferences 1. Denery + + + HErzberger + + The Center-TRACON Automation System: Simulation and Field Testing + Capri, Italy + + Oct. 2-6, 1995. Aug., 1995 + + + Proceedings of the advanced Workshop on ATM (ATM 95). Also published as NASA TM-110366 + IV. REFERENCES 1. Denery, D. G.; and Erzberger, H.: The Center-TRACON Automation System: Simulation and Field Testing. In: Proceedings of the advanced Workshop on ATM (ATM 95), Capri, Italy, Oct. 2-6, 1995. Also published as NASA TM-110366, Aug., 1995. + + + + + Heinz Erzberger of NASA Concerning the Center-TRACON Automation System (CTAS) + + HErzberger + + + + Statement of Dr + Washington, DC + + July 12, 1994 + + + Pesented at the Meeting of the Congressional Capacity Technology Subcommittee + Erzberger, H.:, Statement of Dr. Heinz Erzberger of NASA Concerning the Center-TRACON Automation System (CTAS). Pesented at the Meeting of the Congressional Capacity Technology Subcommittee, FAA Research and Development Advisory Committee, Washington, DC, July 12, 1994. + + + + + + BDSanford + + + KHarwood + + + SNowlin + + + HBergeron + + + HHeinrichs + + + GWells + + + MHart + + Center/TRACON Automation System: Development and Evaluation in the Field. In: 38th Annual Air Traffic Control Association Conference Proceedings + + October, 1993 + + + Sanford, B.D.; Harwood, K.; Nowlin, S.; Bergeron, H.; Heinrichs, H.; Wells, G.; and Hart, M.: Center/TRACON Automation System: Development and Evaluation in the Field. In: 38th Annual Air Traffic Control Association Conference Proceedings, October, 1993. + + + + + Design and Operational Evaluation of the Traffic Management Advisor at the Ft. Worth Air Route Traffic Control Center + + HNSwenson + + + THoang + + + SEngelland + + + DVincent + + + TSanders + + + BSanford + + + KHeere + + + + Proceedings of the 1st US/Europe ATM R&D Seminar + the 1st US/Europe ATM R&D SeminarSaclay, France + + June, 1997 + + + Swenson, H. N.; Hoang, T.; Engelland, S.; Vincent, D.; Sanders, T.; Sanford, B.; and Heere, K.: Design and Operational Evaluation of the Traffic Management Advisor at the Ft. Worth Air Route Traffic Control Center. In: Proceedings of the 1st US/Europe ATM R&D Seminar, Saclay, France, June, 1997. + + + + + Design Principles and Algorithms for Automated Air Traffic Management + + HErzberger + + + + AGARD Lecture Series + + 200 + November, 1995 + Madrid, Spain, Paris, France, and Moffett Field, California, USA + + + Erzberger, H.: Design Principles and Algorithms for Automated Air Traffic Management. In: AGARD Lecture Series 200, Madrid, Spain, Paris, France, and Moffett Field, California, USA, November, 1995. + + + + + The Traffic Management Advisor + + WilliamNedell + + + HeinzErzberger + + + FrankNeuman + + 10.23919/acc.1990.4790788 + + + 1990 American Control Conference + San Diego, CA + + IEEE + May, 1990 + + + Nedell, W.; and Erzberger, H.: The Traffic Management Advisor. Presented at the 1990 American Control Conference, San Diego, CA, May, 1990. + + + + + Descent Advisor preliminary field test + + StevenGreen + + + RobertVivona + + + BeverlySanford + + 10.2514/6.1995-3368 + AIAA-95-3368 + + + Guidance, Navigation, and Control Conference + + American Institute of Aeronautics and Astronautics + August, 1995 + + + Green, S. M.; Vivona, R.; and Sanford, B.: Descent Advisor Preliminary Field Test. AIAA-95-3368, August, 1995. + + + + + Operational Test Results of the Passive Final Approach Spacing Tool + + TJDavis + + + DRIsaacson + + + JERobinson + + + WDen Braven + + + KKLee + + + BSanford + + 10.1016/s1474-6670(17)43820-1 + + + IFAC Proceedings Volumes + IFAC Proceedings Volumes + 1474-6670 + + 30 + 8 + + June, 1997 + Elsevier BV + Chania, Greece + + + Davis, T. J.; Isaacson, D. R.; Robinson III, J. E.; den Braven, W.; Lee, K. K.; and Sanford, B.: Operational Test Results of the Passive Final Approach Spacing Tool. Presented at the IFAC 8th Symposium on Transportation Systems, Chania, Greece, June, 1997. + + + + + Knowledge-based scheduling of arrival aircraft in the terminal area + + KJKrzeczowski + + + ThomasDavis + + + HeinzErzberger + + + IsraelLev-Ram + + + ChristopherBergh + + 10.2514/6.1995-3366 + AIAA-95-3366 + + + Guidance, Navigation, and Control Conference + Baltimore, MD + + American Institute of Aeronautics and Astronautics + August, 1995 + + + Krzeczowski, K. J.; Davis, T. J.; Erzberger, H.; Lev-Ram, I.; and Bergh, C. P.: Knowledge-Based Scheduling of Arrival Aircraft in the Terminal Area. In: Proceedings of the AIAA Guidance, Navigation, and Control Conference, Paper No. AIAA-95-3366, Baltimore, MD, August, 1995. + + + + + THE FINAL APPROACH SPACING TOOL + + TJDavis + + + KJKrzeczowski + + + CPBergh + + 10.1016/b978-0-08-042238-1.50015-x + + + Automatic Control in Aerospace 1994 (Aerospace Control '94) + Palo Alto, CA + + Elsevier + September, 1994 + + + + Davis, T.J.; Krzeczowski, K. J.; and Bergh, C. P.: The Final Approach Spacing Tool. Presented at the IFAC Thirteenth Symposium on Automatic Control in Aerospace, Palo Alto, CA, September, 1994. + + + + + Conflict Probability Estimation for Free Flight + + RussellAPaielli + + + HeinzErzberger + + 10.2514/2.4081 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 20 + 3 + + May-June, 1997 + American Institute of Aeronautics and Astronautics (AIAA) + + + Also published as NASA TM-110411 + Paielli, R. A.; and Erzberger, H.: Conflict Probability Estimation for Free Flight. In: Journal of Guidance, Control, and Dynamics, Vol. 20, No. 3, May-June, 1997. Also published as NASA TM-110411. + + + + + Conflict Detection and Resolution In the Presence of Prediction Error. Presented at the 1st USA + + HErzberger + + + RAPaielli + + + DRIsaacson + + + MMEshow + + + + Europe Air Traffic Management R&D Seminar + + June, 1997 + Paris, France + + + Erzberger, H.; Paielli, R. A.; Isaacson, D. R.; and Eshow, M. M.: Conflict Detection and Resolution In the Presence of Prediction Error. Presented at the 1st USA/Europe Air Traffic Management R&D Seminar, Paris, France, June, 1997. + + + + + Performance of En Route Metering Methods + + BGSokkappa + + + JGSteinbacher + + MTR- 7575 + + June, 1977 + + + MITRE Technical Report + Sokkappa, B. G.; and Steinbacher, J. G.: Performance of En Route Metering Methods. MITRE Technical Report MTR- 7575, June, 1977. + + + + + Time Based Air Traffic Control in an Extended Terminal Area -A Survey of Such Systems + + EPetre + + + + Doc. 912009, Eurocontrol + Bruxelles, Belgium + + June, 1991 + + + Petre, E.: Time Based Air Traffic Control in an Extended Terminal Area -A Survey of Such Systems. Doc. 912009, Eurocontrol, Bruxelles, Belgium, June, 1991. + + + + + An implicit enumeration algorithm for arrival aircraft + + CRBrinton + + 10.1109/dasc.1992.282145 + + + [1992] Proceedings IEEE/AIAA 11th Digital Avionics Systems Conference + Seattle, WA + + IEEE + October, 1992 + + + Brinton, C. R.: An Implicit Enumeration Algorithm for Arrival Aircraft Scheduling. Presented at the Eleventh Digital Avionics Systems Conference, Seattle, WA, October, 1992. + + + + + Trajectory Synthesis for Air Traffic Automation + + RhondaSlattery + + + YiyuanZhao + + 10.2514/2.4056 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 20 + 2 + + March/April, 1997 + American Institute of Aeronautics and Astronautics (AIAA) + + + Slattery, R.; and Zhao, Y.: Trajectory Synthesis for Air Traffic Automation. In: Journal of Guidance, Control, and Dynamics, Vol. 20, No. 2, pp. 232-238, March/April, 1997. + + + + + Terminal area trajectory synthesis for air traffic control automation + + RASlattery + + 10.1109/acc.1995.520941 + + + Proceedings of 1995 American Control Conference - ACC'95 + 1995 American Control Conference - ACC'95 + + American Autom Control Council + June, 1995 + + + Slattery, R.A.: Terminal Area Trajectory Synthesis for Air Traffic Control Automation. Submitted to the American Control Conference, June, 1995. + + + + + + JRumbaugh + + + MBlaha + + + WPremerlani + + + FEddy + + + WLorensen + + Object-Oriented Modeling and Design + Englewood Cliffs, NJ + + Prentice Hall + 1991 + + + Rumbaugh, J.; Blaha, M.; Premerlani, W.; Eddy, F.; and Lorensen, W.: Object-Oriented Modeling and Design. Prentice Hall, Englewood Cliffs, NJ, 1991. + + + + + Fatigue life prediction of an intermetallic matrix composite at elevated temperaturesBartolotta, P.A. NASA Technical Memorandum Report No NASA TM-104494 1991 12 pp + + RGSynnestvedt + + + HNSwenson + + + HErzberger + + 10.1016/0142-1123(92)90375-m + + + International Journal of Fatigue + International Journal of Fatigue + 0142-1123 + + 14 + 3 + + September, 1995 + Elsevier BV + + + Synnestvedt, R. G.; Swenson, H. N.; and Erzberger, H.: Scheduling Logic for Miles-In-Trail Traffic Management. NASA Technical Memorandum TM-4700, September, 1995. + + + + + + + KHarwood + + + BSanford + + + + NASA Contractor Report + + 4554 + October, 1993 + Denver TMA Assessment + + + Harwood, K.; and Sanford, B.: Denver TMA Assessment. NASA Contractor Report 4554, October, 1993. + + + + + + diff --git a/file790.txt b/file790.txt new file mode 100644 index 0000000000000000000000000000000000000000..b64045e2a49a3bc1827658aead944b2cf079217a --- /dev/null +++ b/file790.txt @@ -0,0 +1,249 @@ + + + + +I. IntroductionUnder the NextGen 1 vision, a simulation tool to integrate various surface traffic planning algorithms is required in support of the SESO milestones.This simulation tool requires a modular architecture for scheduling algorithms and fast-time capabilities.There are existing airport simulators 2,3 that have been available to the air transportation community for some time, These simulators typically provide a high level of detail and were designed for analyzing airport capacities under current-day operations.To simulate current-day operations, a non-modular architecture 3 is often times adequate, although some simulators provide rule-based systems2 to compensate for potential changes in traffic controller operations.A rulebased system usually uses localized heuristics and does not have the capability of supporting optimization algorithms.An example of such an optimization algorithm would be formulating the taxi scheduling problem at an airport as a Mixed Integer Linear Program (MILP) and solving the MILP using a generic branch-andbound solver like CPLEX 6 or GLPK. 7Another example of an optimization algorithm would be finding the departure sequence for a given runway that optimizes delay 12 discussed in Section IV.With the development of surface management technologies envisioned by NextGen, SESO requires the capability to rapidly prototype various optimization routines to analyze potential benefits and shortcomings of applying the solutions produced.Rathinam et al 4 showed significant savings to taxi times while maintaining safety through a MILP formulation of a node-link model of the east side of DFW airport.While encompassing all aspects of surface operations into a single optimization model produces the best theoretical solution, the execution-time to solve such models can be prohibitive in application.Dividing the airport into smaller regions, or sub-regions, and decoupling runway management from the taxiway can allow optimization models to execute in fast-time.With such computational gains, application of various optimization models to the sub-regions of the airport could become feasible.Although fast-time capabilities can be achieved when applying optimization models to these sub-regions, the schedules produced still need to be integrated into an efficient solution for the entire airport.This paper describes the development of a fast-time simulation tool called the SOS 2 , a simulation platform with a plug-in architecture that can facilitate the integration of several scheduling algorithms with the surface simulation of the aircraft.In Section II the SOS 2 architecture is described.Using the SOS 2 architecture, implementation of localized schedulers into two different airport schedulers that will be focused on in later sections.The first scheduler described in Section III is Baseline-FCFS, which has pure FCFS heuristics and will be used as a benchmark for comparison of the schedulers currently under development in SESO; a comparison of average taxi times is made between Baseline-FCFS and actual data for calibration of the aircraft dynamics module implemented in SOS 2 .The second scheduler described in Section IV is Enhanced-FCFS, which has pure FCFS heuristics implemented for the taxiway and a dynamic programming module to optimize delay at the departure runways; a comparison is made between Enhanced-FCFS and Baseline-FCFS to demonstrate the modular framework of SOS 2 and improvements upon pure FCFS scheduling heuristics. +II. SOS 2 ArchitectureSOS 2 is a tool for the integration of optimization algorithms with aircraft simulation implemented in C++.The input requirements include a node-link model of the airport, aircraft model information, departure separation times, and runway configuration.Given a list of the aircraft to be scheduled, SOS 2 outputs time-based clearances for each node in aircraft routes.To calculate the time-based clearances, SOS 2 accepts additional input from scheduling algorithms.These algorithms could be local heuristics, such as FCFS described in Sections III and IV, or optimization algorithms such as those described in Sections III.A and IV.A.An alpha version of an OpenGL 10 Graphical User Interface (GUI) used for development and demonstration purposes is shown in Fig. 1 of a SOS 2 simulation at DFW airport; SOS 2 is independent of the GUI.The node-link model used is overlaid on a satellite image from GoogleEarth. 9The GUI uses color to display some aircraft information with unimpeded aircraft within 75% of the nominal speed in green, unimpeded aircraft less than 75% of the nominal speed in yellow, braking aircraft in orange, and non-moving aircraft in red.The size of the rectangle representing an aircraft corresponds to the length and wingspan of the aircraft.To ensure safety, a polygon is created around each node to represent an intersection on the taxiway.The dimensions of this intersection is derived from the maximum expected dimension of any aircraft scheduled in the airport, which is either the maximum expected wingspan or the maximum expected length of the aircraft, and the angle between adjacent links.By creating such a polygon to represent an intersection, any aircraft stopping short of this polygon can be assured that there is no chance of collision when another aircraft passes through the intersection regardless of aircraft model.Another safety mechanism of SOS 2 is a dynamic separation distance when an aircraft is taxiing in trail.This dynamic separation distance is derived from the desired minimum spacing of stationary aircraft on the taxiway, the stopping distance of the aircraft in trail, and an additional safety factor for pilot reaction time; the values for the minimum spacing of stationary aircraft and the pilot reaction time are variable inputs to SOS 2 .A plug-in interface for aircraft dynamics provides many options for modeling aircraft surface movements from simple models using instantaneous acceleration to very detailed models based upon throttle position.For the purposes of this paper an intermediate model has been adopted.Deceleration or braking is modeled using a negative acceleration that remains constant over a single time-frame; a maximum expected deceleration is given for runway exits and for aircraft on the taxiway.Thrust is modeled using a positive acceleration that remains constant over a single-time frame; a maximum expected acceleration for aircraft on the surface is given.Separate functions for aircraft braking and aircraft thrust are provided to SOS 2 .These functions for modeling aircraft dynamics provide a means to generate detailed speed profiles for each aircraft.For aircraft braking, the stopping distance is provided and the braking function calculates the necessary deceleration over the next time-frame.For aircraft acceleration there are two options.If no required time of arrival (RTA) at a node is given, the thrust function returns the maximum expected acceleration.Otherwise if an RTA at a node is given, the thrust function calculates the necessary acceleration to meet that RTA provided it does not exceed the maximum expected acceleration nor the maximum taxi speed of the aircraft.To create a generic framework for scheduling algorithms, SOS 2 hosts a scheduling interface to allow schedulers to input their solutions into SOS 2 control points.On the taxiway, each node is considered a single control point.Input into each control point is a sequence of aircraft passing through the control point ordered on the required time of arrival (RTA) at the control point.Although the aircraft RTAs at the control points is not required by SOS 2 , inputting the RTAs with the sequence allows SOS 2 to attempt meeting the RTA using the plug-in aircraft dynamics.If the RTAs are omitted, then SOS 2 attempts to move the aircraft at nominal speed using the thrust function described above.A runway manager has been implemented in SOS 2 to manage departure and arrival runways.An input requirement is a matrix of departure separation times due to wake vortices as well as separation times between departure and crossings aircraft to ensure safety.Since separation times must be maintained and aircraft can cross runways in parallel, the departure nodes representing departure points and the crossing nodes representing the divide between runway and taxiway must be linked in control.This link in control for a runway causes the sequence input to SOS 2 for these runway nodes to take the form of a single control point that can allow multiple crossings simultaneously.Management of arrival runways is handled slightly differently since ground controllers do not have control over arrival runways.Arrival aircraft are accepted as input to SOS 2 with a time window, a specific time, and the arrival runway.During the time window, SOS 2 has all crossing aircraft hold short of the arrival runway.At the specific time within the time window, the arrival aircraft is considered to have arrived at the taxi exit and is controlled by SOS 2 .For both arrival and departure runways, SOS 2 ensures enough open pavement is available on the taxiway on the opposite side of the crossing to ensure crossing aircraft do not stop on a runway.To integrate aircraft dynamics, scheduling, runway management, and safety parameters, SOS 2 uses a look-ahead distance in front of each aircraft.This look-ahead distance is based upon the maximum stopping distance of the aircraft derived from the maximum airport movement area speed.If a node appears in an aircraft's look-ahead distance, SOS 2 checks whether the aircraft is the next scheduled aircraft through that node.If an aircraft is maintaining proper separation distances and is the next scheduled aircraft through the nodes within the look-ahead distance, then SOS 2 attempts to drive the aircraft to meet the RTA or at the nominal taxiway speed based upon the model of the aircraft and provided aircraft dynamics.If the aircraft is in danger of violating a separation distance or is not the next scheduled aircraft at a node within the look-ahead distance, then SOS 2 applies braking acceleration according to the aircraft dynamics.These calculations are time driven and calculated in a timeframe-to-timeframe fashion similar to Euler's method. 11 +III. Baseline-FCFSA node-link model with 283 nodes of DFW airport in south-flow configuration using all runways and static routes based on the standard taxi routes was input to SOS 2 .Brinton et al 5 showed that almost 90% of aircraft on major taxi routes at DFW airport are handled using FCFS, but only 50% of aircraft are given spot clearances using FCFS.Because only 50% of the clearances at the spots are managed using FCFS, the times of the spot clearances in the actual data was input to SOS 2 as the planned aircraft schedule instead of inputting the times aircraft arrive at the spot.By using the times of spot clearances as input, nearly 100% of spot clearances for departure aircraft in the actual data are aligned with spot clearances in the SOS 2 simulations.For the purposes of these simulations, the controllers at DFW are assumed to use FCFS to sequence aircraft.The Surface Operations Data Analysis and Adaptation (SODAA) 8 tool is used to compile the actual data from DFW airport.Baseline-FCFS described below in Section III.A is used to provide a comparison of the SOS 2 simulations using FCFS heuristics with actual data.Approximate values for the weight classes including small, large, heavy, and B757 were input to SOS 2 as the aircraft models.An acceleration model is input with a constant time and speed values to represent the spool-up time of an aircraft below a speed threshold, where the spool-up time is the amount of time an aircraft waits before acceleration to be applied if it attempts to accelerate and the speed threshold is the point an aircraft's speed has slowed to the point rolling friction takes over and it now must wait for a spoolup time whenever attempting to accelerate.If an aircraft attempts to accelerate, a constant acceleration is applied to the aircraft for the next timeframe if the spool-up time is satisfied or the aircraft is not below the speed threshold, otherwise the aircraft must wait for the spool-up time to be satisfied before the acceleration is applied.The surface data at DFW airport is obtained from SODAA for May 24th and June 11th, 2008.These days were chosen due to clear weather, unrestricted visibility, and low delays.Clear weather was chosen since SOS 2 currently does not use velocity profiles for bad weather days.The same spot times and arrival times from SODAA data were input to SOS 2 as push-back times and runway exit times respectively.Any metering constraints imposed upon departure fixes during the scenarios are not considered since that information was not available in the actual data at the time the simulations were conducted. +III.A. Scheduling AlgorithmsA queue balancing heuristic operates prior to running SOS 2 .Given the aircraft input to SOS 2 for a specified runway, the departure queues are assigned using round-robin scheduling in the order of aircraft spot times.Aircraft on the taxiway are assumed to already have a queue assigned from a previous iteration of the queue balancing heuristic.FCFS on the taxiway is modeled locally at each taxiway node.A list is generated of all aircraft within the look-ahead distance of a node that is not taxiing in trail of another aircraft to the node.If the aircraft already exist in the sequence at the node, then FCFS removes those aircraft from the generated list.The aircraft remaining in the list are then sorted by the estimated time of arrival at the node and appended to the sequence already at the node.FCFS is applied to the departure runways with a similar procedure.All aircraft within the look-ahead distance of a departure node or crossing node are put in a list.If the aircraft are already in the departure runway sequence, then the aircraft are removed from the list.The minimum wait-time of the aircraft in the list is computed where the wait-time is the time which an aircraft requests to use the runway.If the aircraft with the minimum wait-time of the list is a departure aircraft, the aircraft will be appended to the departure runway sequence and removed from the list.If the aircraft with minimum wait-time is a crossing aircraft, the aircraft will be appended with other parallel crossing aircraft to the departure runway sequence and they will be removed from the list.This procedure is repeated until the list is empty. +III.B. ResultsA comparison is made between the actual data derived from SODAA and the Baseline-FCFS output.In Figs.2-5 a notable difference can be seen in average taxi times with arrival aircraft having significantly less taxi times than Baseline-FCFS and departure aircraft having significantly more taxi times than Baseline-FCFS.For the May 24th scenario there is a 10.2% difference on average arrival taxi times and 16.9% difference on average departure taxi times; for the June 11th scenario there is a 10.7% difference on average arrival taxi times and 18.6% difference on average departure taxi times.Some of the difference in arrival taxi times can be attributed to some arrival aircraft landing on departure runways 17R and 18L which have shorter taxi routes than arrival runways 17C and 18R, while all aircraft in Baseline-FCFS landed on 17C and 18R.Since metering constraints were not considered in these simulations, some of the difference in departure taxi times could be contributed to aircraft waiting in the departure queues to meet a miles-in-trail restriction that was not imposed in the Baseline-FCFS simulation.Although the usage of the center runways and metering constraints in the actual data likely contribute to the differences in average taxi times, it does not seem to contribute to all of the difference.A detailed analysis of the data revealed a significant number of arrival aircraft had negative delays and the majority of aircraft with negative delays were arrivals; negative delays indicates an aircraft travelling faster than the average unimpeded speed.The average unimpeded speed was derived from SODAA and determined to be 15 knots for all aircraft on the taxiway and 30 knots for arrival aircraft on a runway exit.The Baseline-FCFS simulation used the average unimpeded speed on the taxiway for both arrival and departure aircraft.These negative delays based upon the average unimpeded speed of aircraft implies that arrival aircraft are likely travelling at a higher average speed than departure aircraft.Indeed, SODAA data for the large weight class, which constitute the majority of the traffic and DFW airport, shows arrival aircraft travel on average at 18 knots on the taxiway and departure aircraft at 16 knots on the taxiway. +IV. Enhanced-FCFSTo demonstrate the ability to quickly integrate various schedulers within SOS 2 , Enhanced-FCFS was developed using the algorithms described below in Section IV.A.The same node-link model, dynamics module, aircraft routes, and scenarios of May 24th and June 11th are used as in Section III.A comparison of the results between Enhanced-FCFS and Baseline-FCFS is made below in Section IV.B.Since the input is identical to both simulations, any gains in delay of Enhanced-FCFS over Baseline-FCFS can be interpreted as potentially realizable savings in the asbsence of departure fixes and if departure and arrival aircraft had the same average speed. +IV.A. Scheduling AlgorithmsEnhanced-FCFS uses the FCFS heuristic on the taxiway and the queue balancing heuristic described in Section III.A.These heuristics are integrated with a dynamic programming module for scheduling departures as well as a crossing heuristic to insert crossings into departure sequences.For departure runway scheduling, the dynamic programming algorithm described by Rathinam et al 12 is implemented for the departure scheduling problem with chain-like queues.The unimpeded time to the departure runway for aircraft in the departure queue and on the taxiway are input to the DP module as the release times.Assignment to the departure queue is provided from the queue balancing heuristic.Since the DP module currently does not have the capability to schedule crossings, a crossing heuristic has been implemented to improve delay.For a specified departure runway, if all departure aircraft estimated times of arrival are more than the estimated time an aircraft takes to cross the runway, then a set of parallel crossings are inserted at the beginning of the runway control point sequence.However, if there are departure aircraft that are waiting or will arrive at the departure runway within the estimated crossing time then decisions branch based upon a few possible states.If the number of aircraft in any crossing queue is three or more, which is close to the capacity of the crossing queues at DFW, then a set of parallel crossings is inserted at the beginning of the sequence for the runway control point.If the count of aircraft in the individual crossing queues are not in danger of violating the capacity of the queues, then the total number of aircraft in the departure queue and the total number of aircraft in the crossings queues are counted.If the number of aircraft in the crossing queues exceeds the number of aircraft in the departure queue, then a set of parallel crossings is inserted at the beginning of the sequence for the runway control point.Otherwise, if no crossing queue is in jeopardy of violating capacity and the total number of aircraft in the crossing queues is not more than the number of aircraft in the departure queue, then a set of parallel crossings is inserted into the runway control point sequence at an estimated time when the total number of aircraft in the crossing queues exceeds the number of aircraft in the departure queue. +IV.B. ResultsIn Fig. 6 there is a large spike in the average taxi delays for arrivals around the ninth hour for Enhanced-FCFS.This spike in the data is a result of the behavior of the crossing heuristic implemented in Enhanced-FCFS.Although this departure push increased arrival delays around the ninth hour of the scenario, a noticeable improvement over Baseline-FCFS can be seen in Fig. 8 in the delay for departure aircraft around the ninth hour.The resulting decrease in delay of departure aircraft over multiple data points surrounding the ninth hour in Fig. 8 while only increasing arrival delays over two data points near the ninth hour in Fig. 6 indicates the crossing heuristic has some ability to balance when a departure push is necessary since arrival delay was temporarily increased to provide a better over system delay over that time period.Examining the average delay shown in Figs.6789, Enhanced-FCFS follows a similar profile to Baseline-FCFS, but does not appear to improve significantly upon the delay from Baseline-FCFS although Enhanced-FCFS consistently has a lower delay than Baseline-FCFS with the exception of a few data points.However, when calculating the saving in delay over the simulation, which was a time window of 7.7 hours for May 24th and 10 hours for June 11th, in both scenarios Enhanced-FCFS improved substantially over Baseline-FCFS on the total delay.For the May 24th scenario Enhanced-FCFS has a 3.37% improvement in total taxi times and a 33.1% improvement in total delay over Baseline-FCFS; for the June 11th scenario Enhanced-FCFS has a 2.34% improvement in total taxi times and a 20.0% improvement in total delay over Baseline-FCFS.The improvements were most notable in the delay of departure aircraft.For the May 24th and June 11th scenarios respectively, Enhanced-FCFS has a 27.7% improvement and 13.2% improvement for arrivals and a 42.0% improvement and 30.6% improvement in delay.The savings in delay and taxi times when using Enhanced-FCFS amounted to 13.8 seconds per aircraft for May 24th and 14.0 seconds per aircraft for June 11th, which is a total system savings of 9475 seconds and 6678 seconds respectively.Although these savings may seem very large, Baseline-FCFS uses pure FCFS behavior for the departure runways, most notably arrival crossings are not queued for departure pushes.When the departure runway is under constant pressure from arrivals and departures, pure FCFS behavior alternates 1-to-1 between giving a departure clearance for a departure aircraft and a crossing clearance for a set of arriving aircraft crossing the runway in parallel.Implementing a heuristic similar to the crossing heuristic implemented in Enhanced-FCFS or a heuristic similar to what traffic controllers use for current day operations could improve the Baseline-FCFS model and prevent overestimating savings as may be the case in the comparison between Enhanced-FCFS and Baseline-FCFS.Although such a heuristic implemented in Baseline-FCFS would reduce the improvement Enhanced-FCFS achieved over Baseline-FCFS, a significant savings in delay is still expected over with Enhanced-FCFS over Baseline-FCFS. +V. ConclusionsEnhanced-FCFS produced a significantly larger than expected savings over Baseline-FCFS, which may be due to the pure FCFS behavior used for scheduling arrivals crossing the departure runway.If the changes to better align the Baseline-FCFS data with actual data suggested in Section III.B are made and there is no significant difference in average delay or taxi times between Baseline-FCFS and actual data, then it can be assumed that the savings seen in delay with Enhanced-FCFS over Baseline-FCFS can be construed as potential savings to current day operation at DFW airport.If there is significantly a better average delay with the actual data over Baseline-FCFS once the changes suggested in Section III.B are made, then the case is likely that FCFS used as a crossing heuristic is not adequate to provide a model for current day operations.It is not certain whether a Baseline-FCFS implementation with FCFS as a crossings heuristic is a valid benchmark of current day operations since the dynamic programming algorithm used in Enhanced-FCFS was able to produce savings in delay up to 30.8% over FCFS during Monte-Carlo simulations 12 for a 3-queue problem, which is the same model for queue structure used for DFW airport, but was only able to produce a savings in delay of 6.7% averaged over all the simulations.Further refinement and testing of SOS 2 and Baseline-FCFS will help remove this uncertainty.The current dynamics module implemented in SOS 2 does not encompass the differing behavior between arrival and departure aircraft.However, the differences between Baseline-FCFS and the actual data for arrival aircraft as well as the differences between Baseline-FCFS and departure aircraft were consistent.These consistent differences implies implementing a new aircraft dynamics module, gathering actual data on metering for departure fixes, and allowing arrivals to use the departure runways during times of low activity at DFW airport will further refine Baseline-FCFS and should give a better alignment of Baseline-FCFS output with the actual data.Additional features are being implemented into SOS 2 for more extensive capabilities such as departure fix metering constraints, uncertainty, and additional airport configurations.Each of the main components of the simulation, such as departure scheduling and taxi scheduling, are being integrated with various optimization algorithms to test how various combinations of schedulers perform under varying conditions.Scenarios and metrics are currently being developed for these simulations.For initial testing, simulations will use a DFW airport model starting at a 1x traffic density and incrementally increase to a 2x traffic density, where 1x and 2x are 1 times and 2 times the current average traffic density at DFW airport respectively.Scenarios will include SOS 2 simulations using Baseline-FCFS compared to DFW airport data gathered from SODAA, similar to that described in Section III.Preliminary testing of Baseline-FCFS provided an initial comparison of the SOS 2 simulation with actual data and indicate the need for further refinement of the aircraft dynamics module to encompass the difference in speeds between arrival and departure aircraft.The integration of a dynamic programming module with FCFS and arrival crossing heuristics demonstrate the ability to use SOS 2 for rapid prototyping and testing combinations of surface schedulers.Additional algorithms will be implemented in varying combinations including generalized Dynamic Programming (DP) for departure scheduling, a MILP formulation solved with CPLEX for taxi scheduling, a MILP formulation solved with CPLEX for departure scheduling and queue balancing, taxi scheduling heuristics, and departure scheduling heuristics.The varying combinations of these algorithms will be compared against the Baseline-FCFS simulations as well as actual data from DFW airport through SODAA where applicable.Figure 1 .1Figure 1.A screen shot of the SOS 2 GUI showing the movements of the aircraft at DFW airport overlaid on a satelite image from Google Earth Mapping Service. +Figure 2 .2Figure 2. Comparison of average taxi times every 15 minutes for arrivals from the SOS 2 Baseline-FCFS simulation with the actual data at DFW airport for May 24th. +Figure 3 .3Figure 3.Comparison of average taxi times every 15 minutes for arrivals from the SOS 2 Baseline-FCFS simulation with the actual data at DFW airport for June 11th. +Figure 4 .Figure 5 .45Figure 4. Comparison of average taxi times every 15 minutes for departures from the SOS 2 Baseline-FCFS simulation with the actual data at DFW airport for May 24th. +Figure 6 .6Figure 6.Comparison of average taxi delay every 15 minutes for arrivals from the SOS 2 Enhanced-FCFS and Baseline-FCFS for the May 24th scenario. +Figure 7 .7Figure 7.Comparison of average taxi delay every 15 minutes for arrivals from the SOS 2 Enhanced-FCFS and Baseline-FCFS for the June 11th scenario. +Figure 8 .8Figure 8.Comparison of average taxi delay every 15 minutes for departures from the SOS 2 Enhanced-FCFS and Baseline-FCFS for the May 24th scenario. +Figure 9 .9Figure 9.Comparison of average taxi delay every 15 minutes for departures from the SOS 2 Enhanced-FCFS and Baseline-FCFS for the June 11th scenario. + of 11 American Institute of Aeronautics and Astronautics + of 11 American Institute of Aeronautics and Astronautics + of 11 American Institute of Aeronautics and Astronautics + + + + + + + + + NASA & The Next Generation Air Transportation System (NextGen) + + Nasa + + + + NASA + + 2008 + + + NASA, "NASA & The Next Generation Air Transportation System (NextGen)", NASA, 2008. + + + + + Experts in Aviation Analysis + + Atac + + 1/28/09 + + + ATAC, "Experts in Aviation Analysis," [www.atac.com/Products Research-a.html. Accessed 1/28/09.] + + + + + TAAM Plus, Fast-Time Gate-to-Gate Simulation, Reference Manual + + ThePreston + + + Group + + + + The Preston Group Pty Ltd + Richmond, Victoria 3121, Australia + + 1999 + + + The Preston Group, "TAAM Plus, Fast-Time Gate-to-Gate Simulation, Reference Manual," The Preston Group Pty Ltd., Richmond, Victoria 3121, Australia, 1999. + + + + + An Optimization Model for Reducing Aircraft Taxi Times at the Dallas Fort Worth International Airport + + SRathinam + + + JMontoya + + + JungY + + + + 26th International Congress of the Aeronautical Sciences + Anchorage, Alaska + + September 2008 + + + Rathinam S., Montoya J., and Jung Y., "An Optimization Model for Reducing Aircraft Taxi Times at the Dallas Fort Worth International Airport," 26th International Congress of the Aeronautical Sciences, Anchorage, Alaska, September 2008. + + + + + Microscopic Analysis of Airport Surface Sequencing + + MBrinton + + + BWood + + + EngellandS + + + + 26th International Congress of the Aeronautical Sciences + Anchorage, Alaska + + September 2008 + + + Brinton M., Wood B., and Engelland S., "Microscopic Analysis of Airport Surface Sequencing," 26th International Congress of the Aeronautical Sciences, Anchorage, Alaska, September 2008. + + + + + ILOG CPLEX 10.0 -Users Manual + + IncIlog + + + 2006 + + + ILOG, Inc., "ILOG CPLEX 10.0 -Users Manual," 2006. + + + + + Inventing the Foss Revolution: Hacker Crisis, GNU, and the Free Software Foundation + 10.7551/mitpress/10803.003.0005 + + + + For Fun and Profit + + The MIT Press + 2/2/09 + + + Free Software Foundation, Inc., "GLPK -GNU Project -Free Software Foundation (FSF)," [http://www.gnu.org/software/glpk/#introduction. Accessed 2/2/09.] + + + + + Surface Operations Data Analysis and Adaptation, Software Package, v.1.8 + + Sodaa + + + + Mosaic ATM + + 2008 + Leesburg, VA + + + SODAA, "Surface Operations Data Analysis and Adaptation, Software Package, v.1.8," Mosaic ATM, Leesburg, VA 2008. + + + + + Google Earth Mapping Service + + Google + + + May 2008 + + + Google, "Google Earth Mapping Service," [Accessed May 2008.] + + + + + OpenGL Reference Manual + + Sgi + + + + 2/3/09 + + + SGI, "OpenGL Reference Manual," [http://www.opengl.org. Accessed 2/3/09.] + + + + + Computer Methods for Ordinary Differential Equations and Differential-Algebraic Equations + + UriMAscher + + + LindaRPetzold + + 10.1137/1.9781611971392 + + 1998 + SIAM + + + Ascher, Uri M., Petzold, Linda Ruth. "Computer methods for ordinary differential equations and differential-algebraic equations," 1998. SIAM. + + + + + A Generalized Dynamic Programming Approach for a Departure Scheduling Problem + + SivakumarRathinam + + + ZacharyWood + + + BanavarSridhar + + + YoonJung + + 10.2514/6.2009-6250 + + + AIAA Guidance, Navigation, and Control Conference + Chicago, Illinois + + American Institute of Aeronautics and Astronautics + August 2009 + + + American Institute of Aeronautics and Astronautics -GNC + + + Rathinam S., Wood Z., Sridhar B., Jung Y., "A Generalized Dynamic Programming Approach for a Departure Scheduling Problem,", American Institute of Aeronautics and Astronautics -GNC, Chicago, Illinois, August 2009. + + + + + + diff --git a/file791.txt b/file791.txt new file mode 100644 index 0000000000000000000000000000000000000000..c54d86f2589ddc9c434ab6cb67435edf703e688c --- /dev/null +++ b/file791.txt @@ -0,0 +1,668 @@ + + + + +Nomenclature +I. IntroductionSuccessful development and validation of performance standards for Detect and Avoid (DAA) Systems serve as a crucial step for the integration of Unmanned Aircraft System (UAS) operations in the National Airspace System (NAS).A DAA system provides alerts and guidance to keep a UAS "Well Clear" of other aircraft. 1,2 ][10] The RTCA Special Committee 228 (SC-228) Working Group I recently finished its Phase I work at the end of 2016, and it is in the process of publishing the Minimum Operational Performance Standards (MOPS) for DAA systems.This Phase I MOPS targets UAS capable of carrying large and highpower sensor systems for operations in non-terminal areas.UAS in this category will be equipped with surveillance systems including Automatic Dependent Surveillance-Broadcast (ADS-B) in, airborne active surveillance, an air-to-air radar, as well as a DAA tracker that processes surveillance data.Phase II work for extending the MOPS to additional UAS categories and operations is underway.The DAA Well Clear (DWC) zone for the UAS targeted in the Phase I MOPS is defined by thresholds of three parameters.It does not have distinct physical boundaries because the definition depends on two aircraft's relative position and velocity during an encounter.Figure 1 zontal Miss Distance (HMD) represents the two aircraft's predicted minimum horizontal distance during an encounter assuming constant velocities (see Appendix A for a mathematical definition).The parameter h represents the two aircraft's current altitude difference.The time metric modified tau, τ mod , is an estimated time taken for the two aircraft to get horizontally close to each other (to be discussed in great detail below).The thresholds, denoted by an asterisk, for the HMD, h, and τ mod are 4000 ft, 450 ft, and 35 sec, respectively.All three parameters must simultaneously fall below their respective thresholds during an encounter for the two aircraft to violate the DWC.Alerting algorithms are designed to reduce the probability of violating DWC to a value required by the MOPS.The MOPS describes alerting requirements in terms of probabilities and test vectors without dictating a specific alerting algorithm.Prior research to inform the MOPS defined raw alerts using the same definitions as the DWC but with larger thresholds to provide buffers for surveillance uncertainties. 5he definition of τ mod is 2τ mod =    -r 2 -D mod 2 r ṙ , r > D mod , 0, r ≤ D mod (1)where r and ṙ are the horizontal range and range rate between the intruding aircraft (referred to as the intruder) and the UAS (referred to as the ownship), respectively.The range rate is negative for closing geometries.The positive incremental distance modifier D mod defines the radius of a "protection" disk around the ownship such that any intruder with a horizontal range less than D mod is always considered "urgent".In this case, τ mod = 0.The time metric τ mod , however, has certain limiting properties:• It does not correspond to the time of any physical event.• It is nonlinear with time, especially in the vicinity of the D mod disk (see Section III for an example).• During a multi-intruder encounter, τ mod cannot be directly used to prioritize intruders.The time metric τ mod has been used by the Traffic Alert and Collision Avoidance System (TCAS) II equipped by many manned aircraft. 11TCAS II is an optional component for the UAS targeted in Phase I MOPS.Nonetheless, interoperability between a DAA system and TCAS II has been a driving force in defining the Phase I MOPS.Ideally, a DWC zone should enclose TCAS's alerting zone for its resolution advisory.Because of this consideration, one argument for using τ mod for the DAA DWC is for potential better interoperability with TCAS II.The choice of τ mod , however, turns out to have minimal effects due to additional differences in TCAS II and DAA's alerting logic such as the following:1. TCAS II does not use the HMD consistently in its alerting logic; and 2. TCAS II's τ mod is based on 3-D range and range rates whereas the τ mod defined for a DAA system is based on horizontal range and range rate.Another argument for using τ mod for DAA systems is that the range-based τ mod avoids bearing measurements and its large measurement error.The bearing measurement errors are large in the active surveillance used by TCAS and, therefore, τ mod is a good choice for TCAS.However, for the DAA system, The bearing measurement error is much less of an issue since the DAA system considered by the MOPS will use ADS-B and an air-to-air radar, which both provide ten times more accurate bearing measurements than the active surveillance used by TCAS.This paper proposes a new, alternative horizontal time metric, the Time to Protected Zone, denoted by t pz .This metric has three advantages over τ mod : it corresponds to a physical event, it is linear with time, and it can be used directly in prioritizing intruders during a multi-intruder encounter.For alerting purposes, the protected zone is extended to be a function of surveillance sensor measurement errors, or surveillance errors.Since the time metric t pz utilizes bearing measurements and, therefore, it may be too sensitive to bearing measurement errors to be effective for alerting.To quantify sensitivities to surveillance errors, simulations of 972 encounters are conducted using two sensitivity metrics to compare τ mod and t pz .The parameters of the encounters are organized into an encounter test matrix.The surveillance sensor errors are modeled by Honeywell's sensor models and proprietary sensor fusion tracker. 12he paper is organized as follows.Section II supplies additional background information and describes the definition of t pz .Section III discusses issues about using τ mod that can be overcome by t pz .Section IV defines the sensitivity criteria for the time metrics.Section V describes the encounter test matrix and the simulation setup.Section VI presents results and discussions.Section VII concludes the paper. +II. Time to Protected Zone +II.A. BackgroundThe development of the MOPS for DAA systems uses a time metric to evaluate the horizontal separation between two aircraft.A fast-approaching intruder 8 nmi away may pose a more urgent threat than a slowapproaching intruder 3 nmi away.With a time metric threshold, the DWC does not have a fixed, physical volume; rather it is a function of the intruder's relative horizontal position and velocity.The time window considered by a DAA system covers up to 3 minutes before the aircraft reach their minimum horizontal separation.A commonly used time metric for pilots' situation awareness and intruder prioritization is the time to the Closest Point of Approach (CPA), denoted as t cpa .This time metric corresponds to a physical event and it is linear with time.However, t cpa underestimates the urgency of an encounter when the intruder and ownship are flying almost parallel trajectories at close range.The metric τ mod mitigates this problem by introducing a disk around the ownship with a radius of D mod .Any intruder within the D mod disk results in zero τ mod and it is always considered an urgent threat.However, τ mod does not correspond to any physical event.Also, τ mod is non-linear with time, especially near the D mod disk.With such behaviors, τ mod cannot be used to prioritize intruders by any criteria (See Section III).A time metric called the Time to Entry Point has been proposed as the predicted time for the intruder to reach the D mod disk. 13,14 ny intruder inside the D mod disk has a zero Time to Entry Point.This time metric maintains the advantages of both t cpa and τ mod and avoids their disadvantages.The Time to Entry Point is left undefined if the intruder is not predicted to enter the D mod disk, i.e., its HMD is greater than D mod . +II.B. DefinitionThis work proposes the Time to Protected Zone, t pz , which extends the Time to Entry Point metric in the following aspects:• The metric t pz is defined as the time to reach a protected around the ownship that does not necessarily take the shape of a disk.• When used for evaluating alerts, the dimensions of the protected zone can be made a function of the surveillance errors of each intruder.• The metric t pz is set to t cpa when the intruder is not predicted to enter the protected zone.This provides an alerting algorithm with a continuous definition of time metric when the projected intruder trajectory moves in and out of the protected zone in real time.Figure 2 demonstrates the definition of the protected zone at any time during an encounter.Consider a relative horizontal reference frame with two aircraft flying in a closing geometry.Without loss of generality, the ownship is placed at the origin of this reference frame.The axes, x and ŷ, are chosen such that the intruder's horizontal velocity vector relative to the ownship, ṙ, points along the negative direction of the ŷ axis.The intruder's relative horizontal position vector from the ownship is denoted by r.This convention of axes is called the collision plane. 15,16 he ownship's true ground heading is not used for choosing the direction of the axes.In this reference frame, the intruder's distance from the ŷ axis is always equal to the HMD.There is flexibility in the choice of the shape and dimensions of the protected zone.Its dimensions can be made to depend on individual intruders' surveillance errors for improved alerting performance.The example protected zone in Figure 2 consists of a disk with radius R 0 and an additional buffer zone shown in blue.The boundary of the protected zone is shown in bold.For brevity, only the right-half plane enclosing the intruder is shown in Figure 2 as the protected zone is symmetric about the ŷ axis.The additional buffer zone in blue is meant to accommodate HMD errors arising from surveillance errors.The metric t pz is defined as the time to reach the protected zone, or the time to reach the bold curve shown in Figure 2. If the intruder's HMD projects outside the protected zone, t pz is defined to be the time to reach the horizontal x axis, i.e., t pz = t cpa outside the protected zone.This definition ensures a smooth change of t pz upon change of an intruder's HMD.With this definition of a protected zone, it follows naturally that, when defining alert criteria, the HMD threshold, denoted as HMD * , should be consistently set to the edge of the protected zone.Let the surface of the protected zone be described by y = f (x).The function y is chosen such that it is symmetric with respect to x, i.e., f (x) = f (-x), where f (x) ≥ 0. Without loss of generality, consider a case in which the x component of the intruder's r is positive.By definition of the reference frame, x = HMD.Therefore, an intruder will be predicted to reach the protected zone at (HMD, y).Since y is the intersection of the the intruder's predicted trajectory with the protected zone, it is also the distance the intruder must fly to reach the CPA.Therefore,t pz = max 0, t cpa - y |ṙ| .(2)Note that t cpa -y | ṙ| < 0 if the intruder is already within the protected zone.The time metric t cpa can be viewed as a special case of t pz , for which the protected zone is defined to have zero area, i.e., y = 0 for every x. +II.C. Dimensions of the Protected ZoneIn this paper, the surface, or boundary, of the protected zone is defined with the following function:x = R 0 2 -y 2 + 1 - y R 0 ∆ H , 0 < y ≤ R 0 and |x| ≤ R 0 + ∆ H (3a) 0 = y, |x| > R 0 + ∆ H (3b)With this choice, the protected zone includes a R 0 disk and an additional buffer outside the disk that increases linearly as y approaches the x axis, reaching ∆ H at y = 0.This parameter ∆ H is user-selected and controls the size of the additional buffer zone.Solving for y in terms of x,y =      1 1+ ∆ H R 0 2 -∆ H R0 (x -∆ H ) 2 + -(x -∆ H ) 2 + R 0 2 + ∆ H 2 , when |x| ≤ R 0 + ∆ H , 0 , when x > R 0 + ∆ H(4) +II.D. Comparison of Time MetricsAll three time metrics described above, τ mod , t pz , and t cpa , are symmetric, 17 meaning their values are preserved upon switching the ownship and the intruder's states.Table 1 compares and contrasts the properties of the three time metrics. +III. Limitations of Modified TauFigure 3 shows the progression of τ mod , t pz , and t cpa during a hypothetical encounter.In this encounter, two aircraft fly at constant velocities with a relative speed of 450 kts (closing) and an HMD of 2000 ft.The The Phase I MOPS defines the DWC using a τ mod = 35 sec.An issue with using τ mod for the DWC is that the non-linearity of τ mod adds complexity to an alerting algorithm.For example, very often an alerting algorithm would use the predicted time to a violation of the DWC.Consider a co-altitude, head-on encounter in which two aircraft have zero altitude difference and zero HMD.Suppose the two aircraft's current τ mod is 70 sec, the predicted time for the two aircraft to violate the DWC, however, is not 35 sec into the future.In fact, it must be computed by solving a quadratic equation using τ mod = 35 sec in Eq. 1.Using t pz would avoid this complexity.If the DWC uses a t pz * = 35 sec, then current value of 70 sec t pz would mean a predicted loss of DWC 35 seconds into the future.For alerting, τ mod also has limitations.Since t cpa and t pz correspond to distinct physical events, they can be used directly for prioritizing intruders.The time metric τ mod , on the other hand, cannot be used for this purpose.Figure 4 illustrates a hypothetical encounter involving three non-accelerating intruders with varying relative horizontal speeds and initial values of t cpa .All three intruders have HMD = 0.Here v ≡ |ṙ| denotes the magnitude of an intruder's relative velocity vector.The intruders' initial ranges are v × t cpa .Therefore, intruders (labeled as Intr in the figure) 1, 2, and 3's initial ranges are 1.18, 1.78, and 10.42 nmi, respectively.Without loss of generality, the encounter begins at t = 0 sec.Recall that a higher time metric value indicates a lower threat.Intruder 1 (Intr 1) has the smallest relative speed v = 50 kts and the largest intial t cpa , t cpa = 85 sec.Intruder 1 is predicted by t cpa to be the least threat.On the other hand, τ mod predicts intruder 1 to be the highest threat.One may argue that the time to the D mod disk serves as a better time metric than t cpa for prioritizing intruders.To test this, consider a protected zone defined by a disk with a radius of D mod with no additional buffer zone. Figure 5 illustrates another hypothetical encounter involving three non-accelerating intruders (different from the previous encounter).All three intruders have HMD = 0. Intruders (labeled as Intr in the figure) 4, 5, and 6's initial ranges are 1.04, 2.67, and 11.81 nmi, respectively.Compared to t pz , τ mod predicts the opposite threat level priorities at the start of the encounter, or time zero.Intruder 4 (Intr 4) has the smallest t pz yet has the largest τ mod 104 sec.Moreover, in this encounter, the intruders' threat level priorities change with time as the τ mod curves intersect.This may be undesirable as it may cause a DAA system's directive guidance to reverse direction during a DAA maneuver. +IV. Sensitivity to Surveillance ErrorsAlthough t pz has advantages over τ mod , its explicit dependency on aircraft velocities makes it likely to be more sensitive to surveillance errors.Fluctuating values of a time metric around the DAA system's alert threshold are undesirable as it may cause the alert type to vary back and forth, posing challenges in providing a stable, consistent alert display to UAS pilots.Inaccurate values of a time metric may advance or delay the onset of an alert and impact the performance of the DAA system.Two metrics are defined in this work to measure the sensitivities of t pz and τ mod to surveillance errors:• Probability of reversal of a time metric during the progression of an non-accelerating encounter, P r • Average absolute error of a time metric as a result of surveillance errors, |∆| avg Both metrics are equal to zero in the absence of surveillance errors and increase with the magnitude of surveillance errors. +V. Simulations of EncountersThe sensitivities of the time metrics t pz and τ mod to the DAA system's surveillance sensor errors are compared by analyzing a set of simulated encounters with varying geometry, speeds, and surveillance equipage.The metric t cpa is not considered here due to its lack of a protected zone.The DAA simulation system uses true trajectories of the intruder and ownship to compute the true relative trajectories between the aircraft, the simulated surveillance sensor measurements, and the intruder track statistics estimated from the simulated surveillance sensor measurements.The sensor errors are simulated using Honeywell's high fidelity surveillance sensor models and the intruders' tracks are estimated using Honeywell's DAA tracking system.The Honeywell DAA tracker is a sub-TRL6 (Technology readiness level 6) tracker that is in its own iterative development cycle.This is one instantiation of the tracker with expected improvements in later versions to meet the developing DAA requirements and help with better alerting and guidance performance.The data are then collected and used to compute the time metrics and perform subsequent time metric sensitivity analysis.A total of 972 encounters are simulated; the parameters are summarized in the encounter test matrix (see Table 4 in Section VI). +V.A. DAA Simulation SystemFigure 6 shows a block diagram of the DAA simulation system.A trajectory generator is used to create the intruder, ownship, and relative intruder trajectories corresponding to the encounters defined in the encounter test matrix.The ownship trajectory is sent to the Ownship Navigation System block where the trajectory is used to provide outputs corresponding to an onboard navigation system.The intruder, ownship, and relative intruder trajectories are sent to the Surveillance Sensor Models block where sensor models are used to simulate ADS-B, Mode-S, Mode-C, and air-to-air radar measurements.The surveillance sensor measurements are sent to the Honeywell Tracking System (HTS).The HTS resolves the surveillance sensor measurements into a common reference frame, uses the H-Fuze system to estimate the track statistics of the intruder relative to the ownship, and computes additional track kinematic statistics required for DAA systems.Finally, the intruder, ownship, and relative intruder trajectories, the ownship navigation solution, and the estimated intruder track statistics are sent to the Time Metric Analysis block for time metric sensitivity analysis.The HTS is a multi-intruder aircraft, multi-sensor fusion system that estimates the track statistics of intruders relative to the ownship. 12The HTS fuses the measurements and statistical information from surveillance sensors in one framework to track intruder aircraft in three dimensions.The HTS features the H-Fuze system that performs Data Association, Track Management, and State (or Track) Estimation of surveillance sensors including ADS-B, TCAS, air-to-air radar, optical systems, and ground based radar.In this encounter trade study, the HTS uses a set of surveillance sensors that includes ADS-B, active surveillance of Mode-S and Mode-C, and air-to-air radar to track intruders.In general, ADS-B sensors provide accurate latitude and longitude measurements and accurate North-South and East-West velocity measurements.Active surveillance sensors provide accurate range measurements, noisy bearing measurements and no velocity measurements.Air-to-air radar provides range measurements with comparable accuracy to active surveillance range measurements and bearing measurements more accurate than active surveillance bearing measurements.Furthermore, air-to-air radar provides North-South and East-West velocity measurements less accurate than ADS-B.Table 2 summarizes the key surveillance sensor parameters used in the encounter trade studies.Note that these parameters were tuned and verified using flight test data. 18,19 he sensor's fields of view are specified using maximum range (R), bearing (B), and elevation (E).The bearing is defined with respect the line pointed to by the ownship's nose (right is positive).The sensors position errors are specified using latitude, longitude, and altitude for ADS-B; range and bearing for Mode-S and Mode-C; and range, bearing, and elevation for air-to-air radar.The sensor's velocity errors are specified using North (N), East (E), and A related consideration is the HMD threshold.For an alerting algorithm that uses an HMD threshold, it seems logical to define an HMD threshold at the edge of the protected zone at y = 0.If the protected zone incorporates the HMD in the buffer zone and the HMD threshold is defined appropriately, then the alerting algorithm can provide consistent raw alerts during the progression of an encounter that would lead to a loss of DWC.The choice of the protected zone, as well as the HMD threshold, plays an important factor in a trade-off between missed/late alerts and false alerts.A conceptually attractive approach is to allow a dynamic protected zone that is a function of the aircraft's estimated tracks as well as surveillance errors.However, in this approach, t pz would no longer be linear with time, and the dimensions of the buffer zone would be a function of encounter geometry and additional sensor parameters.Therefore, estimated tracks are not considered here for defining the protected zone.This work defines the protected zone such that most of the HMD fluctuations due to surveillance errors during an encounter that would lead to loss of DWC fall within the protected zone.To do so, the HMD error for an intruder is estimated at a characteristic range, R c , from the ownship, with a characteristic intruder closure rate, v c .Honeywell sensor model parameters are used for estimating the HMD error.The protected zone is defined by choosing ∆ H to be the estimated HMD error.For ADS-B, the HMD error arising from range measurement error is independent of range (i.e., a constant error).The HMD error arising from velocity measurement error is proportional to range.σ H, ADS-B 2 ≈ σ p. ADS-B 2 + R c σ v, ADS-B v c 2 ,(5)where σ p. ADS-B is the position measurement standand deviation and σ v, ADS-B is the velocity measurement standard deviation.For active surveillance, Mode-S or Mode-C,σ H, MODE-S/C 2 ≈ (R c σ b, MODE-S/C ) 2 ,(6)where σ b, MODE-S/C is the bearing standard deviation of the Mode-S or Mode-C measurement.For radar,σ H, RADAR 2 ≈ R c 2 * σ b, RADAR 2 + σ v, RADAR 2 v c 2 ,(7)where σ b, RADAR and σ v, RADAR are the bearing and speed standard deviations of the radar measurement, respectively.The final value of ∆ H 0 , the superscript denoting a benchmark value, is a function of the combination of sensors, 1∆ H 0 2 = Σ 1 σ H, sensor 2(8)Table 3 lists the sensor error parameters used for estimating the HMD error and the resulting buffer zone size parameters, ∆ H 0 .Note that the values in parentheses for ∆ H 0 are computed from Eq. 8, and the values before the parentheses are the values actually used in the simulations. +V.C. Encounter Test MatrixThe encounter test matrix for the set of encounters is designed to sample various surveillance equipages, relative speeds, HMDs, relative altitudes, and relative headings.The encounters are intended to cover a wide range of representive parameters but not to model a distribution of encounters.The methodology for selecting representative intruder equipage types is as follows.The air-to-air radar detects all intruding aircraft as they enter the radar's field of view, subject to the radar's probability of detection.Some intruders are equipped with ADS-B out and are, therefore, detectable by the ownship's ADS-B sensor.The FAA has mandated that, by 2020, all all aircraft operating above 10,000 ft mean sea level (MSL) must be equipped with ADS-B out. 20Many of the ADS-B out equipped aircraft will also be equipped with a Mode S transponder.Aircraft not equipped with ADS-B may have a Mode S transponder, a Mode C transponder, or no transponder at all.In this work, intruders are categorized by their equipage types in the following way:• ADS-B out: intruder is detectable by ADS-B and radar.• Mode S: intruder is detectable by Mode S and radar.• No equipage: intruder is detectable by radar.Table 4 lists the independent parameters of an encounter defined by the test matrix.The relative speed, altitude, and heading are with respect to the ownship.A relative heading of zero degrees represents a headon encounter in which the two aircraft are flying towards each other.Passing in front (behind) means the intruder passes in front of (behind) the ownship.The test matrix defines a total of 972 encounters.For each encounter, the two time metrics, τ mod and t pz , are computed as follows:• Kinematics: both the true and estimated ownship navigation solutions and intruder tracks; and• Protected Zone: using D mod disk and protected zone parameterized by a scaling factor, ∆ H /∆ H 0 , that scales the benchmark HMD error ∆ H 0 by 0, 1, 2, and 3. Note that D mod = R 0 + ∆ H and R 0 = 4000 ft.The reason that D mod is set to R 0 + ∆ H is because the HMD threshold should be the same as D mod to avoid oscillating alerts. 21 +V.D. Encounter SetupFor each simulated encounter, the ownship is fixed at a specific position and the intruder's initial kinematics are constructed according to the encounter test matrix parameters shown in Table 4, with an initial t cpa = 4 minutes.The intruder trajectories all fly a constant velocity.For analysis of each encounter, both the estimated and true time metrics are computed for tracks with t cpa < 2 minutes.The two minute time window is appropriate for DAA alerts.The time metrics t pz and τ mod are computed for every intruder track estimated by the tracker and the ownship navigation solution with navigation noise.The true time metrics, t pz 0 and τ mod 0 , are computed using the true trajectories.The probability of time reversal, P r , is calculated from Eq. 9:P r = occurrences of increasing time metric in consecutive tracks total number of consecutive tracks (9) The average absolute error of a time metric is calculated by averaging the absolute difference between a computed time metric (from the noisy tracks) and its true value (from the true tracks) over the two minute window of the encounter:|∆| avg = Σ|t pz -t pz 0 | or Σ|τ mod -τ mod 0 | total number of tracks ,(10)where the superscript 0 denotes a time metric computed from the true tracks.An additional metric, P (HMD ≤ HMD * ), calculates the percentage of noisy tracks that result in an HMD within the HMD threshold, HMD * = R 0 + ∆ H .If the protected zone is wide enough to accommodate HMD errors, this metric should be close to 1. +VI. Simulation ResultsFigure 7 shows results aggregated over all encounters of P r , |∆| avg , and P (HMD ≤ HMD * ) at various degrees of extended protected zone, parameterized by the scaling factor ∆ H /∆ H 0 .The general trend is that both P r and |∆| avg decrease as ∆ H increases.The P r and |∆| avg differences between t pz and τ mod are greatest at ∆ H /∆ H 0 = 0, i.e., no additional buffer zone.Values of P r are higher for t pz than for τ mod by 14%(∆ H /∆ H 0 = 1)-50%(∆ H /∆ H 0 = 0).Values of |∆| avg are higher for t pz than for τ mod by 5%(∆ H /∆ H 0 = 1)-33%(∆ H /∆ H 0 = 0).While the performance of t pz is slightly worse than τ mod , the differences may not be significant enough to degrade the DAA system's alerting performance, especially when ∆ H /∆ H 0 = 1.In the following paragraphs, only results for ∆ H /∆ H 0 = 1 are discussed in more detail.The two sensitivity metrics are highly positively correlated as both increase with the standard deviation of the surveillance errors.Figure 8 shows values of the two metrics for all 972 encounters.The same trends are observed for both metrics in various categories of encounters.Therefore, the following discussion will focus on P r only.Figure 9 shows the probability of reversal, P r , categorized by an intruder's relative speed.For both t pz and τ mod , P r decreases noticeably as the relative speed increases.This trend can be explained as follows.The error of t pz and τ mod can be roughly estimated byδt ≈ -t σ r r + σ ṙ ṙ ,(11)where σ r and σ ṙ represent standard deviations of r and ṙ, respectively, corresponding to surveillance errors.All four sensors provide very accurate range measurements, and the range error therefore contributes much less than the speed error to the time error.The standard deviation of the speed error in both ADS-B and radar models is not a function of velocity.Therefore, the speed error term on the right side of Eq. 11 is inversely proportional to the magnitude of ṙ, and therefore is largest when the relative speed itself is small.A secondary contribution to this trend, although to a lesser extent, is the effect of the sensor's field of view.In high relative speed encounters, the intruder does not enter the sensor's field of view at t cpa = 120 seconds but at later times in the encounter.Since the time metric error decreases as the intruder nears the ownship, the later measurement availability lowers the average time metric error.Comparing t pz results to τ mod results in Figure 9, t pz results in slightly larger median values (the line within the box) at every relative speed.Figure 10 shows the probability of reversal, P r , categorized by intruder equipage.Comparing t pz results across categories, intruders with ADS-B out result in the smallest median value of P r and intruders without any equipage result in the largest median value, approximately 30% greater than the smallest median value.The metric τ mod results display the same trend as the t pz results.Comparing t pz results to τ mod results, t pz leads to a slightly larger median value in every equipage category.Figure 11 shows the probability of reversal categorized by planned HMDs.The probability P r (t pz ) increases with HMD, reaching a larger median value and broader distribution at HMD = 4000 ft.This is because the surface of the Protected Zone becomes "steeper" near 4000 ft (see Figure 2) and causes t pz to be more sensitive to HMD errors.On the other hand, P r distribtions for τ mod are fairly stable across the HMD categories.The value of P r shows little variation with relative altitude, relative heading, and the passing direction. +VII. ConclusionsA new time metric called Time to Protected Zone is proposed for use in UAS's Detect and Avoid (DAA) systems.This time metric, denoted as t pz , has three advantages over the currently adopted modified tau, or τ mod .It corresponds to a well-defined physical event, it is linear with the real time during the progression of an encounter, and it can be used directly to prioritize intruders.When used for defining a Well Clear, t pz addresses several limitations of τ mod .For alerting, t pz can be used directly for intruder prioritization while τ mod is limited by its lack of physical interpretation.For alerting, the protected zone can be defined to be a function of surveillance errors for detecting intruding aircraft to provide potentially better alerting performance.To quantify the sensitivity of t pz to surveillance errors, simulations of encounters using realistic sensor and tracker models are performed.Results show that, with adequately selected protected zones, the sensitivity of t pz to surveillance errors is comparable to that of τ mod .The slight increase of time reversal rates and average time errors by using t pz is likely not enough to impact the alerting performance of a DAA system.Therefore, the choice of t pz over τ mod has advantages and no obvious downside, at least not at a fundamental level.Nonetheless, additional simulations such as MIT Lincoln Lab's encounter models 22 or NASA's NAS-wide simultions 23 that capture a wider variety as well as the statistical nature of encounters must be performed in order to assess the overall performance of t pz .Figure 1 .1Figure 1.A schematic representation of the DWC zone. +Figure 2 .2Figure 2.An example protected zone consisting of a disk with a radius R0 and an additional buffer zone colored in blue. +Figure 3 .3Figure 3.The three time metrics for a head-on encounter. +Intr 1 :1t cpa (t = 0) = 85 sec, v = 50 kts Intr 2: t cpa (t = 0) = 80 sec, v = 80 kts Intr 3: t cpa (t = 0) = 75 sec, v = 500 kts +Figure 4 .4Figure 4. τ mod ranks intruder threats in reverse order as compared to tcpa in a multi-intruder encounter. +Intr 4 :4t pz (t = 0) = 75 sec, v = 50 kts Intr 5: t pz (t = 0) = 80 sec, v = 120 kts Intr 6: t pz (t = 0) = 85 sec, v = 500 kts +Figure 5 .5Figure 5. τ mod ranks intruder threat in reverse order as compared to tpz in a multi-intruder encounter. +PPFigure 8 .8Figure 8. Correlation between Pr (vertical) and |∆|avg (horizontal, seconds). +Figure 9 .9Figure 9. Values of Pr for each relative speed +PFigure 10 .10Figure 10.Values of Pr for each intruder equipage category. +PFigure 11 .11Figure 11.Values of Pr for each value of planned HMD. + +Table 1 .1Comparison of the three time metricsTime Metric Physical Event Linear with Time Protecting against Close Intruders mod corresponding to τ mod and R 0 corresponding to t pz are both set to 4000 ft.No additional buffer zone is allocated for the protected zone corresponding to t pz , i.e., ∆ H = 0.At the beginning of the encounter, τ mod is close to t cpa .The metric τ mod drops more rapidly than time towards t pz as the two aircraft approach the CPA.When R 0 = D mod , Apendix A shows that t pz ≤ {t cpa , τ mod } at any predicted time during the encounter.This is a nice property as any zone defined by a t pz threshold in this condition would enclose the corresponding zone defined by a τ mod or t cpa threshold of the same value.t pzYesYesYesτ modNoNoYest cpaYesYesNoD +Table 3 .3Buffer zone parameters and the sensor error parameters used for estimationParameterValue UnitR c5 nmiv c100 ktsσ p, ADS-B2 mσ v, ADS-B2 m/sσ b, MODE-S/C9 degσ b, RADAR0.4 degσ v, RADAR4 m/sσ H, ADSB960 ftσ H, MODE-S/C3880 ftσ H, RADAR1930 ft∆ H, ADS-B0900(860) ft∆ H, MODE-S/C01700(1730) ft∆ H, RADAR01900(1930) ft• Mode C: intruder is detectable by Mode C and radar. +Table 4 .4Parameters of the test matrixParameterValueIntruder EquipageADS-B, Mode S, Mode C, NoneRelative Speed (kts)100, 300, 500HMD (ft)0, 1000, 2000, 3000, 4000Relative Altitude (ft)-500, 0, 500Relative Heading (deg) 0, 45, 90Passingin front, behind (if HMD = 0) + of 17 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on June 9, 2017 | http://arc.aiaa.org| DOI: 10.2514/6.2017-4383 + of 17 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on June 9, 2017 | http://arc.aiaa.org| DOI: 10.2514/6.2017-4383 + of 17 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on June 9, 2017 | http://arc.aiaa.org| DOI: 10.2514/6.2017-4383 + of 17 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on June 9, 2017 | http://arc.aiaa.org| DOI: 10.2514/6.2017-4383 + of 17 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on June 9, 2017 | http://arc.aiaa.org| DOI: 10.2514/6.2017-4383 + of 17 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on June 9, 2017 | http://arc.aiaa.org| DOI: 10.2514/6.2017-4383 + of 17 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on June + 9, 2017 | http://arc.aiaa.org| DOI: 10.2514/6.2017-4383 + Downloaded by NASA AMES RESEARCH CENTER on June 9, 2017 | http://arc.aiaa.org| DOI: 10.2514/6.2017-4383 + of 17 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on June 9, 2017 | http://arc.aiaa.org| DOI: 10.2514/6.2017-4383 + American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on June 9, 2017 | http://arc.aiaa.org| DOI: 10.2514/6.2017-4383 + + + +V.B. Surveillance-Dependent Protected ZoneFor DAA alerting criteria, it is desirable to augment the disk-shaped protected zone with a buffer zone.This additional buffer compensates for surveillance errors and reduces fluctuation of alerts.Surveillance errors vary with the sensor type as well as the intruder's relative range and velocity from the ownship.Therefore, the dimensions of the additional buffer zone can potentially be a function of not just the sensor type, but also of other parameters such as the intruder's range, speed, and time to CPA.a when elevation is within ±10 • b when elevation is outside ±10 • c when bearing is within ±45 • d when bearing is between ±45 • and ±135 • e when bearing is outside ±135 • f when elevation is within ±10 The scalar horizontal range r and range rate ṙ are defined asFor a closing geometry, ṙ < 0 and r • ṙ < 0. Let r, ṙ, r and ṙ represent the values at t = 0, or the beginning of an encounter.The time metric t cpa is the time at which the predicted r, using constant-velocity trajectories, reaches a minimum.Therefore, (r + t cpa ṙ) • ṙ = 0, and (A.3a)If ṙ ≤ 0, t cpa can be conveniently defined to be 0. The HMD is, by definition, the horizontal distance between the intruder and the ownship at t cpa .HMD ≡ |r + t cpa ṙ| (A.4)Taking the square of both sides of Eq.A.4 and substituting Eq.A.3 for t cpa :Eq. A.5 can be rearranged to derive an expression for r • ṙ, valid when r ≥ HMD,where the negative sign indicates a closing geometry.The time metric t pz is defined as the predicted time for the intruder to reach the protected zone.If the predicted trajectory does not intercept the protected zone, then t pz = t cpa .By Eq. ( 2), it is clear that t pz ≤ t cpa .The following paragraphs prove that,The additional buffer zone decreases t pz because the resulting protected zone has dimensions larger than the disk of radius R 0 .Therefore, it suffices to prove that Eq.A.7 holds when the protected zone is a disk.When r ≤ D mod , both t pz and τ mod are zero and, therefore, Eq.A.7 holds because the intruder is already in the protected zone.When r > D mod , rewrite τ mod as a distance divided by the relative speed, |ṙ|:where Eqs.A.2 and A.6 are used to derive Eq.A.8. Similarly, rewrite t pz in Eq. 2 as a distance divided by the relative speed, |ṙ|, to becomeSubtracting Eq.A.9 from Eq. A.8 results in hence the proof of Eq.A.7 is completed. + + + + + + + Defining Well Clear for Unmanned Aircraft Systems + + StephenPCook + + + DallasBrooks + + + RodneyCole + + + DavisHackenberg + + + VincentRaska + + 10.2514/6.2015-0481 + AIAA-2015-0481 + + + AIAA Infotech @ Aerospace + + American Institute of Aeronautics and Astronautics + 2015 + + + Cook, S. P., Brooks, D., Cole, R., Hackenberg, D., and Raska, V., "Defining Well Clear for Unmanned Aircraft Systems," Proceedings of AIAA Infotech@ Aerospace, AIAA-2015-0481, AIAA, 2015. + + + + + Characteristics of a Well Clear Definition and Alerting Criteria for Encounters between UAS and Manned Aircraft in Class E Airspace + + MJohnson + + + ERMueller + + + CSantiago + + + + Europe Air Traffic Management Research and Development Seminar + + + 2015 + + + Eleventh UAS + Johnson, M., Mueller, E. R., and Santiago, C., "Characteristics of a Well Clear Definition and Alerting Criteria for Encounters between UAS and Manned Aircraft in Class E Airspace," Eleventh UAS/Europe Air Traffic Management Research and Development Seminar , 2015, pp. 23-26. + + + + + FAA Position on Building Consensus around the SARP Well-Clear Definition + + DWalker + + + + RTCA Special Committee + + 228 + 2014 + + + Walker, D., "FAA Position on Building Consensus around the SARP Well-Clear Definition," RTCA Special Committee, Vol. 228, 2014. + + + + + Flight Test Overview for UAS Integration in the NAS Project + + JamesRMurphy + + + PeggySWilliams-Hayes + + + SamKKim + + + WayneBridges + + + MichaelMarston + + 10.2514/6.2016-1756 + AIAA-2016-1756 + + + AIAA Atmospheric Flight Mechanics Conference + + American Institute of Aeronautics and Astronautics + 2016 + + + AIAA SciTech + Murphy, J. R., Hayes, P. S., Kim, S. K., Bridges, W., and Marston, M., "Flight Test Overview for UAS Integration in the NAS Project," AIAA Atmospheric Flight Mechanics Conference, AIAA SciTech, AIAA-2016-1756, 2016. + + + + + NATIONAL SYMPOSIUM ON DEVELOPMENTS IN IRRADIATION TESTING TECHNOLOGY, SEPTEMBER 9--11, 1969, NASA LEWIS RESEARCH CENTER, SANDUSKY, OHIO. + + SMLee + + + CPark + + + DPThipphavong + + + DRIsaacson + + + CSantiago + + 10.2172/4165376 + + + NASA/TM-2016-219067, National Aeronautics and Space Administration + + Office of Scientific and Technical Information (OSTI) + 2016 + + + Lee, S. M., Park, C., Thipphavong, D. P., Isaacson, D. R., and Santiago, C., "Evaluating Alerting and Guidance Performance of a UAS Detect-And-Avoid System," NASA/TM-2016-219067, National Aeronautics and Space Administration, Ames Research Center, 2016. + + + + + Analysis of alerting performance for detect and avoid of unmanned aircraft systems + + SamanthaSmearcheck + + + SeanCalhoun + + + WilliamAdams + + + JaredKresge + + + FabriceKunzi + + 10.1109/plans.2016.7479766 + + + 2016 IEEE/ION Position, Location and Navigation Symposium (PLANS) + + IEEE + 2016 + + + + Smearcheck, S., Calhoun, S., Adams, W., Kresge, J., and Kunzi, F., "Analysis of Alerting Performance for Detect and Avoid of Unmanned Aircraft Systems," IEEE/ION Position, Location and Navigation Symposium (PLANS), 2016, pp. 710-730. + + + + + An Evaluation of Detect and Avoid (DAA) Displays for Unmanned Aircraft Systems: The Effect of Information Level and Display Location on Pilot Performance + + LisaFern + + + RConradRorie + + + JessicaPack + + + JayShively + + + MarkDraper + + 10.2514/6.2015-3327 + AIAA-2015-3327 + + + 15th AIAA Aviation Technology, Integration, and Operations Conference + + American Institute of Aeronautics and Astronautics + 2015 + + + Fern, L., Rorie, R. C., Pack, J. S., Shively, R. J., and Draper, M. H., "An Evaluation of Detect and Avoid (DAA) Displays for Unmanned Aircraft Systems: The Effect of Information Level and Display Location on Pilot Performance," Proceedings of 15th AIAA Aviation Technology, Integration, and Operations Conference, AIAA-2015-3327, 2015. + + + + + The Generic Resolution Advisor and Conflict Evaluator (GRACE) for Detect-And-Avoid (DAA) Systems + + MichaelAbramson + + + MohamadRefai + + + ConfesorSantiago + + 10.2514/6.2017-4485 + + + 17th AIAA Aviation Technology, Integration, and Operations Conference + + American Institute of Aeronautics and Astronautics + June 2017 + + + Abramson, M., Refai, M., and Santiago, C., "A Generic Resolution Advisor and Conflict Evaluator (GRACE) in Ap- plications to Detect-And-Avoid (DAA) Systems of Unmanned Aircraft," Proceedings of the 17th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, June 2017. + + + + + DAIDALUS: Detect and Avoid Alerting Logic for Unmanned Systems + + CesarMunoz + + + AnthonyNarkawicz + + + GeorgeHagen + + + JasonUpchurch + + + AaronDutle + + + MariaConsiglio + + + JamesChamberlain + + 10.1109/dasc.2015.7311421 + + + 2015 IEEE/AIAA 34th Digital Avionics Systems Conference (DASC) + + IEEE + 2015 + + + + Muñoz, C., Narkawicz, A., Hagen, G., Upchurch, J., Dutle, A., Consiglio, M., and Chamberlain, J., "DAIDALUS: Detect and Avoid Alerting Logic for Unmanned Systems," 34th Digital Avionics Systems Conference (DASC), IEEE/AIAA, 2015, pp. 5A1-1. + + + + + Development, Integration and Testing of a Stand-alone CDTI with Conflict Probing Support + + BrandonSuarez + + + KevinKirk + + + ErikTheunissen + + 10.2514/6.2012-2487 + + + Infotech@Aerospace 2012 + + American Institute of Aeronautics and Astronautics + 2012. 2012, p. 2487. Feb. 2011 + + + Conflict Prediction and Display System (CPDS). 11. Introduction to TCAS II Version 7.1," HQ-111358, Federal Aviation Administration (FAA + Suarez, B., Kirk, K., and Theunissen, E., "Development, Integration and Testing of a Stand-Alone CDTI with Conflict Probing Support," Infotech@ Aerospace 2012 , 2012, p. 2487, Conflict Prediction and Display System (CPDS). 11 "Introduction to TCAS II Version 7.1," HQ-111358, Federal Aviation Administration (FAA), Feb. 2011. + + + + + Multi-intruder aircraft, multi-sensor tracking system + + VibhorLBageshwar + + + EricAEuteneuer + + 10.1109/dasc.2015.7311425 + + + 2015 IEEE/AIAA 34th Digital Avionics Systems Conference (DASC) + + IEEE + 2015 + + + + Bageshwar, V. L. and Euteneuer, E. A., "Multi-Intruder Aircraft, Multi-Sensor Tracking System," 34th Digital Avionics Systems Conference (DASC), IEEE/AIAA, 2015, pp. 5A5-1-13. + + + + + A Family of Well-Clear Boundary Models for the Integration of UAS in the NAS + + CesarMunoz + + + AnthonyNarkawicz + + + JamesChamberlain + + + MariaCConsiglio + + + JasonMUpchurch + + 10.2514/6.2014-2412 + No. AIAA-2014-2412 + + + 14th AIAA Aviation Technology, Integration, and Operations Conference + Georgia, Atlanta, USA + + American Institute of Aeronautics and Astronautics + June 2014 + + + Muñoz, C., Narkawicz, A., Chamberlain, J., Consiglio, M., and Upchurch, J., "A Family of Well-Clear Boundary Models for the Integration of UAS in the NAS," Proceedings of the 14th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, No. AIAA-2014-2412, Georgia, Atlanta, USA, June 2014. + + + + + A Well-Clear Volume Based on Time to Entry Point + + AJNarkawicz + + + CAMuñoz + + + JMUpchurch + + + JPChamberlain + + + MCConsiglio + + + + NASA/TM-2014-218155, National Aeronautics and Space Administration + + Langley Research Center + 2014 + 15 + + + Narkawicz, A. J., Muñoz, C. A., Upchurch, J. M., Chamberlain, J. P., and Consiglio, M. C., "A Well-Clear Volume Based on Time to Entry Point," NASA/TM-2014-218155, National Aeronautics and Space Administration, Langley Research Center, 2014. 15 + + + + + Aircraft collision avoidance based on GPS position broadcasts + + RYGazit + + + JDPowell + + 10.1109/dasc.1996.559189 + + + 15th DASC. AIAA/IEEE Digital Avionics Systems Conference + + IEEE + 1996 + + + + 15th Digital Avionics Systems Conference + Gazit, R. Y. and Powell, J. D., "Aircraft Collision Avoidance Based on GPS Position Broadcasts," 15th Digital Avionics Systems Conference, IEEE/AIAA, 1996, pp. 393-399. + + + + + The Protected Volume of Airspace Generated by an Airborne Collision Avoidance System + + RLFord + + 10.1017/s0373463300000035 + + + Journal of Navigation + J. Navigation + 0373-4633 + 1469-7785 + + 39 + 2 + + 1986 + Cambridge University Press (CUP) + + + Ford, R., "The Protected Volume of Airspace Generated by an Airborne Collision Avoidance System," Journal of Navi- gation, Vol. 39, No. 02, 1986, pp. 139-158. + + + + + Analysis of Well-Clear Boundary Models for the Integration of UAS in the NAS + + JMUpchurch + + + CAMuñoz + + + AJNarkawicz + + + JPChamberlain + + + MCConsiglio + + + + NASA/TM-2014-218280, National Aeronautics and Space Administration + + Langley Research Center + 2014 + + + Upchurch, J. M., Muñoz, C. A., Narkawicz, A. J., Chamberlain, J. P., and Consiglio, M. C., "Analysis of Well-Clear Boundary Models for the Integration of UAS in the NAS," NASA/TM-2014-218280, National Aeronautics and Space Admin- istration, Langley Research Center, 2014. + + + + + UAS Integration in the NAS Project: Flight Test 3 Data Analysis of JADEM-Autoresolver Detect and Avoid System + + CGong + + + MGWu + + + CSantiago + + + + NASA/TM-2014-218155, National Aeronautics and Space Administration + + Ames Research Center + 2016 + + + Gong, C., Wu, M. G., and Santiago, C., "UAS Integration in the NAS Project: Flight Test 3 Data Analysis of JADEM- Autoresolver Detect and Avoid System," NASA/TM-2014-218155, National Aeronautics and Space Administration, Ames Research Center, 2016. + + + + + Feature article: proof-of-concept airborne sense and avoid system with ACAS-XU flight test + + TatsuyaKotegawa + + + BSuarez + + + CGong + + + VLBageshwar + + + EAEuteneuer + + + CLeeper + + + EMaki + + + ETheunissen + + 10.1109/maes.2016.150163 + + + + IEEE Aerospace and Electronic Systems Magazine + IEEE Aerosp. Electron. Syst. Mag. + 0885-8985 + + 31 + 9 + + 2016. 20 + Institute of Electrical and Electronics Engineers (IEEE) + + + Kotegawa, T., Suarez, B., Gong, C., Bageshwar, V. L., Euteneuer, E. A., Leeper, C., Maki, E., and Theunissen, E., "Proof-of-Concept Airborne Sense and Avoid System with ACAS-X U Flight Test," IEEE Aerospace and Electronic Systems Magazine, Vol. 31, No. 9, 2016, pp. 53-62. 20 https://www.faa.gov/nextgen/programs/adsb/faq/#1. + + + + + Formal Analysis of Extended Well-Clear Boundaries for Unmanned Aircraft + + CésarMuñoz + + + AnthonyNarkawicz + + 10.1007/978-3-319-40648-0_17 + + + Lecture Notes in Computer Science + Lecture Notes in Computer Science + + SRayadurgam + + + OTkachuk + + Minneapolis, MN + + Springer International Publishing + June 2016 + 9690 + + + + Muñoz, C. and Narkawicz, A., "Formal Analysis of Extended Well-Clear Boundaries for Unmanned Aircraft," Proceedings of the 8th NASA Formal Methods Symposium (NFM 2016), edited by S. Rayadurgam and O. Tkachuk, Vol. 9690 of Lecture Notes in Computer Science, Springer, Minneapolis, MN, June 2016, pp. 221-226. + + + + + Extended Airspace Encounter Models for Unmanned Aircraft Sense and Avoid Safety Evaluation + + AndrewJWeinert + + + EricHarkleroad + + + JohnGriffith + + + MatthewWEdwards + + 10.2514/6.2013-5049 + + + AIAA Infotech@Aerospace (I@A) Conference + + American Institute of Aeronautics and Astronautics + 2013 + + + Tech. rep., DTIC Document + Weinert, A. J., Harkleroad, E. P., Griffith, J., Edwards, M. W., and Kochenderfer, M. J., "Uncorrelated Encounter Model of the National Airspace System, Version 2.0," Tech. rep., DTIC Document, 2013. + + + + + Build 8 of the Airspace Concept Evaluation System + + SapaGeorge + + + GoutamSatapathy + + + VikramManikonda + + + KeePalopo + + + LarryMeyn + + + ToddLauderdale + + + MichaelDowns + + + MohamadRefai + + + RichardDupee + + 10.2514/6.2011-6373 + + + AIAA Modeling and Simulation Technologies Conference + + American Institute of Aeronautics and Astronautics + 2011 + 6373 + + + George, S., Satapathy, G., Manikonda, V., Palopo, K., Meyn, L., Lauderdale, T., Downs, M., Refai, M., and Dupee, R., "Build 8 of the Airspace Concept Evaluation System," AIAA modeling and simulation technologies conference, 2011, p. 6373. + + + + + + diff --git a/file792.txt b/file792.txt new file mode 100644 index 0000000000000000000000000000000000000000..2ea8f80e256830da0005f7049cd1483599ec9087 --- /dev/null +++ b/file792.txt @@ -0,0 +1,574 @@ + + + + +This paper investigates effects of limited surveillance volume on the alerting performance of a Detect and Avoid (DAA) system for unmanned aircraft systems (UAS).The surveillance volume accounts for an airborne sensor capable of detecting non-cooperative aircraft.Independent variables include four candidate DAA Well Clear (DWC) definitions and five surveillance volumes.Open-loop alerting performance metrics are computed from the results of running a reference DAA algorithm on a large number of synthesized encounters.The speed range for the UAS traffic considered is between 40 and 100 kts.Results show that, with a 2.5 nmi sensor range, all four candidate DWCs allow at least an average of 25 seconds warning alert times before a loss of DWC.Cumulative distributions of the intruder's bearing and elevation at the first warning alert suggest that ±10 • and ±140 • , respectively, are sufficient for alerting > 95% of the encounters that lead to losses of DWC. +I. Nomenclature +II. IntroductionSuccessful integration of Unmanned Aircraft System (UAS) operations in the National Airspace System (NAS) cannot be realized without adequate Detect and Avoid (DAA) Systems.A DAA system provides surveillance, alerts, and maneuver guidance to keep a UAS "well clear" of other aircraft [1,2].In the United States, simulation tests as well as flight tests have provided supporting information for defining a DAA Well Clear [1,3] (DWC) and requirements for the alerting and maneuver guidance performance [4][5][6][7][8].Prototype DAA algorithms have also been developed for alerting and maneuver guidance (referred to as guidance in this paper) research [9][10][11].These developments enabled the RTCA Special Committee 228 (SC-228) to publish the Minimum Operational Performance Standards (MOPS) for DAA systems [12] and air-to-air radar [13] in 2017.The corresponding Technical Standard Orders (TSO), TSO-C211 and TSO-C212, were published by the Federal Aviation Administration (FAA) in October 2017.These standards, referred to as the Phase 1 MOPS, target UAS operations in non-terminal areas.A DAA system, according to the Phase 1 MOPS, contains surveillance components of Automatic Dependent Surveillance-Broadcast (ADS-B) In, airborne active surveillance, and air-to-air radar that can detect aircraft with or without transponders.Traffic Alert and Collision Avoidance System (TCAS) II [14] is an optional component.Phase 2 work for extending the MOPS to additional UAS categories and operations is underway.One of the Phase 2 objectives is to define requirements for operations by UAS equipped with low cost, size, weight, and power (low C-SWaP) sensors.These UAS missions are envisioned to fly much slower than 200 kts, the maximum UAS airspeed in the Phase 1 MOPS.For these UAS and their missions, a large and high-power radar, as required by the Phase 1 radar MOPS, is physically infeasible and/or economically impractical.Examples of missions in this category are air quality monitoring, aerial imaging and mapping, and flood inundation mapping [15].While low C-SWaP sensors are desirable for these missions, they must provide sufficient surveillance volume and accuracy to ensure the DAA system's capability of maintaining safety.Another Phase 2 objective seeks an alternative DWC for UAS with non-cooperative aircraft, i.e., aircraft without a functioning transponder.The DWC in the Phase 1 work was selected with considerations of interoperability with TCAS-II.To avoid triggering TCAS's resolution advisories during an encounter which leads to DAA maneuvers, the DWC was defined to enclose a vast majority of the TCAS alerting volume [1].The resulting DWC is deemed very safe but may be unnecessarily large for encounters of UAS with non-cooperative aircraft, which TCAS-II cannot detect and therefore need not be considered.Four candidate DWCs were proposed for additinal analyses following an analysis of encounters representing low C-SWaP UAS operations [16].Surveillance Volume DAA Performance OE 1 OE 2Fig. 1 The DAA performance vs. surveillance volume.As a direct support of SC-228 Phase 2 MOPS work, this paper analyzes the effects of surveillance volume on the DAA's alerting performance.Adequate surveillance volume provides sufficient alerting timelines for UAS operators or pilots to maneuver upon DAA's guidance.Additional surveillance volume may provide marginal benefit to the DAA system's overall performance while raising the required C-SWaP of the sensor to a level too high for the feasibility of many UAS operations.Figure 1 depicts a notional plot of the relationship between the DAA system's performance and surveillance volume.The DAA system's performance can be measured by safety metrics, operational suitability, pilot acceptance, etc.Here, an operational environment (OE), defined by parameters such as airspace, mission type, and speed range, can vary the performance of a DAA system.In addition to identifying an adequate surveillance volume, this paper also analyzes the alerting timeline's sensitivity to a DWC by comparing results across four candidate DWCs for non-cooperative aircraft.Results will inform the SC-228 of the selection of a final DWC as well as recommendations to alerting and surveillance requirements.This paper is organized as follows: Section III provides background information about the DWC, alerting, guidance, and operational assumptions.The alerting metrics, encounter set, and the experiment matrix for this work are described in Section IV.For this study, four candidate DWCs and five surveillance volumes serve as independent variables.Section V presents results and discusses variations across DWCs. +III. Background +A. Detect-and-Avoid Well Clear, Alerting, and GuidanceThe DAA system aims to keep the UAS "well clear" of other aircraft.The DWC defines, in a quantitative way, the well clear volume around other aircraft the UAS should avoid.Requirements for alerting and guidance are built upon a DWC definition.For reasons stated in Section II, considerations for an alternative DWC for UAS and non-cooperative aircraft were investigated and four candidate DWCs, two primary and two secondary (backup), were proposed for additional analyses [16].Table 1 On the other hand, the τ mod component in DWC1 can cause a LoDWC when the two aircraft are still 1 nmi apart horizontally if the closure rate between the aircraft is high enough.In general, for most closure rates a LoDWC will occur earlier with DWC1 than with DWC2.The DAA alerting structure consists of three alert types: 1) Preventive: a caution level alert that advises the pilot to maintain the UAS's current altitude in order to avoid conflicts.2) Corrective: a caution level alert that advises the pilot to coordinate with ATC before maneuvering.3) Warning: a warning level alert that requires immediate action from the pilot to start maneuvering in order to maintain DWC.In addition to the three alert types, if the DAA system is equipped with TCAS-II (Equipment Class 2 system only [12]), TCAS alerts may be triggered.The preventive alert is irrelevant for encounters involving non-cooperative aircraft due to the lack of accurate vertical (altitude and vertical speed) surveillance data.The analysis in this paper targets only the corrective and warning alerts.Figure 2 shows the alerting timeline as well as the corresponding guidance.The guidance include ranges of heading and altitude predicted by the DAA system to have a high likelihood of leading to losses of DWC (LoDWC).There is a corresponding guidance for each alert type.Aircraft performance parameters such as turn, climb, and descent rates can be used for computing the ranges of heading and altitude.The DAA MOPS also defines display requirements for alerts and guidance.Figure 3 shows an example of display of a warning alert and guidance, where AC01 represents the position of an intruder and the triangle at the center of the circle represents the position of the ownship (the unmanned aircraft).The ranges of heading and altitude predicted to lead to conflicts are displayed in bands with a red color specifically for the warning alert.If the ownship gets too close to the intruder, a LoDWC may become inevitable even with maneuvers.In this situation, the guidance bands display all red for heading and altitude, but at the same time computes "regain well clear" bands to assist the ownship in maneuvering in order to regain well clear effectively.Regain-well-clear is referred to as well clear recovery (WCR) in this paper.The WCR usually takes place earlier than a LoDWC during an encounter. +B. SurveillanceThe Phase 1 radar, the only sensor that detects non-cooperative aircraft, requires a target declaration volume of 6.7 nmi range, * ±110 • bearing, and ±15 • elevation.This surveillance volume allows more than enough alerting time for warning alerts in an encounter with the highest possible closure rate (370 kts).For detection of cooperative aircraft, both ADS-B and active surveillance provide even greater detection range (> 15 nmi).The Phase 2 work seeks to create requirements for low C-SWaP sensors, which are expected to have smaller surveillance volume than that of the Phase 1 radar.Associated with a low C-SWaP sensor are requirements for an alternative DWC as well as alerting and guidance.ADS-B and active surveillance are still required of the DAA system for detection of cooperative aircraft (those with functioning transponders and/or ADSB-out.)The following observations serve to justify lower sensor requirements than the Phase 1 radar while still maintaining operational safety in the airspace:• Encounters between UAS and non-cooperative aircraft will be relatively infrequent given the fact that, after year 2020, most airspace will mandate ADS-B on aircraft.Even in the airspace outside the ADS-B mandate, i.e., Class E under 10,000 ft MSL, non-cooperative aircraft comprise a small percentage (estimated 15%) of the traffic [12].• An alternative, smaller DWC should give UAS operators more time to maintain DWC.• The Phase 1 operations support UAS speeds up to 200 kts, a speed much higher than the optimal speed for many UAS operations with low C-SWaP sensors.The lower closure rate considered allows for more alerting time.• Even if the surveillance volume is not enough to support correct alerts, a UAS pilot/operator is likely to be able to maintain separation if warning alert and guidance is provided with enough time.Figure 2 also compares the alerting timeline to the detection time provided by surveillance.The detection time is related to the sensor's detection range, and bearing and elevation to a lesser extent, by the probabilistic distribution of closure rates during encounters. +C. UAS Operations with a Low C-SWaP SensorSome of the operational assumptions specific to low C-SWaP operations are given below:• Extended UAS operations in non-terminal Classes D, E, and G airspaces, as well as those transitting Classes B and C airspaces.• UAS mission speed range is between 40 and 100 kts.• UAS is capable of turning horizontally at a rate of 7 deg/s during a maneuver upon DAA guidance.• The non-cooperative aircraft's airspeed is assumed to be at or below 170 kts (95% percentile [17]).• No non-cooperative aircraft exist above 11,000 ft MSL.• Below 500 ft AGL, the airborne sensor for non-cooperative aircraft is not responsible for detecting intruders. +IV. Experiment PlanThe objectives of this analysis are the following: 1) to identify adequate surveillance volume for detecting non-cooperative aircraft that ensures acceptable DAA alerting performance.2) to investigate sensitivity of the alerting timeline to variation of the DWC A surveillance volume is characterized by range (distance), bearing range, and elevation range.For simplicity, bearing and elevation are assumed to be with respect to an aircraft reference frame with zero roll and pitch angles.The following sections discuss the alerting metrics, the DAA algorithm, and the encounter set used for this analysis. +A. Alerting MetricsThe following open-loop (no UAS maneuver) alerting performance metrics are computed:• Average alert time before LoDWC: the average time before the LoDWC at which the alerting system issues an alert.• Average alert time before WCR: the average time before the WCR at which the alerting system issues an alert.• Late Alert probability: a late alert occurs where an intruder has a LoDWC but the alerting system issues an alert less than the required time before LoDWC.The required time is 20 seconds for corrective alerts and 15 seconds for warning alerts.Only encounters that lead to LoDWCs are considered. +B. Detect-and-Avoid AlgorithmThe open-source Detect and AvoID Alerting Logic for Unmanned Systems (DAIDALUS) [10], a reference DAA algorithm for the Phase 1 MOPS, is invoked to generate alerting sequences for the performance analysis.A standard configuration file † containing alerting parameters for DAIDALUS serves as a starting point, while some parameters affine to the DWC are modified according to the candidate DWCs considered.The conflict zone the alerting and guidance protects is based on each DWC with its HMD* buffered by a factor of 1.519 (following the Phase 1 setting).The buffer gives the system a few seconds to alert against aircraft suddenly maneuvering towards the UAS.Corrective and warning alerts are issued if intruders are predicted, with a constant velocity assumption, to enter the conflict zone within 60 and 30 seconds, respectively.DAIDALUS also allows specification of aircraft maneuverability parameters such as the rates of turn, climb, and descent.The turn rate affects the WCR time.For this study the turn rate is set to 7 degrees per second.Only horizontal guidance is considered when calculating the WCR time.The vertical guidance is muct less robust in reality due to the uncertainties of vertical states, including altitudes and vertical speeds. +C. Encounter SetAn entire day's worth of UAS flights are considered for this study.These flights consist of 12 different types of missions considered suitable for UAS with low C-SWaP sensors.The demand and mission profiles were generated based on subject matter experts' opinions and socio-economical analysis [15].These missions cover the entire continental US and amount to a total of 17,100 hours flight time.Details of the twelve missions are described in a previous publication [16].For the intruder traffic, nation-wide VFR flight paths flown in 2012 were extracted from the historical Air Force 84th Radar Evaluation Squadron (RADES) radar data.The VFR track data, including non-cooperative aircraft and cooperative aircraft with 1200 transponder code ‡ were processed to remove measurement noise and generate continuous trajectory data.Due to the limited number of non-cooperative VFR trajectories available, 1200-code cooperative VFR aircraft are used as a surrogate for non-cooperative aircraft for this study.This is a reasonable approach since the flight characteristics of conventional non-cooperative aircraft are similar to those using cooperative VFR aircraft in terms of airspeed, acceleration, and turn rate [18].Figure 4 shows the speed and altitude distributions of UAS and intruder by flight hours.Only data within the speed and altitude ranges considered for low C-SWaP encounters are shown.Encounters are identified when UAS trajectories are overlaid with the VFR traffic.A software suite was developed to detect and produce encounters from these overlaid trajectories [19].To analyze only the encounters that fit the low C-SWaP operational assumptions, the encounter data were filtered by the altitude and speed of ownship and intruder aircraft.Unmanned aircraft whose speed at the closest point of approach (CPA) is between 40 and 100 kts, and altitude at the CPA is below 11,000 ft MSL and above 500 ft AGL, were selected.Non-cooperative intruder aircraft whose speed at CPA is less than 170 kts and altitude at CPA is below 11,000 ft MSL and above 500 ft AGL were selected for the simulation. +D. Experiment MatrixTable 2 shows the full experiment matrix with values of these independent variables.The alerting performance metrics computed from the encounter set can be regarded as optimistic upper bounds.Realistic sensors usually have limited bearing and elevation ranges that will reduce the alerting performance. +V. ResultsEncounters created from overlaying the VFR traffic recorded for twenty-one days in 2012 with the the same one-day UAS trajectories, respectively, were analyzed.The encounter set was analyzed using DAIDALUS for each combination ‡ VFR flights in uncontrolled airspace will "squawk VFR" (1200 in the US, 7000 in Europe). +A. Alerting MetricsFigure 5 shows the average corrective alert time before LoDWC.The general trend is that the more range that is available, the more alerting time.The range of the highest end of 8 nmi is slightly greater of the Phase 1 radar and serves as an upper bound for the low C-SWaP sensor.Going down to a 4 nmi range affects the average corrective alert time minimally for DWC1, DWC2, and DWC3.For DWC4, the average corrective time decreases by a noticeable amount of 5 seconds.Below 4 nmi, the corrective alert time for DWC4 falls below that of the other DWCs.This is expected since DWC4 is the largest and thus more sensitive to surveillance volume reduction.DWC2 yields a consistently higher average alert time than the other three DWCs'.Figure 6 shows the average warning alert time before LoDWC.The values at 4 nmi and 8 nmi are essentially identical, indicating that a 4 nmi range encloses the entire warning alert zone of all four DWCs.The warning alert time for DWC1, DWC2, and DWC3 stays almost the same with a 3 nmi range.The value for DWC4 drops by only 2 seconds at a 3 nmi range.At a 2 nmi range DWC2 is only slightly affected while DWC1 and DWC3's warning alert times drop to 28 seconds.The Phase 1 MOPS expects 25 seconds warning alert time for non-accelerating encounters to support pilot response and maneuver execution.Not all encounters in the encounter set can achieve 25 seconds because some encounters involve maneuvering intruders or ownship.Therefore, an average warning alert time of 25 seconds is likely to be deemed acceptable.This seems to suggest that 2 nmi might be acceptable for DWC1, DWC2, and DWC3.However, these alerting times are optimistic upper bounds and will be reduced by limited bearing and elevation ranges that occur in reality.Therefore, a minimum range of 2.5 nmi seems a more practical requirement.Since the corrective alert is likely to be regarded as optional for operations of UAS with low C-SWaP sensors, the following discussion will focus on only the warning alert metrics.Figure 7 shows the average warning alert time before WCR.This metric indicates the amount of time UAS operators or pilots have upon receiving a warning alert until it is too late to maintain DWC.Prior research indicates that it takes pilots about 10 seconds to respond and execute a maneuver upon a warning alert and guidance [7].With that information, the alert times for ≥ 2 nmi range are deemed acceptable for all four DWCs.Interestingly, DWC2 yields less alert time before WCR compared to the other three DWCs.This is because, in general, alerts with DWC2 start later during an encounter while its WCR time is about the same as that of DWC1. Figure 8 shows late alert percentages for warning alert.A warning alert is regarded late by the Phase 1 MOPS if the first alert starts within 15 seconds of the LoDWC.An late alert happens usually because an intruder maneuvers towards the ownship when the two aircraft are already close.However, if the surveillance volume is very limited a late alert can occur even for a non-accelerating intruder.This is undesirable.DWC2 is more resilient against surveillance range limitation than the other three DWCs, showing a consistent low percentage of late alerts down to 2 nmi range. +B. Distribution of Initial Warning Alert LocationThe surveillance volume should ideally support the warning alert timeline for a majority of the encounters.To investigate what bearing and elevation scan can achieve this, locations of the initial warning alert in the run with an 8 nmi range were further analyzed.Figure 9 shows the cumulative distribution of the range at the start of a warning alert.The 90 percentile range is 2.6, 2.1, 2.5, and 3.1 nmi for DWC1, DWC2, DWC3, and DWC4, respectively.The 95 percentile range is 2.8, 2.3, 2.8, and 3.5 nmi for DWC1, DWC2, DWC3, and DWC4, respectively.This chart provides insight of what the minimum sensor detection range should be to detect a specified percentile of first warning alerts.If the sensor detection range is less than the required minimum range, then the warning alert timeline is likely to be cut short and may leave operators or pilots insufficient time to maintain DWC. Figure 10 shows the cumulative distribution of the bearing at the start of a warning alert.The 90 percentile bearing scan happens at 110 • and 95 percentile at 140 • for all four DWCs.Thus, the minimum required bearing for a sensor needs to be at least 110 • for all four DWCs if the sensor is required to detect more than 90 percentile of warning alerts at the start time.Figure 11 shows the cumulative distribution of the elevation at the start of a warning alert.The 90 percentile elevation scan happens at 6 • and 95 percentile at 10 • for all four DWCs.This chart shows that ±10 • of elevation might be sufficient to detect 95 percentile of warning alerts at the start time for all four DWCs. +VI. Summary and Future WorkThis study analyzes dependency of a DAA system's alerting performance on the surveillance volume of an onboard sensor.The operations considered are those in which an UAS is equipped with a low C-SWaP sensor responsible for detecting non-cooperative aircraft.The independent variables are the four candidate DWCs recommended in prior work and five surveillance volumes.Results show that a 2.5 nmi range can comfortably support an average of 25 second warning alert for all the DWCs.DWC2 among the four DWCs is the least sensitive to limited surveillance range and its alerting time is almost unaffected down to a 2 nmi range.A detection range that covers 95% of the beginning of a warning alert varies across DWCs, ranging from 2.3 (for DWC2) to 3.5 nmi (for DWC4).Cumulative distributions of the warning alert start location indicate that an elevation scan of ±10 • and a bearing scan of ±140 • will ensure that 95% of intruders are detected at the start of a warning alert.Results of this study will provide supporting information for the RTCA Special Committee 228 about determining low C-SWaP sensor requirements as well as changes to DAA's alerting requirements.The surveillance volume required for supporting the alerting performance, derived from this analysis, is an optimistic lower bound.Sensor uncertainties and the time taken by radar to declare a target after detection are expected to raise the bound.Moreover, alerting performance by itself is not enough to ensure effectiveness of a DAA algorithm.As a next step, the guidance algorithm will also be evaluated by closed-loop simulations involving realistic sensor uncertainties and pilots' maneuvers.These analyses will evaluate the DAA algorithm's performance as a whole and provide additional supporting information for the requirements of the low C-SWaP operations. +Appendix: DAA Well Clear and Conflict ZoneThe DAA well clear (DWC) zone for the UAS targeted in the Phase 1 MOPS is defined by thresholds of three parameters.It does not have distinct physical boundaries because the definition depends on two aircraft's relative position and velocity during an encounter.Figure 12 The Horizontal Miss Distance (HMD) represents the two aircraft's predicted minimum horizontal distance during an encounter assuming constant velocities.The parameter h represents the two aircraft's current altitude difference.The time metric modified tau, τ mod , is an estimated time taken for the two aircraft to intersect the "protection" disk.The range rate is negative for closing geometries.The positive incremental distance modifier D mod defines the radius of a "protection" disk around the Ownship such that any intruder with a horizontal range less than D mod is always considered "urgent".In this case, τ mod = 0.The thresholds, denoted by an asterisk, for the HMD, h, and τ mod are 4000 ft, 450 ft, and 35 sec, respectively.All three parameters must simultaneously fall below their respective thresholds during an encounter for the two aircraft to violate the DWC.Alerting algorithms are designed to reduce the probability of violating DWC to a value required by the MOPS.The definition of τ mod is [2]τ mod =        -r 2 -D mod 2r ṙ , r > D mod , 0, r ≤ D mod (1) where r and ṙ are the horizontal range and range rate between the intruding aircraft and the UAS, respectively.The value of D mod must be equal to HMD * to avoid the undesirable on-and-off alert during some constant velocity encounters [20].DAIDALUS's alert conflict zone is defined in a similar way to the DWC, using thresholds of the three variables HMD, h, and τ mod .For this work, the HMD threshold is increased to 1.529 times each DWC's HMD* to account for sensor and intruder intent uncertainties.AGL = above groundlevel C-SWaP = cost, size, weight, and power DAA = detect and avoid DWC = DAA Well Clear D mod = distance modification HMD = horizontal miss distance LoDWC = loss of DWC MOPS = minimum operational performance standards MSL = mean sea level OE = operational environment NMAC = near mid-air collision RADES = radar evaluation squadron SC-228 = special committee 228 TCAS = Traffic Alert and Collision Avoidance System TSO +Fig. 2 Fig. 323Fig. 2 Alerting timeline and the target detection time provided by the surveillance volume. +Fig. 44Fig. 4 Speed and altitude distributions of UAS and VFR traffic. +Fig. 55Fig. 5 Comparison of average corrective alert times before LoDWC. +Fig. 6 Fig. 767Fig. 6 Comparison of average warning alert times before LoDWC. +Fig. 8 Fig. 989Fig. 8 Late alert percentage for warning alert. +Fig. 10 Fig. 111011Fig. 10 Cumulative distribution of the bearing at the start of a warning alert. +Fig. 12 A12Fig. 12 A schematic representation of the DWC zone. +Table 1 Candidate DWCs for non-cooperative aircraft (Phase 1 DWC shown at bottom)1lists the four candidate DWCs.The Phase 1 DWC is also listed at the bottom of the table for comparison.Appendix VI defines the parameters, HMD*, τ mod * , and h * in detail.NameHMD* τ mod (ft) (sec) *h * (ft)CommentDWC1200015450PrimaryDWC222000450PrimaryDWC3150015450SecondaryDWC4250025450SecondaryPhase 1400035450 Phase 1 DWCDWC2 does not have a time component, τ mod , in its definition.The implication of this is, regardless of closure rate, intruders (usually manned aircraft) must be within 2,200 ft (0.36 nmi) of the UAS horizontally to result in a LoDWC. +Table 2 The experiment matrix2VariableValueDWCDWC1, DWC2, DWC3, DWC4Range (nmi)1, 2, 3, 4, 8Surveillance VolumeBearing Range (deg)(-180, 180]Elevation Range (deg)(-90, 90] + * The 6.7 nmi is the range for large non-cooperative intruders.Smaller ranges are required for medium and small intruders[13]. + † https://github.com/nasa/WellClear/blob/master/DAIDALUS/Configurations/WC_SC_228_nom_b.txt + + + + + + + + + Defining Well Clear for Unmanned Aircraft Systems + + StephenPCook + + + DallasBrooks + + + RodneyCole + + + DavisHackenberg + + + VincentRaska + + 10.2514/6.2015-0481 + + + AIAA Infotech @ Aerospace + + American Institute of Aeronautics and Astronautics + 2015 + + + Cook, S. P., Brooks, D., Cole, R., Hackenberg, D., and Raska, V., "Defining Well Clear for Unmanned Aircraft Systems," Proceedings of AIAA Infotech@ Aerospace, AIAA, 2015. + + + + + Characteristics of a Well Clear Definition and Alerting Criteria for Encounters between UAS and Manned Aircraft in Class E Airspace + + MJohnson + + + ERMueller + + + CSantiago + + + + Europe Air Traffic Management Research and Development Seminar + + + 2015 + + + Eleventh UAS + Johnson, M., Mueller, E. R., and Santiago, C., "Characteristics of a Well Clear Definition and Alerting Criteria for Encounters between UAS and Manned Aircraft in Class E Airspace," Eleventh UAS/Europe Air Traffic Management Research and Development Seminar, 2015, pp. 23-26. + + + + + FAA Position on Building Consensus Around the SARP Well-Clear Definition + + DWalker + + + + RTCA Special Committee + + 228 + 2014 + + + Walker, D., "FAA Position on Building Consensus Around the SARP Well-Clear Definition," RTCA Special Committee 228, 2014. + + + + + Flight Test Overview for UAS Integration in the NAS Project + + JamesRMurphy + + + PeggySWilliams-Hayes + + + SamKKim + + + WayneBridges + + + MichaelMarston + + 10.2514/6.2016-1756 + + + AIAA Atmospheric Flight Mechanics Conference + + American Institute of Aeronautics and Astronautics + 2016 + + + Murphy, J. R., Hayes, P. S., Kim, S. K., Bridges, W., and Marston, M., "Flight Test Overview for UAS Integration in the NAS Project," AIAA Atmospheric Flight Mechanics Conference, AIAA SciTech, 2016. + + + + + Evaluating Alerting and Guidance Performance of a UAS Detect-And-Avoid System + + SMLee + + + CPark + + + DPThipphavong + + + DRIsaacson + + + CSantiago + + + + NASA Ames Research Center + + 2016 + + + Lee, S. M., Park, C., Thipphavong, D. P., Isaacson, D. R., and Santiago, C., "Evaluating Alerting and Guidance Performance of a UAS Detect-And-Avoid System," NASA Ames Research Center, 2016. + + + + + Analysis of alerting performance for detect and avoid of unmanned aircraft systems + + SamanthaSmearcheck + + + SeanCalhoun + + + WilliamAdams + + + JaredKresge + + + FabriceKunzi + + 10.1109/plans.2016.7479766 + + + 2016 IEEE/ION Position, Location and Navigation Symposium (PLANS) + + IEEE + 2016 + + + + Smearcheck, S., Calhoun, S., Adams, W., Kresge, J., and Kunzi, F., "Analysis of Alerting Performance for Detect and Avoid of Unmanned Aircraft Systems," IEEE/ION Position, Location and Navigation Symposium (PLANS), 2016, pp. 710-730. + + + + + The Impact of Suggestive Maneuver Guidance on UAS Pilot Performing the Detect and Avoid Function + + RobertCRorie + + + LisaFern + + + JayShively + + 10.2514/6.2016-1002 + + + AIAA Infotech @ Aerospace + + American Institute of Aeronautics and Astronautics + 2016 + + + Rorie, R. C., Fern, L., and Shively, J., "The Impact of Suggestive Maneuver Guidance on UAS Pilot Performing the Detect and Avoid Function," AIAA InfoTech @ Aerospace Conference, AIAA SciTech, 2016. + + + + + An Evaluation of Detect and Avoid (DAA) Displays for Unmanned Aircraft Systems: The Effect of Information Level and Display Location on Pilot Performance + + LisaFern + + + RConradRorie + + + JessicaPack + + + JayShively + + + MarkDraper + + 10.2514/6.2015-3327 + + + 15th AIAA Aviation Technology, Integration, and Operations Conference + + American Institute of Aeronautics and Astronautics + 2015 + + + Fern, L., Rorie, R. C., Pack, J. S., Shively, R. J., and Draper, M. H., "An Evaluation of Detect and Avoid (DAA) Displays for Unmanned Aircraft Systems: The Effect of Information Level and Display Location on Pilot Performance," Proceedings of 15th AIAA Aviation Technology, Integration, and Operations Conference, 2015. + + + + + The Generic Resolution Advisor and Conflict Evaluator (GRACE) for Detect-And-Avoid (DAA) Systems + + MichaelAbramson + + + MohamadRefai + + + ConfesorSantiago + + 10.2514/6.2017-4485 + + + 17th AIAA Aviation Technology, Integration, and Operations Conference + + American Institute of Aeronautics and Astronautics + 2017 + + + Abramson, M., Refai, M., and Santiago, C., "A Generic Resolution Advisor and Conflict Evaluator (GRACE) in Applications to Detect-And-Avoid (DAA) Systems of Unmanned Aircraft," Proceedings of the 17th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, 2017. + + + + + DAIDALUS: Detect and Avoid Alerting Logic for Unmanned Systems + + CesarMunoz + + + AnthonyNarkawicz + + + GeorgeHagen + + + JasonUpchurch + + + AaronDutle + + + MariaConsiglio + + + JamesChamberlain + + 10.1109/dasc.2015.7311421 + + + 2015 IEEE/AIAA 34th Digital Avionics Systems Conference (DASC) + + IEEE + 2015 + + + + Muñoz, C., Narkawicz, A., Hagen, G., Upchurch, J., Dutle, A., Consiglio, M., and Chamberlain, J., "DAIDALUS: Detect and Avoid Alerting Logic for Unmanned Systems," 34th Digital Avionics Systems Conference (DASC), IEEE/AIAA, 2015, pp. 5A1-1. + + + + + Development, Integration and Testing of a Stand-alone CDTI with Conflict Probing Support + + BrandonSuarez + + + KevinKirk + + + ErikTheunissen + + 10.2514/6.2012-2487 + + + + Infotech@Aerospace 2012 + + American Institute of Aeronautics and Astronautics + 2012. 2012 + 2487 + + + Suarez, B., Kirk, K., and Theunissen, E., "Development, Integration and Testing of a Stand-Alone CDTI with Conflict Probing Support," Infotech@ Aerospace 2012, 2012, p. 2487. doi:10.2514/6.2012-2487, URL https://arc.aiaa.org/doi/abs/ 10.2514/6.2012-2487. + + + + + Minimum Operational Performance Standard (MOPS) for Helicopter Hoist Systems + 10.4271/as6342 + + 2017 + SAE International + + + Minimum Operational Performance Standards (MOPS) for Detect and Avoid (DAA) Systems, DO-365, RTCA. Inc., 2017. + + + + + AeroMACS minimum operational performance standards (MOPS) compliance field trials for Hitachi prototype + + RafaelApaza + + 10.1109/icnsurv.2015.7121331 + + + 2015 Integrated Communication, Navigation and Surveillance Conference (ICNS) + + IEEE + 2017 + + + Minimum Operational Performance Standards (MOPS) for Air-to-Air Radar for Traffic Surveillance, DO-366, RTCA. Inc., 2017. + + + + + Fixed-wing group petitions FAA for advisory circular + 10.1016/s0740-8315(88)80079-2 + + + + Hospital Aviation + Hospital Aviation + 0740-8315 + + 7 + 7 + + 2011 + Elsevier BV + + + 1%20Intro%20booklet.pdf + "Introduction to TCAS II Version 7.1," Federal Aviation Administration (FAA), 2011. URL http://www.faa.gov/ documentLibrary/media/Advisory_Circular/TCAS%20II%20V7.1%20Intro%20booklet.pdf. + + + + + UAS Demand Generation Using Subject Matter Expert Interviews and Socio-economic Analysis + + SricharanKAyyalasomayajula + + + RohitSharma + + + FrederickWieland + + + AntonioTrani + + + NicolasHinze + + + ThomasSpencer + + 10.2514/6.2015-3405 + + + 15th AIAA Aviation Technology, Integration, and Operations Conference + + American Institute of Aeronautics and Astronautics + 2015 + + + Ayyalasomayajula, S., Sharma, R., Wieland, F., Trani, A., Hinze, N., and Spencer, S., "UAS Demand Generation Using Subject Matter Expert Interviews and Socio-Economic Analysis," Proceedings of the AIAA Aviation Conference, 2015. + + + + + Well Clear Trade Study for Unmanned Aircraft System Detect And Avoid with Non-Cooperative Aircraft + + MinghongGWu + + + AndrewCCone + + + SeungmanLee + + + ChristineChen + + + MatthewWEdwards + + + DevinPJack + + 10.2514/6.2018-2876 + + + 2018 Aviation Technology, Integration, and Operations Conference + + American Institute of Aeronautics and Astronautics + 2018 + + + Wu, M. G., Cone, A. C., Lee, S., Chen, C., Edwards, E. W. M., and Jack, D. P., "Well Clear Trade Study for Unmanned Aircraft System Detect And Avoid with Non-Cooperative Aircraft," 18th AIAA Aviation Technology, Integration, and Operations Conference, 2018. + + + + + A Bayesian Approach to Aircraft Encounter Modeling + + MykelKochenderfer + + + JamesKuchar + + + JGriffith + + + LeoEspindle + + 10.2514/6.2008-6629 + + + AIAA Guidance, Navigation and Control Conference and Exhibit + Lexington, Massachusetts + + American Institute of Aeronautics and Astronautics + 2008 + + + MIT Lincoln Laboratory + + + Tech. rep + Kochenderfer, M. J., Kuchar, J. K., Espindle, L. P., and Griffith, J., "Uncorrelated Encounter Model of the National Airspace System, Version 1.0," Tech. rep., MIT Lincoln Laboratory, Lexington, Massachusetts, 2008. + + + + + Extended Airspace Encounter Models for Unmanned Aircraft Sense and Avoid Safety Evaluation + + AndrewJWeinert + + + EricHarkleroad + + + JohnGriffith + + + MatthewWEdwards + + 10.2514/6.2013-5049 + + + + AIAA Infotech@Aerospace (I@A) Conference + Lexington, Massachusetts + + American Institute of Aeronautics and Astronautics + Aug. 2013 + + + MIT Lincoln Laboratory + + + Tech. rep. + Weinert, A. J., Harkleroad, E. P., Griffith, J., Edwards, M. W., and Kochenderfer, M. J., "Uncorrelated Encounter Model of the National Airspace System, Version 2.0," Tech. rep., MIT Lincoln Laboratory, Lexington, Massachusetts, Aug. 2013. URL www.dtic.mil/cgi-bin/GetTRDoc?Location=U2&doc=GetTRDoc.pdf&AD=ADA589697. + + + + + Encounter-Based Simulation Architecture for Detect-And-Avoid Modeling + + MAbramson + + + MRefai + + + SLee + + + MGWu + + + + AIAA Science and Technology Forum and Exposition + + 2019. 2019 + + + Abramson, M., Refai, M., Lee, S., and Wu, M. G., "Encounter-Based Simulation Architecture for Detect-And-Avoid Modeling," AIAA Science and Technology Forum and Exposition 2019, 2019. + + + + + Formal Analysis of Extended Well-Clear Boundaries for Unmanned Aircraft + + CésarMuñoz + + + AnthonyNarkawicz + + 10.1007/978-3-319-40648-0_17 + + + Lecture Notes in Computer Science + + Springer International Publishing + 2016 + 9690 + + + + Muñoz, C., and Narkawicz, A., "Formal Analysis of Extended Well-Clear Boundaries for Unmanned Aircraft," Proceedings of the 8th NASA Formal Methods Symposium, Vol. 9690, Springer, 2016, pp. 221-226. + + + + + + diff --git a/file793.txt b/file793.txt new file mode 100644 index 0000000000000000000000000000000000000000..61c685e0560e0ffaaa1d5b03a8399d08089f57af --- /dev/null +++ b/file793.txt @@ -0,0 +1,774 @@ + + + + +I. IntroductionDetect and Avoid (DAA) Systems are a critical component for successful integration of Unmanned Aircraft System (UAS) operations in the National Airspace System (NAS).A DAA system provides surveillance, alerts, and guidance to keep a UAS "Well Clear" of other aircraft. 1,2 ][10] These developments enabled the RTCA Special Committee 228 (SC-228) to publish the Minimum Operational Performance Standards (MOPS) for DAA systems 11 and air-to-air radar 12 in 2017.The corresponding Technical Standard Orders (TSO), TSO-C211 and TSO-C212, were published by the Federal Aviation Administration (FAA) in October 2017.These standards, referred to as the Phase 1 MOPS, target UAS operations in non-terminal areas.UAS shall be equipped with surveillance systems containing Automatic Dependent Surveillance-Broadcast (ADS-B) In, airborne active surveillance, and air-to-air radar that can detect aircraft without transponders.Traffic Alert and Collision Avoidance System (TCAS) II 13 is an optional component.Phase 2 work for extending the MOPS to additional UAS categories and operations is underway.One of the Phase 2 MOPS objectives is to enable operations for UAS equipped with low cost, size, weight, and power sensors (low C-SWaP).These UAS missions are envisioned to take place at altitudes below 10,000 ft MSL with UAS speeds much lower than 200 kts, the maximum UAS airspeed in the Phase 1 MOPS.For these UAS and their missions, a large and high-power radar required by the Phase 1 radar MOPS is physically infeasible and/or economically impractical.Missions falling in this category include air quality monitoring, aerial imaging and mapping, and flood inundation mapping. 14Compared to the radar meeting the Phase 1 MOPS, low C-SWaP sensors provide less range for the UAS to detect and remain Well Clear from aircraft without transponders, i.e., non-cooperative aircraft.To find a path to enable these operations, assumptions and performance metrics utilized in Phase 1 MOPS are re-examined in Phase 2 work.One of the Phase 1 requirements reexamined in Phase 2 is the DWC definition.Phase 1 DWC was driven largely by interoperability with TCAS II.The form and size of the DWC were chosen such that the DWC encloses almost the entire volume inside which TCAS II would issue Resolution Advisories (RA).TCAS cannot detect non-cooperative aircraft, and therefore the TCAS RA consideration is irrelevant for non-cooperative aircraft, for which the Phase 1 DWC volume is likely to be unnecessarily large.It is desirable by SC-228 to seek an alternative DWC definition between an UAS and non-cooperative aircraft.The UAS considered include both those with low C-SWaP sensors and UAS defined in the Phase 1 MOPS (carrying a high-power radar).The alternative DWC does not necessarily take the same form as the Phase 1 MOPS DWC.It is, nonetheless, expected to be smaller than the Phase 1 DWC.As a direct support of SC-228 Phase 2 MOPS work, this paper presents a trade study of potential DWC definitions using two metrics related to loss of DWC (LoDWC).The two metrics are the unmitigated collision risk and the maneuver initiation range.Multiple types of DWC are explored and results are compared.Low C-SWaP operations are assumed for computing the metrics, although the resulting candidate DWCs are expected to be applicable to Phase 1 UAS as well.Section II briefly reviews the approach adopted in Phase 1 for selecting a DWC.The approach in this work is described in Section III.Section IV describes the two encounter sets used for computing the unmitigated collision risk.The maneuver initiation range is described in Section V. Results are presented in Section VI, and the down-selection process is discussed in Section VII. +II. BackgroundThe DWC in Phase 1 was initially selected from three types of DWC definitions using eight performance metrics. 1 The unmitigated collision risk, denoted as P , was used for tuning the DWC threshold parameters so that all three DWCs yield the same value of P . 1 P stands for the conditional probability of a near mid-air collision (NMAC) given a LoDWC without mitigation (by ownship maneuver):P = P (NMAC|LoDWC)(1)An earlier study recommended a P value of 5% for consideration of a DWC. 15During the Phase 1 MOPS work, the target value of P was set to 1.5% initially so as to expand the DWC volume to enclose most of the TCAS II Resolution Advisory alerting volume. 1 The selected DWC, however, had an undesirable vertical separation threshold of 700 ft, above the vertical separation of 500 ft required by visual flight rules (VFR). 3he vertical separation threshold was changed to 450 ft, and the final DWC resulted in a P of 2.2%.Maneuver initiation range (MIR) was one of the eight performance metric considered in Phase 1 MOPS. 16It is defined by a stressing case of head-on encounter, as the range between aircraft when the UAS must start maneuvering away in order to maintain DWC.Constant heading rate turns are assumed for horizontal maneuvers and constant climb and descent rates for vertical maneuvers.The other performance metrics adopted in the Phase 1 DWC selection process are not considered for this work.Some of these metrics are irrelevant (TCAS interoperability for example), and some others require additional assumptions about an alerting and guidance algorithm that would make it hard to tie the results directly to the DWC.The DWC in Phase 1 does not map to distinct physical boundaries because it depends on two aircraft's relative position and velocity.Figure 1 illustrates a DWC zone defined by the three parameters.The Horizontal Miss Distance (HMD) represents the two aircraft's predicted minimum horizontal distance (in the future) assuming constant velocities.The parameter h represents the two aircraft's current altitude difference.The time metric modified tau, τ mod , is an estimated time taken for the two aircraft to get horizontally close to each other.The thresholds, denoted by an asterisk, for the HMD, h, and τ mod are 4000 ft, 450 ft, and 35 sec, respectively.All three parameters must simultaneously fall below their respective thresholds during an encounter for the two aircraft to violate the DWC.The use of HMD and τ mod in a DWC definition was meant to facilitate interoperability with TCAS II, which uses similar definitions of HMD and τ mod in its alerting algorithm.The definition of τ mod is 2τ mod =    -r 2 -D mod 2 r ṙ , r > D mod , 0, r ≤ D mod (2)where r and ṙ are the horizontal range and range rate between the intruding aircraft (referred to as the intruder) and the UAS (referred to as the ownship), respectively.The range rate is negative for closing geometries.The distance modification D mod defines the radius of a protection disk around the ownship such that any intruder with a horizontal range less than D mod is always considered an urgent threat.In this case, τ mod = 0. +III. ApproachThis trade study aims to recommend candidate DWC definitions for the SC-228 to consider.The envisioned workflow diagram is shown in Figure 2.The independent variable is a DWC definition, • Encounters above 500 ft AGL and below 10,000 ft mean sea level (MSL) are considered.The upper bound of 10,000 ft comes from the FAA's rule that requires Mode C transponders for aircraft flying above 10,000 ft MSL.• Encounters in which the non-cooperative aircraft's airspeed is at or below 170 kts (95% percentile 20 ) are considered.• UAS speed range is between 40 and 100 kts.The types of DWC definitions considered are described in Table 1.Here t pz is the predicted time to protected D mod disk. 21This time metric corresponds to a physical event and changes linearly with time during a non-accelerating encounter.Compared to τ mod , t pz can be potentially a better time component for DWC or alerting algorithms.DWC3 is a conventional static cylinder adopted by air traffic control as a type of separation standard.DWC4 is a dynamic cylinder that is conceptually simpler than DWC1 and DWC2 because it does not have a HMD threshold.Note DWC3 is the limiting case of other DWC types when their time component reduces to zero. +IV. Encounter Sets +IV.A. ACES Generated EncountersEncounters are needed for this study for interactions between UAS (assumed to fly under IFR) and noncooperative VFR aircraft that do not have a transponder.ACES fast-time simulation was used for generating UAS trajectories.ACES provides a capability to simulate NAS-wide, gate-to-gate air traffic operations at local, regional, and national levels with medium-fidelity aircraft flight dynamics models. 17It simulates flight trajectories using aircraft models represented in a form similar to those from the Base of Aircraft Data (BADA). 22Aircraft models tailored for Aerosonde Mark 4.7, Shadow RQ-7, Ikhana/Predator B, Reaper MQ-9, and Global Hawk have been implemented and verified in ACES.For this study, an entire day's worth of UAS flights conducting 12 different types of missions considered suitable for UAS with low C-SWaP sensors are analyzed.The demand and mission profiles were generated based on subject matter experts opinions and socio-economical analysis. 14Table 2 shows the 12 UAS missions.These missions amount to a total of 17,100 hours' flight time.For the intruder traffic, nation-wide VFR flight paths flown on 21 days in 2012 were extracted from the historical Air Force 84th Radar Evaluation Squadron (RADES) radar data.The RADES data include primary-only radar returns (for non-cooperative aircraft) as well as cooperative transponder returns, providing track updates every 5 or 12 seconds.The instrument flight rules (IFR) track data, i.e., those with non-1200 discrete transponder codes, were excluded from the analysis.This is because air traffic controllers have a responsibility for separating IFR aircraft from other IFR (including UAS) aircraft.The remaining VFR track data, including non-cooperative aircraft and those cooperative aircraft with 1200 transponder code were processed to remove measurement noise and generate continuous trajectory data.Due to the limited number of non-cooperative VFR trajectories available, 1200-code cooperative VFR aircraft are used as a surrogate for non-cooperative aircraft.This is a reasonable approach since the flight characteristics of conventional non-cooperative aircraft are similar to those using a 1200-code cooperative VFR aircraft in terms of airspeed, acceleration, and turn rate. 18he encounters between UAS and VFR aircraft were simulated using SAA Control, a client of the Java Architecture for DAA Extensibility and Modeling (JADEM). 24JADEM employs a model of a DAA system that provides functions to evaluate potential LoDWC, to declare a DAA alert, and to optionally execute a maneuver to avoid predicted LoDWC.In this study, both the UAS trajectories and the actual VFR flight tracks were played back from files to create encounters.No maneuvers were applied to the UAS to avoid LoDWC.To analyze only the encounters that fit the operational assumptions, the encounter data were filtered by the altitude and speed of ownship and intruder aircraft.Unmanned aircraft whose speed at the CPA is less than 100 kts, and altitude at the CPA is below 10,000 ft MSL and above 500 ft AGL, were selected.Non-cooperative intruder aircraft whose speed at CPA is less than 170 kts and altitude at CPA is below 10,000 ft MSL and above 500 ft AGL were selected for the simulation results data analysis in this study.A total of two million encounters were constructed from the UAS trajectories against 21 days of traffic data.Since all VFR traffic within a large volume of 20 nmi horizontal distance and 10,000 ft altitude with respect to an UAS are considered for encounters, only a small fraction of the encounters resulted in LoDWC.An even smaller fraction of encounters led to NMAC.A total of 500 NMACs were identified. +IV.B. Uncorrelated Encounter ModelThe MIT Lincoln Laboratory's uncorrelated encounter model (referred to as the encounter model in the remainder of this paper) is used for this work to complement and validate results from the ACES generated encounters.This encounter model has been used in Phase 1 for tuning the parameters of candidate DWC types for a target P . 1 This model "is representative of encounters between a cooperative aircraft and a conventional non-cooperative aircraft similar those using a 1200 transponder code." 18,25 ayesian networks are used to create random aircraft trajectories that are statistically representative of VFR trajectories observed from radar data.The radar data used for developing this model came from radar returns of 1200-code aircraft from over 200 RADES radar sites across the NAS.As with the VFR traffic used in generating ACES encounters, the 1200-code aircraft were used to represent similar, conventional non-cooperative aircraft due to the lack of altitude information for non-cooperative aircraft.Altitude estimation was based on Mode C transponder reports of pressure altitude from the same radar data.The encounter model can be used to create trajectories up to 5 minutes before the closest point of approach, and so is well suited for studies involving collision avoidance and/or loss of separation for UAS encountering non-cooperative aircraft.The trajectories created from this model are also suitable for use in fast-time Monte Carlo simulations.The encounters are all pair-wise by default, but there is a capability for producing multiple aircraft encounters if required.Since both the UAS and intruder trajectories are sampled from the same Bayesian networks, the UAS trajectory has the same characteristics as the intruder trajectory in terms of climb, descent, and turn rates.This distinguishes the encounter model from ACES encounters, in which UAS trajectories have different characteristics than the intruder trajectories because the UAS trajectories are constructed from flight data and specific aircraft performance models.About one million encounters, defined when aircraft come within 3 nmi horizontally and 1000 ft vertically, were generated from the encounter model for the computation of P . +V. Maneuver Initiation RangeIn addition to P computed from encounters, MIR is computed for each DWC definition using 2PAIRS. 16he MIR is the distance between two aircraft when the UAS must maneuver away from the intruder to maintain DWC during a stressing head-on encounter.Figure 4 shows how an MIR is computed from a hypothetical, head-on encounter.Although the UAS maneuver can be horizontal, vertical, or a combination of both, only horizontal maneuvers are considered for this study.This is because vertical maneuvers are less robust in the presence of significant vertical surveillance errors typical for non-cooperative aircraft.A constant turn rate following a transient period of constant roll rate is used for modeling turns.It is assumed that smaller MIRs are preferable so as to potentially reduce the associated sensor requirements. +Intruder OwnshipTurn Rate (deg/sec) 234 This work defines the MIR as the maximum value resulting from the range of UAS speed considered.Either the low speed of 40 kts or the high speed of 100 kts can be the stressing case. 26,27 t turns out that 40 kts MIR governs the low time metric (τ mod or t pz less than 20 sec) parameter space for DWC1 and DWC2.The 100 kts MIR governs the high time metric area of the contour. +Maneuver Initiation RangeA turn rate of 7 deg/sec is considered suitable for the range of UAS speed (40 to 100 kts) and is thus assumed for modeling the UAS maneuver.This turn rate is higher than the standard turn rate of 3 deg/sec considered for Phase 1 UAS, which has a speed range of 40 to 200 kts.Increasing the turn rate beyond 7 deg/sec has diminishing returns in the UAS's ability to maintain WC 26 (meaning it hardly reduces the MIR further). +VI. Results +VI.A. Parameter RangesTable 3 lists ranges of DWC parameters considered for this work.No altitude separation other than h * = 450 ft is explored, because raising h * over the legal VFR traffic separation of 500 ft is not an option, and reducing h * will undesirably drive the required horizontal separation up (and the associated sensor range requirements).Although the VFR traffic from which intruder trajectories are sampled is similar in both sets of encounters, the UAS trajectories are different and, most likely, result in the slight difference of P between the two sets.For the remaining sections, only ACES results are discussed. +VI.C. Low C-SWaP UASFigures 7,8, and 9 show results of P and MIR for each DWC type.For each DWC type, a DWC that yields the minimum MIR on the contour of 5% P is selected.Such a DWC is marked with a red asterisk.The MIR for DWC1 and DWC2 becomes independent of τ mod or t pz when their values are small, because HMD * becomes the limiting parameter as the ownship must maneuver away to maintain the HMD * before it goes below the threshold of τ mod or t pz .While DWC1 and DWC2 contours are very similar, DWC2 yields marginally favorable MIR than DWC1.DWC3 results have already been shown in Figure 6 repeated here.Recall that DWC3's parameter space is a subset of DWC1, DWC2, and DWC4, when their respective time parameter thresholds reduce to zero.The points at which the contours intersect with the vertical axis in Figures 7,8, and 9 represent DWC3 results and are consistent with the values in Figure 6.DWC4 with t * > 0 leads to unfavorable MIR, and the minimum MIR occurs at t * = 0. +VII. DWC Down-SelectionThe DWCs that yield minimum MIR with 5% P are considered as primary candidate DWCs going forward for additional analysis.In addition, it is desirable to select a couple of secondary (backup) candidate DWCs with P under and above 5%, because of the following considerations.The target value of 5% for P is based on a open-loop risk recommendation, not from safety metric evaluation (which will be follow-up work).Without additional analysis, there is no guarantee that 5% for P is low enough.On the other hand, the DAA system might be able to achieve required safety with even higher values of P .Such an option should be explored as an opportunity to enable more UAS operations (with reduced sensor requirements).Table 4 shows the potential candidate DWCs being considered.The first two definitions are from the DWC1 and DWC2 types that yield the minimum MIR on the 5% contour of P .The third DWC is of the DWC3 type, a cylinder.This DWC is simple since it does not have a time metric.DWC4 is not selected because the added time component worsens MIR performance.Although the second DWC provides the minimum MIR, and it can be argued that t pz serves as a more intuitive time metric for DAA, the second DWC be dropped due to the SC-228 being less familiar with the behavior of t pz .The first and third DWCs are selected as the primary DWCs for future analysis.The fourth and fifth DWCs 28 are two candidate DWCs proposed for terminal area UAS operations.The fourth DWC has been accepted by SC-228 as the terminal WC.However it is fairly small and results in 10.0% of P, deemed too high by the authors.The sixth DWC arises from the authors' consideration to carry forward a DWC with P lower than 5%.This DWC is selected to have a lower P of 3.7% while raising the MIR slighly to 2.3 nmi.The fifth and sixth DWCs are selected as the secondary candidate DWCs.The four candidate DWCs are shown on Figure 10 as red (primary) and blue (secondary) asterisks in terms of the DWC1 parameter, HMD and τ mod . +VIII. Conclusion and Future WorkThis analysis evaluated potential Detect-and-Avoid (DAA) Well Clear (DWC) using two performance metrics, the unmitigated collision risk P and the maneuver initiation range (MIR).Two sets of encounters, one from projected UAS and historical VFR traffic and another from the MIT Lincoln Laboratory uncorrelated encounter model, were used to compute the unmitigated collision risk.The MIR was computed with the anticipated aircraft speed range and turn rate.Based on reasonable values of P and minimum MIR among other considerations, a set of four candidate DWCs, two primary and two secondary, are selected.Additional analyses will be performed to further differentiate these four candidate DWCs.The final DWC is expected to be applicable to Phase 1 UAS as well, when encountering non-cooperative aircraft.These additional analyses include:• A trade study between the required sensor surveillance volume and alerting timeline;• A closed-loop study (using a DAA algorithm to maneuver the UAS away from the intruder) to compute safety metrics such as the risk ratio.The ultimate goal of this line of research is to provide supporting information for the requirements of low C-SWaP sensors as well as that of alerting and guidance for UAS in encounters with non-cooperative aircraft.* +Figure 1 .1Figure 1.A schematic representation of the DWC zone. +Figure 2 .2Figure 2. Workflowshown at the top of this diagram.Each DWC is defined by a set of threshold parameters.Encounters between UAS and non-cooperative aircraft, generated from simulations or models with representative distributions, are analyzed to estimate P .The simulation suites for generating encounters are the Airspace Concept Evaluation System (ACES)17 and MIT Lincoln Laboratory's uncorrelated encounter model.18Target values of P (NMAC|LoDWC) are applied to down-select DWC definitions.The 2degree-of-freedom Prototyping Airplane Interaction Research Simulation (2PAIRS)19 is used to compute the MIR.Section IV describes the simulation setup in details.A few variables that set the scope of the low C-SWaP operations considered, and likewise the scope of the encounter data considered, are given here:• Extended UAS operations in non-terminal Classes D, E, and G airspaces, as well as those occurring during transit UAS operations in Classes B and C airspaces, are considered. +a) UAS Speed Distribution c) VFR Speed Distribution b) UAS Altitude Distribution d) VFR Altitude Distribution +Figure 3 .3Figure 3. Speed and altitude distribution of UAS and VFR traffic +Figure 4 .4Figure 4. Head-on encounter and ownship maneuver considered for an MIR. +Figure 5 Figure 5 .55Figure5shows results of P computed from ACES encounters and from the encounter model.The computation is based on DWC1.The contours are produced by MATLAB from values of P at grid points.The contours agree well, with a difference of 0.1% or less for τ mod < 25.The difference increases slightly for larger values of τ mod . +Figure 6 shows6Figure 6 shows P computed using DWC3.The encounter model yields slightly higher values of P, although the values are very close. +Figure 6 .6Figure 6.P for DWC3 from ACES encounters and the encounter model. +Figure 7 .Figure 8 .78Figure 7. P and MIR for DWC1.The minimum MIR at 5% P is about 1.8 nmi at 15 sec τ mod and 2000 ft HMD. +Figure 9 .9Figure 9. P and MIR for DWC4.The minimum MIR at 5% P is about 2.2 nmi at 0 sec t * and 2200 ft r * . +Figure 10 .10Figure 10.The four candidate DWCs (red for primary and blue for secondary) +Table 1 .1Types of DWC consideredTypeParametersLoDWC ConditionCommentDWC1 h* , HMD * , τ mod * h < h * and HMD < HMD * and τ mod < τ mod * Phase 1 DWC DWC2 h * , HMD * , tpz * h < h * and HMD < HMD * and tpz < tpz * Alternative time metric DWC3 h * , r * h < h * and r < r * Static cylinder DWC4 h * , r * , t * h < h * and r < r * -min( ṙ, 0) × t * Dynamic cylinder +Table 2 .2Missions for UAS with low C-SWaP sensorsMission TypeUAS TypeCruise Altitude (AGL)Cruise Speed (KTAS)Flight PatternAerial Imaging and MappingAerosonde Mk 4.73000 ft.44 to 51Radiator-grid pattern or circular patternAir Quality MonitoringShadow(RQ-7B)/ NASA SIERRA 234k, 5k, and 6k ft.74 to 89Radiator-grid patternAirborne Pathogen TrackingShadow(RQ-7B)/ NASA SIERRA3,000 ft., 10,000 ft. 5,000 ft. and72 to 97Radiator-grid patternFlood Inundation MappingAerosonde Mk 4.74,000 ft.46 to 51Grid patternFlood Stream FlowAerosonde Mk 4.74,000 ft.46 to 51Grid pattern and/or along stream directionThree types of pattern:Law EnforcementAerosonde Mk 4.73,000 ft.44 to 511) grid pattern, 2) random,3) outward spiralPoint Source EmissionShadow(RQ-7B)3,000 ft.72 to 80Grid pattern and/or along stream directionSpill MonitoringShadow(RQ-7B)/ SIERRA3,000 ft. to 13k ft. (at every 1,000 ft.)72 to 93Up and down-wind flights radiator-grid pattern, round-the-clockTactical Fire MonitoringScanEagle/ Shadow(RQ-7B)3,000 ft.72 to 75Circular flight path following the perimeter of a wildfireTraffic MonitoringShadow(RQ-7B)/ SIERRA1,500 ft.58 to 84Geo-spatial monitoring flight pathWildlife MonitoringAerosonde Mk 4.73,000 ft.44 to 51Radiator-grid patternNews GatheringAerosonde Mk 4.71,500 ft. to 3,000 ft.44 to 51Random-path: e.g., police-chase; circular orbit: +Table 3 .3Parameter space considered for DWCsTypeParameter SpaceDWC1 h * = 450 ft; HMD * ∈ [1000, 6000] ft; τ mod* ∈ [0, 35] secDWC2 h * = 450 ft; HMD * ∈ [1000, 6000] ft; t pz* ∈ [0, 35] secDWC3 h * = 450 ft; r * ∈ [1000, 10000]DWC4 h * = 450 ft; r * ∈ [1000, 6000] ft; t * ∈ [0, 35] sec +Table 4 .4DWC down-selectionTypeID HMD* (or r * ) τ * mod or t * pz or t * (ft) (sec)P (%) (nmi) MIRCommentSelectionDWC1120001551.8PrimaryDWC2219501551.8Not selectedDWC332200052.0simplePrimaryDWC341200010.01.25terminal WC 1 Not selectedDWC1515001571.7terminal WC 2SecondaryDWC162500253.62.3smaller PSecondary + + + + + + + + + Defining Well Clear for Unmanned Aircraft Systems + + StephenPCook + + + DallasBrooks + + + RodneyCole + + + DavisHackenberg + + + VincentRaska + + 10.2514/6.2015-0481 + AIAA-2015-0481 + + + AIAA Infotech @ Aerospace + + American Institute of Aeronautics and Astronautics + 2015 + + + Cook, S. P., Brooks, D., Cole, R., Hackenberg, D., and Raska, V., "Defining Well Clear for Unmanned Aircraft Systems," Proceedings of AIAA Infotech@ Aerospace, AIAA-2015-0481, AIAA, 2015. + + + + + Characteristics of a Well Clear Definition and Alerting Criteria for Encounters between UAS and Manned Aircraft in Class E Airspace + + MJohnson + + + ERMueller + + + CSantiago + + + + Europe Air Traffic Management Research and Development Seminar + + + 2015 + + + Eleventh UAS + Johnson, M., Mueller, E. R., and Santiago, C., "Characteristics of a Well Clear Definition and Alerting Criteria for Encounters between UAS and Manned Aircraft in Class E Airspace," Eleventh UAS/Europe Air Traffic Management Research and Development Seminar , 2015, pp. 23-26. + + + + + FAA Position on Building Consensus Around the SARP Well-Clear Definition + + DWalker + + + + RTCA Special Committee + + 228 + 2014 + + + Walker, D., "FAA Position on Building Consensus Around the SARP Well-Clear Definition," RTCA Special Committee 228 , 2014. + + + + + Flight Test Overview for UAS Integration in the NAS Project + + JamesRMurphy + + + PeggySWilliams-Hayes + + + SamKKim + + + WayneBridges + + + MichaelMarston + + 10.2514/6.2016-1756 + AIAA-2016-1756 + + + AIAA Atmospheric Flight Mechanics Conference + + American Institute of Aeronautics and Astronautics + 2016 + + + AIAA SciTech + Murphy, J. R., Hayes, P. S., Kim, S. K., Bridges, W., and Marston, M., "Flight Test Overview for UAS Integration in the NAS Project," AIAA Atmospheric Flight Mechanics Conference, AIAA SciTech, AIAA-2016-1756, 2016. + + + + + Evaluating Alerting and Guidance Performance of a UAS Detect-And-Avoid System + + SMLee + + + CPark + + + DPThipphavong + + + DRIsaacson + + + CSantiago + + NASA/TM-2016-219067 + + 2016 + NASA Ames Research Center + + + Lee, S. M., Park, C., Thipphavong, D. P., Isaacson, D. R., and Santiago, C., "Evaluating Alerting and Guidance Performance of a UAS Detect-And-Avoid System," NASA/TM-2016-219067, NASA Ames Research Center, 2016. + + + + + Analysis of alerting performance for detect and avoid of unmanned aircraft systems + + SamanthaSmearcheck + + + SeanCalhoun + + + WilliamAdams + + + JaredKresge + + + FabriceKunzi + + 10.1109/plans.2016.7479766 + + + 2016 IEEE/ION Position, Location and Navigation Symposium (PLANS) + + IEEE + 2016 + + + + Smearcheck, S., Calhoun, S., Adams, W., Kresge, J., and Kunzi, F., "Analysis of Alerting Performance for Detect and Avoid of Unmanned Aircraft Systems," IEEE/ION Position, Location and Navigation Symposium (PLANS), 2016, pp. 710-730. + + + + + An Evaluation of Detect and Avoid (DAA) Displays for Unmanned Aircraft Systems: The Effect of Information Level and Display Location on Pilot Performance + + LisaFern + + + RConradRorie + + + JessicaPack + + + JayShively + + + MarkDraper + + 10.2514/6.2015-3327 + AIAA-2015-3327 + + + 15th AIAA Aviation Technology, Integration, and Operations Conference + + American Institute of Aeronautics and Astronautics + 2015 + + + Fern, L., Rorie, R. C., Pack, J. S., Shively, R. J., and Draper, M. H., "An Evaluation of Detect and Avoid (DAA) Displays for Unmanned Aircraft Systems: The Effect of Information Level and Display Location on Pilot Performance," Proceedings of 15th AIAA Aviation Technology, Integration, and Operations Conference, AIAA-2015-3327, 2015. + + + + + The Generic Resolution Advisor and Conflict Evaluator (GRACE) for Detect-And-Avoid (DAA) Systems + + MichaelAbramson + + + MohamadRefai + + + ConfesorSantiago + + 10.2514/6.2017-4485 + AIAA-2017-4485 + + + 17th AIAA Aviation Technology, Integration, and Operations Conference + + American Institute of Aeronautics and Astronautics + June 2017 + + + Abramson, M., Refai, M., and Santiago, C., "A Generic Resolution Advisor and Conflict Evaluator (GRACE) in Ap- plications to Detect-And-Avoid (DAA) Systems of Unmanned Aircraft," Proceedings of the 17th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, AIAA-2017-4485, June 2017. + + + + + DAIDALUS: Detect and Avoid Alerting Logic for Unmanned Systems + + CesarMunoz + + + AnthonyNarkawicz + + + GeorgeHagen + + + JasonUpchurch + + + AaronDutle + + + MariaConsiglio + + + JamesChamberlain + + 10.1109/dasc.2015.7311421 + + + 2015 IEEE/AIAA 34th Digital Avionics Systems Conference (DASC) + + IEEE + 2015 + + + + Muñoz, C., Narkawicz, A., Hagen, G., Upchurch, J., Dutle, A., Consiglio, M., and Chamberlain, J., "DAIDALUS: Detect and Avoid Alerting Logic for Unmanned Systems," 34th Digital Avionics Systems Conference (DASC), IEEE/AIAA, 2015, pp. 5A1-1. + + + + + Development, Integration and Testing of a Stand-alone CDTI with Conflict Probing Support + + BrandonSuarez + + + KevinKirk + + + ErikTheunissen + + 10.2514/6.2012-2487 + + + Infotech@Aerospace 2012 + + American Institute of Aeronautics and Astronautics + 2012 , 2012, p. 2487. 2017. 2017 + + + 11 Minimum Operational Performance Standards (MOPS) for Detect and Avoid (DAA) Systems, DO-365 + Suarez, B., Kirk, K., and Theunissen, E., "Development, Integration and Testing of a Stand-Alone CDTI with Conflict Probing Support," Infotech@ Aerospace 2012 , 2012, p. 2487. 11 Minimum Operational Performance Standards (MOPS) for Detect and Avoid (DAA) Systems, DO-365, RTCA. Inc., 2017. 12 Minimum Operational Performance Standards (MOPS) for Air-to-Air Radar for Traffic Surveillance, DO-366, RTCA. Inc., 2017. + + + + + Practice for Application of Federal Aviation Administration (FAA) Federal Aviation Regulations Part 21 Requirements to Unmanned Aircraft Systems (UAS) + 10.1520/f2505-07 + + + Federal Aviation Administration (FAA) + + ASTM International + Feb. 2011 + + + HQ-111358 + "Introduction to TCAS II Version 7.1," HQ-111358, Federal Aviation Administration (FAA), Feb. 2011. + + + + + UAS Demand Generation Using Subject Matter Expert Interviews and Socio-economic Analysis + + SricharanKAyyalasomayajula + + + RohitSharma + + + FrederickWieland + + + AntonioTrani + + + NicolasHinze + + + ThomasSpencer + + 10.2514/6.2015-3405 + AIAA-2015- 3405 + + + 15th AIAA Aviation Technology, Integration, and Operations Conference + + American Institute of Aeronautics and Astronautics + 2015 + + + Ayyalasomayajula, S., Sharma, R., Wieland, F., Trani, A., Hinze, N., and Spencer, S., "UAS Demand Generation Using Subject Matter Expert Interviews and Socio-Economic Analysis," Proceedings of the AIAA Aviation Conference, AIAA-2015- 3405, 2015. + + + + + Establishing a Risk-Based Separation Standard for Unmanned Aircraft Self Separation + + RolandWeibel + + + MatthewEdwards + + + CarolineFernandes + + 10.2514/6.2011-6921 + + + 11th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + + American Institute of Aeronautics and Astronautics + 2011 + 16 + + + Ninth USA + Weibel, R. E., Edwards, M. W., and Fernandes, C. S., "Establishing a Risk-Based Separation Standard for Unmanned Aircraft Self Separation," Ninth USA/Europe Air Traffic Management Research & Development Seminar , 2011. 16 + + + + + An Assessment of Unmanned Aircraft System Level-Turn Maneuver Performance Requirements In Relation to a Quantified Well-Clear Definition + + DevinPJack + + + KeithDHoffler + + + SallyJohnson + + 10.2514/6.2015-2394 + AIAA-2015-2394 + + + AIAA Atmospheric Flight Mechanics Conference + + American Institute of Aeronautics and Astronautics + June 2015 + + + Jack, D. P., Hoffler, K. D., and Johnson, S., "An Assessment of Unmanned Aircraft System Level-Turn Maneuver Performance Requirements in Relation to a Quantified Well-Clear Definition," 14th AIAA Atmospheric Flight Mechanics Conference, AIAA-2015-2394, June 2015. + + + + + Build 8 of the Airspace Concept Evaluation System + + SapaGeorge + + + GoutamSatapathy + + + VikramManikonda + + + KeePalopo + + + LarryMeyn + + + ToddLauderdale + + + MichaelDowns + + + MohamadRefai + + + RichardDupee + + 10.2514/6.2011-6373 + AIAA-2011-6373 + + + AIAA Modeling and Simulation Technologies Conference + + American Institute of Aeronautics and Astronautics + Aug. 2011 + + + George, S. E., Satapathy, G., Manikonda, V., Refai, M., and Dupee, R., "Build 8 of the Airspace Concept Evaluation System," AIAA Modeling and Simulation Technologies Conference, AIAA-2011-6373, Aug. 2011. + + + + + Extended Airspace Encounter Models for Unmanned Aircraft Sense and Avoid Safety Evaluation + + AndrewJWeinert + + + EricHarkleroad + + + JohnGriffith + + + MatthewWEdwards + + 10.2514/6.2013-5049 + + + AIAA Infotech@Aerospace (I@A) Conference + Lexington, Massachusetts + + American Institute of Aeronautics and Astronautics + Aug. 2013 + 404 + + + MIT Lincoln Laboratory + + + Tech. rep + Weinert, A. J., Harkleroad, E. P., Griffith, J., Edwards, M. W., and Kochenderfer, M. J., "Uncorrelated Encounter Model of the National Airspace System, Version 2.0," Tech. rep., MIT Lincoln Laboratory, Lexington, Massachusetts, ATC-404, Aug. 2013. + + + + + Exploration of the Trade Space Between Unmanned Aircraft Systems Descent Maneuver Performance and Sense-and Avoid System Performance Requirements + + DevinPJack + + + KeithDHoffler + + 10.2514/6.2014-2288 + + + 14th AIAA Aviation Technology, Integration, and Operations Conference + + American Institute of Aeronautics and Astronautics + May 2014. 20 + + + NASA CR-2014-218264 + Jack, D. P., Hoffler, K. D., and Johnson, S., "Exploration of the Trade Space Between UAS Maneuver Performance and SAA System Performance Requirements," NASA CR-2014-218264, NASA Langley Research Center, May 2014. 20 + + + + + A Bayesian Approach to Aircraft Encounter Modeling + + MykelKochenderfer + + + JamesKuchar + + + JGriffith + + + LeoEspindle + + 10.2514/6.2008-6629 + + + AIAA Guidance, Navigation and Control Conference and Exhibit + Lexington, Massachusetts + + American Institute of Aeronautics and Astronautics + 2008 + + + MIT Lincoln Laboratory + + + Tech. rep + Project Report ATC-345 + Kochenderfer, M. J., Kuchar, J. K., Espindle, L. P., and Griffith, J., "Uncorrelated Encounter Model of the National Airspace System, Version 1.0," Tech. rep., MIT Lincoln Laboratory, Lexington, Massachusetts, Project Report ATC-345, 2008. + + + + + An Alternative Time Metric to Modified Tau for Unmanned Aircraft System Detect And Avoid + + MinghongGWu + + + VibhorLBageshwar + + + EricAEuteneuer + + 10.2514/6.2017-4383 + AIAA-2017-4383 + + + 17th AIAA Aviation Technology, Integration, and Operations Conference + + American Institute of Aeronautics and Astronautics + June 2017 + + + Wu, M. G., Bageshwar, V. L., and Euteneuer, E. A., "An Alternative Time Metric to Modified Tau for Unmanned Aircraft System Detect And Avoid," 17th AIAA Aviation Technology, Integration, and Operations Conference, AIAA-2017-4383, June 2017. + + + + + BADA: An advanced aircraft performance model for present and future ATM systems + + AngelaNuic + + + DamirPoles + + + VincentMouillet + + 10.1002/acs.1176 + + + International Journal of Adaptive Control and Signal Processing + Int. J. Adapt. Control Signal Process. + 0890-6327 + + 24 + 10 + + 2010 + Wiley + + + Nuic, A., Poles, D., and Mouillet, V., "BADA: An Advanced Aircraft Performance Model for Present and Future ATM Systems," Internaitonal Journal of Adaptive Control and Signal Processing, Vol. 24, No. 10, 2010, pp. 850 866. 23 + + + + + SIERRA Team Flight of Zephyr UAS at West Virginia Wild Land Fire Burn + + RobertCharvat + + + RodgerOzburn + + + ScottBushong + + + KellyCohen + + + ManishKumar + + 10.2514/6.2012-2544 + + + Infotech@Aerospace 2012 + + American Institute of Aeronautics and Astronautics + 2012. 2012-2544, 2012 + 2544 + + + Charvat, R., Ozburn, R., Bushong, S., Cohen, K., and Kumar, M., "SIERRA Team Flight of Zephyr UAS at West Virginia Wild Land Fire Burn," Infotech@ Aerospace 2012 , AIAA-2012-2544, 2012, p. 2544. + + + + + The Generic Resolution Advisor and Conflict Evaluator (GRACE) for Detect-And-Avoid (DAA) Systems + + MichaelAbramson + + + MohamadRefai + + + ConfesorSantiago + + 10.2514/6.2017-4485 + NASA/TM-2017-219507 + + + 17th AIAA Aviation Technology, Integration, and Operations Conference + + American Institute of Aeronautics and Astronautics + 2017 + + + Abramson, M., Refai, M., and Santiago, C., "The Generic Resolution Advisor and Confict Evaluator (GRACE) for Unmanned Aircraft Detect-And-Avoid Systems," NASA/TM-2017-219507, NASA Ames Research Center, 2017. + + + + + A Bayesian Approach to Aircraft Encounter Modeling + + MykelKochenderfer + + + JamesKuchar + + + JDGriffith + + + LeoEspindle + + 10.2514/6.2008-6629 + + + AIAA Guidance, Navigation and Control Conference and Exhibit + + American Institute of Aeronautics and Astronautics + 2008 + 17 + + + Kochenderfer, M. J., Espindle, L. P., Kuchar, J. K., and Griffith, J. D., "A Comprehensive Aircraft Encounter Model of the National Airspace System," Lincoln Laboratory Journal, Vol. 17, No. 2, 2008. + + + + + Analysis of Influence of UAS Speed Range and Turn Performance on Detect and Avoid Sensor Requirements + + DevinPJack + + + JeremyHardy + + + KeithDHoffler + + 10.2514/6.2018-3507 + AIAA-2018 + + + 2018 Aviation Technology, Integration, and Operations Conference + + American Institute of Aeronautics and Astronautics + June 2018 + + + Jack, D. P., Hardy, J., and Hoffler, K. D., "Analysis of Influence of UAS Speed Range and Turn Performance on Detect and Avoid Sensor Requirements," 18th AIAA Aviation Technology, Integration, and Operations Conference, AIAA-2018, June 2018. + + + + + Sensitivity Analysis of Detect and Avoid Well Clear Parameter Variations on UAS DAA Sensor Requirements + + JeremyHardy + + + DevinPJack + + + KeithDHoffler + + 10.2514/6.2018-3505 + AIAA-2018 + + + 2018 Aviation Technology, Integration, and Operations Conference + + American Institute of Aeronautics and Astronautics + June 2018 + + + Hardy, J., Jack, D. P., and Hoffler, K. D., "Sensitivity Analysis of Detect and Avoid Well Clear Parameter Variations on UAS DAA Sensor Requirements," 18th AIAA Aviation Technology, Integration, and Operations Conference, AIAA-2018, June 2018. + + + + + A Recommended DAA Well-Clear Definition for the Terminal Environment + + MichaelJVincent + + + AnnaTrujillo + + + DevinPJack + + + KeithDHoffler + + + DimitriosTsakpinis + + 10.2514/6.2018-2873 + AIAA-2018 + + + 2018 Aviation Technology, Integration, and Operations Conference + + American Institute of Aeronautics and Astronautics + June 2018 + + + Vincent, M., Trujillo, A., Jack, D., Hoffler, K., and Tsakpinis, D., "A Recommended DAA Well-Clear Definition for the Terminal Environment," 18th AIAA Aviation Technology, Integration, and Operations Conference, AIAA-2018, June 2018. + + + + + + diff --git a/file794.txt b/file794.txt new file mode 100644 index 0000000000000000000000000000000000000000..d16c1654677a09375474bbb9a7dca700bc5cbd7c --- /dev/null +++ b/file794.txt @@ -0,0 +1,595 @@ + + + + +for JFK to those for DFW and investigates major factors that contributed to their differences; and, finally, Section VII summarizes the findings in this work. +II. Three FPA Selection StrategiesIn previous work, three strategies were proposed to define the fixed FPA for small jet arrivals under metering conditions. 9,10 he three strategies were as follows:1. Universal FPA -defines an FPA for all flights 2. Descent-Speed FPA -defines the FPA as a function of the descent speed 3. Min-Fuel FPA -computes a custom FPA for each flight Universal FPA defines a universal FPA for all small jets arriving to an airport, gate, or route.This is akin to a glide slope extending from the first altitude constraint on the Standard Terminal Arrival Route (STAR) back up to the top of descent (TOD) in the en-route airspace.The advantage of Universal FPA is its simple form, which allows it potentially to be published as part of the arrival procedure.The disadvantage is that it does not account for the effects of descent speed and winds on the fuel efficiency or the range of flyable FPAs.Descent-Speed FPA takes into account the effects of the descent speed on the range of fuel-efficient and flyable FPAs.During periods of high density traffic, controllers would issue speed clearances for each arrival flight, and EDA supports clearances that give both cruise and descent speeds.One the descent speed is relevant to the selection of FPA.This FPA function reduces the FPA by 0.1 • for every 10 knot increase of descent calibrated airspeed (CAS).Note the FPA is negative for descents, and reducing the FPA makes it steeper.It was observed that this rate of change of FPA with respect to the descent CAS captures the variation of fuel-efficient FPAs with speed. 10The change of 0.1 • matches the precision of FPA-selection increments typical of small jet avionics. 11The FPA function is defined by Equation 1:γ =              γ 0 , if DCAS < 255 γ 0 -0.1 , if 255 ≤ DCAS < 265 γ 0 -0.2 , if 265 ≤ DCAS < 275 . . . ,(1)where γ is the FPA in degrees, DCAS is the descent CAS in knots, and the adaptive parameter γ 0 stands for the value of γ at 250 knots descent CAS, a typical speed expected at the metering fix.Note the selection of the parameter γ 0 defines the entire FPA function.Descent-Speed FPA is expected to be more fuel-efficient than Universal FPA by accounting for the descent speed.Both Universal FPA and Descent-Speed FPA have the disadvantage that they do not account for the effects of the direction of the winds aloft on the range of fuel-efficient and flyable FPAs.While Universal FPA and Descent-Speed FPA define the FPA and the FPA function ahead of time, Min-Fuel FPA computes the minimum-fuel FPA to be communicated explicitly to the pilot of each flight just prior to the TOD.The computation is based on the route, winds and temperature aloft, and the speed profile necessary to meet the STA. Figure 1 sketches the fuel burn as a function of the descent FPA for a typical flight.While the requirement of communicating the FPA to the pilot in real time may make Min-Fuel FPA infeasible to implement without controller/pilot data-link, it serves as a reference point for fuel burn comparison with the first two strategies. +II.A. Planned Speed-Brake UsageIn addition to predicted fuel burn, the selection of FPA also takes into account uncertainties in vertical profile planning.To keep the aircraft on the planned vertical path, power adjustment is preferred to speedbrake deployment for reasons of passenger comfort and the desire of pilots to reserve the use of a speed brake for rare occasions.Therefore, the analysis of trajectories was limited to those FPAs that do not require planned speed-brake usage.The universally fixed FPA in Universal FPA, denoted as γ univ , and the γ 0 in Descent-Speed FPA are parameters that must be selected carefully, with consideration for the prevailing winds aloft and, to a lesser extent, the anticipated traffic demand.The following selection criteria are applied: 9,10 1.For the parameter considered, at least 99% of the flights have feasible trajectories, meaning trajectories that have speeds within the performance envelope and do not require speed brake usage.2. The parameter should result in the least average fuel burn per flight.The first criterion defines the steepest parameter that can be selected, and the second criterion selects a minimum-fuel FPA from those parameters no steeper than the steepest parameter allowed by the first criterion. +II.C. Adaptation of Universal FPA and Descent-Speed FPAWhile steeper FPAs are typically more fuel-efficient for arrival flights in a headwind, they can be unflyable for flights in the opposite direction.A shallow FPA would guarantee flyability for both directions, but will be fuel-inefficient for flights in a headwind.These observations motivated the adaptation of γ univ and γ 0 to the direction of arrival.Compared to a static implementation, adaptation to the direction of arrival achieves a greater degree of "customization" and can improve fuel efficiency.While adaptation to the direction of arrival reduces the variation of along-track winds due to directions, it does not mitigate the variation of winds over time.The fuel efficiency of the Universal FPA and Descent-Speed FPA strategies can be further improved by adapting γ univ and γ 0 to seasonal norms, monthly norms, or even daily predictions.A set of systematic, temporal and location/directional adaptations was proposed in previous work 9 and reviewed here in Table 1.The columns represent different levels of adaptation for different airspaces, starting with a basic "one size fits all" adaptation for all airports across the National Airspace System (NAS).Moving to the right, each column represents a progressively finer adaptation of the FPA strategy to a specific airport, individual arrival gates (corner posts) feeding an airport, all the way down to specific arrival routes feeding each arrival gate.The rows represent a temporal scale starting at the top with the simplest option of a static adaptation.Moving down, each row represents a progressively finer adaptation to account for changes in the prevailing winds as a function of season, month, day or even hour.The table illustrates the overall approach and potential scope.For the purposes of this paper, the analysis will assess the eight types highlighted and numbered in the table.The analysis of adaptations at the NAS-wide level, or at the level of specific arrival routes and/or hours of the day, are left for future work. +III. Selection of Case Study: JFKThe selection of John F. Kennedy Airport (JFK) was motivated by multiple factors.The major factor was the search for an airport that has strong wind variations.Wind along the route acts as a major discriminator in selecting the descent FPA for an arrival flight.Stronger variation of winds among directions of arrival increase the benefits of adapting to the directions of arrival.Stronger variation of winds across timespans increases the fuel-burn merit of adapting the FPA selection to season, month, and day.Previous analysis of the DFW airport showed fairly strong seasonal variation of winds.However, other major airports in the United States may have even stronger directional and temporal variation of winds, and JFK was expected to be an example of such.To compare the variation of winds along the route among some major airports in the United States, winds along the route were estimated for four hypothetical arrival routes constructed for each airport.Four hypothetical points in space at four corners around an airport, NE, NW, SE, and SW, were selected at 150 nmi each from the airport and at 35,000 ft in altitude.The vectors connecting each point to the airport defines the four hypothetical directions of arrival.Wind components along the vectors were estimated using the two-hour, 40 km Rapid Update Cycle (RUC) weather forecast 12 for year 2011.The wind forecast was generated on an hourly basis, amounting to more than 8660 forecast winds (a few days of RUC were not available).Two quantities were calculated from the wind components to represent the wind variations across directions and timespans.Let N r denote the number of available RUC wind estimates.The standard deviation of the average wind along each of the four directions is defined by Equation 2:σ dir 2 = 1 4 4 i=1 (W i -µ) 2 , where µ = 1 4 4 i=1 W i .(2)Here W i is the average of the N r wind components along a direction i, defined by Equation 3:W i = 1 N r N r j=1 W i,j ,(3)and W i,j denotes the wind component estimated using RUC wind file j along direction i.The average of Copyright © 2013 by the American Institute of Aeronautics and Astronautics, Inc.The U.S. Government has a royalty-free license to exercise all rights under the copyright claimed herein for Governmental purposes.the standard deviation of wind for each direction is defined by Equation 4:σ time 2 = 1 4 4 i=1 σ i 2 , where σ i 2 = 1 N r N r j=1 (W i,j -W i ) 2 .(4)Note that, although only standard deviations were calculated, the wind distribution may not be exactly Gaussian and outliers may have effects on the fuel merits of the FPA selection strategies.Further investigation of the wind distribution is left for future work.While σ dir is indicative of the benefits of adapting the descent FPA to flight directions, σ time is indicative of the benefits of adapting the descent FPA to timespans.Figure 2 from east to west and from north to south.JFK has the highest variation of both types and hence was chosen as the subject of this analysis.LAX and MIA experience significantly weaker winds and can be the subject of another study.The airport analyzed in previous work, DFW, ranked 8th in σ dir and 11th in σ time among the twelve airports.The rank for σ time was a bit surprising since the jet stream passed through DFW in winter and resulted in clear seasonal variations.Nonetheless, the definition of σ time accounts for wind variation on a per-hour basis and the seasonal variation was not directly captured by this definition.Figure 3 shows the wind components along the four hypothetical directions of arrival leading to JFK as a function of time of year, 2011.While flights from the NW and SW clearly would experience tailwind most of the time, flights from the NE and SE would experience mostly headwinds.Extremely strong winds were observed for SW and NE during some early hours of February 7, when the SW direction had up to 166 knots of tailwind and the NE had up to 172 knots of headwind.It will be shown later that these strong winds on this day had interesting effects on the selected FPA and FPA function.Two of the four STARs leading to JFK have restrictions at high altitudes due to the congested airspace in the New York metroplex area.Figure 4 shows arrival tracks of small jets on January 7th, 2011.Each arrival jet aircraft enters the Terminal Radar Approach Control Facilities (TRACON) through one of the four arrival STARs, characterized by the waypoints CCC, CAMRN, HARTY, and LOLLY, respectively.These four points are the most upstream points in the STARs that have altitude restrictions, which are summarized in Table 3.Note that the speed restrictions are defined in terms of calibrated airspeed (CAS).Flights from the N and NW must capture the high-altitude restrictions of 20,000 and 21,000 ft, respectively, when they cross LOLLY and HARTY.Additional downstream altitude and speed restrictions apply, making the vertical profile beyond LOLLY and HARTY very constrained.For this work, only the portion of the flight trajectory from cruise to one of these four points is considered for varying the FPA.This reduction in descent altitude change is expected to diminish the fuel burn differences observed among different values of the descent FPA, and hence diminish the difference in the relative fuel-burn merits of the three selection strategies.Unlike DFW studied in prior work, which has two-thirds of its small jet arrivals from the east, 70% to 80% of JFK's small jet arrivals are from the west, in directions of NW, W, or SW.Although JFK has a small number of arrival flights from NE, very few small jet flights approach from SE, which is over the Atlantic Ocean.Since the fuel burn of the trajectory is less sensitive to the selected FPA in the headwind than in the tailwind, 10 the difference in the distribution of directions of arrival between JFK and DFW will change the relative fuel-burn merits of the three strategies.Besides, the uneven distribution of the directions of arrival for JFK may diminish the benefits of adapting the selection of descent FPAs to the flight directions when compared to DFW. +IV. Analysis ApproachThe methodologies for selecting the FPAs and the modeling schemes have been described in detail previously. 9The following sections briefly review them for completeness of this paper. +IV.A. Calibrating and Comparing the Three StrategiesThe methodology that compares the benefits of the three strategies consists of two conceptual parts: first select the parameters for the Universal FPA and Descent-Speed FPA strategies, and then compare the fuel burn between the three strategies for a set of traffic and wind conditions.The first part of the methodology determines the parameters of Universal FPA and Descent-Speed FPA using estimates of winds and temperatures aloft, traffic demand, and modeled metering delays.The second part of the methodology estimates and compares the fuel burn of the strategies, using the estimates of winds and temperatures aloft, actual traffic data, and modeled or actual metering delays.While the estimates of winds, temperatures aloft, and traffic demand for the second part need not be the same as that used for the first step, they are the same in this analysis, being the RUC data and the track data of JFK recorded in 2011.Since the same estimates of winds and temperatures aloft and traffic demand were used in both parts of the methodology, the fuel burn and planned speed-brake usage recorded in the first part for selecting the parameters of γ univ and γ 0 can be directly used in the second part without having to recalculate them.The implementation of the methodology combines the two conceptual parts into one run of a fast-time simulation as described in detail in the following sections.For each arrival flight and a modeled metering delay, the analysis computes a set of meet-time trajectories with varying FPA and descent CAS profile.Because the FPA selected by each strategy must be from these trajectories, the analysis of these trajectories was sufficient for the fuel burn comparison between the three strategies.For adaptations to longer timespans, the relative fuel-burn merits of the Universal FPA and Descent-Speed FPA strategies would be optimistic compared to those attainable in actual implementation.This is because the high-fidelity weather, and flight plan information accessible in real time to Min-Fuel FPA were used for selecting their parameters, while in actual implementation only crude forecasts are available for longer timespans. +IV.B. Route and Vertical ProfileTrajectories are constructed for arrival flights to JFK that enter the terminal airspace through one of the four points, CCC, LOLLY, HARTY, and CAMRN, shown in Figure 4, that were treated as "pseudo" metering fixes.While JFK currently does not meter the arrival flights, controllers issue speed clearances to sequence the arrival flights while maintaining their spacings during periods of high traffic density.Therefore, arrival flights into JFK can effectively be modeled as if they were in an metered environment.In this analysis, the terms "metering fix" and "gate" have a one-to-one mapping and are used interchangeably.A distance-based freeze horizon of 160 nmi was assumed, inside of which TMA would fix the STA for the aircraft. 13The selection of 160 nmi guaranteed that the TOD is inside the freeze horizon and each trajectory has cruise and descent phases.The initial condition was selected at a corresponding track point nearest the freeze horizon.For simplification, direct trajectories from the initial point to the metering fix were assumed without actually parsing the flight plans for the waypoints.The crossing restrictions shown in Table 3 for the four pseudo metering fixes were modeled at the end of the trajectories.Figure 5 illustrates a typical vertical profile that has five segment types distinguished by altitude changes and pilot procedures.Individual trajectories will contain all or a subset of these segments depending on the speed profile needed to meet the STA.Each segment is modeled kinetically by fixing two control parameters.One of the parameters is the FPA; the second depends on the segment.For a cruise segment the model fixes the airspeed or the engine control for acceleration or deceleration.For the constant-speed descent segments, the model fixes an airspeed in Mach or CAS. +IV.C. Aircraft and Fuel Burn ModelingThe Trajectory Synthesizer (TS) component 14,15 of the Center-TRACON Automation System (CTAS) 16 was used to compute trajectories, their associated fuel burn, and the planned speed-brake usage.While a detailed performance model of small jet types would have been desirable, one was not available.Instead, a high-fidelity, CTAS model for a mid-size, narrow-body, twin-jet airliner with a typical descent weight of 170,000 lbs was used.The speed envelopes were selected within the ranges of small jets.To capture variation of descent performances due to weight differences, the weight for each flight was selected randomly from a normal distribution with a standard deviation of 8,400 lbs.To account for the variation of the fuel-burn rate among aircraft types, the fuel burn was scaled by the empirical formulaf i = f 0 * N i + 30 230 ,(5)where f i is the scaled fuel burn; f 0 is the raw fuel-burn rate calculated by CTAS for the mid-size, narrowbody, twin-engine jet; and N i is the number of passenger seats typical of the aircraft type i.This empirical formula was derived by taking the linear regression of the nominal cruise fuel-burn rate of eight small jets plus the mid-sized twin engine jet, using the Base of Aircraft Database (BADA) 3.8 performance model. 17lthough BADA provides modeling parameters for small jets, the calibration of these parameters focused on nominal flight conditions only.Since the analysis in this work explored a wide range of the speeds, containing both nominal and off-nominal ones, it was decided that the high-fidelity CTAS model with scaled fuel burn was more appropriate for the fuel-burn analysis. +IV.D. Metering DelayThe delay at the metering fix for each flight was selected randomly from a uniform distribution between zero and the maximum delay that can be absorbed by speed reductions.The delay time was added to the nominal time in order to specify the STA at the metering fix for a flight.By definition, a flight with nominal cruise and descent speeds would arrive at the metering fix with zero delay.A flight with minimum cruise and descent speeds would arrive at the metering fix with maximum delay.Idle thrust descent was assumed for these two trajectories. +IV.E. Fixed-FPA Meet-Time TrajectoriesA set of meet-time trajectories with varying FPA and descent CAS profile was computed for each flight using a modeled metering delay.Only the meet-time trajectories that do not have planned usage of speed brakes are included for the analysis.Fuel burn was calculated for each meet-time trajectory.These meettime trajectories provided all the fuel-burn data needed for comparison of the three strategies, because each of the three strategies must select an FPA from these meet time trajectories.For descent FPAs ranging from -1.8 • to -5.5 • , with an increment of 0.1 • , a meet-time algorithm attempted to compute a fixed-FPA trajectory for each value of the FPA.The meet-time algorithm iterated cruise and descent speeds until the trajectory met the desired time-to-fly within a tolerance of 2 seconds.Cruise and descent speeds were related by the Cruise-Equals-Descent speed mode developed for EDA. 18The resulting meet-time trajectories have different combinations of FPA-descent-CAS pairs.All three strategies were applied to select FPAs.If no meet-time trajectories had the FPA defined by Universal FPAor satisfy the FPA-descent-CAS relationship defined by Descent-Speed FPA, a failure was recorded for this parameter of the strategy.The number of failures, usually resulting from speed brake usage or speeds going out of bounds, was used to determine whether the candidate parameters γ univ and γ 0 should be rejected (See Section II.B). +IV.F. Simulation and Data AnalysisA fast-time Monte Carlo simulation was performed to generate the meet-time trajectories for small jet arrivals to JFK.For each arrival flight, a descent weight and a delay time were sampled from the distributions described in Sections IV.C and IV.D. The meet-time algorithm computed a set of fixed-FPA meet-time trajectories for the specified test condition, using the RUC data in the TS calculation of the trajectories.The fuel burn and speed-brake usage were recorded for further analysis to be described below.All data were categorized by gates and days.The selection criteria described in Section II.B were used to select γ univ and γ 0 for each of the eight adaptation types listed in Section II.C.The analysis selected γ univ for Universal FPA from values between -1.8 • and -5.5 • with increments of 0.1 • .The analysis selected γ 0 for Descent-Speed FPA from values between -1.8 • and -3.7 • with increments of 0.1 • .The steepest parameters were selected such that all speed-brake-free trajectories in any wind condition were analyzed.The shallowest parameters were selected such that the route has enough path distances for the cruise phase.For each γ univ and each γ 0 , the average fuel burn per flight and feasibility rate were computed from results of all flights.The feasibility rate was defined as the ratio of the flights with flyable FPAs (total number of "success") to the total flights analyzed.It must be 99% or better for γ univ or γ 0 to be selected.For the Airport-Static adaptation, all flights were analyzed for the selection.For other adaptations, a subset of the flights was analyzed to select γ univ or γ 0 for a gate and/or a timespan.For example, a total of sixteen pairs of γ univ and γ 0 were selected for the Gate-Season adaptation (four gates times four seasons), each using the flights into a specific gate during a specific season.Selection of the parameters based on the feasibility rate ensures that the vast majority of flights will have flyable FPAs, but it does not consider the variation of winds that can make the flights through some gates on some days particularly difficult to fly.To ensure the feasibility rate for any given day and gate was "tolerable" for adaptations that have longer timespans or are airport specific, another feasibility rate of 80% or better for any pair of gate and day was required.Once the parameters were selected, each strategy would select an FPA for each flight.The extra fuel burn for the FPAs selected by Universal FPA and Descent-Speed FPA were computed for all flights as a metric for the relative fuel-burn merits between the three strategies. +V. ResultsResults of the fast-time simulation are presented as follows: Section V.A summarizes the statistics of the number and aircraft types of the arrival flights used in the simulation.The following sections discuss results for each strategy, starting from Min-Fuel FPA because it revealed distributions of the individually selected FPAs among flights.Section V.B shows the selected FPAs for flights using Min-Fuel FPA, the minimum-fuel strategy, and discusses their correlation with winds.Sections V.C and V.D show the selected FPA and FPA function for Universal FPA and Descent-Speed FPA, respectively, in the gate-specific adaptation types.Section V.E compares the fuel-burn merits of the three strategies. +V.A. Arrival FlightsTable 4 shows the total number of arrival flights of small jets identified and used in this analysis.Due to occasional periods when the data feed was unavailable, the track data was missing for 13 of the 365 days.Among the days where the track data was available, some data were not recorded due to short interruptions of the data feed.The CAMRN gate was the busiest, accounting for 36% of the small jet arrival flights.Figure 6 shows the most frequent small jet aircraft types observed among the 51,591 arrival flights.The Embraer ERJ 190 and Bombardier CRJ 200 accounted for more than 48% of the fleet, with the other 58 aircraft types making up the rest.E190 CRJ2 CRJ9 E145 E135 CRJ7 CRJ1 H25B C56X BE40 CL60 F2TH GLF4 LJ35 C750 E170 C560 C680 DC95 F900 CL30 +V.B. Min-Fuel FPAMin-Fuel FPA selects a minimum-fuel FPA for each flight.Figure 7 presents distributions of the FPAs selected for flights using Min-Fuel FPA on a per-day basis.Among the four gates, the steepest FPAs were selected for CCC, ranging from -2.9 • to -3.6 • in winter time and from -2.7 • to -3.2 • in summer time.The other three gates have shallower FPAs selected, with the fluctuations stronger for flights to LOLLY.This is most likely due to a wider variation of the headings of the arrival flights to LOLLY (See Figure 4).Since steeper FPAs are typically selected in the presence of strong headwinds, 9 the results suggest that flights into CCC experience mostly strong headwinds throughout the year.This is in strong correlation to the wind along the route for the NE direction of JFK shown in Figure 3.In addition, seasonal variation of the selected FPAs was clear for CCC in Figure 7.The same variation is also visible in the other three plots, although subject to more fluctuation.It is interesting to see that the very shallow -1.8 • was the minimum-fuel FPA for some flights, even in the presence of headwinds as for CCC and LOLLY on Feb. 7, 2011 (shown curve).This inversion of minimum-fuel FPA towards a shallow value can happen for certain combinations of headwinds along with favorable wind gradients with respect to altitude and has been discussed previously. 9 +V.C. Universal FPAUniversal FPA selects the FPA based on the average fuel burn per flight and the feasibility rate.Figure 8 shows values of γ univ selected for the gate-specific adaptations of Universal FPA.The airport-specific FPAs are not shown but their fuel burn will be analyzed in Section V.E.Compared to Min-Fuel FPA, similar trends were observed in the selected FPA using Universal FPA as the steepest FPAs were selected for CCC.The FPAs selected for HARTY are the shallowest of all gates.Regarding the Gate-Day adaptation on February 7, -1.9 • was selected for CAMRN while -1.8 • was selected for CCC.The former FPA was selected because any steeper FPA would result in too many speed-brake demanding trajectories while the latter was selected because it results in better fuel burn than any steeper FPAs.The same shallow angle of -1.8 • was also selected for CCC on January 12, February 3, February 4, and March 30.For CCC on February 7, the nineteen flights analyzed all have flyable trajectories for all the values of FPA from -1.8 • to -3.4 • .The extra fuel burn computed for each FPA ranged from 2.5 lbs to 6.9 lbs, with the lowest extra fuel burn of 2.5 lbs yielded by the FPA of -1.8 • .The FPA that yielded the second lowest extra fuel burn was -3.3 • with an extra fuel burn of 3.0 lbs, and can be a reasonable choice of FPA as well. +V.D. Descent-Speed FPAFigure 9 shows values of γ 0 selected for the gate-specific adaptations of Descent-Speed FPA.The airportspecific FPA functions are not shown here, but their fuel burn will be analyzed in Section V.E.The selected values of γ 0 are very similar to γ univ selected for Universal FPA.The steepest values of γ 0 were selected for CCC.Similar to Universal FPA, a shallow value of -1.8 • was selected for γ 0 for both CCC and CAMRN for the Gate-Day adaptation on February 7 for reasons described in previous sections. +V.E. Fuel Burn ComparisonComparison of the three strategies in terms of their fuel burn merits is a necessary step in a benefit assessment.Other aspects of the comparison of the three strategies include the cost of implementation, which is beyond the scope of this paper.Another interesting comparison would be the fuel burn merit with any strategy against current-day operations.Since the strategies are expected to support EDA, their fuel burn benefit tends to be compounded with EDA's fuel benefit.An experimental design that would separate the benefit of the strategies from the benefit of EDA is left for future work.To facilitate the fuel-burn comparison, results are presented in terms of the average fuel burn per flight, over the year's worth of JFK traffic data, relative to the minimum-fuel solution of Min-Fuel FPA.In this way, the results will show how close the simpler strategies, Universal FPA and Descent-Speed FPA, and their adaptations can come to the minimum-fuel solution without requiring real-time pilot-controller communication of FPA just prior to the top of descent.The fuel burn comparison was based on trajectories spanning from the freeze horizon before the top-ofdescent to the metering fix, and therefore has contributions from both the cruise and the descent segments.Figure 10 shows the extra fuel burn per flight computed for Universal FPA and Descent-Speed FPA.Overall, two interesting observations can be made.First, even the simplest strategy, a single static FPA adapted for JFK, has the potential to come within 21.5 lbs of the minimum-fuel solution for each flight on average.This 21.5 lbs represents the potential benefit of the Min-Fuel FPA solution.To put this into perspective, this represents approximately 4% of the fuel burn for a typical small-jet arrival transitioning over a 160 nmi segment from cruise to the TRACON boundary.This of course assumes that the modeled winds aloft and traffic conditions used for parameter selection are reasonably accurate.To see the potential impact of directional and temporal adaptations on the fuel efficiency of Universal FPA and Descent-Speed FPA, consider the Universal FPA results on Figure 10.Adapting the universal "airport" FPA to season, month and day has the potential of reducing that extra 21.5 lbs of fuel per flight by 12%, 21%, and 38%, respectively.Relative to the Airport-Static adaptation, adaptation to the arrival gate only slightly reduces the "extra" 21.5 lbs of fuel per flight by 6%.In other words, publishing a universal FPA for each of the four gates will recover only 6% of the difference between the Aircraft-Specific adaptation of Universal FPA and the minimum fuel solution.Adaptation to gates becomes more effective when combined with finer granularity of timespan, as the Gate-Season, Gate-Month, and Gate-Day adaptations reduce the extra fuel burn from the corresponding airport-specific adaptations by 11%, 14%, and 38%, respectively.The combined adaptation of a universally fixed FPA for each arrival gate, for each day of operations, recovers about 61% of the fuel savings of the minimum-fuel solution for each flight.By "adapting" to the descent speed, Descent-Speed FPA is essentially a surrogate adaptation for metering delay, one of the primary factors being considered in this analysis.As such, this strategy was anticipated to yield a fair amount of benefit under metering conditions.The results indicate that Descent-Speed FPA contributes a 21% reduction in the 21.5 lbs of extra fuel burn over Universal FPA for the Airport-Static case.In considering the overall results for Descent-Speed FPA shown in Figure 10, both the directional and temporal adaptations yield similar improvements in fuel efficiency ranging from 22% to 29% among the types of adaptation.The combined effect of Descent-Speed FPA with both directional and temporal adaptation has the potential to achieve 74% of the Min-Fuel FPA benefit. +VI. Comparison of JFK and DFW ResultsComparison of the extra fuel burn between JFK and DFW sheds light on the major factors that affect the relative merits of the three strategies.The following discussion focuses on results for Universal FPA for JFK and DFW, with the results for DFW taken from previous work. 9Descent-Speed FPA's fuel burn follows the same pattern and hence is not discussed here.Figure 11 airport-specific adaptation types for JFK are consistently less than those for DFW by 15% to 20%.A second observation is the relative ineffectiveness of the gate-specific adaptations for JFK, especially for the Gate-Static adaptation.While the gate-specific adaptations for DFW reduce the extra fuel burn from the corresponding airport-specific adaptations by 25% to 34%, the Gate-Static adaptation for JFK reduced only 6% of the Airport-Static extra fuel burn.To determine each gate's contribution to the ineffectiveness of the Gate-Static adaptation for JFK, the extra fuel burn for each gate was examined and listed in Table 5.An average of the four values of extra fuel 11 and10.The extra fuel burn for CAMRN is above 40 lbs, more than three times higher than the extra fuel burn for any of the other three gates.This high extra fuel burn is definitely related to the very shallow FPA of -1.9 • selected for CAMRN.However, HARTY also had -1.9 • selected but resulted in an extra fuel burn of 8 lbs only.The difference between the extra fuel burn of CAMRN and HARTY was due to the short descent segments required for capturing the high-altitude constraints for HARTY and LOLLY, which reduced the extra fuel burn significantly.To demonstrate the effects of the descent segment on the extra fuel burn, a modification to the JFK configuration/constraints, denoted as +AL, was used in another simulation that lowered the altitude constraint for HARTY and LOLLY from 21,000 ft and 20,000 ft, respectively, to 12,000 ft.The results are shown on the right side of Table 5.The modification increased the extra fuel burn for HARTY and LOLLY by more than three times, approaching that for CAMRN.Another factor to consider is that the directions of the arrival flights to JFK are very different than those to DFW.While three quarters of the arrival flights to JFK are from the west, two-thirds of the arrival flights to DFW are from the east. 9This means most arrival flights to JFK experience tailwinds, while most arrival flights to DFW experience headwinds.The selected FPA for a tailwind-dominant gate is usually shallower and farther from the minimum-fuel FPA of some individual flights, because it had to be shallow enough to be flyable for 99% of flights.The selected FPA for a headwind-dominant gate, on the contrary, is usually close to the minimum-fuel FPA of many flights.Therefore, the extra fuel burn for a tailwind-dominant gate is usually higher than that for a headwind-dominant gate.Since tailwinds dominate the gates CAMRN, HARTY, and LOLLY, the extra fuel burn for these three gates is much higher (in the +AL configuration) than the value for CCC, the only gate where headwinds dominate.In addition to the +AL configuration, two more modifications of the configuration were used to further isolate the contributions of various factors contributing to the difference of extra fuel burn between JFK and DFW.All three modifications are listed in Table 6.The +CR aligns the average ground course of the +Name Comment ++ALReduced altitude constraints for LOLLY and HARTY +CR Aligned average ground course for each gate to that of DFW +WD Swapped the RUC wind out for that for DFW arrival flights for a specific gate to JFK with a gate to DFW, by moving the aircraft's initial position while fixing its distance to the gate.With this modification, the flights to CCC, CAMRN, HARTY, and LOLLY are aligned in terms of their average ground courses with those to the NE, SE, SW, and NW gates of DFW, respectively.The +WD swaps out the RUC wind estimates for JFK for the RUC wind estimates for DFW.This configuration would reveal the effect of the wind field on the relative fuel-burn merits of the three strategies.Figure 12 shows the differences of extra fuel burn between JFK and DFW, averaged over the eight adaptation types, for various modified JFK configurations.The average percentage difference uses each of the eight extra fuel burn values for DFW as the reference and is defined by Equation 6.∆P = 100 × 1 8 8 k=1 ∆f k,JFK -∆f k,DFW ∆f k,DFW ,(6)where ∆P is the average percentage difference and ∆f k is the extra fuel burn for adaptation k.The vertical error bar represents variation of the percentages of differences in the eight adaptation types.The increase in extra fuel burn in the +AL configuration was contributed largely by HARTY and LOLLY, with one example shown for the Gate-Static adaptation in Table 5. Adding +CR on top of +AL reduces the average difference from 50% above DFW to 24% above DFW.A closer look at the Gate-Static extra fuel burn with +AL+CR revealed that the extra fuel burn for CAMRN reduced significantly from 40.6 lbs to 17.0 lbs.This is largely due to the change of the average ground course of the arrival flights from 58 • to 290 • , turning the wind along the route from predominantly tailwinds to predominantly headwinds.Adding +WD on top of +AL and +CR further reduces the average difference from 24% above DFW to 12% above DFW.This is in agreement with the fact that the winds for DFW are weaker than those for JFK.Other factors contribute to the difference of the extra fuel burn between JFK and DFW but are not further investigated.One noticeable factor is that the aircraft arriving at JFK are, on average, larger than those arriving at DFW.For example, the most frequent aircraft at JFK, E190, is larger than both the most frequent E135 and E145 observed for DFW.The resulting extra fuel burn per flight for JFK, therefore, will be larger than that for DFW with other factors being equal.To summarize the discoveries in the section, three factors were investigated and found to contribute significantly to the relative fuel-burn merits of the three strategies:• High-altitude constraints at JFK diminish the difference of fuel-burn merits between the three selection strategies.• Arrival directions at JFK, predominantly from west, increase the difference of fuel-burn merits between the three selection strategies.• Stronger winds at JFK increase the difference of fuel-burn merits between the three selection strategies. +VII. Conclusion and Future WorkThis paper applied three strategies for choosing/defining the descent flight-path angles (FPAs) for small jets in transition airspace under metering conditions to the John F. Kennedy Airport (JFK).The three FPA selection strategies are:1. Universal FPA -defines a universal FPA for all small jets arriving to an airport, gate, or route.2. Descent-Speed FPA -defines different FPAs for different descent speeds.3. Min-Fuel FPA -computes a minimum-fuel FPA for each flight, but requires communication of the FPA to the pilot in real time.The three strategies vary in operational complexity.The Min-Fuel FPA strategy served as a reference point for the fuel-burn merits.One year's worth of traffic data arriving at JFK during 2011 was used for the analysis of the three strategies.The FPA of Universal FPA and the FPA function of Descent-Speed FPA were selected based on fuel burn and planned speed-brake usage of the meet-time trajectories computed for the flights.Results showed that the Universal FPA strategy with its FPA adapted to the JFK Airport had 21.5 lbs of extra fuel burn relative to the Min-Fuel FPA solution for each flight on average.Adaptation of the FPA to the arrival-gate reduced the extra fuel burn per flight by 6%.Adaptation of the FPA to each day reduced the extra fuel burn per flight by up to 38%.Combining the directional and temporal adaptations reduced the extra fuel burn by 61%.The Descent-Speed FPA stratey, considered as a surrogate adaptation to the descent speed, reduced the extra fuel burn per flight by 21%.The combined effect of the directional, temporal, and speed adaptation recovered 74% of extra fuel burn relative to the Min-Fuel FPA solution.Three factors were investigated and found to contribute significantly to the difference of relative fuel-burn merits between the three strategies between JFK and DFW: The high-altitude constraints at JFK, directions of arrival, and winds.While the high-altitude constraints diminish the difference of the fuel-burn merits of the three strategies, the specific directions of arrival and winds of JFK augment the difference.The effects of these airport-specific conditions on the relative fuel-burn merits show the complexity of analyzing the economics of an FPA selection, and should assist in the design of a generic FPA selection procedure for a variety of airports.The planned future work includes generalization of the FPA selection process to a third disparate airport that has weaker wind variations; possibly the Los Angeles International Airport (LAX).Another relevant analysis orthogonal to the airport adaptation is to define a "normalized" fuel-burn metric that removes the effects of high-altitude restrictions and aircraft size on the fuel-burn merits.It would be interesting to see if such a metric yields more universal results than the absolute fuel-burn difference.Figure 1 .1Figure 1.The Min-Fuel FPA strategy selects the minimum-fuel FPA. +Figure 2 .2Figure 2. Variation of winds aloft at twelves major airports in the United States for 2011. +Figure 3 .3Figure 3. Wind along the hypothetical routes for the JFK airport.A positive wind is a tailwind. +Figure 4 .4Figure 4. Tracks of arrival small jets to JFK on January 7, 2011. +Figure 5 .5Figure 5.A general vertical profile that contains five segment types. +Figure 6 .6Figure 6.Most frequent arrival small jet types observed arriving at the JFK in 2011. +- +Figure 9 .9Figure 9. Values of γ0 selected for the Gate-Static, Gate-Season, Gate-Month, and Gate-Day adaptations of Descent-Speed FPA. +Figure 10 .10Figure 10.Extra fuel burn calculated for eight adaptations of Universal FPA and Descent-Speed FPA. +Figure 11 .11Figure 11.Comparison of Universal FPA's extra fuel burn between DFW and JFK. +Figure 12 .12Figure 12.Universal FPA's average extra fuel burn difference relative to DFW using modified JFK configurations. +Table 22lists the names given to these eight adaptation types analyzed in this work: +Table 1 .1The types of adaptations categorized by their granularity in location/direction and timeAirspace/DirectionTimeNAS Airport Arrival Gate Arrival RouteStatic15Season26Month37Day48Hour +Table 2 .2The eight adaptation types analyzed for the JFK traffic# NameFPA/FPA Function Defined for1 Airport-Staticthe airport2 Airport-Season the airport by season3 Airport-Month the airport by month4 Airport-Daythe airport for each day5 Gate-Staticeach arrival gate6 Gate-Seasoneach arrival gate by season7 Gate-Montheach arrival gate by month8 Gate-Dayeach arrival gate for each day +Table 3 .3The first altitude and/or speed restrictions along the STARs for JFKWaypoint Altitude (ft) Speed (CAS in knots)CCC12,000250CAMRN12,000250LOLLY20,000-HARTY21,000- +Table 4 .4Total number of arrival flights of small jetsMonthDaysCCCLOLLYHARTYCAMRNMissing DaysJan3154111339061464Feb202906585368471, 21-25, 27-28Mar31598119610101675Apr30555114711091626May2949112071110148121-22Jun30635128511921496Jul3082412581229158116Aug2874311491132141521, 27-28Sep30681137112341669Oct31603128412061747Nov30469103911241719Dec314059409711610Total3516835136671275918330 +in the dip of the first quartileLOLLYCCC-3.4-3.4-3.0-3.0-2.6-2.6-2.2-2.2-1.8-1.8JANFEBMARAPRMAYJUNJULAUGSEPOCTNOVDECJANFEBMARAPRMAYJUNJULAUGSEPOCTNOVDEC1st QuartileHARTYCAMRN3rd Quartile Median-3.4-3.0-3.0-2.6-2.6-2.2-2.2-1.8-1.8JANFEBMARAPRMAYJUNJULAUGSEPOCTNOVDECJANFEBMARAPRMAYJUNJULAUGSEPOCTNOVDECFigure 7.The first quartile, third quartile, and median FPAs selected using Min-Fuel FPA. +Table 5 .5The extra fuel burn for each gate of JFK for the Gate-Static adaptation burn, weighted by the number of flights (See Table4), would yield the average extra fuel burn of 20.2 lbs in FiguresJFKJFK+ALGate Extra Fuel Burn (lbs) FPA ( • ) Extra Fuel Burn (lbs) FPA ( • )CCC11.82-2.411.82-2.4CAMRN40.47-1.940.47-1.9HARTY7.97-1.938.63-1.9LOLLY8.55-2.031.45-2.0 +Table 6 .6Modifications to the JFK configuration used for calculating the extra fuel burn + of 18 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on August 14, 2013 | http://arc.aiaa.org| DOI: 10.2514/6.2013-4250Copyright© 2013 by the American Institute of Aeronautics and Astronautics, Inc.The U.S. Government has a royalty-free license to exercise all rights under the copyright claimed herein for Governmental purposes. + of 18 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on August 14, 2013 | http://arc.aiaa.org| DOI: 10.2514/6.2013-4250Copyright© 2013 by the American Institute of Aeronautics and Astronautics, Inc.The U.S. Government has a royalty-free license to exercise all rights under the copyright claimed herein for Governmental purposes. + of 18 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on August 14, 2013 | http://arc.aiaa.org| DOI: 10.2514/6.2013-4250Copyright© 2013 by the American Institute of Aeronautics and Astronautics, Inc.The U.S. Government has a royalty-free license to exercise all rights under the copyright claimed herein for Governmental purposes. + American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on August 14, 2013 | http://arc.aiaa.org| DOI: 10.2514/6.2013-4250Copyright© 2013 by the American Institute of Aeronautics and Astronautics, Inc.The U.S. Government has a royalty-free license to exercise all rights under the copyright claimed herein for Governmental purposes. + Downloaded by NASA AMES RESEARCH CENTER on August 14, 2013 | http://arc.aiaa.org| DOI: 10.2514/6.2013-4250Copyright© 2013 by the American Institute of Aeronautics and Astronautics, Inc.The U.S. Government has a royalty-free license to exercise all rights under the copyright claimed herein for Governmental purposes. + + + + + + + + + + + Nasa Ames ResearchCenter On + + 10.2514/6.2013-4250 + + + August 14. 2013 + American Institute of Aeronautics and Astronautics + + + American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on August 14, 2013 | http://arc.aiaa.org | DOI: 10.2514/6.2013-4250 + + + + + Stability Limits and Critical Size of Structures in Premixed Flames Presented at the 9th ICODERS, Poitiers, France, July 3–8, 1983. Copyright American Institute of Aeronautics and Astronautics, Inc., 1984. All rights reserved. + + JQuinard + + + GSearby + + + LBoyer + + 10.1016/b978-0-08-092523-3.50047-2 + + + Dynamics of Curved Fronts + + Elsevier + + + + + Copyright © 2013 by the + Copyright © 2013 by the American Institute of Aeronautics and Astronautics, Inc. The U.S. Government has a royalty-free license to exercise all rights under the copyright claimed herein for Governmental purposes. References + + + + + Field Evaluation of the Tailored Arrivals Concept for Datalink-Enabled Continuous Descent Approach + + RichardACoppenbarger + + + RobWMead + + + DouglasNSweet + + 10.2514/1.39795 + + + Journal of Aircraft + Journal of Aircraft + 0021-8669 + 1533-3868 + + 46 + 4 + + July-August 2009 + American Institute of Aeronautics and Astronautics (AIAA) + + + Coppenbarger, R. A., Mead, R. W., and Sweet, D. N., "Field Evaluation of the Tailored Arrivals Concept for Datalink- Enabled Continuous Descent Approach," Journal of Aircraft, Vol. 46, No. 4, July-August 2009, pp. 1200-1209. + + + + + Development, Design and Flight Test Evaluation of Continuous Descent Approach Procedure in FIR Zagreb + + DorisNovak + + + TinoBucak + + + TomislavRadišić + + 10.7307/ptt.v21i5.247 + + + PROMET - Traffic&Transportation + PROMET + 0353-5320 + 1848-4069 + + 21 + 5 + + Sept. 2009 + Faculty of Transport and Traffic Sciences + + + Novak, D., Bucak, T., and Radisić, T., "Development, Design and Flight Test Evaluation of Continuous Descent Approach Procedure in FIR Zagreb," PROMET -Traffic & Transportation, Vol. 21, No. 5, Sept. 2009, pp. 319-329. + + + + + Continuous Descent Approach: Design and Flight Test for Louisville International Airport + + John-Paul B.Clarke + + + NhutTHo + + + LilingRen + + + JohnABrown + + + KevinRElmer + + + Kwok-OnTong + + + JosephKWat + + 10.2514/1.5572 + + + Journal of Aircraft + Journal of Aircraft + 0021-8669 + 1533-3868 + + 41 + 5 + + 2004 + American Institute of Aeronautics and Astronautics (AIAA) + + + Clarke, J.-P. B., Ho, N. T., Ren, L., Brown, J. A., Elmer, K. R., Tong, K.-O., and Wat, J. K., "Continuous Descent Approach: Design and Flight Test for Louisville International Airport," Journal of Aircraft, Vol. 41, No. 5, 2004, pp. 1054-1066. + + + + + FAA Aviation Forecasts: Fiscal Years 1981-1992. Federal Aviation Administration, U.S. Department of Transportation, Washington, D.C. 20591. 1980. 69p + 10.1177/004728758102000159 + + + Journal of Travel Research + Journal of Travel Research + 0047-2875 + 1552-6763 + + 20 + 1 + + 2011 + SAGE Publications + + + "FAA Aerospace Forecast Fiscal Years 2012-2032," Federal Aviation Administration, 2011. + + + + + Relative Significance of Trajectory Prediction Errors on an Automated Separation Assurance Algorithm + + TALauderdale + + + ACCone + + + ARBowe + + + + Proceedings of the 9th USA/Europe Air Traffic Management R&D Seminar + the 9th USA/Europe Air Traffic Management R&D Seminar + + June 2011 + + + Lauderdale, T. A., Cone, A. C., and Bowe, A. R., "Relative Significance of Trajectory Prediction Errors on an Automated Separation Assurance Algorithm," Proceedings of the 9th USA/Europe Air Traffic Management R&D Seminar , June 2011. + + + + + Predictability of Top of Descent Location for Operational Idle-Thrust Descents + + LaurelStell + + 10.2514/6.2010-9116 + + + 10th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + + American Institute of Aeronautics and Astronautics + June 2011 + + + Stell, L., "Prediction of Top of Descent Location for Idle-thrust Descents," Proceedings of the 9th USA/Europe Air Traffic Management R&D Seminar , June 2011. + + + + + The Efficient Descent Advisor: Technology Validation and Transition + + RichardCoppenbarger + + + MiwaHayashi + + + GauravNagle + + + DouglasSweet + + + RenanSalcido + + 10.2514/6.2012-5611 + + + 12th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference and 14th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference + + American Institute of Aeronautics and Astronautics + 2012-5611, Sept. 2012. 8 Swenson,. June 1997 + + + The Efficient Descent Advisor: Technology Validation and Transition + Coppenbarger, R. A., Nagle, G., Sweet, D., and Hayashi, M., "The Efficient Descent Advisor: Technology Validation and Transition," Proceedings of the AIAA Aircraft Technology, Integration, and Opterations Conference, AIAA-2012-5611, Sept. 2012. 8 Swenson, H. N., Hoang, T., Engelland, S., Vincent, D., Sanders, T., Sanford, B., and Heere, K., "Design and Operational Evaluation of the Traffic Management Advisor at the Fort Worth Air Route Traffic Control Center," Proceedings of the 1st USA/Europe Air Traffic Management R&D Seminar , June 1997. + + + + + Strategies for Choosing Descent Flight-Path Angles for Small Jets + + MinghongWu + + + StevenGreen + + 10.2514/6.2012-4817 + + + AIAA Guidance, Navigation, and Control Conference + + American Institute of Aeronautics and Astronautics + 2012-4817, Aug. 2012 + + + Wu, M. G. and Green, S. M., "Strategies for Choosing Descent Flight Path Angles for Small Jets," Proceedings of the AIAA Guidance, Navigation, and Control Conference, AIAA-2012-4817, Aug. 2012. + + + + + Experimental Approach for Efficient Arrival Procedures with Fixed-Flight Path Angle Descent + + MGWu + + + SMGreen + + 10.1109/icnsurv.2019.8735138 + + + 2019 Integrated Communications, Navigation and Surveillance Conference (ICNS) + + IEEE + 2011-215992, Nov. 2011 + + + Wu, M. G. and Green, S. M., "Analysis of Fixed Flight Path Angle Descents for the Efficient Descent Advisor," NASA/TM-2011-215992, Nov. 2011. + + + + + Continuous Descent Approaches with Variable Flight-Path Angles under Time Constraints + + RichardSopjes + + + PaulDe Jong + + + ClarkBorst + + + RenéVan Paassen + + + MaxMulder + + 10.2514/6.2011-6219 + AIAA- 2011-6219 + + + AIAA Guidance, Navigation, and Control Conference + + American Institute of Aeronautics and Astronautics + Aug. 2011 + + + Sopjes, R., De Jong, P., Borst, C., Van Paassen, M., and Mulder, M., "Continuous Descent Approaches with Variable Flight-Path Angles under Time Constraints," Proceedings of the AIAA Guidance, Navigation, and Control Conference, AIAA- 2011-6219, Aug. 2011. + + + + + An Hourly Assimilation–Forecast Cycle: The RUC + + StanleyGBenjamin + + + DezsöDévényi + + + StephenSWeygandt + + + KevinJBrundage + + + JohnMBrown + + + GeorgAGrell + + + DongsooKim + + + BarryESchwartz + + + TatianaGSmirnova + + + TracyLorraineSmith + + + GeoffreySManikin + + 10.1175/1520-0493(2004)132<0495:ahactr>2.0.co;2 + + + Monthly Weather Review + Mon. Wea. Rev. + 0027-0644 + 1520-0493 + + 132 + 2 + + 1998 + American Meteorological Society + + + Benjamin, S. G., Brown, J. M., Brundage, K. J., Schwartz, B., Smirnova, T., Smith, T. L., Morone, L. L., and Dimego, G., "The Operational RUC-2. Preprints," Proceedings of the 16th Conference on Weather Analysis and Forecasting, Amer. Meteor. Soc., 1998, pp. 249-252. + + + + + Distributed Scheduling Architecture for Multi-Center Time-Based Metering + + StevenLandry + + + ToddFarley + + + JohnFoster + + + StevenGreen + + + TyHoang + + + GregoryWong + + 10.2514/6.2003-6758 + AIAA-2003-6758 + + + AIAA's 3rd Annual Aviation Technology, Integration, and Operations (ATIO) Forum + + American Institute of Aeronautics and Astronautics + Nov. 2003 + + + Landry, S., Farley, T., Foster, J., Green, S., Hoang, T., and Wong, G. L., "Distributed Scheduling Architecture for Multi-Center Time-Based Metering," Proceedings of the AIAA Aviation Technology, Integration, and Opterations Conference, AIAA-2003-6758, Nov. 2003. + + + + + The Trajectory Synthesizer Generalized Profile Interface + + AlanLee + + + XavierBouyssounouse + + + JamesMurphy + + 10.2514/6.2010-9138 + AIAA-2010-9138 + + + 10th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + + American Institute of Aeronautics and Astronautics + Sept. 2010 + 15 + + + Lee, A. G., Bouyssounouse, X., and Murphy, J. R., "The Trajectory Synthesizer Generalized Profile Interface," Proceed- ings of the 10th AIAA Aviation Technology, Integration, and Operations Conference, AIAA-2010-9138, Sept. 2010. 15 + + + + + Trajectory Synthesis for Air Traffic Automation + + RhondaSlattery + + + YiyuanZhao + + 10.2514/2.4056 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 20 + 2 + + March 1997 + American Institute of Aeronautics and Astronautics (AIAA) + + + Slattery, R. and Zhao, Y., "Trajectory Synthesis for Air Traffic Automation," Journal of Guidance, Control and Dynam- ics, Vol. 20, No. 2, March 1997, pp. 232-238. + + + + + Design of Center-TRACON Automation System + + HErzberger + + + TJDavis + + + SMGreen + + + + Machine Intelligence in Air Traffic Management, Andre Benoit + + Oct. 1993 + + + + AGARD CP-538 + Erzberger, H., Davis, T. J., and Green, S. M., "Design of Center-TRACON Automation System," Machine Intelligence in Air Traffic Management, Andre Benoit, ed., AGARD CP-538 , Oct. 1993, pp. 11-1-11-12. + + + + + Advanced Aircraft Performance Modeling for ATM: Enhancements to the Bada Model + + ANuic + + + CPoinsot + + + M.-GIagaru + + + EGallo + + + FANavarro + + + CQuerejeta + + 10.1109/dasc.2005.1563320 + + + 24th Digital Avionics Systems Conference + + IEEE + Nov. 2005 + + + Nuic, A., Poinsot, C., Iagaru, M.-G., Gallo, E., Navarro, F. A., and Querejeta, C., "Advanced Aircraft Performance Modeling for ATM: Enhancements to the BADA Model," Proceedings of the IEEE/AIAA 24th Digital Avionics Systems Conference, Nov. 2005. + + + + + Design and Development of the En Route Descent Advisor (EDA) for Conflict-Free Arrival Metering + + RichardCoppenbarger + + + RichardLanier + + + DougSweet + + + SusanDorsky + + 10.2514/6.2004-4875 + AIAA- 2004-4875 + + + AIAA Guidance, Navigation, and Control Conference and Exhibit + + American Institute of Aeronautics and Astronautics + Aug. 2004 + + + Coppenbarger, R. A., Lanier, R., Sweet, D., and Dorsky, S., "Design and Development of the En Route Descent Advisor (EDA) for Conflict-Free Arrival Metering," Proceedings of the AIAA Guidance, Navigation, and Control Conference, AIAA- 2004-4875, Aug. 2004. + + + + + + diff --git a/file795.txt b/file795.txt new file mode 100644 index 0000000000000000000000000000000000000000..ffd9bbe9ffba9f3e785f86bb304966c13216d21d --- /dev/null +++ b/file795.txt @@ -0,0 +1,810 @@ + + + + +I. IntroductionTrajectory management of metered arrivals into high-density terminal airspace is a critical component for NextGen Trajectory Based Operations concepts. 1,2 ignificant research has focused on the utilization of modern flight management systems (FMS) to enable continuous descent planning, at least from cruise to a metering fix within the the Terminal Radar Approach Control Facilities (TRACON) airspace in the United States or the Terminal Control airspace in other countries.4][5] However, little attention has been paid to "small" (regional, business and light) jet types, which comprise a large and potentially high-growth portion of NextGen traffic operations. 6nlike the larger aircraft types, which are equipped with performance-based FMS systems that attempt to optimize the vertical profile with near-idle descents, the smaller jets are equipped with simpler Vertical Navigation (VNAV) capabilities.Descent planning for these types typically involves a fixed-flight path angle (FPA) descent that is either based on a company-programmed default or a pilot-selected value.For example, the "standard operating procedure" of one large regional carrier called for an indicated airspeed of 320 knots for descent, initiated at the cruise Mach number, using a default FPA of -3.8 • .This works fine for nominal conditions with light to moderate winds and no Air Traffic Control interruptions to the descent.However, when speed restrictions are issued by controllers for metering and spacing, the nominal descent plan can become inefficient and difficult, if not sometimes impossible, to fly in strong tailwinds.In addition, random observations of regional jet operations and pilot interviews revealed that a large variety of descent-planning techniques are used by pilots, even for the same equipment.These techniques vary in terms of the selection of descent angle, bottom-of-descent planning, and top-of-descent transition.Sometimes they take into account winds aloft and weight, but rarely descent speed.It is important to develop and standardize the procedures for establishing efficient descent FPAs for small jets.Such standardized procedures lead to better trajectory predictability and provide benefits for separation assurance. 7However, selecting the FPA is a non-trivial task, as the most fuel-efficient and "flyable" FPAs a can vary significantly as a function of aircraft type, weight, speed profile, and, particularly, winds and wind gradient. 80][11][12] Under some conditions, a steep descent may be the most fuel efficient and yet be operationally unacceptable to pilots.Even if a steep descent is achievable with the utilization of speed brakes, many pilots are reluctant, if not unwilling, to use them because of noise and ride discomfort.Given the significant variation in the winds aloft from one area of the National Airspace System (NAS) of the United States to another, and from one day, week or month to another, the FPA procedure may need to be "adaptive".The purpose of this paper is to propose several candidate methodologies for selecting descent FPAs for small jets, and analyze their impact on the FPA and fuel efficiency.In this work, three candidate strategies are applied to a full year's worth of arrival traffic and recorded winds data for Dallas/Fort Worth (DFW) Airport from the year 2011.DFW was selected because of its large seasonal variation in winds aloft, the eastwest symmetry of it's arrival routes which accentuates head/tail-wind differences, and the density of regional jet operations.The benefits of various adaptations of these strategies are investigated as a function of the arrival gate and periods of time (categorized by season, month and day).The rest of the paper is organized as follows: Section II describes the metered environment; Section III outlines the three proposed FPA strategies and the methodology for parameter selection; Section IV describes the adaptation of Strategies 1 and 2 as a function of location (arrival gate) and time; Section V describes the approach taken to model the descent conditions and the metering-constrained arrival trajectories; Section VI presents results of the selected FPAs in various adaptations and compares fuel burn among the three strategies; Section VII investigates the effect of wind on the results, focusing on data from a few selected days for further insight; and, finally, Section VIII summarizes the findings. +II. BackgroundIn the United States, the Traffic Management Advisor (TMA) 13,14 computes metering-fix scheduled times of arrival (STA) at the TRACON boundary in order to control throughput of en-route traffic arriving at a high-density airport.][17][18][19][20] Consider an arrival that is guided by a controller using EDA to plan and execute a continuous descent in order to cross a TRACON metering fix at the STA specified by TMA.During periods of congestion, this STA will result in a small delay at the metering fix to keep the TRACON arrival traffic manageable. 14epending on the conditions for each aircraft, speed reductions are typically able to absorb three to four minutes of delay for flights about 20 minutes or 150 nmi from the metering fix. 18While other maneuvers such as temporary altitude clearances or path stretches are required for larger delays, it is the speed advisories that are relevant to the analysis presented below.The previous development and testing of EDA focused primarily on descent procedures for large jets equipped with a performance-based FMS.The last field trial conducted at Denver Center in the fall of 2010 (results yet to be published) began to explore the issues related to small jets.While a simple, fixed-FPA descent procedure using prescribed clearances was introduced for the purposes of that field trial, EDA itself still lacks a defined descent procedure and corresponding algorithm for defining the descent FPA for small jets.The strategies presented here for selecting descent FPAs will complement the current EDA capabilities and support fuel-efficient and predictable arrival-metering operations for small jets. +III. Three FPA Selection StrategiesIn a previous paper, three strategies were proposed to define the fixed FPA for small jet arrivals under metering conditions. 8Monte Carlo simulations using simplified wind distributions were performed to evaluate these strategies.The three strategies were as follows:1. Universally fixed FPA 2. FPA as a function of descent speed 3. Minimum-fuel FPA for every flight (Referred to as "Custom FPA" in the previous paper) Each strategy is discussed in more details in the following sections. +III.A. Strategy 1: Universally Fixed FPAStrategy 1 defines a universal FPA for all small jets arriving to an airport, gate, or route.This is akin to a glide slope extended from the arrival metering fix back up to the top of descent (TOD) in the en-route airspace.The idea of a universally fixed-FPA descent procedure was explored by Tong et al. at Boeing, with a focus on modifications required for the performance-based FMS equipped jets. 9The difference between Strategy 1 and the work of Tong et al. is twofold.First, Strategy 1 applies to small jets whose VNAV can guide fixed-FPA descents, while the work of Tong et al. attempts to enable a fixed FPA procedure for jets equipped with a performance-based FMS.Second, methods are developed in this paper to adapt the universally fixed FPA in Strategy 1 to season, month, day, or arrival gate.The advantage of Strategy 1 is its simple form, which allows it potentially to be published as part of the arrival procedure.The disadvantage is that it does not account for the effects of descent speed and winds on the fuel efficiency or the range of flyable FPAs. +III.B. Strategy 2: FPA as a Function of the Descent SpeedStrategy 2 takes into account the effects of the descent speed on the range of fuel-efficient and flyable FPAs.Table 1 summarizes an example of the FPA as a function of the descent calibrated airspeed (CAS), designed for an EDA flight trial held in Denver Center in 2010 for regional jets operated by Skywest Airlines.Note that the descent CAS in Table 1 ranges from 250 knots to 320 knots.The low end is the minimum speed 1 was static and not adapted for winds aloft and the direction of flight.In our previous work, a finer-grained, step-wise function with an adaptive parameter was used. 8In contrast to the FPA function used in the field trial, this FPA function changes the FPA by 0.1 • for every 10 knots of descent CAS.It was observed that this rate of change of FPA with respect to the descent CAS reflects the variation of fuel-efficient FPAs with speed. 8Moreover, the small increment of 0.1 • leads to smaller gaps in the achievable ranges of time-to-fly.The FPA function is defined by Equation 1:γ =              γ 0 , if DCAS < 255 γ 0 -0.1 , if 255 ≤ DCAS < 265 γ 0 -0.2 , if 265 ≤ DCAS < 275 . . . ,(1)where γ is the FPA in degrees, DCAS is the descent CAS in knots, and the adaptive parameter γ 0 stands for the value of γ at 250 knots of the descent CAS.Note that the increment of 0.1 • matches the precision of FPA-selection increments typical of small jet avionics.Also note the selection of the parameter γ 0 defines the entire FPA function.To select an FPA for an arrival flight, the range of time (i.e., time-to-fly) achievable by the speeds mapping to each FPA is computed, and the FPA whose resulting range of time covers the STA is selected.One advantage of Strategy 2, like Strategy 1, is that the function can be determined ahead of time and therefore published as part of the arrival procedure.Moreover, by taking into account the descent speed, Strategy 2 is expected to be more fuel-efficient than Strategy 1.However, like Strategy 1, Strategy 2 has the disadvantage that it does not account for the effects of the direction of the winds aloft on the range of fuel-efficient and flyable FPAs. +III.C. Strategy 3: Minimum-Fuel FPAWhile Strategies 1 and 2 define the FPA and the FPA function ahead of time, Strategy 3 selects a computed minimum-fuel FPA to be communicated explicitly to the pilot of each flight just prior to the TOD.The computation is based on the route, winds and temperature aloft, and the speed profile necessary to meet the STA. Figure 1 sketches the fuel burn as a function of the descent FPA for a typical flight.Strategy 3, by definition, selects the minimum-fuel FPA for each arrival flight and therefore is better than Strategies 1 and 2 in terms of fuel saving.It serves as a reference of the fuel burn to compare with the first two strategies.However, the requirement of communicating the FPA to the pilot in real time may make it too complex to implement in the near term. +III.D. Planned Speed-Brake UsageIn addition to predicted fuel burn, the selection of FPA should also take into account uncertainties in vertical profile planning.To keep the aircraft on the planned vertical path, power adjustment is preferred to speed-brake deployment for reasons of passenger comfort and the desire of pilots to reserve the use of a speed brake for rare occasions.Besides, some aircraft types have less effective speed brakes than others.Two levels of planned-speed-brake usage are modeled in our analysis of feasible meet-time trajectories to explore the impact of speed-brake usage on the selection of descent FPA and resulting fuel burn.In one condition, FPAs are limited to those that do not require speed-brake usage.In the other, any FPA is considered valid as long as the estimated amount of speed-brake usage is within the modeled speed brake capacity for that aircraft.These conditions are imposed on the acceptable predicted trajectories in the analysis.Actual speed-brake usage required in flight might vary based on the difference between the actual and the predicted state of the flight and atmosphere. +III.E. The Parameters in Strategies 1 and 2The universally fixed FPA in Strategy 1 and the γ 0 in Strategy 2 are parameters that must be selected carefully, with consideration for the prevailing winds aloft and, to a lesser extent, the anticipated traffic demand.Our approach is to use fast-time simulation to analyze the fuel efficiency and flyability of the FPAs resulting from the parameters selected for each strategy.On the one hand, the resulting FPA should not be so shallow as to waste fuel unnecessarily.On the other hand, the FPA should not be so steep as to require excessive use of speed brakes to maintain both the assigned descent speed and FPA throughout the descent.For the purposes of this work, it is assumed that reasonably accurate estimates of the prevailing winds and arrival traffic demand are available.This assumption will be further discussed in Section IV.The following criteria have been applied in previous work 8 to select the universally fixed FPA, denoted as γ univ , in Strategy 1, and γ 0 in Strategy 2:1.The parameter must be selected such that at least 99% of the flights do not exceed the limit for the planned speed-brake usage.2. The parameter must be selected such that, without violating the previous criterion, it results in the least average fuel burn per flight.The first criterion defines the steepest parameter that can be selected, and the second criterion selects a minimum-fuel FPA from those parameters no steeper than the steepest parameter allowed by the first criterion.These same criteria are used for the analysis in this work.IV. Adaptation of Strategies 1 and 2 +IV.A. Adaptation to Gates and TimespansA practical implementation of Strategies 1 or 2 would be to adapt their parameters, γ univ and γ 0 , to specific airports, in order to account for the prevailing winds and local traffic flows.As discussed in Section III.E, the selection of the parameter(s) should be such that the resulting descent FPAs are flyable across significant variations of wind along the route.For airports with opposing arrival directions, particularly the classical four-corner-post configuration, strong winds become problematic.While steeper FPAs are typically more fuel-efficient for flights in a headwind, they may be unflyable for flights in the opposite direction.To ensure that γ univ or γ 0 leads to flyable descents over a wide range of arrival routes and wind conditions, the static implementation must be conservative in its selection.This implies selecting a shallow, relatively fuel-inefficient γ univ or γ 0 .These observations motivated the adaptation of γ univ and γ 0 to the direction of flight.Such an adaptation would define a distinct γ univ or γ 0 for each arrival gate feeding an adapted airport.Compared to the static implementation, adaptation to the direction of flight provides a greater degree of "customization" and the potential to improve fuel efficiency.While adaptation to the arrival direction reduces the variation of along-track winds due to directions, it does not mitigate the variation of winds over time.In many areas of the United States that are in the path of the jet stream, seasonal variation of winds aloft can be significant in both magnitude and direction.Therefore, the fuel efficiency of Strategies 1 and 2 can be further improved by adapting γ univ and γ 0 to each season, month, or even each day.A good selection of γ univ and γ 0 for a period of time relies primarily on the quality of the estimated (or forecast) winds aloft and, to a lesser extent, the estimated traffic demand.While quality forecasts of winds aloft beyond 24 hours may be problematic, estimates of the winds aloft over a period of time, such as a season or a month, may be possible through climatological and historical analysis.The approach here is to assume the availability of reliable estimates or forecasts of winds aloft for the selection of the FPA or FPA function for various adaptations.Although the actual fuel benefits for adaptations to longer time horizons will be less than that estimated by this analysis, the trends are insightful. +IV.B. Dallas/Fort Worth AirportA full year's worth of Fort Worth Air Route Traffic Control Center's arrival traffic during 2011 was selected for our analysis.Figure 2 shows tracks of arrival flights of small jets on July 28th, 2011.Each arrival flight enters the TRACON through one of the NE, SE, SW, and NW gates.All flights of the small jets were assumed to arrive at the Dallas/Fort Worth (DFW) airport.b The primary metering fixes for these four gates are KARLA (NE), HOWDY (SE), DEBBB (NW), and FEVER (SW).Although there are other metering fixes for each gate, it is assumed in this analysis that the arrival flights all go through the primary metering fixes.Therefore, the names KARLA, HOWDY, DEBBB, and FEVER will be used interchangeably with the names of the gates.The Rapid Update Cycle (RUC) weather forecast 21 serves as an estimate of the "true" winds and temperatures aloft.Figure 3 illustrates a typical variation of wind across the four arrival gates.This figure presents the probability distribution of tailwinds during the year of 2011 along a straight path to each gate starting from a central location, 150 nmi upstream of the metering fix, at 35,000 ft.The four paths selected for measuring the probability are shown as red lines in Figure 2. In general, the winds in winter are stronger.Winds experienced in the eastern gates have similar distributions; winds experienced in the western gates also have similar distributions.Between January and March, winds for KARLA and FEVER show slightly wider distribution than the winds for HOWDY and DEBBB, respectively.Note that the distribution of winds for KARLA in the JAN-MAR timespan shows a small probability of a tailwind.Similarly, a small probability of a headwind for FEVER is observed.Independent inspection of some of the RUC data files revealed a predominantly west wind that prevailed in the first three and the last two months of 2011.These conditions result in strong headwinds for flights coming through the NE and SE gates during these periods of time, and strong tailwinds for the western gates.The strong west winds during the winter months are typical of the subtropical jet stream at high altitudes.The jet stream moves north in the summer, and so the wind magnitude reduces in July, August, and September, to 50 knots or less most of the time for all gates. +IV.C. Adaptation TypesWe propose a set of adaptations, organized as in Table 2.The columns represent different levels of adaptation for different airspaces, starting with a basic "one size fits all" adaptation for all airports across the NAS.Moving to the right, each column represents a progressively finer adaptation of the FPA strategy to a specific airport, individual arrival gates (corner posts) feeding an airport, all the way down to specific arrival routes feeding each arrival gate.The rows represent a temporal scale starting at the top with the simplest option of a static adaptation.Moving down, each row represents a progressively finer adaptation to account for changes in the prevailing winds as a function of season, month, day or even hour.The table illustrates the overall approach and potential scope.For the purposes of this paper, the analysis will assess the eight types highlighted and numbered in the table.The analysis of adaptations at the NAS-wide level, or at the level of specific arrival routes and/or hours of the day, are left for future work.Table 3 lists the names of these eight adaptation types analyzed in this work: Section V describes the methodologies for the analysis, modeling details, and methods for selecting the FPA and FPA function for each adaptation. +V. Analysis Approach +V.A. Calibrating and Comparing the Three StrategiesA methodology was developed in the previous work 8 to compare the benefits of the three strategies of selecting the descent FPA.Conceptually, this methodology consists of two parts: first select the parameters for Strategies 1 and 2, and then compare the fuel burn between the three strategies for a set of traffic and wind conditions.The first part of the methodology determines the parameters of Strategies 1 and 2 using estimates of winds and temperatures aloft, traffic demand, and modeled metering delays:• For each flight and each candidate value of the parameter, compute the (predicted) fuel burn and planned speed-brake usage.• Select the value of the parameter based on the results, using the criteria defined in Section III.E for speed-brake usage and fuel burn.• Repeat the above steps for both Strategies 1 and 2.The second part of the methodology estimates and compares the fuel burn of the strategies, using the estimates of winds and temperatures aloft, actual traffic data, and modeled or actual metering delays:• Compute the fuel burn and planned speed-brake usage for each flight using Strategies 1 and 2, respectively.• Compute the fuel burn and planned speed-brake usage for each flight using Strategy 3.• Compare the fuel burn of Strategies 1 and 2 relative to Strategy 3.• Iterate over all flights, compute the average fuel burn penalty of Strategies 1 and 2.While the estimates of winds, temperatures aloft, and traffic demand for the second part need not be the same as that used for the first step, they are the same in this analysis, being the RUC data and the track data of DFW recorded in 2011.This means the parameters of Strategies 1 and 2 are selected as best as they can be, using the same wind and track information accessible in real time to Strategy 3. The fuel benefits of Strategies 1 and 2 obtained in this way can be viewed as an upper bound of the actual relative benefits.Since the same estimates of winds and temperatures aloft and traffic demand were used in both parts of the methodology, the fuel burn and planned speed-brake usage recorded for the first part can be directly used in step 1 of the second part without having to recalculate them.The implementation of the methodology further simplifies the data collection process by combining its two conceptual parts into one run of a fast-time simulation as described in detail in the following sections.The basic idea is that, for each arrival flight and a modeled metering delay, the analysis computes a set of meet-time trajectories with varying FPA and descent CAS profile.Because the FPA selected by each strategy must be from these trajectories, the analysis of these trajectories was sufficient for the fuel burn comparison among the three strategies.The following sections describe modeling of the trajectories, metering delays, meet-time trajectories, parameter selection for the adaptations of Strategies 1 and 2, and fuel burn comparison. +V.B. Route and Vertical ProfileA distance-based freeze horizon of 160 nmi was assumed, inside of which TMA would fix the STA for the aircraft. 13The initial condition was selected at a corresponding track point.For flights following shorter arrival routes from the north and the south, the first track point in the Center was selected.For simplification, Direct-To trajectories from the initial point to the primary metering fix of a gate were assumed without actually parsing the flight plans for the waypoints.The initial altitude must be at least 16,000 ft for the flight to be considered for the analysis.Metering fix crossing conditions of 11,000 ft and 250 knots in CAS were assumed.Figure 4 illustrates a typical vertical profile that has five segment types.Individual trajectories will contain all or a subset of these segments depending on the speed profile needed to meet the STA.Each segment is modeled by fixing two control parameters.One of the parameters is the FPA; the second depends on the segment.For a cruise segment the model fixes the airspeed or the engine control for acceleration or deceleration.For the constant speed descent segments, the model fixes an airspeed in Mach or CAS. +V.C. Aircraft ModelingThe Trajectory Synthesizer (TS) component 22,23 of the Center-TRACON Automation System (CTAS) 24 was used to compute the trajectories and their associated fuel burn and planned speed-brake usage.While a detailed performance model of small jet types would have been desirable, one was not available.Instead, a high-fidelity, CTAS model for a mid-size, narrow-body, twin-jet airliner with a typical descent weight of 170,000 lbs was used.The speed envelopes were selected within the ranges of small jets.An empirical constant was used to model the maximum drag coefficient resulting from speed-brake deployment. 8Weight uncertainty was modeled by a normal distribution with a deviation of 8,400 lbs.To account for the variation of the fuel-burn rate among aircraft types, the fuel burn was scaled by the empirical formulaf i = f 0 * N i + 30 230 ,(2)where f i is the scaled fuel burn; f 0 is the raw fuel-burn rate calculated by CTAS for the mid-size, narrowbody, twin-engine jet; and N i is the number of passengers' seats typical of the aircraft type i.This empirical formula was derived by taking the linear regression of the nominal cruise fuel-burn rate of eight small jets plus the mid-sized twin engine jet, using the Base of Aircraft Database (BADA) 3.8 performance model. 25lthough BADA provides modeling parameters for small jets, the calibration of these parameters focused on nominal flight conditions only.Since the analysis in this work explored a wide range of the speeds, containing both nominal and off-nominal ones, it was decided that the high-fidelity CTAS model with scaled fuel burn was more appropriate for the fuel-burn analysis. +V.D. Metering DelayThe delay at the metering fix was modeled by a uniform distribution between zero and the maximum delay that can be absorbed by speed reductions.The delay time was added to the nominal time in order to specify the STA.In the absence of a "standard" FPA for defining the nominal trajectory, the idle-thrust descent was used for defining the nominal and slow-limit trajectories.The descent CAS of 320 knots was assumed as the airline-preferred descent CAS and was used, along with an idle thrust descent and the aircraft's initial cruise speed in defining the nominal time-to-fly.To define the slowest time-to-fly, the minimum cruise and descent speeds of along with the idle thrust descent were used.The maximum delay is the difference between the slowest time-to-fly and the nominal time-to-fly. +V.E. Fixed-FPA Meet-Time TrajectoriesA set of meet-time trajectories with varying FPA and descent CAS profile was computed for each flight using a modeled metering delay.All the meet-time trajectories must satisfy the allowed planned speed-brake usage.Fuel burn and planned speed-brake usage was calculated for each meet-time trajectory.It turned out that these meet-time trajectories provided all the fuel-burn data needed for comparison of the three strategies, because each of the three strategies must select an FPA from these meet time trajectories.For descent FPAs ranging from -1.8 • to -5.5 • , with an increment of 0.1 • , a meet-time algorithm attempted to compute a fixed-FPA trajectory for each value of the FPA.The meet-time algorithm iterated cruise and descent speeds until the trajectory met the desired time-to-fly within a tolerance of 2 seconds.Cruise and descent speeds were related by the Cruise-Equals-Descent speed mode developed for EDA, 18 designed based on operational considerations.Note that this speed mode does not guarantee the best fuel efficiency in the trajectories. 26igure 5 shows a typical range of FPA-descent-speed combinations defining the set of meet-time trajectories for a flight meeting an STA, labeled as "STA1".Each symbol represents the pair of descent CAS and FPA of a meet-time trajectory.The steepest FPA in this set of meet-time trajectories is -3.6 • .Trajectories steeper than -3.6 • were not included, because their predicted speed-brake usage exceeded the allowed planned speed brake usage.To illustrate how the descent FPA may be selected differently by the three strategies for the "STA1" flight, suppose Strategy 3 is used for selecting the FPA.Strategy 3 selects the FPA of the minimum-fuel trajectory, which is found to be -2.8 • .Now suppose Strategy 1 with an γ univ of -3.4 • is used.A schematic representation of Strategy 1 is shown as the dashed vertical line in Figure 5.The resulting meet-time trajectory with the selected γ univ of -3.4 must belong to the set of meet-time trajectories.Therefore, the fuel burn as a result of the selection by Strategy 1 for this flight is computed from the meet-time trajectory with an FPA of -3.4 • .This fuel burn is generally higher than that of the minimum-fuel trajectory with an FPA of -2.8 • .Now suppose Strategy 2 with an γ 0 of -2.3 • is used.While the selected FPA must be from one of the meet-time trajectories, the relationship between the descent CAS and the FPA of the trajectory must satisfy the FPA function defined in Eq. 1.This FPA function is represented by the solid vertical steps shown in Figure 5.The fuel burn as a result of the selection by Strategy 2 is computed from the trajectory that intercepts with the steps representing the FPA function.This trajectory has an FPA of -2.6 • .Hence all three strategies select distinct FPAs for this flight.If no meet-time trajectories have the FPA defined by Strategy 1 or satisfy the FPA-descent-CAS relationship defined by Strategy 2, then a failure is recorded for this parameter of the strategy.The number of failures is used to determine whether the parameter should be rejected.The meet-time trajectories labeled "STA2" in Figure 5 show an example of a flight with an STA2 later than STA1, in which the meet-time trajectories cross over the gap of the vertical steps defining the FPA function of strategy 2. This case is still considered a "success," and the vertical step with a closer achievable time-to-fly to the metering fix is selected as the FPA. +V.F. Simulation and Data AnalysisA fast-time Monte Carlo simulation was performed to generate the meet-time trajectories for small jet arrivals to DFW.For each arrival flight, a delay time and a descent weight were modeled from the distributions described in Sections V.D and V.C.The meet-time algorithm computed a set of fixed-FPA meet-time trajectories for the test condition specified, using the RUC data in the TS calculation of the trajectories.The fuel burn and speed-brake usage were recorded for further analysis to be described below.All data were categorized by gates and days.Analysis of the meet-time trajectories for all the flights achieves both parts of the methodology described in Section V.A.The following paragraphs detail how the first part of the methodology, which is selection of the parameters, was achieved.The analysis selects γ univ for Strategy 1 from values between -1.8 • and -5.5 • .The analysis selects γ 0 for Strategy 2 from -1.8 • to -3.7 • .For each γ univ and γ 0 , the average fuel burn per flight and feasibility rate are computed for all flights.The feasibility rate is the ratio of the flights with flyable FPAs (total number of "success") to the total flights analyzed.It must be 99% or better for γ univ or γ 0 to be selected (see Section III.E). Figure 6 sketches a notional average fuel burn per flight and feasibility rate as a function of γ univ or γ 0 .The γ univ or γ 0 that yields the least average fuel burn per flight while having a feasibility rate of at least 99% is selected.This figure shows the dramatic effect the planned-speed-brake usage on the feasibility rate and the selection of the parameters.Recall that two levels of planned speed-brake usage were described in Section III.D.Here Speed-Brake-Any (SBANY) allows any FPA to be considered for a flight as long as the modeled speed-brake usage is within the speed-brake capacity for that aircraft.The lenient SBANY condition very often allows the minimum-fuel FPA to be selected, as the computed cutoff γ univ or γ 0 (using the criterion of 99% feasibility rate) for SBANY is frequently steeper than the minimum-fuel γ univ or γ 0 .Speed-Brake-Zero (SB0) represents the case that FPAs for a flight are limited to those that require no speed-brake usage.The SB0 condition has a steeper cutoff FPA that would restrict the acceptable γ univ or γ 0 to shallower, less fuel-efficient values.The preceding selection criteria were used to select γ univ and γ 0 for each of the eight adaptation types listed in Section IV.C.For the Airport-Static adaptation, all flights were analyzed for the selection.For the other adaptations, a subset of the flights was analyzed to select γ univ or γ 0 for a gate and/or a timespan.For example, a total of sixteen pairs of γ univ and γ 0 were selected for the Gate-Season adaptation (four gates times four seasons), each using the flights into a specific gate during a specific season.Selection of the parameters based on the feasibility rate ensures that the vast majority of flights will have flyable FPAs, but it does not consider the variation of winds that can make the flights through some gates on some days particularly difficult to fly.To ensure the feasibility rate for any given day and gate is "tolerable," a feasibility rate of 80% or better for any pair of gate and day is required for all adaptations.For achieving the second part of the methodology, the fuel-burn benefits of trajectories by applying the three strategies are compared.For each flight, a fuel-burn penalty associated with a selected FPA is defined as the extra fuel burn of the trajectory with the selected FPA with respect to the fuel burn of the trajectory with the the minimum-fuel FPA.By definition, Strategy 3 results in zero fuel-burn penalty.The fuel-burn penalties for the FPAs selected by Strategy 1 and Strategy 2 are computed for all flights as a metric for the difference of fuel-burn benefits between the three strategies. +VI. ResultsResults of the fast-time simulation are presented as follows: Section VI.A summarizes the statistics of the number and aircraft types of the arrival flights used in the simulation.Section VI.B shows the selected FPAs for flights using Strategy 3 and discusses their correlation with winds.Sections VI.C and VI.D show the selected FPA and FPA function for Strategy 1 and Strategy 2, respectively, in eight adaptation types and two levels of planned speed-brake usage.Section VI.E compares the fuel-burn benefits of the three strategies. +VI.A. Arrival FlightsTable 4 shows the total number of arrival flights of small jets identified and used in this analysis.Due to occasional periods when the data feed was unavailable, the track data was missing for 18 of the 365 days.Among the days where the track data was available, some data were not recorded due to short interruptions of the data feed.The NE gate was the busiest gate, accounting for 38% of the arrival flights.Figure 7 shows the most frequent small jet aircraft types observed among the arrival flights.The Embraer ERJ 145 and 135 accounted for more than 63% of the fleet, with the other 70 aircraft types making up the rest. +! " #$ % & ' () +VI.B. Strategy 3Strategy 3 selects a minimum-fuel FPA for each flight.Figure 8 shows the distribution of the FPAs selected for all flights under the SBANY condition.In the first four and last two months, steeper FPAs are selected for KARLA and HOWDY while shallower FPAs are selected for DEBBB and FEVER.This is expected since DEBBB and FEVER arrivals experienced mostly tailwinds in these months, while flights entering KARLA and HOWDY experienced mostly headwinds.In these months, both larger fluctuations of the FPAs across days and wider variation of FPAs among flights within a day were observed.This is due to stronger magnitudes and variations of winds, and is consistent with Figure 3.For KARLA (NE), the third quartile FPA were as steep as -3.5 • on Dec. 6th and 20th.For HOWDY (SE), the third quartile of the selected FPAs reached the steepest -3.6 • on Apr.15th.In July and August, the FPAs selected for all four gates are close to -2.6 • and -2.7 • , implying weaker winds with more or less uniformly distributed directions.This observation is again consistent with Figure 3.The direction and magnitude of wind is apparently the strongest discriminator, causing correlated fluctuations of the selected FPAs for all four gates.Although steeper FPAs are typically selected for strong headwinds, a few exceptions occurred, as shown for KARLA on Jan. 31st, Feb. 28th, Dec. 21th, and Dec. 23th, where the first quartile FPA was found to be -1.8 • .In fact, inspection of the data on Feb. 28th revealed that the shallowest FPA of -1.8 • was selected for more than 45% of the flights (not shown in figure).This suggests that very shallow FPAs may be the most fuel-efficient in some conditions of strong headwinds, and will be further investigated in Section VII.In contrast, the shallow first quartile FPA of -2.2 • observed for KARLA on Nov. 27th, nonetheless, is the result of a unusual tailwind on that day.The tailwind corresponds to a headwind for FEVER and resulted in the third quartile FPA peaking at -3.5 • .The shallow first quartile FPA of -2.1 • observed for KARLA on Dec. 6th was possibly a mixed case between the exceptions and the normal days, and was not further investigated.Figure 9 presents distributions of the FPAs selected for flights using Strategy 3 under the SB0 condition.In general, the values in Figure 9 are very close to those in Figure 8.As expected from SB0, some values of FPAs are shallower by 0.1 • and occasionally 0.2 • , because the use of the speed-brake drag is forbidden during descent.These results suggest that the SB0 condition has little impact on the selection of the minimum-fuel FPA on a flight by flight basis. +VI.C. Strategy 1Strategy 1 selects the FPA based on the average fuel burn per flight and the feasibility rate.Although the feasibility rate falls off with steeper FPAs, for SBANY the FPAs satisfying the feasibility rate criterion usually allows the minimum-fuel FPA and even steeper values.Therefore, the fuel-burn criterion is the limiting factor for SBANY.Figure 10 shows the values of γ univ selected for the airport-specific adaptations of Strategy 1.In general, the γ univ selected for an adaptation of shorter timespan fluctuates mostly at or above the γ univ selected for a longer timespan.For the γ univ selected for the Airport-Day adaptation, the fluctuation is larger in winter and smaller in summer.The SB0 condition had a much stronger effect on Strategy 1 than on Strategy 3. Results of the SB0 condition shown on the right hand side of Figure 10 had a huge impact on γ univ , shifting γ univ towards the shallow end by 0.3 • to 0.4 • .The impact of the SB0 condition is further discussed in Section VI.D.Figure 11 shows values of γ univ selected for the gate-specific adaptations of Strategy 1 under the SBANY condition.A similar trend was observed in Figure 11 as in Figure 8.The values of Gate-Day FPA are close to the values of median FPA in Figure 8. Steeper values of γ univ for KARLA and HOWDY are observed in the first four and last two months.On some days, the γ univ selected for KARLA and HOWDY were as steep as -3.4 • .As observed for Strategy 3, the anomaly of -1.8 • selected for KARLA on Feb. 28th was due to the fact that almost half of the flights on that day had -1.8 • as their minimum-fuel FPA.This is investigated further in Sec.VII.Generally speaking, the selected values of γ univ for KARLA and HOWDY are very close, with HOWDY having slightly steeper values of γ univ .The FPAs for DEBBB and FEVER are very close, with DEBBB having slightly shallower values of γ univ .An unusually shallow γ univ of -2.3 • was selected for KARLA on Nov. 27th.The other values of γ univ selected for HOWDY, FEVER, and DEBBB on this day are -2.7 • , -3.2 • , and -2.4 • , respectively.These values suggest an average direction of the wind from the north.Inspection of the RUC data revealed a fairly strong wind of 85 knots from the NNE at altitudes between 20,000 ft and 25,000 ft.Similar values of FPA were observed on Nov. 28th and the wind again came from NNE, although the wind was strongest between 28,000 ft and 35,000 ft.Figure 12 shows values of γ univ selected for the gate-specific adaptations of Strategy 1 under the SB0 condition.The SB0 condition reduced the number of feasible trajectories for each flight and made the feasibility rate fall off.In contrast to the SBANY condition, the feasibility rate of 99% becomes the limiting factor instead of the fuel burn.The overall effect is a shift of 0.3 • to 0.4 • of the γ univ towards shallow values compared to Figure 11.The variation of γ univ from one gate to another reached 1.3 • on Apr.15th, when -3.2 • was selected for HOWDY and -1.9 • was selected for DEBBB.The variation γ univ from one day to another reached 1.4 • for KARLA, when -1.8 • was selected on Feb. 28th and -3.2 • was selected on Nov. 14th.Note that the values of γ univ for the airport-specific adaptations shown in Figure 10 are very close to the values of γ univ selected for DEBBB shown in Figures 11 and12.This is because the selected γ univ must ensure a feasibility rate of 80% for any given gate on any day.Therefore, the γ univ selected for airport adaptations is "almost" completely constrained by the shallowest FPAs selected for DEBBB. +VI.D. Strategy 2Recall that the family of FPA functions described in Section III changes the selected FPA by 0.1 • for every 10 knots of the descent CAS.A selected FPA function is defined by the FPA it yields at 250 knots, denoted as γ 0 .Figure 13 shows γ 0 selected for the airport-specific adaptations of Strategy 2. Similar to the values of γ univ in Figure 10, values of γ 0 for an adaptation of shorter timespan fluctuate mostly at or above values of γ 0 selected for a longer timespan.For values of γ 0 selected for the Airport-Day adaptation, the fluctuation is larger in winter and smaller in summer.Results of the SB0 condition, shown on the right hand side of Figure 13, again had a huge impact on the selected FPA, shifting the values of γ 0 towards the shallow end by 0.3 • to 0.4 • .Figure 14 shows γ 0 selected for the gate-specific adaptations of Strategy 2 under the SBANY condition.The values of γ 0 are roughly shallower than the values of γ univ for Strategy 1 in Figure 11 by 0.1 • .The trends are apparently similar between the two strategies.Observing the values of the parameter for the Gate-Static adaptation for these two strategies, the gates ordered in terms of the steepness of their FPAs are HOWDY, KARLA, FEVER, and DEBBB.Strategy2 Adapted to Gate with SBANY - For the SB0 condition, Figure 15 shows values of γ 0 selected for the gate specific adaptations of Strategy 2. Similar to Strategy 1, the SB0 condition had a huge impact on the selected FPA function, shifting γ 0 to shallower values by 0.4 • to 0.5 • .Similar to Strategy 1, γ 0 selected for KARLA on Feb. 28th was -1.8 • at 250 knots for both SBANY and SB0 conditions.Similar to Strategy 1, the selection of γ 0 in Figure 13 is constrained by the results for DEBBB in Figures 14 and15.Therefore the values of γ 0 in Figure 13 close to those in Figures 14 and15. +VI.E. Fuel Burn ComparisonThe culmination of the preceding analysis is the comparison of fuel consumption across the strategies and the adaptation types studied.While it would be interesting to estimate the actual fuel burn of flights under actual arrival-metering operations, it would not contribute to the choice of a FPA-defining strategy for EDAassisted CDA operations.Instead, the salient question is: "what FPA strategy would provide the best value for implementation?"To facilitate the comparison, results are presented in terms of the average fuel burn per flight, over the year's worth of DFW traffic data, relative to the minimum-fuel solution of Strategy 3. In this way, the results will show how close the simpler strategies (1 and 2), and their adaptations can come to the minimum-fuel solution without requiring real-time pilot-controller communication of FPA just prior to top of descent.The fuel burn comparison was based on trajectories from the freeze horizon before the top-of-descent to the metering fix, and therefore has contributions from both the cruise and the descent segments.Figure 16 shows the fuel-burn penalty per flight computed for Strategies 1 and 2. Overall, two interesting observations can be made.First, even the simplest strategy, a single static FPA adapted for DFW, has the potential to come within 26 lbs of the minimum-fuel solution for each flight on average.This 26 lbs represents the potential benefit of the minimum-fuel (Strategy 3) solution.To put this into perspective, this represents approximately 5% of the typical small-jet arrival transitioning over a 160 nmi segment from cruise to the TRACON boundary.This of course assumes that the modeled winds aloft and traffic conditions used for parameter selection are reasonably accurate.The second observation relates to the potential impact of speed brake usage.While the planned routine use of speed brakes in such an arrival procedure is problematic from a flight operation's point of view, it is very interesting to note that the use of speed brakes has the potential to recover upwards of two-thirds of the difference in fuel burn between the minimum fuel solution and a static adaptation of Strategy 1 or 2. In other words, the planned use of up to the full speed brake drag modeled in this study (SBANY) is more effective at reducing fuel consumption than any of the studied adaptations of Strategy 1 or 2 under the SB0 condition.This highlights the sensitivity of the fuel burn to FPA limits in Strategies 1 and 2 stemming from strong wind conditions.Without the use of speed brakes, the strong tailwind conditions would otherwise prevent Strategy 1 or 2 from selecting FPAs that are far more fuel efficient during the rest of the year, month, or season.For the remainder of this section, only the SB0 condition (no planned use of speed brakes) will be discussed.Returning to the primary focus of this paper, the potential impact of directional and temporal adaptations on the fuel efficiency of Strategies 1 and 2, consider the Strategy1 results on the left side of Figure 16.Relative to the Airport-Static adaptation, adaptation to arrival-gate direction has the potential to reduce the "extra" 26 lbs of fuel per flight by 27%.In other words, publishing a universal FPA for each of the four gates will recover a little more than a quarter of the way to the minimum fuel solution.By comparison, adapting the universal "airport" FPA to season, month and day has the potential of reducing that extra 26 lbs of fuel per flight by 11%, 24%, and 34%, respectively.When combined, the directional and temporal adaptations together have the potential for reducing the 26 lbs of extra fuel burn by 57%.Essentially, the combined adaptation of a universally fixed FPA for each arrival gate, for each day of operations, recovers more than half of the fuel savings of the minimum-fuel solution for each flight.In considering the overall results for Strategy 2 shown in Figure 16, both the directional and temporal adaptations yield similar improvements in fuel efficiency.By "adapting" to the descent speed, Strategy 2 is essentially a surrogate adaptation for metering delay, one of the primary factors being considered in this analysis.As such, this strategy was anticipated to yield a fair amount of benefit under metering conditions.The results indicate that Strategy 2 contributes a 17% reduction in the 26 lbs of extra fuel burn over Strategy 1 for the Airport-Static case, and anywhere from 15-23% depending on the extent of adaptation.The combined effect of Strategy 2 with both directional and temporal adaptation has the potential to achieve 67% of the minimum-fuel (Strategy 3) benefit. +VII. Discussion +VII.A. A Very Shallow Minimum-Fuel FPAThe selection of -1.8 • , the shallowest FPA analyzed, for Strategy 1 for KARLA (NE) on Feb. 28th raised questions as to what kind of headwinds can result in a fuel-efficient shallow FPA.Strong headwinds for the NE gate were observed on many days, but only the calculation for Feb. 28th showed -1.8 • as the overall minimum-fuel FPA.To illustrate this "anomalous" behavior, the meet-time trajectories of two flights were compared below.Figure 17 demonstrates the fuel burn of the meet-time trajectories for two E145 flights entering the NE gate on two distinct days.EGF3271 entered the NE gate on Feb. 1st and had a minimum-fuel FPA at -3.4 • .Such minimum-fuel FPA in the middle of the range of sampled FPAs is characteristic of the meet-time trajectories of most arrival flights.Head winds typically shift the minimum-fuel FPA to a steeper value, while tailwinds usually shift the minimum-fuel FPA to a shallower value.However, a small fraction of flights in the presence of strong headwinds had an inversion of slope at shallow FPAs, causing -1.8 • to be the most fuel-efficient of all FPAs.EGF3272 on Feb. 28th in Figure 17 was one of them.This inversion had been observed in previous work, 8 and seemed to occur most frequently in the presence of strong headwinds.Careful examination of the 114 flights entering the NE gate on Feb. 28th revealed 54 flights (45%) that had -1.8 • as their minimum-fuel FPA.The accumulated fuel-burn advantage of -1.8 • in the 54 flights overpowered the fuel-burn disadvantage in the remaining 60 of the 114 flights and resulted in the selection of -1.8 • as the overall minimum-fuel FPA.A detailed examination of flights of the entire year entering the NE gate showed about 1200 flights that had -1.8 • as their minimum-fuel FPA.The headwinds at their cruise altitudes were typically strong, although mild headwinds and even mild tailwinds were observed for some flights.The results suggest that a strong headwind promotes, but is not a necessary condition for, the inversion of slope.To look for special characteristics of the wind on Feb. 28th, the wind along the route of each flight entering the NE gate on Feb. 28th was further analyzed.It was observed that, for many of the 54 flights that had an inversion, the headwind weakened significantly along the descent segment.Therefore, it was possible that the inversion of slope was highly correlated with the wind gradient with respect to the altitude.Take the two flights shown in Figure 17 as an example.The headwind experienced by EGF2708 decreased from 110 knots at 36,000 ft to 90 knots at 25,000 ft, while the headwind experienced by EGF3272 decreased from 120 knots at 36,000 ft to 60 knots at 25,000 ft.The faster decay of the headwind for EGF3272 might contribute to favoring an early descent.To support the above conjecture, the wind gradients were examined systematically, using a measure of the change of along-the-route wind experienced by a flight defined by Eq. 3:W i = (W (x i , y i , z i , t i , Ψ i ) -W (x i , y i , z i -∆z, t i , Ψ i )) ∆z ,(3)where the subscript i denotes the index of the flight, W represents a measure of the change of the alongthe-route wind with respect to altitude, x, y, z, t represent the flight's initial 4D position for the analysis, Ψ represents the flight's heading, and ∆z represents a characteristic change of altitude.A value of 8,000 ft was chosen for ∆z.Calculation of W i for all flights entering the NE gate showed that, on average, the flights on Feb. 28th experienced the greatest change of the along-the-route wind.The average value of W for the flights entering the NE gate was 4.4 knots per 1,000 ft.Feb. 27th had the second highest average of W of 3.6 knots per 1,000 ft.Among the 125 flights entering the NE gate on Feb. 27th, 31 of them had -1.8 • as their minimum-fuel FPA while the remaining 94 flights had other minimum-fuel FPAs.However, the fuel-burn advantages of -1.8 • for these 31 flights were not enough to make -1.8 • the overall minimum-fuel FPA for all flights.The other days had smaller W and smaller fractions of flights with selected minimum-fuel FPAs of -1.8 • .Although the results showed strong correlation between W and the selection of a shallow FPA, W was by no means the only parameter that could have resulted in a fuel-efficient shallow FPA.In fact, a small fraction of flights that had -1.8 • as their minimum-fuel FPA had very small W i .Future detailed analysis needs to better identify the combination of parameters that promote a shallow minimum-fuel FPA. +VII.B. The Form of the FPA FunctionThe speed-dependent FPA function in Strategy 2 was intended to account for the effect of the descent speed on the minimum-fuel FPA.However, for some gates on certain days, the selected FPA function resulted in more predicted fuel burn per flight than a universal FPA.The following discussion contrasts the relative fuel benefits of Strategies 1 and 2 with two examples.The results were based on the SBANY condition, although results of the SB0 condition followed a similar pattern.Figure 18 shows a specific gate-day pair for which Strategy 2 worked relatively well.The average fuel-burn penalties of the meet-time trajectories of the flights entering the SW gate on Feb. 28th under the SBANY condition were shown using a color map.Due to the fact that the metering delay, chosen from a uniform distribution, maps to a distribution denser in the low end of the descent CAS, trajectories were denser in the low descent CAS region than the high descent CAS region.No trajectories with FPAs steeper than -4.0 • were found in Figure 18 because those trajectories all exceeded the speed-brake drag capacity.The "valley" area shifts roughly by 0.1 • for every 10 knots of the descent CAS, making it possible to "pick up" minimum-fuel trajectories with low fuel burn with the vertical steps of an FPA function illustrated in Figure 5.The resulting fuel-burn penalty is 4.1 lbs with a selected γ 0 of -2.4 • .The fuel-burn penalty of applying Strategy 1 is a less favorable 7.4 lbs with a selected γ univ of -2.4 • .On the other hand, Strategy 2 did not work as well as Strategy 1 for the flights entering KARLA on Feb. 28th, as shown in Figure 19.It can be observed that a universal FPA of -1.8 would pick up the trajectories with slightly lower fuel-burn penalty on the left vertical line of the figure.An FPA function must pick up those trajectories in the middle that have higher fuel-burn penalty than those on the left, therefore resulting in higher overall fuel burn penalty.The selected γ 0 is -1.8 • , and the associated fuel-burn penalty is 9.6 lbs, compared to a slightly lower fuel-burn penalty of 9.0 lbs for a γ univ of -1.8 • selected for Strategy 1.The function form for Strategy 2 can potentially be extended to have as a second parameter the ratio of the change of DCAS and the change of FPA.Both γ 0 and this ratio must be selected via the methodology described in Section V.A.With this more general definition of the FPA function, Strategy 1 becomes a special case of Strategy 2 with a ratio of infinity. +VII.C. Further Adaptation to RoutesGiven the wide range of arrival-route courses feeding a typical arrival gate, there is potential for additional benefit from adapting to specific arrival routes.Figure 2 clearly shows that variation of the ground course for each gate can be as wide as 90 • (see DEBBB, for example).Therefore, individual flights can experience very different wind along their routes even if they enter the same gate at roughly the same time.Particularly for the SB0 condition, few flights in tailwind would shift the selected FPA or FPA function to the shallow end.With adaptation to the routes, different FPAs or FPA functions are selected for different branches of the Standard Terminal Arrival Route.This adaptation is expected to further improve fuel-efficiency. +VII.D. Implementation of the StrategiesLooking towards the operational implementation of continuous descents under metering conditions, the Federal Aviation Administration (FAA) will have to decide upon an approach for defining continuous descent FPAs for small jets.While this paper compares and contrasts the relative fuel efficiency of candidate strategies for defining descent FPAs, other factors will enter the implementation decision, not the least of which will be the complexity and cost of implementing and supporting an FPA procedure.Of the three strategies proposed and analyzed in this paper, Strategy 3 saves the most fuel but is the most costly and complex to implement, due to the requirement of communicating the FPA to the pilot in real time.Strategies 1 and 2 are both simple enough to lend themselves to off-line analysis to select their parameters, publication and dissemination before flight.Strategy 1 has the advantage of defining a single descent FPA for an airport.While appealing in its simplicity, this limits the potential fuel efficiency gained from adjusting the descent profile for operational factors such as speed and prevailing winds.Strategy 2, while slightly more complex in terms of defining FPA as a function of descent speed, has the potential to capture much of the fuel efficiency related to the descent speed.Both strategies lend themselves to fast-time analysis that can support the selection and publication of their parameters a day or more before flight.This would allow the FAA to adapt the appropriate parameters to each airport and disseminate those parameters through the aeronautical information network.Depending on the time horizon chosen for the adaptation (annual, seasonal, monthly, weekly or daily), this information may be made available in several ways.For longer periods of time, the parameters could be included as part of published arrival procedures or flight manual amendments.Alternatively, this information could be provided to pilots as part of their standard pre-flight planning and weather briefing.At the very least, this dissemination approach would be necessary for adaptations performed on a more frequent (e.g., daily) basis.Perhaps the most critical implementation aspect associated with Strategies 1 and 2 is the adaptation of the FPA parameter(s) to specific airports and time horizons.Given the sensitivity of the selected FPA to prevailing winds and the lack of precise wind forecast for longer time horizons, analysis based on historical data or climatological data would lend itself to parameter adaptation on an annual, seasonal or monthly basis.Historical data from previous years may be used for such prediction, assuming a similar pattern of weather and wind.However, the validity of such an assumption remains to be verified.Such uncertainties would raise the lowest achievable fuel-burn penalty of the static, per-season, and per-month adaptation.For shorter adaptation horizons, on the order of daily updates, numerical weather prediction such as the RUC weather forecast would provide precise and relatively accurate forecasts.This would not only improve the fuel efficiency of the FPA or FPA-function adaptation, it would also provide the precision needed to support the adaptation of the FPA or FPA function to specific arrival routes feeding an arrival gate. +VIII. ConclusionThis paper proposed three strategies for choosing the descent flight-path angles (FPAs) for small jets in transition airspace under metering conditions.The three FPA selection strategies are:1. Universally fixed FPA -defines a universal FPA for all small jets arriving to an airport, gate, or route.2. FPA as a function of descent speed -defines different FPAs for different descent speeds.3. Minimum-fuel FPA -computes minimum-fuel FPA for each flight, but requires communication of the FPA to the pilot in real time.The three strategies vary in operational complexity.The minimum-fuel FPA strategy served as the reference point for the fuel-burn benefits.A full year's worth of traffic data arriving at the Dallas/Fort Worth (DFW) airport during 2011 was used for the analysis of the three strategies.The FPA of Strategy 1 and the FPA function of Strategy 2 were selected based on fuel burn and planned speed-brake usage of the meet-time trajectories computed for the flights.Two levels of planned speed-brake usage were analyzed to explore the impact of speed-brake usage on the selected FPA, FPA function, and the resulting fuel burn.Results showed that the universally fixed FPA adapted for the DFW Airport had only 26 lbs of extra fuel burn relative to the minimum-fuel solution for each flight on average.To improve fuel-efficiency by reducing wind variation along the route, adaptations of Strategies 1 and 2 to the airport, the arrival gate (direction), and various timespans, were defined and analyzed.The adapted FPAs vary significantly, up to 1.3 • from one arrival gate to another, and up to 1.4 • from one day to another.Adaptation to the arrival-gate reduced the extra fuel burn per flight by up to 27%.Adaptation to the day alone reduced the extra fuel burn per flight by up to 34%.Combining the directional and temporal adaptations reduced the extra fuel burn by 57%.The FPA function strategy, considered as a surrogate adaptation to the descent speed, reduced the extra fuel burn per flight by 17%.The combined effect of the directional, temporal, and speed adaptation recovered 67% of extra fuel burn relative to the minimum-fuel solution.In conclusion, the simple forms of the universally-fixed FPA Strategy and FPA function Strategy, together with the fact that they do not require explicit ground-air communication of the FPA in real time, make them favorable for implementation.Adaptations to direction and timespans reduced up to two-thirds of the extra fuel burn relative to the minimum-fuel solution.For future work, it would be interesting to analyze the adaptations of these strategies to other airports and/or routes.It would also be interesting to analyze the variation of winds with respect to years, which would help determine if historical data can provide adequate estimates of the winds for longer time horizons for the purpose of selecting FPAs.Figure 1 .1Figure 1.Strategy 3 selects the minimum-fuel FPA. +Figure 2 .2Figure 2. Tracks of arrival small jets into DFW on July 28th, 2011. +Figure 3 .3Figure 3. Wind along typical arrival routes at 35,000 ft in Fort Worth Center, 2011.Note that a negative tailwind represents a headwind. +/Decel to cruise Mach 2. Cruise at Mach 3. Descent at constant Mach 4. Descent at constant CAS 5. Final speed adjustment +Figure 4 .4Figure 4.A general vertical profile that contains five segment types. +Figure 5 .5Figure 5. Applying Strategies 1 and 2 to the selection of FPAs. +Figure 6 .6Figure 6.Notional representation of the average fuel burn and feasibility rate.The feasibility rate falls off at steeper values of γ univ or γ0. +Figure 7 .7Figure 7.Most frequent arrival small jet types observed in Dallas/Fort Worth Airport, 2011. +Figure 8 .8Figure 8.The first quartile, third quartile, and median FPAs selected using Strategy 3 under the SBANY condition. +Figure 9 .9Figure 9.The first quartile, third quartile, and median FPAs selected using Strategy 3 under the SB0 condition. +Figure 11 .11Figure 11.Values of γ univ selected for the gate-specific adaptations of Strategy 1 under the SBANY condition. +Figure 12 .12Figure 12.Values of γ univ selected for the Gate-Static, Gate-Season, Gate-Month, and Gate-Day adaptations of Strategy 1 under the SB0 condition. +Figure 13 .13Figure 13.Values of γ univ selected for the airport-specific adaptations of Strategy 2. +Figure 14 .14Figure 14.Values of γ0 selected for the Gate-Static, Gate-Season, Gate-Month, and Gate-Day adaptations of Strategy 2 under the SBANY condition. +Figure 15 .15Figure 15.Values of γ0 selected for the Gate-Static, Gate-Season, Gate-Month, and Gate-Day adaptations of Strategy 2 under the SB0 condition. +Figure 16 .16Figure 16.Fuel burn penalty calculated for various adaptations of Strategies 1 and 2. +Figure 17 .17Figure 17.Fuel burn of the meet-time trajectories for two flights entering the NE gate.While EGF2708 had a minimum-fuel FPA of -3.4 • , EGF3272 had a minimum-fuel FPA of -1.8 • . +Figure 18 .18Figure 18.Fuel-burn penalty for the meet-time trajectories entering FEVER (SW) on Feb. 28th under the SBANY condition. +Table 1 .1The FPA function used in the flight trial at Denver Center in 2010 to issue without consulting the pilot.The high end represents the Skywest Airlines' preferred descent speed for their Bombardier CRJ aircraft.Participating Skywest pilots determined the FPA to fly by using the descent CAS issued by the EDA clearance.Table1was developed in collaboration with Skywest, and validated in a piloted simulation at their training facility.The specific values were selected with consideration of flyability, avoiding the use of speed brake or relatively high power settings, and fuel efficiency.Although it was adequate for the purpose of the flight trial, the table's sparse values of FPA leave large gaps in the achievable ranges of time-to-fly.Also, the FPA function in TableRange of Descent CAS (knots) FPA( • )250-260-2.8270-280-3.1290-300-3.4310-320-3.8controllers are allowed +Table 2 .2The types of adaptations categorized by their granularity in location and timeAirspace/DirectionTimeNAS Airport Arrival Gate Arrival RouteStatic15Season26Month37Day48Hour +Table 3 .3The eight adaptation types analyzed for the DFW traffic# NameFPA/FPA Function Defined for1 Airport-Staticthe airport2 Airport-Season the airport by season3 Airport-Month the airport by month4 Airport-Daythe airport for each day5 Gate-Staticeach arrival gate6 Gate-Seasoneach arrival gate by season7 Gate-Montheach arrival gate by month8 Gate-Dayeach arrival gate for each day +Table 4 .4Total number of arrival flights of small jets in the analysisMonth DaysNESENWSWMissing DaysJan2833222515120411331, 19, 24Feb2124491761971909 11, 16, 18-20, 23, 25Mar3036292644147412656Apr302963268616141468May29290824561428126521, 22Jun303465261518151393Jul29290823111791118916, 23Aug28282324041772121321, 27, 28Sep303256247116431313Oct313422246515701409Nov303424255516051454Dec313335262016281399Total347 37904 29503 18515 15410 +Figure 10.Values of γ univ selected for the airport-specific adaptations of Strategy 1.Strategy1 Adapted to AirportAirport-StaticAirport-SeasonAirport-MonthAirport-DaySBANYSB0-3.4-3.4univ ( )-3.0 -2.6-3.0 -2.6-2.2-2.2-1.8-1.8JANFEBMARAPRMAYJUNJULAUGSEPOCTNOVDECJANFEBMARAPRMAYJUNJULAUGSEPOCTNOVDECStrategy1 Adapted to Gate with SBANYGate-StaticGate-SeasonGate-MonthGate-DayDEBBB (NW)KARLA (NE)-3.4-3.4univ ( )-3.0 -2.6-3.0 -2.6-2.2-2.2-1.8-1.8JANFEBMARAPRMAYJUNJULAUGSEPOCTNOVDECJANFEBMARAPRMAYJUNJULAUGSEPOCTNOVDECFEVER (SW)HOWDY (SE)-3.4-3.4univ ( )-3.0 -2.6-3.0 -2.6-2.2-2.2-1.8-1.8JANFEBMARAPRMAYJUNJULAUGSEPOCTNOVDECJANFEBMARAPRMAYJUNJULAUGSEPOCTNOVDEC +are veryStrategy2 Adapted to Gate with SB0Gate-StaticGate-SeasonGate-MonthGate-DayDEBBB (NW)KARLA (NE)-3.4-3.40 ( )-3.0 -2.6-3.0 -2.6-2.2-2.2-1.8-1.8JANFEBMARAPRMAYJUNJULAUGSEPOCTNOVDECJANFEBMARAPRMAYJUNJULAUGSEPOCTNOVDECFEVER (SW)HOWDY (SE)-3.4-3.40 ( )-3.0 -2.6-3.0 -2.6-2.2-2.2-1.8-1.8JANFEBMARAPRMAYJUNJULAUGSEPOCTNOVDECJANFEBMARAPRMAYJUNJULAUGSEPOCTNOVDEC +•Figure 19.Fuel-burn penalty for the meet-time trajectories entering KARLA (NE) on Feb. 28th under the SBANY condition. + a In this paper, an FPA is considered flyable if the aircraft can maintain both the FPA and the descent speed necessary to meet its assigned metering time. + b The contribution of the arrival small jets to the other airports in the TRACON is negligible. + + + + +AcknowledgmentsWe thank Richard Coppenbarger and Harry Swenson for helpful discussions. + + + + + + + + + Practice for Application of Federal Aviation Administration (FAA) Federal Aviation Regulations Part 21 Requirements to Unmanned Aircraft Systems (UAS) + 10.1520/f2505-07 + + March 2011 + ASTM International + + + Federal Aviation Administration + "FAA's NextGen Implementation Plan," Federal Aviation Administration, March 2011. + + + + + The Next-Generation Air Transportation System's Joint Planning Environment: A Decision Support System + + EdgarWaggoner + + + ScottGoldsmith + + + JoshElliot + + 10.2514/6.2009-7011 + + + 9th AIAA Aviation Technology, Integration, and Operations Conference (ATIO) + + American Institute of Aeronautics and Astronautics + Oct. 2009 + + + "Concept of Operations for the Next Generation Air Transportation System," Joint Planning and Development Office (JPDO), Oct. 2009. + + + + + Field Evaluation of the Tailored Arrivals Concept for Datalink-Enabled Continuous Descent Approach + + RichardACoppenbarger + + + RobWMead + + + DouglasNSweet + + 10.2514/1.39795 + + + Journal of Aircraft + Journal of Aircraft + 0021-8669 + 1533-3868 + + 46 + 4 + + July-August 2009 + American Institute of Aeronautics and Astronautics (AIAA) + + + Coppenbarger, R. A., Mead, R. W., and Sweet, D. N., "Field Evaluation of the Tailored Arrivals Concept for Datalink- Enabled Continuous Descent Approach," Journal of Aircraft, Vol. 46, No. 4, July-August 2009, pp. 1200-1209. + + + + + Development, Design and Flight Test Evaluation of Continuous Descent Approach Procedure in FIR Zagreb + + DorisNovak + + + TinoBucak + + + TomislavRadišić + + 10.7307/ptt.v21i5.247 + + + PROMET - Traffic&Transportation + PROMET + 0353-5320 + 1848-4069 + + 21 + 5 + + Sept. 2009 + Faculty of Transport and Traffic Sciences + + + Novak, D., Bucak, T., and Radisić, T., "Development, Design and Flight Test Evaluation of Continuous Descent Approach Procedure in FIR Zagreb," PROMET -Traffic & Transportation, Vol. 21, No. 5, Sept. 2009, pp. 319-329. + + + + + Continuous Descent Approach: Design and Flight Test for Louisville International Airport + + John-Paul B.Clarke + + + NhutTHo + + + LilingRen + + + JohnABrown + + + KevinRElmer + + + Kwok-OnTong + + + JosephKWat + + 10.2514/1.5572 + + + Journal of Aircraft + Journal of Aircraft + 0021-8669 + 1533-3868 + + 41 + 5 + + 2004 + American Institute of Aeronautics and Astronautics (AIAA) + + + Clarke, J.-P. B., Ho, N. T., Ren, L., Brown, J. A., Elmer, K. R., Tong, K.-O., and Wat, J. K., "Continuous Descent Approach: Design and Flight Test for Louisville International Airport," Journal of Aircraft, Vol. 41, No. 5, 2004, pp. 1054-1066. + + + + + FAA Aviation Forecasts: Fiscal Years 1981-1992. Federal Aviation Administration, U.S. Department of Transportation, Washington, D.C. 20591. 1980. 69p + 10.1177/004728758102000159 + + + Journal of Travel Research + Journal of Travel Research + 0047-2875 + 1552-6763 + + 20 + 1 + + 2011 + SAGE Publications + + + "FAA Aerospace Forecast Fiscal Years 2012-2032," Federal Aviation Administration, 2011. + + + + + Relative Significance of Trajectory Prediction Errors on an Automated Separation Assurance Algorithm + + TALauderdale + + + ACCone + + + ARBowe + + + + Proceedings of the 9th USA/Europe Air Traffic Management R&D Seminar + the 9th USA/Europe Air Traffic Management R&D Seminar + + June 2011 + + + Lauderdale, T. A., Cone, A. C., and Bowe, A. R., "Relative Significance of Trajectory Prediction Errors on an Automated Separation Assurance Algorithm," Proceedings of the 9th USA/Europe Air Traffic Management R&D Seminar , June 2011. + + + + + Experimental Approach for Efficient Arrival Procedures with Fixed-Flight Path Angle Descent + + MGWu + + + SMGreen + + 10.1109/icnsurv.2019.8735138 + + + 2019 Integrated Communications, Navigation and Surveillance Conference (ICNS) + + IEEE + 2011-215992, Nov. 2011 + + + Wu, M. G. and Green, S. M., "Analysis of Fixed Flight Path Angle Descents for the Efficient Descent Advisor," NASA/TM-2011-215992, Nov. 2011. + + + + + Descent Profile Options for Continuous Descent Arrival Procedures within 3d Path Concept + + K.-OTong + + + ESchoemig + + + DBoyle + + + JScharl + + + AHaraldsdottir + + + + Proceedings of the IEEE/AIAA 26th Digital Avionics Systems Conference + the IEEE/AIAA 26th Digital Avionics Systems Conference + + Oct. 2007 + + + Tong, K.-O., Schoemig, E., Boyle, D., Scharl, J., and Haraldsdottir, A., "Descent Profile Options for Continuous Descent Arrival Procedures within 3d Path Concept," Proceedings of the IEEE/AIAA 26th Digital Avionics Systems Conference, Oct. 2007, pp. + + + + + Sensitivity Studies of 4D Descent Strategies in an Advanced Metering Environment + + KennethHIzumi + + 10.23919/acc.1986.4789026 + + + 1986 American Control Conference + + IEEE + 1986. June 1986 + + + + Izumi, K. H., "Sensitivity Studies of 4D Descent Strategies in an Advanced Metering Environment," Proceedings of the American Control Conference, 1986 , June 1986, pp. 687-692. + + + + + Sensitivity Studies of 4D Descent Strategies in an Advanced Metering Environment + + KennethHIzumi + + + RWSchwab + + + JLGroce + + + ACoote + + 10.23919/acc.1986.4789026 + + + 1986 American Control Conference + + IEEE + Dec. 1986 + + + Tech. Rep. NASA CR-178093 + Izumi, K. H., Schwab, R. W., Groce, J. L., and Coote, A., "An Evaluation of Descent Strategies for TNAV-Equipped Aircraft in an Advanced Metering Environment ATOPS," Tech. Rep. NASA CR-178093, Boeing Commercial Airplane Company, Dec. 1986. + + + + + 4 D fuel-optimal guidance in the presence of winds + + AChakravarty + + 10.2514/6.1983-2242 + AIAA-1983-2242 + + + Guidance and Control Conference + Gatlinburg, Tenn + + American Institute of Aeronautics and Astronautics + 1983 + + + Chakravarty, A., "Four-Dimensional Fuel-Optimal Guidance in the Presence of Winds," Proceedings of the AIAA Guid- ance and Control Conference, AIAA-1983-2242, Gatlinburg, Tenn., 1983. + + + + + Distributed Scheduling Architecture for Multi-Center Time-Based Metering + + StevenLandry + + + ToddFarley + + + JohnFoster + + + StevenGreen + + + TyHoang + + + GregoryWong + + 10.2514/6.2003-6758 + AIAA-2003-6758 + + + AIAA's 3rd Annual Aviation Technology, Integration, and Operations (ATIO) Forum + + American Institute of Aeronautics and Astronautics + Nov. 2003 + + + Landry, S., Farley, T., Foster, J., Green, S., Hoang, T., and Wong, G. L., "Distributed Scheduling Architecture for Multi-Center Time-Based Metering," Proceedings of the AIAA Aviation Technology, Integration, and Opterations Conference, AIAA-2003-6758, Nov. 2003. + + + + + Design and Operational Evaluation of the Traffic Management Advisor at the Fort Worth Air Route Traffic Control Center + + HNSwenson + + + THoang + + + SEngelland + + + DVincent + + + TSanders + + + BSanford + + + KHeere + + + + Proceedings of the 1st USA + the 1st USA + + June 1997 + + + Swenson, H. N., Hoang, T., Engelland, S., Vincent, D., Sanders, T., Sanford, B., and Heere, K., "Design and Operational Evaluation of the Traffic Management Advisor at the Fort Worth Air Route Traffic Control Center," Proceedings of the 1st USA/Europe Air Traffic Management R&D Seminar , June 1997. + + + + + The Efficient Descent Advisor: Technology Validation and Transition + + RichardCoppenbarger + + + MiwaHayashi + + + GauravNagle + + + DouglasSweet + + + RenanSalcido + + 10.2514/6.2012-5611 + + + 12th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference and 14th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference + + American Institute of Aeronautics and Astronautics + + + + Sept. 2012 (to be published + Coppenbarger, R. A., Nagle, G., Sweet, D., and Hayashi, M., "The Efficient Descent Advisor: Technology Validation and Transition," Proceedings of the AIAA Aircraft Technology, Integration, and Opterations Conference, Sept. 2012 (to be published). + + + + + Human-in-the-Loop Simulation of Three-Dimensional Path Arrival Management with Trajectory Error + + GauravNagle + + + DouglasSweet + + + GregoryCarr + + + ValentinoFelipe + + + AndrewTrapani + + + RichardCoppenbarger + + + MiwaHayashi + + 10.2514/6.2011-6877 + AIAA-2011-6877 + + + 11th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Virginia Beach, VA + + American Institute of Aeronautics and Astronautics + 2011 + + + Nagle, G., Sweet, D., Carr, G., Felipe, V., Trapani, A., Coppenbarger, R., and Hayashi, M., "Human-in-the-Loop Simulation of the Efficient Descent Advisor for 3D Path Arrival Management," Proceedings of the AIAA Aircraft Technology, Integration, and Opterations Conference, AIAA-2011-6877, Virginia Beach, VA, 2011. + + + + + Development and Testing of Automation for Efficient Arrivals in Constrained Airspace + + RCoppenbarger + + + GDyer + + + MHayashi + + + RLanier + + + LStell + + + DSweet + + ICAS2010-11.11.3 + + + Proceedings of the 27th International Congress of the Aeronautical Sciences + the 27th International Congress of the Aeronautical SciencesNice, France + + 2010 + + + Coppenbarger, R., Dyer, G., Hayashi, M., Lanier, R., Stell, L., and Sweet, D., "Development and Testing of Automation for Efficient Arrivals in Constrained Airspace," Proceedings of the 27th International Congress of the Aeronautical Sciences, ICAS2010-11.11.3, Nice, France, 2010. + + + + + Design and Development of the En Route Descent Advisor (EDA) for Conflict-Free Arrival Metering + + RichardCoppenbarger + + + RichardLanier + + + DougSweet + + + SusanDorsky + + 10.2514/6.2004-4875 + AIAA- 2004-4875 + + + AIAA Guidance, Navigation, and Control Conference and Exhibit + + American Institute of Aeronautics and Astronautics + Aug. 2004 + + + Coppenbarger, R. A., Lanier, R., Sweet, D., and Dorsky, S., "Design and Development of the En Route Descent Advisor (EDA) for Conflict-Free Arrival Metering," Proceedings of the AIAA Guidance, Navigation, and Control Conference, AIAA- 2004-4875, Aug. 2004. + + + + + Improved Aircraft Path Stretch Algorithms for the En Route Descent Advisor (EDA) + + TysenMueller + + + DavidSchleicher + + + RichardCoppenbarger + + 10.2514/6.2003-5571 + AIAA-2003-5571 + + + AIAA Guidance, Navigation, and Control Conference and Exhibit + + American Institute of Aeronautics and Astronautics + Aug. 2003 + + + Mueller, K. T., Schleicher, D. R., and Coppenbarger, R. A., "Improved Aircraft Path Stretch Algorithms for the En Route Descent Advisor (EDA)," Proceedings of the AIAA Guidance, Navigation, and Control Conference, AIAA-2003-5571, Aug. 2003. + + + + + En route Descent Advisor concept for arrival metering + + StevenGreen + + + RobertVivona + + 10.2514/6.2001-4114 + AIAA-2001-4114 + + + AIAA Guidance, Navigation, and Control Conference and Exhibit + + American Institute of Aeronautics and Astronautics + Aug. 2001 + + + Green, S. M. and Vivona, R. A., "En Route Descent Advisor Concept for Arrival Metering," Proceedings of the AIAA Guidance, Navigation, and Control Conference, AIAA-2001-4114, Aug. 2001. + + + + + An Hourly Assimilation–Forecast Cycle: The RUC + + StanleyGBenjamin + + + DezsöDévényi + + + StephenSWeygandt + + + KevinJBrundage + + + JohnMBrown + + + GeorgAGrell + + + DongsooKim + + + BarryESchwartz + + + TatianaGSmirnova + + + TracyLorraineSmith + + + GeoffreySManikin + + 10.1175/1520-0493(2004)132<0495:ahactr>2.0.co;2 + + + Monthly Weather Review + Mon. Wea. Rev. + 0027-0644 + 1520-0493 + + 132 + 2 + + 1998 + American Meteorological Society + + + Benjamin, S. G., Brown, J. M., Brundage, K. J., Schwartz, B., Smirnova, T., Smith, T. L., Morone, L. L., and Dimego, G., "The Operational RUC-2. Preprints," Proceedings of the 16th Conference on Weather Analysis and Forecasting, Amer. Meteor. Soc., 1998, pp. 249-252. + + + + + The Trajectory Synthesizer Generalized Profile Interface + + AlanLee + + + XavierBouyssounouse + + + JamesMurphy + + 10.2514/6.2010-9138 + AIAA-2010-9138 + + + 10th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + + American Institute of Aeronautics and Astronautics + Sept. 2010 + + + Lee, A. G., Bouyssounouse, X., and Murphy, J. R., "The Trajectory Synthesizer Generalized Profile Interface," Proceed- ings of the 10th AIAA Aviation Technology, Integration, and Operations Conference, AIAA-2010-9138, Sept. 2010. + + + + + Trajectory Synthesis for Air Traffic Automation + + RhondaSlattery + + + YiyuanZhao + + 10.2514/2.4056 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 20 + 2 + + March 1997 + American Institute of Aeronautics and Astronautics (AIAA) + + + Slattery, R. and Zhao, Y., "Trajectory Synthesis for Air Traffic Automation," Journal of Guidance, Control and Dynam- ics, Vol. 20, No. 2, March 1997, pp. 232-238. + + + + + Design of Center-TRACON Automation System + + HErzberger + + + TJDavis + + + SMGreen + + + + Machine Intelligence in Air Traffic Management, Andre Benoit + + Oct. 1993 + + + + AGARD CP-538 + Erzberger, H., Davis, T. J., and Green, S. M., "Design of Center-TRACON Automation System," Machine Intelligence in Air Traffic Management, Andre Benoit, ed., AGARD CP-538 , Oct. 1993, pp. 11-1-11-12. + + + + + Advanced Aircraft Performance Modeling for ATM: Enhancements to the Bada Model + + ANuic + + + CPoinsot + + + M.-GIagaru + + + EGallo + + + FANavarro + + + CQuerejeta + + 10.1109/dasc.2005.1563320 + + + 24th Digital Avionics Systems Conference + + IEEE + Nov. 2005 + + + Nuic, A., Poinsot, C., Iagaru, M.-G., Gallo, E., Navarro, F. A., and Querejeta, C., "Advanced Aircraft Performance Modeling for ATM: Enhancements to the BADA Model," Proceedings of the IEEE/AIAA 24th Digital Avionics Systems Conference, Nov. 2005. + + + + + Comparison of Fuel Consumption of Descent Trajectories under Arrival Metering + + TasosNikoleris + + + GanoChatterji + + + RichardCoppenbarger + + 10.2514/6.2012-4818 + AIAA-2012- 4818 + + + AIAA Guidance, Navigation, and Control Conference + + American Institute of Aeronautics and Astronautics + Aug. 2012 + + + Nikoleris, A. N., Chatterji, G. B., and Coppenbarger, R. A., "Comparison of Fuel Consumption of Alternative Descent Trajectories under Arrival Metering," Proceedings of the AIAA Guidance, Navigation, and Control Conference, AIAA-2012- 4818, Aug. 2012. + + + + + + diff --git a/file796.txt b/file796.txt new file mode 100644 index 0000000000000000000000000000000000000000..c8e3124727d83cc367eaccb806c87a25f9e5a4ff --- /dev/null +++ b/file796.txt @@ -0,0 +1,751 @@ + + + + +IntroductionThe Efficient Descent Advisor (EDA) [1][2][3][4][5] is being developed by NASA as a key capability for the Next-Generation Air Transportation System (NextGen) [6][7][8].A trajectory-based decision-support tool intended for use by controllers working in FAA en-route air traffic control facilities, EDA is capable of generating dynamic Continuous Descent Approach (CDA) trajectories for arrivals transitioning from en-route to terminal airspace.The advisories generated by EDA take into account airspace restrictions, aircraft performance, atmospheric conditions, conflict avoidance, and the required time of arrival satisfying the time-based metering schedule computed by the Traffic Management Advisor (TMA) [9].TMA specifies at which time each airplane is required to cross a meter fix located at the Terminal Radar Approach Control Facilities (TRACON) boundary for optimal arrival throughput.A series of human-in-the-loop (HITL) simulations were conducted in 2009 through 2011 for testing the concept of EDA [1,2].These simulations focused on aircraft equipped with performance-based flight management systems (FMS).With a descent speed given by the EDA advisory, the vertical navigation (VNAV) of a performance-based FMS generates an idle-or near-idle thrust descent trajectory for that descent speed.However, these HITL simulations did not incorporate fixed flight path angle (FPA) descent procedures employed by regional (RJ) and business jet (BJ) types; these represent about one-third of aircraft operations today.There are no standards for such procedures to define the FPA to be flown, and actual operations can vary greatly.Any practical EDA clearances for BJs or RJs must allow for fixed-FPA descent procedures.Limited analysis of fixed-FPA descents exists in the literature [10][11][12], and the sensitivity of feasible, fuelefficient FPAs to varying environmental conditions is far from understood.This work attempts to obtain insight into fixed-FPA descents and to design practical fixed-FPA descent procedures.It investigates the effect of FPA on the fuel burn of the trajectory, analyzes the sensitivity of the fueloptimal FPA to flight conditions, and proposes three FPA selection strategies.Three different speed brake conditions are imposed on each strategy to model different levels of robustness requirements.Monte Carlo simulations of test conditions are performed to optimize and evaluate these selection strategies.This paper is organized as follows: Section 2 gives background information on fixed-FPA descents, reviewing past work and discussing factors to consider in the design of descent procedures.Section 3 describes modeling schemes and methods for solving the intrinsic meet-time problem in EDA.Section 4 examines in depth a representative EDA test condition and investigates the sensitivity of the fuel burn and speed brake usage to the FPA.Section 4 further studies the sensitivity of the fuel-optimal FPA to the variation of wind and target time.Section 5 presents three FPA selection strategies, three speed brake conditions, and discusses the sampling of realistic test conditions defined by wind, weight, and target time.Results of comparison of the three FPA selection strategies are also presented in Section 5. Section 6 discusses in more detail a few related observations of the results given in Sections 4 and 5. Finally, we conclude in Section 7 and discuss future work to be done. +BackgroundLarger jets equipped with performance-based flight management systems (FMS) are capable of generating and flying idle-thrust descents.Idle-thrust descents are intrinsically sensitive to the aircraft's performance parameters, the descent speed profile, and atmospheric conditions [13,14], and predictions of the idle-thrust descents have proved challenging [15,16].Fixed-FPA descents, on the other hand, have the potential of being more predictable by the ground automation tools.Most BJs and RJs are equipped with a kinematic FMS that can provide guidance for fixed-FPA descents.This type of FMS, however, cannot guide idle-thrust descents.Large commercial jets equipped with a performance-based FMS, on the contrary, do not have built-in capabilities for executing fixed-FPA descents.Nonetheless, potential procedures to execute fixed-FPA descents using performance-based FMS have been suggested [10].Field tests for fixed-FPA procedures were conducted at Denver Center in 2010 using the FAA Global 5000 test aircraft and participating Skywest regional flights to understand the feasibility of pilot/controller procedures and determine trajectory prediction accuracies.Pilots determined the FPA to fly using a published function of the descent calibrated airspeed (CAS).However, no attempt was made to optimize the function for the tests.Many factors should be considered for the design of fixed-FPA descent procedures, but robustness, i.e., the ability to consistently execute a continuous descent, out-weighs all other factors.Steep descent angles will not be flyable under certain combinations of wind, speed, and weight conditions.Even if a steep descent is achievable with the use of speed brakes, many pilots are reluctant, if not unwilling, to use them because of noise and ride discomfort.On the other hand, shallow descent angles burn more fuel, increasing cost and environmental impact.Pilot procedures, airspace restrictions, aircraft equipage, and traffic separation should also be taken into account.A fixed-FPA descent procedure must also provide a way to define the FPA for both controllers and pilots.Analysis of various descent strategies has been done in a pioneering work [11,12] by Izumi et al. at Boeing.They compared idle-thrust descent, fixed-FPA descent, and fuel-optimal descent strategies in a metering environment in terms of their fuel burn and the mixed traffic throughput they achieve.Although constant time to fly is assumed, their work did not explore the variation of wind or the FPA.Tong et al. have considered the design of fixed-FPA descent strategies within the 3D-PAM concept [10], which is enabled by EDA capabilities.Their work touched upon the variation of FPA on fuel burn, but the trajectories were compared on a basis of identical descent speeds instead of identical times to fly.This work will explore the variation of wind, weight, and time-to-fly, quantifying their effect on the fuel-optimal FPA.Idle-thrust descents have been frequently referred to as fuel-optimal descents in the literature [10,17,18].Although idle thrust is fuel-optimal for the descent segment alone, it does not necessarily achieve fuel-optimal trajectories overall.Izumi et al. [12] compared computed fuel-optimal trajectories using the singular perturbation theory.They made a clear distinction between a fuel-optimal trajectory and an idle-thrust trajectory.The fuel-optimal trajectory for B747 showed less fuel burn and a much earlier top-of-descent than the idle-thrust trajectory.In this work we compare the fueloptimal fixed-FPA descent to the corresponding idle-thrust descent under the same test condition in the context of a constant-time-to-fly problem.These comparisons will show that a fixed-FPA descent can actually burn less fuel than an idle-thrust descent. +Modeling Schemes and MethodsConsider a typical arrival flight in en-route airspace that has just entered the TMA freeze horizon, where the TMA schedules and freezes the flight's scheduled time of arrival (STA) at the meter fix [19].In heavy traffic conditions, TMA imposes flow management by ensuring sufficient time spacing between aircraft at the metering fix.In such conditions the STA usually delays the aircraft with respect to its predicted nominal arrival time, denoted as the estimated time of arrival (ETA).EDA attempts to absorb the time delay by computing a speed advisory that meets the STA by reducing the cruise and descent speeds of the aircraft's trajectory.If speed changes are not enough to absorb the delay time, EDA stretches the horizontal route in addition to reducing speeds.During the HITL simulations altitude changes are occasionally issued in combination with speed changes to avoid conflicts [2].In this work, we constrain the solution space to those aircraft requiring speed changes only.While the analysis in this work is based on a simple, direct route to the meter fix, it can be readily extended to general horizontal routes. +Vertical Profile ModelingThe model of the vertical profile for an arrival flight in the en-route airspace, when described in terms of the altitude's status, consists of a cruise segment, a descent segment and, if necessary, a level deceleration segment to the meter fix.Taking into account the distinct control variables applied during the flight, we further break down the cruise and descent segments to a combination of some or all of the following five segment types:1.An acceleration/deceleration cruise segment at the cruise altitude 2. A constant airspeed cruise segment at the cruise altitude 3. A constant Mach descent segment 4. A constant calibrated airspeed (CAS) descent segment 5. A deceleration level segment at the meter-fix crossing altitude.Figure 1 demonstrates a general vertical profile that has all the five segment types.Note that segment 1 exists only if the cruise speed in the advisory is different than the aircraft's current speed.Segment 3 exists only if the descent CAS is greater than the cruise CAS in the advisory.Segment 5 exists only if the descent CAS in the advisory is greater than the meter-fix crossing speed.For this work each segment is modeled by fixing two control parameters.For a cruise segment the model fixes a parameter, which is engine control or an airspeed, and sets the FPA to zero.For the constant speed descent segments, the model fixes an airspeed in Mach or CAS and fixes another parameter, which can be engine control or FPA. +Trajectory SynthesisFor the purpose of this study, the Center-TRACON Automation System (CTAS) Trajectory Synthesizer (TS) [20][21][22] is used to simulate the trajectories to be analyzed.TS takes as its input a flight's current position and velocity, flight plan, airspace restrictions, a model of pilot's intent, wind and temperature aloft, and aircraft performance model.The output is a trajectory defined by the aircraft's 3-D position and velocity as a function of time along with computed forces and fuel burn.TS uses an aircraft performance model database that has been validated and improved by various research projects [23].We selected a model representative of a mid-size, narrow-body, twin-engine jet airliner with a typical descent weight of approximately 170,000 lbs.Although this model represents an aircraft larger than RJs and BJs, it has been validated against other tools such as Boeing's INFLT and should provide high-fidelity results.The minimum cruise and descent CAS speeds for delay advisories are set at 250 knots and the maximum cruise mach and descent CAS are set at 0.84 and 350 knots, respectively.The maximum drag that can incur as a result of speed brake deployment, defined as the speed brake capacity, is modeled as a maximum speed brake drag coefficient.The chosen aircraft model in TS did not have parameters for speed brake.We asked a few pilots to estimate the maximum speed reduction the full speed brake deployment can effect at a few typical descent conditions.Their estimates of speed reductions were used to derive an empirical maximum speed brake drag coefficient. +Initial and Environmental ConditionsThroughout the rest of the modeling work a hypothetical flight is heading toward the meter fix from a distance of 150 nautical miles (nmi), a typical distance for the TMA freeze horizon [19].The aircraft is initially level at 35,000 ft and its initial airspeed is 0.80 Mach, typical for aircraft of this type.The meter-fix crossing restrictions are an altitude of 10,000 ft and a CAS of 250 knots or less.Wind is modeled as a function linear in altitude that has intercept zero at sea level.The direction of wind is the same at all altitudes and has only horizontal components.This linear approximation is reasonable within the range of altitude of interest but becomes unrealistic above 35,000 ft or below 6,000 ft [24].Standard atmospheric conditions are assumed for temperature and pressure as functions of altitude.The target time is chosen within the time range achievable by speed changes of the aircraft.Idle thrust descents were used in defining this achievable time range.We neglect the slight expansion of this achievable time range that fixed FPA descents can do.This ensures that the comparison of fixed-FPA descents to the idle-thrust descents can be made at any test condition. +Meet-Time AnalysisA meet-time analysis computes a family of trajectories of varying FPAs for a specific time-to-fly to the meter fix.Both FPA and the speed profile may be varied in the process of computing a trajectory to meet the target arrival time at the meter fix.Suppose we fix the FPA.Iterating the speeds for a trajectory that meets the target time, we could have varied the cruise and descent speeds independently.This simplistic approach can produce operationally impractical speed changes.One extreme example is a speed-up in cruise followed by a speed-down in descent, while the same target time can be achieved by maintaining the current CAS in cruise and descent.Taking into account practical issues, EDA supports three distinct speed modes, referred to as Descent-Only, Cruise-Only, and Cruise-Equals-Descent [3].The Cruise-Equals-Descent mode can absorb the most delay and therefore is used in this work.The Cruise-Equals-Descent mode attempts to identify solutions where the cruise CAS and descent CAS are adjusted to minimize the difference between the two speeds.Some operational considerations are accommodated in this mode and the resulting cruise and descent speeds are not always equal.For example, this speed mode attempts to maintain the aircraft's current airspeed and use the nominal descent speed.If the target time can be achieved by just varying one of the two, EDA does so without enforcing that the cruise and descent speeds be equal.Further modeling details can be found in [3].Although both the FPA and the speeds can be computationally represented as continuous parameters, we have restricted the FPAs to multiples of 0.1 • , using negative values to represent descents.The choice of one decimal place for the FPA matches the precision of the FPAs published to the pilots in the flight test conducted at Denver Center in 2010.It also matches the precision in typical flight deck automation.Speeds are solved for the meet-time problem as precisely as the computer's precision allows for this work, although the actual EDA clearance gives the speeds as multiples of knots.In the following sections the symbol γ i represents the FPA value of a descent.Spanning FPAs from γ i = -1.8• to γ i = -6.0• , the meet-time analysis iterates in the Cruise-Equals-Descent speed mode to obtain the correct speed profile for that FPA and the STA.The minimum cruise and descent speeds restrict the solution space on the slow end where delays are large.The maximum cruise and descent speeds limit the solution space where the STA is earlier than the ETA.Speed brake capacity defines the steepest FPA for many conditions studied.The path distance may limit the shallower descent if its range is too short to accommodate the speed changes and the descent segment.For the choice of 150 nmi, initial Mach of 0.80, and modeled winds, we find that the shallowest FPA of -1.8 • used in the study was always achievable. +Sensitivity AnalysisFuel burn and speed brake usage are two major factors to consider in the efficient and robust choice of FPA.How sensitive are these factors to the selection of an FPA in a typical test condition?Does a fuel-optimal fixed-FPA trajectory require speed brakes?Is it always less fuel-efficient than an idlethrust descent trajectory?To shed some light on these questions, Section 4.1 picks a representative test condition and analyzes a family of fixed-FPA trajectories in terms of their fuel burn, speed brake usage, and descent speed profiles.Wind has strong influence on the vertical profile, and Section 4.2 studies wind effects on the variation of idle-thrust and fuel-optimal fixed-FPA trajectories.The effect of the target time on the variation of the fuel-optimal FPA is also investigated in Section 4.2. +Sensitivity of Fuel Burn and Speed Brake to FPAA meet-time analysis at standard atmospheric conditions without wind was performed.The time to fly was selected to be 1,311 seconds, 40 seconds more than the nominal time the aircraft would have flown based on the nominal trajectory with the nominal speed profile of 0.8 Mach in cruise and 0.8/290 knots in descent.Figure 2 depicts the fuel burn of trajectories flying with different FPAs.The fuel-optimal FPA is approximately γ i = -2.75• for this test condition.Analysis of the trajectories indicates that those with FPAs steeper than γ i = -2.7 • would require speed brake usage, and trajectories with FPAs steeper than γ i = -3.6 • exceed the speed brake capacity during part or all of the descent.Even with full speed brake deployment the aircraft is unable to descend steeper than γ i = -3.6 • .Note that the fuel burn of trajectories between γ i = -2.6 • and γ i = -2.9• is less than the fuel burn of the corresponding idle-thrust descent, whose fuel burn is represented by the gray horizontal line on Figure 2.This is not specific to this test condition as later analysis will show that fixed-FPA descents can be more fuel-efficient than idle-thrust descents on average.Figure 3 shows the corresponding altitude profiles with different descent FPAs.The idle-thrust descent profile is close to the γ i = -3.0• fixed-FPA descent.The idle-thrust descent operates on the boundary of the speed brake "region," and any descents steeper than the idle-thrust descent are expected to require speed brake usage.However, the γ i = -2.8• descent requires speed brake usage even though its descent is shallower than the idle-thrust descent.This counter-intuitive result is explained later in Figure 5 after we examine the speed profile of each trajectory.Figure 4 shows the CAS profile for each trajectory.Recall that the Cruise-Equals-Descent speed mode in EDA is applied in determining the cruise and descent CAS values for each trajectory.In general, shallower descents require higher cruise and/or descent CAS values to meet the same time to compensate for the lower true airspeed (TAS) associated with the CAS at lower altitudes.The shallowest fixed-FPA descent shown in Figure 4, with the FPA γ i = -2.0• , requires a cruise at 0.84 Mach (287 knots CAS) and a descent at 0.84 Mach and 291 knots CAS.The vertical profile of this trajectory contains the five distinct segments described in section 3.1: an initial acceleration in cruise, a constant speed in cruise, a constant mach descent, a constant CAS descent, and a deceleration level.The trajectory with γ i = -2.2• has a longer cruise portion of flight and thus Figure 5 provides insight to the need for speed brakes during the descent.The vertical axis represents one of two events that do not happen simultaneously in the modeling scheme: the additional thrust above idle or speed-brake usage.Its value is meaningful only for the descent segments of the trajectory.Positive values represent the excess thrust required above idle to maintain the aircraft on the FPA at the designated speed.Negative values represent the drag required by the speed brake to keep the aircraft on the FPA at the designated speed.For shallow descents such as γ i = -2.0• , the whole descent requires excess thrust, thus consuming more fuel during descent.The trajectory with γ i = -2.8• uses speed brake at a path distance of -84 nmi and needs excess thrust above idle before and after.At -84 nmi the trajectory is at 34,000 ft altitude.This need for speed brake occurs right below the Mach-CAS transition.Above the Mach-CAS transition altitude, the local FPA flown by an idle-thrust descent at 0.8 Mach is much steeper than γ i = -2.8• .Therefore, thrust is required in the constant-Mach descent segment to maintain γ i = -2.8• .Upon transition to a constant CAS segment, the FPA flown by an idle thrust descent at 278 knots CAS becomes slightly shallower than γ i = -2.8• .Therefore, speed brake usage is required during this part of the descent.As the aircraft descends to lower altitudes, the local FPA of the idle thrust descent gradually becomes steeper, crossing over the angle of -2.8 • again.Therefore the aircraft transitions from speed brake usage to excess thrust above idle.This double crossover is observed for the γ i = -3.0• trajectory too.For trajectories steeper than -3.2 • , the power is on idle throughout the descent, with speed brake deployed for the entire descent.Trajectories steeper than -3.6 • require more speed brake than available.Figure 6 shows the rate of fuel burn as a function of time.The area under a curve equates to the total fuel burn for the entire flight.Note, the idle-thrust fuel burn curve is largely coincident with the γ i = -3.0• fixed-FPA descent curve.The γ i = -2.0• trajectory has an acceleration segment in cruise from 0.80 Mach to 0.84 Mach characterized by a burn rate that well exceeds the nominal rate in cruise.This acceleration segment is followed by a constant-mach cruise segment with a fuel burn rate of about 9,500 lbs/hr.The top of descent for the γ i = -2.0• trajectory happens at 215 seconds when the flight transitions to a constant-Mach/constant-FPA descent segment that lasts for 18 seconds.This segment has an average fuel burn rate of about 5,300 lbs/hr.The flight then transitions to a long constant-CAS/constant-FPA descent segment at 233 seconds until it reaches the bottom of descent at about 1,272 seconds.In all trajectories this segment has a fuel burn rate that slowly increases as the flight descends.The flight then levels off and decelerates to 250 knots at the meter fix at 1,311 seconds.This final decelerating segment has an average low fuel burn rate of about 1,400 lbs/hr.Similar segments are observed for the γ i = -2.2• trajectory.The trajectories with γ i values of -2.4 • , -2.6 • , -2.8 • , and -3.0 • do not have an accelerating cruise segment at the beginning.Instead, they cruise at the initial Mach to their top of descent points.The trajectories with γ i values of -3.2 • , -3.4 • , and -3.6 • have deceleration cruise segments to lower cruise speed.They do not have a constant Mach descent segment because they cruise and descend at the same CAS.Of the trajectories shown in Figure 6, the γ i = -2.8• descent consumes 1,351 lbs of fuel, 38 lbs less than the value of 1,389 lbs for the idle-thrust descent.Both trajectories have the same fuel burn rate in cruise due to their identical cruise speed.The γ i = -2.8• fixed-FPA trajectory starts descending 45 seconds earlier than the idle-thrust trajectory, and thus consumes less fuel in the cruise segment.However in the descent segment, the γ i = -2.8• trajectory consumes more fuel than the idle-thrust trajectory, especially near the bottom of descent.Nonetheless, the fuel benefit gained by the γ i = -2.8• trajectory in the cruise segment exceeds the fuel burn penalty in the descent, resulting in an overall fuel burn advantage of 38 lbs.The same observation is made on Figure 7 regarding the fuel burn trade-offs between cruise and descent.The fuel burn values at the end of the curves correspond to the fuel burn values in Figure 2. The local slope of each curve represents the rate of fuel burn plotted in Figure 6.The trajectory with the shallowest descent, γ i = -2.0• , has the most overall fuel burn due to its high cruise speed and long, less fuel-efficient descent phase.The most fuel-efficient fixed-FPA trajectories are those near γ i = -2.8• .When compared with the idle-thrust trajectory, the γ i = -2.8• trajectory gains an advantage in fuel with its earlier descent.The difference between idle-thrust and the γ i = -2.8• trajectory diminishes in the descent phase, but in the end the γ i = -2.8• trajectory burns 38 lbs less.The trajectories with FPA steeper than γ i = -3.0• all have very fuel-efficient descent phases, but they spend more time in cruise and therefore burn too much fuel in their cruise phase, resulting in an overall fuel burn penalty. +Sensitivity of FPA to Wind and Target TimeThree representative wind conditions, no wind, strong head wind, and strong tail wind, are applied to the analysis of meet-time trajectories.The slope of the wind function is chosen so that the wind magnitude at 35,000 ft is 100 knots for both head wind and tail wind conditions.This magnitude of wind is strong but realistic.The variation of the fuel-optimal FPA due to typical target times is also probed.The three targettime conditions are fast time, nominal time, and slow time.The nominal time is defined as the time flown by the aircraft to the meter fix under the wind condition considered, using a speed of 0.8 Mach in cruise and 290 knots CAS in descent.For this specific comparison, the fast time is defined as 90 seconds earlier, while the slow time is defined as 120 seconds later than the nominal time.The time intervals are selected so that they are far enough from the nominal time and still within the time range achievable by speed maneuvers.Figure 8 shows the fuel-optimal FPA resulting from the three wind conditions and three target time conditions.As a comparison the effective FPA of the idle-thrust trajectory is computed for each test condition.The effective FPA of an idle-thrust descent is defined as the angle between the level flight and the line connecting the top-of-descent point in the space with the bottom-of-descent point in the space.As expected, the idle-thrust descent is steeper in the head-wind and shallower in the tail-wind.The idle-thrust descent is steeper for higher descent speeds (fast) and shallower for lower descent speeds (slow).The fuel-optimal FPA also shows similar trends, steeper in a head wind or for faster times and shallower in a tail wind or for slower times.One exception to these trends is the value at -1.8 • , denoted by an orange square at the bottom, which we discuss in the next paragraph.Aside from this anomaly, the other fuel-optimal FPAs display less variation over the wind change when compared with the effective FPAs for idle-thrust.For example, the fueloptimal FPA for nominal time varies from -3.2 • in the head-wind to -2.3 • in the tail-wind, while the effective FPA of the idle-thrust descent for nominal time varies from -4.19 • in the head-wind to -2.58 • in the tail-wind, as shown in the diamonds of Figure 8.The exception for fuel-optimal FPA into a head wind with fast time is further investigated.At the test condition, the fuel burn continues to decrease as the FPA becomes shallower.In fact, the value of -1.8 • shown in Figure 8 at this test condition is the shallowest FPA computed in this program.The program may find even shallower trajectories until the cruise segment becomes too short for the acceleration to finish.A closer look at the trajectories at this test conditions reveals that most of these shallow trajectories cruise at the highest cruise speed of 0.84 Mach.This high cruise speed is required to meet the time in the strong head wind.The fuel-burn rate at this cruise speed is about 9,500 lbs/hr.Such a high cruise fuel burn rate would tip the trade-off between cruise and descent and favors early descent for the reduction of the overall fuel burn.Figure 9 shows the fuel burn of fixed-FPA trajectories as a function of time to fly.The time-to-fly is denoted as ∆t in the legend.No wind is applied in these test conditions.The symbols on the curves highlight the trajectories that do not require speed brake usage.The FPA for the steepest speed-brake-free trajectory ranges from -3.5 • for a fast time of 1,171 seconds to -2.5 • for a slow time of 1,416 seconds.On the fast end of the time range shown at the top with ∆t = 1, 171 seconds, Figure 6.Fuel rate for different FPAs as a function of time.The total path distance is 150 nmi.shallow descents are bounded by maximum cruise and descent speeds, causing no trajectories to be found for FPA shallower than -2.6 • .On the slow end of the time range with ∆t = 1, 416 seconds, steep descents are bounded by minimum cruise and descent speeds, causing no trajectories to be found for FPA steeper than -2.9 • .In between the ends, the speed brake capacity also bounds the steepest descent a trajectory can have, e.g., γ i = -3.8 is the steepest trajectory for a time to fly of 1,271 seconds.The variation in fuel burn with FPA has an interesting change of slope for ∆t = 1, 271 seconds and ∆t = 1, 281 seconds near the FPA of -2.9 • .The rapid change of overall fuel burn is mostly due to change of cruise speed.For example, in the curve for ∆t = 1, 271 seconds the cruise speed changes from 0.839 Mach at -2.6 • to 0.815 Mach at -2.9 • .Within this range of FPAs the descent CAS is held at 290 knots.This apparent change of slope is specific to the Cruise-Equals-Descent speed mode of EDA, which may speed up the aircraft in cruise in order to meet time.Since the fuel burn rate is very sensitive to the speed in cruise, rapid change of fuel burn is observed among trajectories that have varying cruise speeds.The same analysis was performed using the Descent-Only speed mode of EDA, which varies only the descent speed in order to meet the time.The resulting fuel burn change with FPA showed more "homogeneous" curves without distinct regions of different behaviors. +Comparison of Three Descent Strategies +Custom FPA for every clearanceStrategy 1 issues advisories to all flights coming through a specific meter fix based on a universally fixed FPA akin to a glide slope for an Instrument Landing System (ILS).Strategy 1 is inspired by the early work of Izumi et al. [11].Tong et al. also explored implementation of a procedure using universally fixed-FPAs for all arrival flights.When this strategy is applied, EDA fixes this FPA and iterates the speeds in order to find the meet-time trajectory for a specific arrival flight.Strategy 2 defines the FPA as a function of the descent CAS issued in the advisory and is motivated by the descent-CAS-dependent FPA function used in the flight test in Denver Center in 2010, as shown in Table 1.Participating Skywest pilots determined the value of FPA to fly by referencing the look-up table using the descent CAS issued in the EDA clearance.The FPA function forγ i = -0.1 * floor DCAS -245 10 + γ 0 i ,(1)Figure 8.The fuel-optimal FPA of the fixed-FPA trajectories and the effective FPA for the corresponding idle-thrust trajectories.where γ i is in degrees, DCAS is the descent CAS in knots, and the parameter γ 0 i is in degrees.Here the "floor" function returns the maximum integer that is no greater than its argument.The parameter γ 0 i represents the value of γ 0 i at 250 knots of descent CAS.Later analysis results in Sec.5.4.1 justify the adequacy of this family of functions.When Strategy 2 is applied, EDA computes a set of meet-time trajectories and identify a trajectory that has a descent CAS-FPA relationship described by the FPA function.Section 5.2 discusses in detail how this is done.Both strategies 1 and 2 could presumably be published in the airport's arrival procedures.Strategy 3 issues a custom FPA as part of the advisory to adapt the FPA to weather, wind, and time to fly.For each arrival flight, EDA computes a set of fixed FPA, meet-time trajectories that satisfy the speed brake condition and picks the fuel-optimal trajectory as its advisory.Strategy 3 contains no parameters for optimization, but requires EDA to analyze the family of candidate meet-time trajectories in real time.Also, the FPA must be communicated explicitly between the ground and pilots. +Speed Brake ConditionsWhile efficiency in the model can be measured approximately by the fuel burn of the constant time-to-fly trajectory, flyability and robustness are difficult to quantify.For simplicity, we define flyability and robustness as the ability to maintain the aircraft on the desired trajectory upon uncertainty.Here uncertainty can come from errors in the predicted wind, weather, and pilots' execution.A robust procedure should also take into account controllers' interruptions.In all cases, power adjustment is preferred to speed brake deployment for reasons mentioned in Section 1.Since speed brake usage may still be acceptable for some aircraft types, the approach here is to compare strategies under varying degrees of acceptable speed brake usage.For each strategy, one of three speed brake conditions is used to determine the set of feasible meet-time trajectories.1. Any speed brake usage (SBANY) is allowed to maintain the descent FPA, as long as the speed brake usage is within the speed brake capacity defined in Section 3.2. +2.No more than 20% of the speed brake capacity (SB20) is allowed at any point during the descent.3. No speed brake usage (SB0) is allowed during the descent.For Strategy 3, multiple trajectories can satisfy the speed brake condition.In this case, the trajectory with the least fuel burn is selected.Note that these conditions are imposed on the analysis.Actual speed brake usage required in flight would vary based on the difference between the actual and the predicted state of the flight and atmosphere. +Parameterization of StrategiesBoth strategies 1 and 2 have parameters that need to be optimized for the operational conditions.The choice of the universal fixed FPA from Strategy 1, and the FPA as a function of descent CAS for Strategy 2, are key decisions that will strongly impact the efficiency, flyability and robustness of the arrival operations.Figure 10 illustrates how strategies 1 and 2 can select different trajectories from a set of meet-time trajectories.The value of the universal FPA parameterizes Strategy 1 and has the effect of shifting the selection line left and right on the descent CAS vs. FPA plot, as shown in Figure 10.A typical set of meet-time trajectories labeled "Trajectories 1" is shown in Figure 10, in which each square represents a trajectory's values of descent CAS and FPA.The trajectories must all satisfy the speed brake condition of interest.The trajectory that lies on the vertical line representing strategy 1 is selected as the EDA advisory.If no meet-time trajectory is found, that particular time to fly is considered impossible to achieve based on the strategy and flight conditions.Strategy 2 selects from the set of meet-time trajectories the one that has a descent CAS-FPA pair that satisfies its FPA function.As shown in Figure 10, this family of functions yields shallower FPA for lower descent CAS.For a set of meet-time trajectories, the one trajectory that intercepts with this function is selected as the EDA advisory.The trajectory must also satisfy any additional speed brake condition such as SB20 or SB0.Otherwise, that particular time to fly is considered impossible to achieve based on the strategy and flight conditions.The trajectories labeled "Trajectories 2" in Figure 10 can occur during the sampling of the test conditions, in which no trajectory falls right on the vertical lines defining strategy 2. This condition will be discussed in more detail in section 5.4.1 and treated approximately in the statistics. +Sampling of Test ConditionsWe consider an experiment in which EDA advisories are issued to many arrival aircraft over a wide variety of test conditions.Among the factors defining the test conditions are the variation of wind, aircraft weight, and the amount of time to delay from the nominal arrival time.For each test condition, TS computes a set of meet-time trajectories for each strategy and accumulates statistics.Wind is modeled as a linear function of altitude.The actual distribution of the wind strongly depends on the location and time of the year.Also, the direction of wind is far from random.Considering a generic arrival route, a random direction of the wind is sampled using a 2-D normal distribution.To estimate the width of the normal distribution, we turned to the Rapid Update Cycle (RUC) 2-hour weather prediction [25] for realistic wind distribution.We estimated the root-mean-square value of the wind speed at 35,000 ft using (RUC) 2-hour weather predictions for Denver Center from Oct. 25 to Nov. 10 in 2010.These dates were chosen because complete data sets were available and flight tests were being conducted during that time.Nine horizontal locations in the Denver Center airspace were chosen for sampling the wind predictions in RUC data.The minimum distance between any two locations is 100 nmi.The root-mean-square value was found to be 61 knots.Analysis of the RUC data for Fort Worth Center from Sep. 18 to Sep. 23 of 2011 yields a much smaller root-mean-square value of 35 knots.For the simulation, the magnitude of the wind at 35,000 ft is sampled from a normal distribution with a standard deviation of 40 knots in both X and Y directions.The standard deviation is chosen to yield a root-mean-square wind magnitude of 57 knots, a value close to but below the value of Denver Center.Since the variation of wind along a specific route may be smaller than the variation of wind averaged over locations of a Center, the choice of magnitude may qualify as an upper bound of the actual wind variation.A systematic analysis of wind data is required to support this claim.The weight of the aircraft is sampled with a standard deviation of 8,400 lbs and a mean of 170,000 lbs.These values are estimated from the distribution of landing weights of thousands of arrival flights of this aircraft type.Although TMA can speed up the arrival aircraft, in actual operations TMA almost always delays the aircraft.Therefore in most of this work, we sample the target time uniformly between the nominal time and the slow limit, unless noted otherwise.EDA issues a speed advisory when the delay time to absorb is small, usually less than four minutes [3].The maximum delay time is determined by the difference in meter-fix crossing times between a trajectory flying nominal speeds and a trajectory flying minimum speeds.Note that EDA issues a clearance only when the difference between ETA and STA is greater than about 15-30 seconds.This tolerance is related to the meter flow rate specified by TMA.For this analysis we neglect this small tolerance and allow the target time to be sampled uniformly between the nominal time and the slow time.When the target time is near the slow end of the range, the steepest trajectory could be flying at the slowest speeds flyable in the aircraft's performance model.Trajectories with steeper descents may still satisfy the speed brake condition, but they are rejected by the program because they do not absorb enough of the delay.In actual operation such descents can be made to meet the target time with path stretches.Therefore in this case, we attempt steeper descents using the slowest speeds until the descent exceeds speed brake capacity.Conceptual path stretches are added to these trajectories in order to meet the time.These path stretches are "conceptual" because we add them just to meet the time and calculate the fuel burn.We do not need to define the turn out and turn back points for them.Assuming a conceptual path stretch before the top of descent, we add to each trajectory an amount of fuel burn equal to the cruise fuel burn rate multiplied by the delay time to absorb. +ResultsWe conducted a Monte Carlo simulation that sampled 50,000 test conditions that varied in wind, weight, and targeted time-to-fly to the meter fix.The statistics accumulated in the simulation are used to optimize Strategy 1 and Strategy 2, and therefore enable comparison of the three strategies.For each test condition, a set of trajectories with their values of FPA ranging from -1.8 • to -6.0 • are computed by TS.One of three speed brake conditions was applied to the set of meet-time trajectories to reject steep descents that do not meet the speed brake condition.For each trajectory, five to ten iterations over the speeds were required to converge the solution to within the criterion of time, which is chosen to be 0.5 second.Data were accumulated during the simulation and statistics calculated.After sampling 5,000 test conditions, the statistics stabilized and did not show noticeable change.The three fixed-FPA descent strategies defined in section 5 were applied to select a trajectory from the set of trajectories.Up to nine distinct selections can be made for each test condition as a result of the three strategies and three speed brake conditions.Strategy 3, the custom FPA approach, selects a fuel-optimal trajectory for each one of the three speed brake conditions.Strategies 1 and 2, however, have parameters that need to determined.To determine the parameters for these two strategies, the results of applying these two parametrized strategies are stored in the simulation and analyzed for optimization. +Optimizing StrategiesStrategies 1 and 2 should choose parameters so that the selected advisories are close to the fueloptimal fixed-FPA descent.On the other hand, if speed brake is utilized to a great extent in the fuel-optimal fixed FPA descent, it may be desirable to pick a shallower fixed-FPA trajectory to reduce speed brake usage, albeit at the sacrifice of certain fuel efficiency.Both factors should be considered in optimizing the parameters of the strategies with a certain level of trade-offs.We define the fuel-burn penalty of the selected trajectory as the extra fuel burn this trajectory incurs relative to the fuel-optimal fixed-FPA trajectory computed in the same test condition.The selected trajectory should satisfy the speed brake condition of interest.The fuel-optimal fixed-FPA trajectory is chosen using SBANY.For example, the selection of any trajectory other than the γ i = -2.7 • trajectory, as shown in Figure 2, has a positive fuel-burn penalty.Results of a Monte Carlo simulation are used to identify the preferred solution space that satisfies both low fuel burn and low speed brake.This simulation samples both fast and slow target times.Figure 11 shows the fuel burn penalty as a function of FPA and descent CAS resulting from a Monte Carlo simulation of 50,000 test conditions.The SBANY condition is applied in choosing the trajectories.Since most of the reference fuel-optimal trajectories utilize very little speed brake, as shown later in Section 5.4.2, the choice of a speed brake condition does not change the average fuel burn penalty by more than 5 lbs.A stricter speed brake condition, however, rejects steeper descents and therefore results in reduced or even empty samples in the area of high speed brake usage.All meet-time trajectories for each test condition are included in the statistics.Since each set of trajectories covers a different range of FPA and a different range of descent CAS, each descent CAS-FPA pair in the map is sampled with unequal frequencies.Roughly speaking, the low fuel penalty, dark-blue area that spans from about -2.5 • near 250 knots descent CAS to -3.5 • near 350 knots descent CAS was sampled with highest frequencies.Strategies 1 and 2 should be parametrized to sample descent CAS-FPA pairs in this dark blue "valley" in order to minimize fuel burn penalty.Figure 11.Fuel burn penalty averaged over 50,000 test conditions.Figure 12 shows the speed brake usage as a function of FPA and descent CAS averaged from the same Monte Carlo simulation used for Figure 11.The speed brake usage is defined as the maximum fraction of speed brake drag coefficient relative to the speed brake capacity along the descent.The SBANY condition is applied in choosing the trajectories for comparison.Selection of the universal FPA for Strategy 1 and the FPA function for Strategy 2 should stay in the low speed brake usage region, denoted by dark blue colors.The choice of a universal FPA for Strategy 1 and FPA function for Strategy 2 should also consider the probability of failure.If the speed brake usage of a selected trajectory exceeds the maximum allowable speed brake defined by a speed brake condition, this trajectory is rejected and the strategy fails to create an advisory in this test condition.A stringent tolerance of failure can make the procedure more robust but would be less fuel-efficient.For the purpose of this analysis, the tolerance of failure is chosen to be 1% for both strategies.In other words, if during the Monte Carlo analysis for a specific strategy, more than 1% of the conditions sampled failed to yield a trajectory, then the parameter used for this strategy is considered unacceptable.Since both Strategy 1 and Strategy 2 have one parameter each, the failure rate can be plotted as a function of this parameter.To determine the optimal universal FPA for Strategy 1, we performed a Monte Carlo simulation that samples only the slow end of the target time, i.e., the time between nominal and slow.Figure 13 shows the resulting average fuel burn penalty and speed brake usage as a function of the universal FPA.The FPA that yields the least average fuel burn penalty is γ i = -2.7 • .Since the probability of failure at γ i = -2.7 • is less than 1%, this FPA is acceptable under the SBANY condition.However at γ i = -2.7 • , 24% of trajectories require speed brake usage between 0% and 20%, 14% Figure 12. speed brake usage averaged over 50,000 test conditions. of trajectories require speed brake usage between 20% and 40%, 5% of trajectories require speed brake usage between 40% and 60%, and 1% of trajectories require speed brake usage between 60% and 80%.Therefore, 20% and 44% of the trajectories at -2.7 • do not satisfy the SB20 and SB0 conditions, respectively.These high probabilities of failure are unacceptable for SB20 and SB0, and shallower FPAs must be chosen such that the failure rate goes below 1%.The resulting optimal universal FPAs when imposing SB20 and SB0 are -2.3 • and -2.2 • , respectively.To determine the optimal FPA function defined in Eq. 1 for Strategy 2, results of the same Monte Carlo simulation are used.Figure 14 shows the average fuel burn penalty and speed brake usage vs. the choice of the FPA function.The optimal FPA function for SBANY isγ i = -0.1 * floor DCAS -245 10 -2.5 • ,(2)where a value of -2.5 • is assigned to the γ 0 i in Eq. 1 .Shallower values of γ 0 i must be used for more stringent speed brake conditions to keep the probability of failure under 1%.Therefore, the optimal FPA function for SB20 isγ i = -0.1 * floor DCAS -245 10 -2.2 • ,(3)and the the FPA function for SB0 isγ i = -0.1 * floor DCAS -245 10 -2.0 • . (4)During the simulation a small fraction of the test conditions resulted in meet-time trajectories that cross over one of the "gaps" of the FPA function without intercepting with any of the vertical segments.An example is shown in "Trajectories 2" of Figure 10.This is because the step-wise FPA Figure 13.Fuel burn penalty and speed brake usage for Strategy 1.function leaves gaps of three to seven seconds in the range of target time with its discrete values of FPA.Since the typical tolerance of time for EDA is 15-30 seconds [3], such gaps should cause no problem in delivering aircraft to the meter fix within the tolerance.In this case an approximate trajectory is defined by selecting the FPA in which the difference between the trajectory and function in descent CAS is the smallest.The fuel burn and speed brake usage are interpolated from the two closest trajectories using the target time.Although the solution space of trajectories near the latest arrival time are slightly expanded to steeper FPAs by the introduction of the conceptual path stretch in Section 5.3, we find very little contribution from these trajectories in Figure 13 and Figure 14. +Fuel-Burn BenefitsTable 2 summarizes the fuel burn penalty for the three descent strategies under the three speed brake conditions.Again, the results are from the Monte Carlo simulation that produced Figure 13 and Figure 14.Note that all the values of fuel burn penalty are relative to the fuel-optimal fixed-FPA trajectory, which is selected with SBANY.Also, note that the values of the fuel burn penalty for strategies 1 and 2 are observed in the fuel burn penalty curves of figures 13 and 14, respectively.For each speed brake condition, Strategy 1 consumes more fuel than Strategy 2, while Strategy 2 consumes more fuel than Strategy 3.For all strategies, more stringent speed brake conditions lead to an increase of the fuel burn penalty.Compared to the SBANY condition, the SB20 condition results in extra fuel burn of 69 lbs and 34 lbs in strategies 1 and 2, respectively.The most stringent SB0 condition creates about 23 lbs and 41 lbs of extra fuel penalty than SB20 for Strategy 1 and Strategy 2, respectively.Also, shallower FPAs are selected under more stringent speed brake conditions, resulting in more fuel burn.This shift to a shallower FPA is required to accommodate the trajectories at the slow end, which required speed brake usage even for very shallow FPAs shown in the bottom part of Figure 12.The custom FPA strategy, denoted as Strategy 3 in Section 5, is relatively insensitive to the speed brake condition, as the fuel burn penalty relative to SBANY goes up to only 5 lbs for SB0.This implies that most of the fuel-optimal fixed-FPA descents for the test conditions sampled utilize very little speed brake.Table 2.The fuel burn penalty for the three strategies under the three speed brake conditions averaged over 50,000 test conditions.Strategy\Fuel burn penalty (lbs) and FPA( 2, fixed-FPA trajectories can be more fuel-efficient than the idle-thrust descent.The bottom row of Table 2 shows that, of all the sampled test conditions, a custom FPA strategy with SBANY can save 41 lbs of fuel relative to the idle-thrust descents on average.This should not be surprising.The results of Izumi et al. for Boeing 747 showed fuel-optimal descent trajectory burns 50 to 60 lbs less than the idle-thrust descent [12].It is possible that, on average of the test conditions, a fuel-optimal fixed-FPA trajectory for this aircraft type is closer to the fuel-optimal trajectory than the idle-thrust descent is, resulting in the average fuel burn benefit. +DiscussionThe cruise and descent speeds computed in both Section 4 and Section 5 are considered continuous.In actual operations, the cruise and descend speeds issued in an EDA clearance may consider the precision of the equipage of the aircraft.For some small aircraft, the CAS values may need to be constrained to increments of 10 knots.The latest human-in-the-simulation of EDA attempts to model this limitation (unpublished).The effect of such limitations on the accuracy of the delivered meet time remains to be investigated.We believe the descent strategies should be able to accommodate these limitations.Figure 6 of Section 4.1 shows that the fuel burn rate is very sensitive to an increase in cruise speed.This can hardly be desirable in practice except for the purpose of conflict resolution.If EDA needs to speed up an aircraft, we believe it should increase the descent speed first.Only when the descent speed cannot be fast enough to meet the target time should EDA speed up the aircraft in cruise.Similarly, EDA should decrease the cruise speed before decreasing the descent speed to slow down the aircraft.The current Cruise-And-Descent mode can potentially be modified to adopt this behavior.The aircraft's preferred descent speed was modeled as 290 knots in this work.This choice defines the nominal trajectory and the nominal time-to-fly.In practice this speed is not only determined by the aircraft type but also by the airline.For Section 5, another Monte Carlo simulation using 300 knots as the nominal descent was conducted, and the results showed that this choice of speed had only minor effects on the statistics in Table 2.Figure 12 of Section 5.4.1 shows that speed brake usage occurs even for very shallow FPAs in the low descent CAS region.This is due to the fact that strong tail winds prevail in the statistics for this region.Therefore, speed brakes are used to great extent to compensate for the tail wind.The high speed brake usage of these test conditions shifts the optimal FPAs in strategies 1 and 2 towards shallower FPAs for the SB20 and SB0 conditions.This shift tends to increase the average fuel burn penalty of the strategy.A possible improvement of the fuel-burn benefit of a strategy is to treat trajectories near the slow end with a different FPA function that uses shallower FPAs.The original FPA function, be it a universal fixed FPA or a function given in Eq. 1, can be combined with this second FPA function at the 265 knots descent CAS boundary.Although Strategy 3 provides most fuel-efficient descent profile, it requires FPA to be communicated to the pilot before top-of-descent via voice or data-link.Table 2 also shows that, for a meet-time constraint, fixed-FPA descents can burn less fuel than idle-thrust descents when averaged over the test conditions.However, individual test conditions may favor either the fixed-FPA descent or the idle-thrust descent.Section 4.1 investigated the fuel burn and speed brake usage for a specific test condition.The results showed 38 lbs fuel benefits of the fuel-optimal fixed-FPA trajectory over the idle-thrust trajectory.Limited amount of analysis under different test conditions revealed that such fuel benefit is greatest in a head wind and diminishes in a tail wind.In fact, with a strong tail wind the idle-thrust trajectory seems to always have fuel burn benefits.In this case the idle-thrust trajectory has an earlier top-of-descent than the fuel-optimal fixed-FPA trajectory in the tail wind condition.More analysis is needed to provide a systematic description of wind effects on the competition of a fuel-optimal fixed-FPA trajectory and an idle-thrust trajectory.In addition to fixed-FPA descents, other types of low-power descents can potentially also be more fuel-efficient than idle-thrust descents.It would be interesting to to see if similar fuel benefits can be observed in smaller aircraft types such as business or regional jets.The optimal parameters for strategies 1 and 2 will vary for every specific arrival route.The prevailing wind along each arrival route has a major influence in the optimization of these parameters.The wind variation sampled in the simulation has complete random directions, and we believe the choice of wind distribution may be somewhat larger than a typical wind variation along a specific arrival route.This remains to be confirmed.When strategies 1 and 2 are customized for an arrival route, the smaller variation of wind may lead to less fuel burn penalties.However, variation of performance among aircraft types has not been modeled, and is expected to increase the fuel burn penalty for strategies 1 and 2. The effects of these factors remain to be investigated. +ConclusionThis work modelled fixed-flight-path-angle (fixed-FPA) descents, performed sensitivity analysis of the input parameters, and compared three FPA selection strategies that were applied to realistic test conditions for the En-route Descent Advisor (EDA).A high-fidelity performance model of a mid-size twin-engine jet was chosen from the aircraft performance database of the Center-TRACON Automation System for computing trajectories.The first part of this work studied in depth the sensitivity of fuel burn and speed brake usage to FPA for a typical operational condition.Sensitivity of the fuel-optimal FPA to winds and target times was also investigated.The second part of this work proposed three FPA selection strategies given below:1. Universally fixed FPA 2. FPA as a function of descent speed 3. Custom FPA for every clearance.The sensitivity of fuel burn to the selection strategies and speed brake usage was studied.To take into account robustness requirements in the optimization of the FPA selection strategies, three speed brake conditions were imposed on the selection of the FPA:1. Any speed brake usage (SBANY) is allowed to maintain the descent FPA, as long as the speed brake usage is within the speed brake capacity.2. No more than 20% of the speed brake capacity (SB20) is allowed at any point during the descent.3. No speed brake usage (SB0) is allowed during the descent.Monte Carlo simulations were performed to sample 50,000 realistic test conditions.The results showed that Strategy 1 and Strategy 2 burn more fuel than Strategy 3. Strategy 3 was the most fuel-efficient, but requires explicit communication of the FPA between the ground and the pilot prior to the top-of-descent.Stringent speed brake conditions, although they may increase robustness of the procedure, do reduce the fuel efficiency of the procedures.Strategy 1 and Strategy 2 were more sensitive to speed brake conditions and can have up to 112 lbs and 90 lbs of fuel burn penalty when speed brake usage was disallowed.Contrary to some literature that referred to an idle-thrust descent as a fuel-optimal descent, results of this work showed that idle-thrust descents burned 41 lbs more fuel than the fuel-optimal fixed-FPA descents on average of the test conditions sampled.In some test conditions particularly head-wind and no-wind, fuel can be saved by a top-of-descent earlier than that of an idle-thrust descent.The extra fuel burn arising from the inefficient descent segment is more than compensated by the reduced fuel burn in the cruise segment.Our next steps are to analyze other aircraft types to understand sensitivity/variation of the optimal FPA for descent in aircraft types.We also plan to develop adequate models for typical RJ/BJ types.The Base of Aircraft Data (BADA) [26] may provide a good starting point for such models.Another future direction is to include path stretches and step-downs in our fuel-burn analysis.Figure 1 .1Figure 1.A general vertical profile that contains five segment types. +Figure 2 .2Figure 2. Fuel burn variation with flight path angle.The time to fly is 1,311 seconds. +Figure 3 .3Figure 3. Altitude along the path distance.The time to fly is 1,311 seconds. +Figure 4 .4Figure 4. Calibrated airspeed along the path distance.The time to fly is 1,311 seconds. +Figure 5 .5Figure 5. Power above idle or speed brake usage along the descent.The time to fly is 1,311 seconds. +Three strategies are proposed to determine the fixed FPA for EDA advisories at run time.Their advantages and disadvantages are discussed in Section 5.4.The three strategies are 1.Universally fixed FPA 2. FPA as a function of descent speed +Figure 7 .7Figure 7. Accumulated fuel burn as a function of fly time.The total path distance is 150 nmi. +Figure 9 .9Figure 9. Fuel burn as a function of FPA and time to fly.No wind is applied.The symbols represent speed-brake-free trajectories. +Figure 10 .10Figure 10.Demonstrating how a trajectory is selected using strategies 1 and 2 for a specific time to fly. +Figure 14 .14Figure 14.Fuel burn penalty, probability of speed brake usage, and failure rate for different FPA functions used for Strategy 2. The X-axis FPA represents the intercept of the function with 250 knots descent CAS, the value of γ 0 i in Eq. 1. + + + + +Table 1 .1The FPA function used for Skywest flights in the flight test conducted at Denver Center in 2010.Range of Descent CAS (knots) FPA( • )250-260-2.8270-280-3.1290-300-3.4310-320-3.8Strategy 2 is arbitrary, although in practice it should stay in the fuel-efficient and flyable regionof the descent CAS-FPA space. Considering the fact that steeper FPAs are preferred for fasterdescent speeds, we choose a parametric family of simple, step-wise functions of this form: +Table 2 of2Section 5.4.2 shows that Strategy 1 and Strategy 2 lead to more fuel burn than Strategy 3, especially for stricter speed brake conditions.Strategy 1 and Strategy 2 do not require FPA to be communicated to the aircraft and can presumably be published in the arrival procedures.Strategy 1 is slightly simpler to implement than strategy 2, and may result in fewer pilot execution errors. + + + + +AcknowledgmentsWe thank Richard Coppenbarger and David Williams for helpful discussions. + + + + + + + + + Human-in-the-Loop Simulation of Three-Dimensional Path Arrival Management with Trajectory Error + + GauravNagle + + + DouglasSweet + + + GregoryCarr + + + ValentinoFelipe + + + AndrewTrapani + + + RichardCoppenbarger + + + MiwaHayashi + + 10.2514/6.2011-6877 + AIAA-2011-6877 + + + 11th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Virginia Beach, VA + + American Institute of Aeronautics and Astronautics + 2011 + + + Nagle, G.; Sweet, D.; Carr, G.; Felipe, V.; Trapani, A.; Coppenbarger, R.; and Hayashi, M.: Human-in-the-Loop Simulation of the Efficient Descent Advisor for 3D Path Arrival Manage- ment. Proceedings of the AIAA Aircraft Technology, Integration, and Opterations Conference, AIAA-2011-6877, Virginia Beach, VA, 2011. + + + + + Development and Testing of Automation for Efficient Arrivals in Constrained Airspace + + RCoppenbarger + + + GDyer + + + MHayashi + + + RLanier + + + LStell + + + DSweet + + ICAS2010-11.11.3 + + + Proceedings of the 27th International Congress of the Aeronautical Sciences + the 27th International Congress of the Aeronautical SciencesNice, France + + 2010 + + + Coppenbarger, R.; Dyer, G.; Hayashi, M.; Lanier, R.; Stell, L.; and Sweet, D.: Development and Testing of Automation for Efficient Arrivals in Constrained Airspace. Proceedings of the 27th International Congress of the Aeronautical Sciences, ICAS2010-11.11.3, Nice, France, 2010. + + + + + Design and Development of the En Route Descent Advisor (EDA) for Conflict-Free Arrival Metering + + RichardCoppenbarger + + + RichardLanier + + + DougSweet + + + SusanDorsky + + 10.2514/6.2004-4875 + AIAA-2004-4875 + + + AIAA Guidance, Navigation, and Control Conference and Exhibit + + American Institute of Aeronautics and Astronautics + Aug. 2004 + + + Coppenbarger, R. A.; Lanier, R.; Sweet, D.; and Dorsky, S.: Design and Development of the En Route Descent Advisor (EDA) for Conflict-Free Arrival Metering. Proceedings of the AIAA Guidance, Navigation, and Control Conference, AIAA-2004-4875, Aug. 2004. + + + + + Improved Aircraft Path Stretch Algorithms for the En Route Descent Advisor (EDA) + + TysenMueller + + + DavidSchleicher + + + RichardCoppenbarger + + 10.2514/6.2003-5571 + AIAA-2003-5571 + + + AIAA Guidance, Navigation, and Control Conference and Exhibit + + American Institute of Aeronautics and Astronautics + Aug. 2003 + + + Mueller, K. T.; Schleicher, D. R.; and Coppenbarger, R. A.: Improved Aircraft Path Stretch Algorithms for the En Route Descent Advisor (EDA). Proceedings of the AIAA Guidance, Navigation, and Control Conference, AIAA-2003-5571, Aug. 2003. + + + + + En route Descent Advisor concept for arrival metering + + StevenGreen + + + RobertVivona + + 10.2514/6.2001-4114 + AIAA-2001-4114 + + + AIAA Guidance, Navigation, and Control Conference and Exhibit + + American Institute of Aeronautics and Astronautics + Aug. 2001 + + + Green, S. M.; and Vivona, R. A.: En Route Descent Advisor Concept for Arrival Metering. Proceedings of the AIAA Guidance, Navigation, and Control Conference, AIAA-2001-4114, Aug. 2001. + + + + + Practice for Application of Federal Aviation Administration (FAA) Federal Aviation Regulations Part 21 Requirements to Unmanned Aircraft Systems (UAS) + 10.1520/f2505 + + + Federal Aviation Administration + + ASTM International + Mar. 2011 + + + Federal Aviation Administration, FAA's NextGen Implementation Plan. Mar. 2011. + + + + + A Concept for Robust, High Density Terminal Air Traffic Operations + + DougIsaacson + + + JohnRobinson + + + HarrySwenson + + + DallasDenery + + 10.2514/6.2010-9292 + AIAA-2010-9292 + + + 10th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + + American Institute of Aeronautics and Astronautics + Sept. 2010 + + + Isaacson, D. R.; Robinson, J. E.; Swenson, H.; and Genery, D. G.: A Concept for Robust, High Density Terminal Air Traffic Operations. Proceedings of the 10th AIAA Aviation Technology, Integration, and Opterations Conference, AIAA-2010-9292, Sept. 2010. + + + + + Concept for Next Generation Air Traffic Control System + + HeinzErzberger + + + RussellAPaielli + + 10.2514/atcq.10.4.355 + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 10 + 4 + + Aug. 2004 + American Institute of Aeronautics and Astronautics (AIAA) + + + Proceedings of the 24th International Congress of the Aeronautical Sciences, ICAS 2004-8.1.2 (I.L. + Erzberger, H.: Transforming the NAS: The Next Generation Air Traffic Control System. Pro- ceedings of the 24th International Congress of the Aeronautical Sciences, ICAS 2004-8.1.2 (I.L.), Aug. 2004. + + + + + Design and Operational Evaluation of the Traffic Management Advisor at the Fort Worth Air Route Traffic Control Center + + HNSwenson + + + THoang + + + SEngelland + + + DVincent + + + TSanders + + + BSanford + + + KHeere + + + + Proceedings of the 1st USA + the 1st USA + + June 1997 + + + Swenson, H. N.; Hoang, T.; Engelland, S.; Vincent, D.; Sanders, T.; Sanford, B.; and Heere, K.: Design and Operational Evaluation of the Traffic Management Advisor at the Fort Worth Air Route Traffic Control Center. Proceedings of the 1st USA/Europe Air Traffic Management R&D Seminar , June 1997. + + + + + Descent Profile Options for Continuous Descent Arrival Procedures within 3d Path Concept + + K.-OTong + + + ESchoemig + + + DBoyle + + + JScharl + + + AHaraldsdottir + + + + Proceedings of the IEEE/AIAA 26th Digital Avionics Systems Conference + the IEEE/AIAA 26th Digital Avionics Systems Conference + + Oct. 2007 + + + 3.A.3-1 -3.A.3-11 + Tong, K.-O.; Schoemig, E.; Boyle, D.; Scharl, J.; and Haraldsdottir, A.: Descent Profile Op- tions for Continuous Descent Arrival Procedures within 3d Path Concept. Proceedings of the IEEE/AIAA 26th Digital Avionics Systems Conference, Oct. 2007, pp. 3.A.3-1 -3.A.3-11. + + + + + Sensitivity Studies of 4D Descent Strategies in an Advanced Metering Environment + + KennethHIzumi + + 10.23919/acc.1986.4789026 + + + 1986 American Control Conference + + IEEE + 1986. June 1986 + + + + Izumi, K. H.: Sensitivity Studies of 4D Descent Strategies in an Advanced Metering Environ- ment. Proceedings of the American Control Conference, 1986 , June 1986, pp. 687-692. + + + + + Sensitivity Studies of 4D Descent Strategies in an Advanced Metering Environment + + KennethHIzumi + + + RWSchwab + + + JLGroce + + + ACoote + + 10.23919/acc.1986.4789026 + + + 1986 American Control Conference + + IEEE + Dec. 1986 + + + Izumi, K. H.; Schwab, R. W.; Groce, J. L.; and Coote, A.: An Evaluation of Descent Strate- gies for TNAV-Equipped Aircraft in an Advanced Metering Environment ATOPS. NASA CR- 178093, Boeing Commercial Airplane Company, Dec. 1986. + + + + + Impact of Mismodeled Idle Engine Performance on Calculation and Tracking of Optimal 4-D Descent Trajectories + + DavidHWilliams + + 10.23919/acc.1986.4789025 + + + 1986 American Control Conference + + IEEE + 1986 + 2 + + + + Williams, D. H.: Impact of Mismodeled Idle Engine Performance on Calculation and Tracking of Optimal 4-D Descent Trajectories. Proceedings of the 5th American Control Conference, vol. 2, 1986, pp. 681-686. + + + + + Flight Evaluation of Center-TRACON Automation System Trajectory Prediction Process. NASA/TP + + DHWilliams + + + SMGreen + + + 1998-208439, July 1998 + + + Williams, D. H.; and Green, S. M.: Flight Evaluation of Center-TRACON Automation System Trajectory Prediction Process. NASA/TP-1998-208439, July 1998. + + + + + Analysis of flight management system predictions of idle-thrust descents + + LaurelStell + + 10.1109/dasc.2010.5655506 + + + 29th Digital Avionics Systems Conference + + IEEE + Oct. 2010 + + + Stell, L.: Analysis of Flight Management System Predictions of Idle-thrust Descents. Proceed- ings of the IEEE/AIAA 29th Digital Avionics Systems Conference, Oct. 2010. + + + + + Predictability of Top of Descent Location for Operational Idle-Thrust Descents + + LaurelStell + + 10.2514/6.2010-9116 + + + 10th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + + American Institute of Aeronautics and Astronautics + June 2011 + + + Stell, L.: Prediction of Top of Descent Location for Idle-thrust Descents. Proceedings of the 9th USA/Europe Air Traffic Management R&D Seminar , June 2011. + + + + + Potential Benefits of User-Preferred Descent Speed Profile + + TWeidner + + + TGDavidson + + + LBirtcil + + NASS AATT TO26 Seagull TR98188.26-02 + + Dec. 2000 + Seagull Technologies + + + Weidner, T.; Davidson, T. G.; and Birtcil, L.: Potential Benefits of User-Preferred Descent Speed Profile. NASS AATT TO26 Seagull TR98188.26-02, Seagull Technologies, Dec. 2000. + + + + + Analysis of Continuous Descent Benefits and Impacts During Daytime Operation + + SShresta + + + DNeskovic + + + SSWilliams + + + + Proceedings of the 8th USA/Europe Air Traffic Management R&D Seminar + the 8th USA/Europe Air Traffic Management R&D Seminar + + June 2009 + + + Shresta, S.; Neskovic, D.; and Williams, S. S.: Analysis of Continuous Descent Benefits and Im- pacts During Daytime Operation. Proceedings of the 8th USA/Europe Air Traffic Management R&D Seminar , June 2009. + + + + + Distributed Scheduling Architecture for Multi-Center Time-Based Metering + + StevenLandry + + + ToddFarley + + + JohnFoster + + + StevenGreen + + + TyHoang + + + GregoryWong + + 10.2514/6.2003-6758 + AIAA-2003-6758 + + + AIAA's 3rd Annual Aviation Technology, Integration, and Operations (ATIO) Forum + + American Institute of Aeronautics and Astronautics + Nov. 2003 + + + Landry, S.; Farley, T.; Foster, J.; Green, S.; Hoang, T.; and Wong, G. L.: Distributed scheduling architecture for multi-center time-based metering. Proceedings of the AIAA Aviation Technol- ogy, Integration, and Opterations Conference, AIAA-2003-6758, Nov. 2003. + + + + + En-route descent trajectory synthesis for air traffic control automation + + RASlattery + + + YZhao + + 10.1109/acc.1995.532248 + + + Proceedings of 1995 American Control Conference - ACC'95 + 1995 American Control Conference - ACC'95 + + American Autom Control Council + June 1995 + 5 + + + + Slattery, R.; and Zhao, Y.: En-route Descent Trajectory Synthesis for Air Traffic Control Automation. Proceedings of the American Control Conference, vol. 5, June 1995, pp. 3430- 3434. + + + + + Trajectory Synthesis for Air Traffic Automation + + RhondaSlattery + + + YiyuanZhao + + 10.2514/2.4056 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 20 + 2 + + Mar. 1997 + American Institute of Aeronautics and Astronautics (AIAA) + + + Slattery, R.; and Zhao, Y.: Trajectory Synthesis for Air Traffic Automation. Journal of Guid- ance, Control and Dynamics, vol. 20, no. 2, Mar. 1997, pp. 232-238. + + + + + The Trajectory Synthesizer Generalized Profile Interface + + AlanLee + + + XavierBouyssounouse + + + JamesMurphy + + 10.2514/6.2010-9138 + AIAA-2010-9138 + + + 10th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + + American Institute of Aeronautics and Astronautics + Sept. 2010 + + + Lee, A. G.; Bouyssounouse, X.; and Murphy, J. R.: The Trajectory Synthesizer Generalized Profile Interface. Proceedings of the 10th AIAA Aviation Technology, Integration, and Opera- tions Conference, AIAA-2010-9138, Sept. 2010. + + + + + Using Flight Manual Data to Derive Aero-Propulsive Models for Predicting Aircraft Trajectories + + ChesterGong + + + WilliamChan + + 10.2514/6.2002-5844 + AIAA-2002-5844 + + + AIAA's Aircraft Technology, Integration, and Operations (ATIO) 2002 Technical Forum + + American Institute of Aeronautics and Astronautics + Oct. 2002 + + + Gong, C.; and Chan, W. N.: Using Flight Manual Data to Derive Aero-propulsive Models for Predicting Aircraft Trajectories. Proceedings of the AIAA Aircraft Technology, Integration, and Opterations Conference, AIAA-2002-5844, Oct. 2002. + + + + + Physics of Climate + + JoséPPeixoto + + + AbrahamHOort + + + CurtCovey + + + KarlTaylor + + 10.1063/1.2809772 + + + Physics Today + Physics Today + 0031-9228 + 1945-0699 + + 45 + 8 + + 1992 + AIP Publishing + + + Peixoto, J. P.; and Oort, A. H.: Physics of Climate, American Institute of Physics. 1992, pp. 154-155. + + + + + An Hourly Assimilation–Forecast Cycle: The RUC + + StanleyGBenjamin + + + DezsöDévényi + + + StephenSWeygandt + + + KevinJBrundage + + + JohnMBrown + + + GeorgAGrell + + + DongsooKim + + + BarryESchwartz + + + TatianaGSmirnova + + + TracyLorraineSmith + + + GeoffreySManikin + + 10.1175/1520-0493(2004)132<0495:ahactr>2.0.co;2 + + + Monthly Weather Review + Mon. Wea. Rev. + 0027-0644 + 1520-0493 + + 132 + 2 + + 1998 + American Meteorological Society + + + Benjamin, S. G.; Brown, J. M.; Brundage, K. J.; Schwartz, B.; Smirnova, T.; Smith, T. L.; Morone, L. L.; and Dimego, G.: The Operational RUC-2. Preprints. Proceedings of the 16th Conference on Weather Analysis and Forecasting, Amer. Meteor. Soc., 1998, pp. 249-252. + + + + + Advanced Aircraft Performance Modeling for ATM: Enhancements to the Bada Model + + ANuic + + + CPoinsot + + + M.-GIagaru + + + EGallo + + + FANavarro + + + CQuerejeta + + 10.1109/dasc.2005.1563320 + + + 24th Digital Avionics Systems Conference + + IEEE + Nov. 2005 + + + Nuic, A.; Poinsot, C.; Iagaru, M.-G.; Gallo, E.; Navarro, F. A.; and Querejeta, C.: Advanced Aircraft Performance Modeling for ATM: Enhancements to the BADA Model. Proceedings of the IEEE/AIAA 24th Digital Avionics Systems Conference, Nov. 2005. + + + + + + diff --git a/file797.txt b/file797.txt new file mode 100644 index 0000000000000000000000000000000000000000..6f488ba55d81023214b7d68e990c78afa981105a --- /dev/null +++ b/file797.txt @@ -0,0 +1,1023 @@ + + + + +I. Introduction2][3] Significant research in the past two decades has focused on the utilization of modern flight management systems (FMS) to enable continuous descent planning, at least from cruise to a metering fix at the boundary of the terminal area.5][6][7] Fuel benefits for continuous descents compared to current-days' operations have also been evaluated. 1,4,8 Hwever, little attention has been paid to "small" (regional, business and light) jet types, which comprise a large and potentially high-growth portion of NextGen traffic operations in the United States. 9redicting the descent profiles of small jets has been challenging.Unlike larger aircraft types, which are equipped with performance-based FMS systems that attempt to optimize the vertical profile with idle or near-idle descents, most small jets are equipped with simple Vertical Navigation (VNAV) capabilities that cannot guide idle-thrust descents.Observations of regional jet operations and pilot interviews revealed that a large variety of descentplanning techniques are used by pilots, even for the same equipment and air carrier.These techniques vary in terms of the selection of descent angle or vertical speed, bottom-of-descent planning, and top-of-descent transition.The intended descent procedure for a specific flight is generally unknown to Air Traffic Control (ATC), and therefore cannot be utilized by ground automation desicion support tools to ensure vertical separation from aircraft at different altitudes.Moreover, the inflexibility of company preferred descent procedures can result in trajectories that are fuel-inefficient or difficult to fly under certain conditions.For example, the "standard operating procedure" of one large regional carrier called for a calibrated airspeed (CAS) of 320 knots for descent, initiated at the cruise Mach number, using a default flight-path-angle (FPA) of -3.8 • (negative indicates descent).This works reasonably well for nominal conditions with light to moderate winds and no ATC interruptions to the descent.However, when the winds are strong and speed clearances are issued by controllers for metering and spacing, the nominal descent FPA can become inefficient or difficult, if not sometimes impossible, to fly.It is desirable to develop a standard descent procedure for arriving small jets in the transition airspace leading to congested terminal areas.This paper proposes a standard fixed-FPA descent procedure that is supported by the simple VNAV of almost all small jets.The selection of an FPA, which is simultaneously known by the pilots and ATC, is expected to realize better trajectory predictability, which in turn would better support Trajectory-Based Operations and increase airspace throughput.Lack of a descent procedure for small jets can hamper the performance of all trajectory-based ground automation decision support tools for scheduling and spacing.Specifically, a standard descent procedure for small jets was required by the Efficient Descent Advisor (EDA), which was developed to assist en-route controllers achieve fuel-efficient continuous descent arrivals (CDA) 1 that meet the Traffic Management Advisor's 10 (TMA) scheduled times of arrival (STAs) at the metering fix and maintain separation during congested operations.While a fixed-FPA descent improves trajectory predictability, selecting the FPA is a nontrivial task.The most fuel-efficient and "flyable" FPAs can vary significantly with the winds along the route, wind gradient, descent speed, aircraft type, and aircraft weight. 11][14][15][16] Given the significant variation of the winds aloft from one area of the National Airspace System (NAS) of the United States to another, and from one day, week or month to another, the FPA selection may need to be "adaptable."This paper presents three candidate strategies for selecting descent FPAs for small jets in the transition airspace of a metered environment, where arrival flights are expected to cross the metering fix each at a specific STA.While the three strategies are expected to achieve the same level of trajectory predictability, they vary in operational complexity and fuel efficiency.The minimum-fuel strategy has the best fuel efficiency but highest operational complexity.A methodology was developed to select parameters for the two other simpler strategies and compare their fuel efficiency to the minimum-fuel strategy.Benefits of various adaptations of the two simple strategies to the airport, arrival gate, and periods of time categorized by season, month and day, are estimated.Three major airports of the United States with different levels of wind variation and disparate arrival configurations are analyzed.The analysis reveals how close in fuel efficiency the two simple strategies can get to the minimumfuel strategy, and helps evaluate which strategy provides the best value for implementation.The rest of the paper is organized as follows: Section II describes the metered environment and defines the fixed-FPA descent; Section III outlines the three strategies, the methodology for selecting parameters of the strategies, and adaptation of the simple strategies to airspace and time; Section IV analyzes the variation of winds aloft for twelve major US airports and selects three for detailed analysis; Section VI describes the modeling schemes for the metering-constrained arrival trajectories; Section VII presents and compares results of the selected FPAs and relative fuel burn merits between the three strategies for the three airports; Section VII also explores the effect of planned speed brake usage on fuel burn as well as adaptations of the simple strategies to the NAS; Section VIII discusses the implication of the results and considerations for implementation of the simple strategies.Finally, Section IX summarizes the methodology and the findings. +II. BackgroundThe fixed-FPA descent procedure defines an FPA for the descent profile of a flight in the transition airspace.Consider an arrival flight that is about 150 to 200 nmi away from the destination airport and is transitioning from cruise to descent in 10 to 15 minutes.Figure II depicts a flight that follows a fixed-FPA procedure to descend to the metering fix at the boundary of the terminal area.During periods of congestion, ATC would issue speed clearances to maintain separation and absorb delay.Depending on the difference between the issued descent speed and the metering fix crossing speed, some flights may utilize a short level segment before the metering fix for deceleration.Note that the FPA, γ, the cruise altitude of the arrival flight, and the arrival route define the top-of-descent point in space (ignoring the short level segment before the metering fix).This shared definition is expected to achieve better trajectory predictability by significantly reducing the top-ofdescent uncertainty, which has a big impact on the performance of ground-based decision support tools that involve automated separation assurance. 17This is contrasted with the top-of-descent point for an idle-thrust descent, which is sensitive to the aircraft's performance parameters, aircraft weight, and winds aloft. 18,19 hile the fixed-FPA descent procedure is applicable to any level of traffic, it is most beneficial in highly congested traffic scenarios when ATC cannot afford to reserve large airspace buffers to accommodate the uncertainty of the arrival flight's vertical profile.The following discussion and the rest of the paper focuses on the high-density traffic scenario in which arrival traffic is metered and ATC issues speed clearances for all arrival flights.During periods of congestion in the terminal area, Traffic Management Coordinators in the United States use TMA to plan a schedule that delivers the flights to the runway at a rate that will meet the capacity of the airport.This plan consists of arrival sequences and scheduled times of arrival (STA) at the metering fix, which are typically published points along an aircraft's route of flight and may lie at the boundary of the Terminal Radar Approach Control (TRACON).TMA computes each flight's STA at the metering fix according to the desired throughput specified by the terminal area control. 10,20 he STA is constantly updated until the arrival flight enters the TMA freeze horizon, where the STA is fixed (frozen) without further adjustment. 20The definition of a freeze horizon can be time-based or distance-based, and is typically 15-25 minutes or 130-180 nmi away from the metering fix. 20][23][24][25] During periods of congestion, TMA's STA requires delay with respect to the flight's estimated time of arrival at the metering fix to keep the TRACON arrival traffic manageable. 10For flights that are about to transition from the cruise phase to the descent phase in a few minutes, EDA utlizes speed reductions in both cruise and descent to absorb up to four minutes of delay. 23If speed reductions alone are not enough for absorbing the delay, other types of maneuvers such as path stretches are combined with speed reductions.Early development and simulations of EDA focused primarily on large jets equipped with a performance-based FMS.Later flight tests conducted at Denver Center in the fall of 2010 began to address the descent procedures for small jets.Analysis of 44 fixed-FPA descent trajectories of the Federal Aviation Administration's (FAA) Global 5000 test aircraft showed that errors in the top-of-descent location were 0.5±1.0nmi. 26The Global 5000 analysis results suggested that fixed-FPA decents were much more predictable than idle-thrust descents typically performed by large commercial jets, of which the ground predictions had top-of-descent errors around 5 to 10 nmi. 18,27 hile a simple, fixed-FPA descent procedure using prescribed clearances was introduced for the purposes of the flight tests, EDA itself lacks a defined descent procedure and corresponding algorithm for small jets.This hampers EDA's capability to deliver small jet arrivals to the metering fix on time and impacts EDA's ability to maintain separation involving the arrival aircraft.A standard descent procedure with proper selection of the descent profile would aid EDA in delivering its full benefits.Although the analysis in this paper focuses on small jets, the methodology applies to any jet that can fly fixed-FPA descents.In fact, in recent years many Airbus types and new Boeing types have been equiped with VNAV that has fixed-FPA descent guidance. 28 +III. Selection and Definition of the Descent FPAThe criteria for the selection of an FPA consider both fuel efficiency and flyability.A shallow descent may have long and fuel-inefficient descent segments that require excessive thrust.A steep descent may require extensive usage of speed brakes that are operationally unacceptable to pilots.Many pilots are reluctant, if not unwilling, to use speed brakes on a routine basis because of passenger comfort and pilots' desire to reserve the use of speed brakes for rare occasions.The range of fuel-efficient and flyable FPAs varies with winds, wind gradient, descent speed, and aircraft performance specifics.The fixed-FPA must be shared between both the pilot and ATC to have benefits of trajectory predictability.Questions arise regarding whether additional infrastructure or equipage would be required for disseminating such information.When considering a selection strategy's potential benefits in fuel and airspace throughput, the complexity of the required changes to the infrastructure and procedures for its implementation should also be considered. +III.A. Three Selection StrategiesThree strategies for defining the descent FPAs are defined below':• Min-Fuel FPA: computes a flight-specific minimum-fuel FPA• Universal FPA: selects a universally fixed FPA• Descent-Speed FPA: defines the FPA as a function of the issued descent speed While all three are expected to achieve the same level of trajectory predictabilty, they vary in operational complexity and fuel efficiency.Each strategy is further described in the following paragraphs.The Min-Fuel FPA computes a minimum-fuel FPA to be communicated explicitly to the pilot of each flight just prior to the top of descent (TOD).The computation takes into account the arrival route, weather and wind forecast, aircraft performances, and traffic conditions in real time.Figure 2 sketches the typical fuel burn as a function of the descent FPA for a flight.The Min-Fuel FPA, by definition, selects the minimum-fuel FPA for each arrival flight and therefore achieves the best fuel efficiency.However, the requirement of communicating the FPA to the pilot in real time makes it too complex to implement in the near term without datalink.Nonetheless, it serves as a reference point for the fuel burn comparison.The following paragraphs describes two simple strategies that are less fuel efficient but easier to implement.The Universal FPA defines a single FPA for a large group of flights.This is akin to a glide slope extended from the arrival metering fix back up to the top of descent (TOD) in the en-route airspace.The idea of a universally fixed-FPA descent procedure was explored by Tong et al. at Boeing, with a focus on modifications required for the performance-based FMS equipped jets. 13The advantage of the Universal FPA is its simple form, which allows it potentially to be published as part of the arrival procedure.The disadvantage is that it does not account for the effects of descent speed, winds aloft along the direction of flight, and other factors, on the range of fuel efficient and flyable FPAs.The Descent-Speed FPA defines the FPA as a function of ATC-issued descent speed for a large group of flights.This is motivated by the aerodynamical observation that, without considering wind varations, an aircraft flying a slower descent speed on idle thrust would have a shallower descent.The Descent-Speed FPA captures this trend by defining a shallower FPA for a slower descent speed to keep the required power low during descent.The proposed FPA function in this work changes by 0.1 • for every 10 knots of descent CAS, a rate observed in a series of test conditions that captures the variation of fuel-efficient FPAs with speed. 11he FPA function is defined by Equation 1:γ =                  γ 0 , if V d < V 0 d + 5 γ 0 -0.1 , if V 0 d + 5 ≤ V d < V 0 d + 15 γ 0 -0.2 , if V 0 d + 15 ≤ V d < V 0 d + 25 . . . ,(1)where γ is the FPA in degrees, V d is the descent CAS in knots, V 0 d is the slowest descent CAS that can be issued by ATC without consulting with the pilots, and the adaptable parameter γ 0 is the prescribed value of γ at V 0 d .V 0 d is fixed at 250 knots CAS for the analysis in this paper.All values of FPA are negative for the descent, with steeper descents represented by more negative angles.The increment of 0.1 • matches the precision of FPA-selection of small jet avionics.Note the selection of the parameter γ 0 defines the entire FPA function.One advantage of the Descent-Speed FPA, like the Universal FPA, is that the function can be determined ahead of time and therefore published as part of the arrival procedure.Moreover, by taking into account the descent speed, the Descent-Speed FPA is expected to be more fuel-efficient than Universal FPA.The Descent-Speed FPA will not be as fuel-efficient as the Min-Fuel FPA, since it does not account for variation of the winds aloft along the flight direction and other factors on the range of fuel-efficient and flyable FPAs. +III.B. Planned Speed-Brake UsageIn addition to fuel burn, the selection of an FPA should also take into account uncertainties in vertical profile planning.To keep the aircraft on the planned vertical path, power adjustment is preferred to speed-brake deployment for reasons of passenger comfort and the desire of pilots to reserve the use of a speed brake for rare occasions.Besides, some aircraft types have less effective speed brakes than others.Therefore, an upper bound for the planned speed brake usage should be imposed to ensure the flyability of the descent profile.Two levels of planned-speed-brake usage are considered in this analysis to explore the impact of speed-brake usage on the selection of descent FPA and resulting fuel burn.In one condition, FPAs are limited to those that do not require planned speed-brake usage.This condition , named Speed-Brake-Zero (SB0), is assumed for most of the analysis in this paper.In the other, any FPA is considered valid as long as the estimated amount of speed-brake usage is within the modeled speed brake capacity for that aircraft.This condition, named Speed-Brake-Any (SBANY), is explored for some of the data analysis to understand its effect on the fuel efficiency of strategies. +III.C. Parameters in Universal FPA and Descent-Speed FPAThe FPA of Universal FPA, denoted as γ univ , and γ 0 of Descent-Speed FPA are the parameters that must be selected carefully with consideration for fuel efficiency and flyability.The parameter is sensitive to the prevailing winds aloft and, to a lesser extent, the anticipated traffic demand.The approach in this paper constructs candidate fixed-FPA trajectories from historical flight plans, radar track data, and weather forecasts, and evaluates the aggregated fuel efficiency and flyability as a function of the parameters selected for each strategy.The following criteria are used for selecting γ univ and γ 0 :1.For the parameter considered, at least 99% of the flights must have feasible trajectories, meaning trajectories that have speeds within the performance envelope and satisfy the speed brake usage limits.2. The parameter should result in the lowest average fuel burn per flight.The first criterion constrains the steepest γ univ or γ 0 that can be selected, and the second criterion selects the parameter that yields the best fuel efficiency.III.D. Adaptation of Universal FPA and Descent-Speed FPAThe simplest implementation of the Universal FPA and Descent-Speed FPA would be to apply a static FPA or FPA function to the entire NAS.However, large variation of the environmental conditions such as prevailing winds aloft and local traffic flows impact the fuel efficiency of such procedure, and the magnitude of the impact is not understood.A refined implementation of Universal FPA and Descent-Speed FPA is to adapt their parameters to specific airports.As discussed in Section III.C, the selection of the parameter(s) should be such that the resulting descent FPAs are flyable across significant variations of wind along the route.An airport's arrival traffic configuration also impacts the fuel efficiency of the simple strategies.For airports with opposing arrival directions, particularly the classical four-corner-post configuration, strong winds become problematic.While steeper FPAs are typically more fuel-efficient for arrival flights in a headwind, they can be unflyable for flights in the opposite direction.A shallow FPA would guarantee flyability from both directions, but can be fuel-inefficient for flights in a headwind.These observations motivated the adaptation to the direction of arrival.Compared to a static implementation, adaptation to the direction of arrival achieves a greater degree of customization by mitigating the impact of the variation of winds between flights and thus improves fuel efficiency.While adaptation to the direction of arrival reduces the impact of the variation of along-track winds due to directions, it does not mitigate the impact of the variation of winds over time.The fuel efficiency of the Universal FPA and Descent-Speed FPA strategies can be further improved by adapting to seasonal norms, monthly norms, or even daily predictions.A set of systematic, temporal and airspace adaptation is presented in Table 1.The columns represent different levels of adaptation for different airspaces, starting with a basic "one-size-fits-all" adaptation for all airports across the NAS.Moving to the right, each +IV. Variation of Winds Aloft at AirportsThe benefits of adapting the simple FPA selection strategies to airspace/direction and time increases with the variation of winds aloft, the degree of which is different from airport to airport.While some airports may require finer adaptations for the desired level of fuel efficiency, others may achieve reasonable fuel efficiency with a simple Airport-Static adaptation.It is therefore desireable to examine the variations of winds aloft across airports in order to select a set of airports with different degrees of wind variation for application of the FPA selection strategies.In the rest of the paper, reference to winds always refer to the winds aloft, not the winds on the ground.Twelve major airports of the United States were analyzed for the variation of winds with respect to directions and time of the year.Winds along the route were estimated for four hypothetical arrival routes constructed for each airport.Four hypothetical points in space at four corners around an airport, NE, NW, SE, and SW, were selected at 150 nmi each from the airport and at 35,000 ft in altitude.The vectors connecting each point to the airport defined the four hypothetical routes of arrival.The four points and their routes were used for the estimates of winds along the route.No airport-specific arrival routes were considered here.A year's worth of wind components along the routes of arrival were estimated using the two-hour, 40-km Rapid Update Cycle (RUC) weather forecast 29 for year 2011.The hourly wind forecast amounted to more than 8660 forecast winds.a Two quantities, σ dir and σ time , that represented the wind variations with respect to directions and time of the year, respectively, were calculated from the wind components.Let N r denote the number of available RUC wind estimates.The standard deviation of the average wind along each of the four directions, σ dir , is computed byσ dir 2 = 1 4 4 i=1 (W i -µ) 2 , where µ = 1 4 4 i=1 W i .(2)a About 1% of the RUC forecast files were not available.Here W i is the average of the N r wind components along route i,W i = 1 N r N r j=1 W i,j ,(3)and W i,j denotes the wind component estimated using RUC wind file j along direction i.The average of the standard deviation of wind for each route, σ time , is computed byσ time 2 = 1 4 4 i=1 σ i 2 , where σ i 2 = 1 N r N r j=1 (W i,j -W i ) 2 . (4)While σ dir is indicative of the benefits of adapting the descent FPA to flight directions, σ time is indicative of the benefits of adapting the descent FPA to time.north to south.JFK has the highest values of σ dir and σ time and hence was chosen as an airport representing large wind variations for the analysis.LAX and MIA have the least wind variations among the twelve airports.LAX was chosen to represent an airport with small wind variations for its large traffic volume and the fact that it is on the West Coast, in constrast to JFK that is on the East Coast.The third airport chosen for the analysis was DFW, which ranks 7th in σ dir and 9th in σ time among the twelve airports and represents an airport with medium wind variation.DFW was chosen also for its large volume of small jet traffic and its classical four-corner-post configuration.Figure 4 contrasts the winds in summer to those in winter at the LAX, DFW, JFK airports, using the Rapid Update Cycle (RUC) weather forecast in 2011 as an estimate.The wind vectors were computed at 30,000 ft right above the airports, and the coordinates W E and W N stand for the east and north components of the wind vector, respectively.Each dot represents a wind vector made by RUC's two-hour forecast, pointing in the direction that the wind is blowing to.Wind vectors from January to March were grouped in the "winter" category, shown in green, while wind vectors from July to September were grouped in the "summer" category, shown in red.Two observations are made regarding the wind distributions at all three airports.Firstly, the winds in winter were stronger, and were predominantly westerly winds.The winds in summer were weaker and more or less isotropic at least for LAX and DFW.Secondly, JFK has stronger winds than DFW, which in turn has stronger winds than LAX.This was consistent with Figure 3 since airports with strong winds are more likely to have strong variations of wind. +V. Small Jet Arrival Traffic for DFW, JFK, and LAXThis section describes the arrival traffic and procedures for small jets in the transition airspace for DFW, JFK, and LAX.DFW is described first because its classical four-cornerpost configuration serves as a "textbook" example.Table 2 summarizes the three airports' arrival gates and their associated speed and altitude restrictions.Note that the speed restrictions are defined in CAS.Note that the term "gate" in this paper is used interchangeably with, and has a one-to-one relationship with, the term "metering fix."Therefore, the gate adaptation of the simple strategies should be thought of as an adaptation to a metering fix.This approach ignores the fact that, in actual airspace configuration, a gate can contain more than one metering fix.For example, the east gate for LAX contains both KONZL and GRAMM as metering fixes.The Dallas/Fort Worth International Airport has the classical four-corner-post configuration, where an arrival flight enters the TRACON through either of the NE, SE, SW, or NW gate.About two-thirds of its small jet arrivals come from the east.Figure 5 shows radar tracks of arrival flights of small jets on July 28, 2011.The primary metering fixes for the four gates are KARLA, HOWDY, DEBBB, and FEVER.Although there are other fixes for each gate, it was assumed in this analysis that the arrival flights all go through the primary metering fixes.Therefore, the metering fixes KARLA, HOWDY, DEBBB, and FEVER represent the gates.The altitude and speed restrictions at these metering fixes are listed in Table 2.In actual operations, these altitude and speed constraints could depend on a flight's runway assignment and an airport's runway configuration.Such dependencies were not considered in this analysis.Consider the winds at DFW shown in Figure 4 acting on the arrival flights shown in Figure 5.The prevailing westerly winds in winter resulted in strong headwinds for the eastern gates as well as strong tailwinds for the western gates.In summer, the winds reduced to 60 knots or less most of the time and was somewhat isotropic.Therefore, all four gates experienced similar wind distributions. +V.B. John F. Kennedy International AirportFigure 6 shows arrival tracks of small jets on January 7, 2011.About 85% of JFK's small jet arrivals were from NW, W, or SW.The remaining arrival flights are from NE. Very have restrictions at high altitudes due to the congested airspace in the New York metroplex area.Flights from the N and NW must capture the high-altitude restrictions of FL200 and FL210 ft when they cross LOLLY and HARTY, respectively.Additional downstream altitude and speed restrictions apply, making the vertical profile beyond LOLLY and HARTY very constrained.For this work, only the portion of the flight trajectory from cruise to one of these four points was considered for varying the FPA.Further desecent segments beyond the first altitude constraints were not considered.This shorter descent segment is expected to diminish the fuel burn differences observed among different values of the descent FPA, and hence diminish the difference in the relative fuel-burn merits of the three selection strategies.Comparing the wind directions with the arrival directions showed that most flights entering LOLLY, HARTY, and CAMRN, about 80% of the total arrivals, experienced tailwinds.This is in contrast to DFW, in which two-thirds of the flights are from the east and experience mostly headwinds. +V.C. Los Angeles International AirportFigure 7 shows arrival tracks of small jets to LAX on January 7, 2011.About half of the small jet arrivals were from the NW and the other half from NE and E. Only three arrival metering fixes, characterized by the waypoints GRAMM, KONZL, and SYMON, were found to be utilized by small jets.Note GRAMM and KONZL were treated as different gates in this work although they belong to the same east gate in current days' operation.The altitude and speed restrictions at these three waypoints are summarized in Table 2. Two of the three arrival routes, GRAMM and KONZL, have high altitude constraints at 19,000 ft and 17,000 ft, respectively.Note that GRAMM is part of a newly implemented Optimized Profile Descent (OPD) procedure for LAX. 4 This new procedure actually defines for GRAMM an altitude range, which is between FL210 and 17,000 ft.To keep the altitude restrictions in similar form between gates, the altitude restriction was modeled with a hard value of FL190.The magnitude of winds at LAX were smaller than those at DFW and JFK.Flights from the east experienced mild headwinds while flights from the NW experienced mild tailwinds in winter. +VI. Modeling and Analysis Approach +VI.A. Calibrating and Comparing the Three StrategiesComparison of the simple strategies to Min-Fuel FPA in terms of their fuel merits is a necessary step in a benefit assessment.Other aspects of the comparison of the three strategies include the cost of implementation, which is beyond the scope of this paper.The salient question asked here is: "what FPA strategy would provide the best value for implementation?"To facilitate the fuel-burn comparison, results are presented in terms of the average fuel burn per flight relative to the minimum-fuel solution of Min-Fuel FPA.In this way, the results will show how close the simpler strategies, Universal FPA and Descent-Speed FPA, and their adaptations can come to the minimum-fuel solution without requiring real-time pilot-controller communication of FPA just prior to the top of descent.A methodology was developed in previous work 11 to compare the fuel benefits of the three strategies of selecting the descent FPA.Conceptually, this methodology consists of two parts: select the parameters for Universal FPA and Descent-Speed FPA; and compare the extra fuel burn of the simple strategies to Min-Fuel FPA for a set of traffic and wind conditions.The first part of the methodology selects the parameters of Universal FPA and Descent-Speed FPA by analyzing traffic forecast using estimates of winds and temperatures aloft:• For each arrival flight in the projected traffic demand, evaluate the predicted fuel burn and planned speed-brake usage as a function of the FPA by computing a set of fixed-FPA trajectories.• Select the value of the parameter based on the aggregated results of speed-brake usage and fuel burn, using the criteria given in Section III.C.The second part of the methodology estimates and compares the fuel burn of the strategies, using the estimates of winds and temperatures aloft, actual traffic data, and modeled or actual metering delays:• For each flight, compute the fuel burn and planned speed-brake usage for each flight using all three strategies, which in general select different FPAs.For the Min-Fuel FPA, a set of fixed-FPA trajectories are computed and the FPA with the lowest fuel burn while satisfying the planned speed brake usage is selected.The extra fuel burn of Universal FPA and Descent-Speed FPA strategies relative to Min-Fuel FPA is recorded.• Iterate over all flights and compute the aggregated extra fuel burn of Universal FPA and Descent-Speed FPA relative to Min-Fuel FPA.While the estimates of winds, temperatures aloft, and traffic demand for the second part are, in general, not necessarily the same as that used for the first step, they were the same in the analysis presented in this paper.The RUC data and the radar track data from the Air Route Traffic Control Centers (ARTCC) recorded in 2011 along with a simple metering delay model were used.This means the selection process for the parameters of Universal FPA and Descent-Speed FPA in this analysis used the same wind and track information accessible in real time to Min-Fuel FPA, while in operation the selection process can only be based on forecast or historical data, which may be low-fidelity for longer periods of adaptation time.The fuel benefits of Universal FPA and Descent-Speed FPA obtained in this way should be regarded as an upper bound of the actual benefits.The implementation of the methodology further simplifies the data collection process by combining its two conceptual parts into one run of a fast-time simulation as described in detail in the following sections.For each arrival flight and a modeled metering delay, the analysis computes a set of trajectories with varying FPA and speed profile that meet the STA.Because the FPA selected by each strategy must be from these trajectories, analysis of these trajectories was sufficient for the fuel burn comparison among the three strategies.The following sections describe the aircraft modeling, metering delay modeling, meet-time trajectory construction, and data analysis. +VI.B. Route and Vertical ProfileThe modeled trajectories considered an arrival flight just entering the freeze horizon towards the transition airspace for metering.A distance-based freeze horizon of 160 nmi was assumed, inside of which TMA would fix the STA for the aircraft. 20The initial position and speed of the trajectory was picked from a corresponding radar track point.The metering fix that the flight crossed was derived from the track data, and Direct-To trajectories from the initial point to the metering fix were assumed without actually parsing the flight plans for the waypoints.The vertical profile of an arrival flight was modeled as consisting of up to five segments, as shown in Figure 8. Individual trajectories will contain all or a subset of these segments depending on the speed profile needed to meet the STA.Each segment is modeled by fixing two control parameters.One of the parameters is the FPA; the second depends on the segment.For a cruise segment, the model fixes the airspeed or the engine control for acceleration or deceleration.For the constant speed descent segments, the model fixes an airspeed in Mach or CAS. +VI.C. Aircraft ModelingThe Trajectory Synthesizer (TS) component 30,31 of the Center-TRACON Automation System (CTAS) 32 was used to compute trajectories and their fuel burn and planned speedbrake usage.While a detailed performance model of small jet types would have been desirable, one was not available.Instead, a high-fidelity CTAS model for a mid-size, narrow-body, twin-jet airliner with a typical descent weight of 170,000 lbs was used.The speed envelopes were selected within the ranges of small jets.Weight uncertainty was modeled by a normal distribution with a 5% deviation of the typical descent weight.To account for the variation of the fuel-burn rate among aircraft types, the fuel burn was scaled by this empirical formulaf i = f 0 * N i + 30 230 ,(5)where f i is the scaled fuel burn; f 0 is the raw fuel-burn rate calculated by CTAS for the mid-size, narrow-body, twin-engine jet; and N i is the number of passengers' seats typical of the aircraft type i.This empirical formula was derived by taking the linear regression of the nominal cruise fuel-burn rate of eight small jets plus the mid-sized twin engine jet, using the Base of Aircraft Database (BADA) 3.8 performance model. 33Although BADA 3.8 provides modeling parameters for small jets, the calibration of these parameters focused on nominal flight conditions only.Since the analysis in this work explored a wide range of the speeds, containing both nominal and off-nominal ones, it was decided that the high-fidelity CTAS model with scaled fuel burn was more appropriate for the fuel-burn analysis. +VI.D. Speed-Brake Usage ModelingTwo levels of planned speed brake usage were considered, Speed-Brake-Zero (SB0) and Speed-Brake-Any (SBANY).SB0 represents the case that FPAs for a flight are limited to those that require no speed-brake usage; SBANY allows any FPA to be considered for a flight as long as the modeled speed-brake usage is within the speed-brake capacity for that aircraft.An empirical constant was used to model the maximum drag coefficient resulting from speed-brake deployment. 11The results presented in the analysis of the three airports in Section VII assumed the SB0 condition, that is, no planned speed brake usage was allowed in any of the computed trajectories used.The effect of the SBANY on the extra fuel burn of the simple strategies is explored in Section VII.E. +VI.E. Metering DelayA simple metering delay model was applied to each arrival flight by assigning a random delay time that was independent of the delay times of previous or subsequent flights.The delay at the metering fix was modeled by a uniform distribution between zero and the maximum delay that can be absorbed by speed reductions.No path stretches were considered in this analysis.The delay time was added to the nominal time in order to specify the STA.During the Monte-Carlo simulation (to be described below), two trajectories were computed to define the STA window for random sampling of the delay, using RUC weather forecast and the modeled route described in VI.B.The descent CAS of 320 knots was assumed as the airline-preferred descent CAS and was used together with the aircraft's initial cruise speed in defining the nominal time.The minimum cruise and descent speeds of 220 knots and 250 knots, respectively, were used in defining the slow time.In the absence of a "standard" FPA for defining the nominal trajectory, the two trajectories assumed idle-thrust descent. +VI.F. Fixed-FPA Meet-Time TrajectoriesA set of meet-time trajectories with varying FPA and speed profiles was computed for each flight using a modeled metering delay.Here the term "meet-time" refers to the requirement of the trajectories that cross the metering fix at a specific STA.For descent FPAs ranging from -1.8 • to -5.5 • , with an increment of 0.1 • , a meet-time algorithm attempted to compute a fixed-FPA trajectory for each value of the FPA.Fuel burn and planned speedbrake usage was calculated for each meet-time trajectory.These trajectories provided all the fuel-burn data needed for comparison of the three strategies.The algorithm iterated cruise and descent speeds until the trajectory met the desired time-to-fly within a tolerance of 2 seconds.Cruise and descent speeds were related by the Cruise-Equals-Descent speed mode developed for EDA and designed based on operational considerations. 23igure 9 shows a typical range of FPA-descent-speed combinations defining the set of trajectories for a flight meeting an STA.Each square represents the pair of descent CAS and FPA of a trajectory.Trajectories steeper than -3.6 • were not shown, because their predicted speed-brake usage exceeded the allowable planned speed brake usage.In most wind conditions, lower descent speeds are required for steeper descent FPAs in order to meet the time.This is because an arrival flight flying a steeper descent FPA would have a longer cruise segment, whose higher ground speed must be compensated for by slowing down in descent.To illustrate how the descent FPA may be selected differently by the three strategies for the flight, suppose Min-Fuel FPA is used for selecting the FPA.Min-Fuel FPA selects the FPA of the minimum-fuel trajectory, which is found to be -2.8 • .Now suppose a set of flights are used for determining the parameters of Universal FPA and Descent-Speed FPA.This set of flights could be on a specific airport or an arrival gate, and during a specific period of time, depending on the type of adaptation.Analysis of this set of flights, determines an γ univ of -2.5 • for Universal FPA.A schematic representation of Universal FPA is shown as the dashed vertical line in Figure 9.The resulting meet-time trajectory with the selected γ univ of -2.5 must belong to the set of meet-time trajectories.Therefore, the fuel burn as a result of the selection by Universal FPA for this flight is computed from that trajectory with an FPA of -2.5 • .This fuel burn is generally higher than that of the Min-Fuel FPA trajectory with an FPA of -2.8 • .Now suppose analysis of the same set of flights determines an γ 0 of -2.3 • for Descent-Speed FPA.While the selected FPA must correspond to one of the meet-time trajectories, the relationship between the descent CAS and the FPA of the trajectory must satisfy the FPA function defined in Eq. 1.This FPA function is represented schematically by the solid vertical steps.The fuel burn as a result of the selection by Descent-Speed FPA is computed from the trajectory that intercepts with the steps representing the FPA function.This trajectory has an FPA of -2.6 • .Hence all three strategies select distinct FPAs for this flight.If no meet-time trajectories have the FPA defined by Universal FPA or satisfy the FPAdescent-CAS relationship defined by Descent-Speed FPA, a failure is recorded for this parameter of the strategy.The number of failures is used to determine whether the parameter is feasible for selection. +VI.G. Simulation and Data AnalysisA fast-time Monte Carlo simulation was performed for each of the airports being analyzed to generate all the meet-time trajectories for small jet arrivals.During the simulation, the metering delay and aircraft weight for each flight were sampled randomly.The fuel burn and speed-brake usage were recorded.All data were categorized by gates and days.Analysis of the meet-time trajectories for all the flights allowed the parameters of Universal FPA and Descent-Speed FPA to be selected.The γ univ for Universal FPA was selected from values between -1.8 • and -5.5 • , and the γ 0 for Descent-Speed FPA was selected from values between -1.8 • and -3.7 • .These ranges covered the most fuel efficient values of the parameter under the conditions analyzed.For each γ univ and γ 0 , the average fuel burn per flight and feasibility rate were computed for all flights.The feasibility rate was the ratio of the flights with flyable FPAs (total number of success) to the total flights analyzed.It must be 99% or better for γ univ or γ 0 to be selected (see Section III.C).The same selection criteria in Section III.C were used to select γ univ and γ 0 for each of the adaptation types.For the Airport-Static adaptation, all flights arriving at the destination airport were analyzed for the selection.For the gate-specific and Airport-seasonal, -monthly, and -daily adaptations, a subset of the flights arriving at the destination airport was analyzed to select γ univ or γ 0 for a gate and/or a timespan.For example, a total of sixteen pairs of γ univ and γ 0 were selected for the Gate-Season adaptation (four gates times four seasons), each using the flights crossing a specific gate during a specific season.Selection of the parameters based on the feasibility rate ensured that the vast majority of flights would have flyable FPAs, but it did not consider the variation of winds that can make the flights through some gates on some days particularly difficult to fly.To ensure the feasibility rate for any given day and gate was tolerable, a feasibility rate of 80% or better for any pair of gate and day was required for all adaptations.The fuel-burn benefits of trajectories were compared by applying the three strategies.Each trajectory consumed additional fuel with respect to the fuel burn of the minimum-fuel solution of Min-Fuel FPA.By definition, Min-Fuel FPA resulted in zero extra fuel burn.The extra fuel burn for the FPAs selected by Universal FPA and Descent-Speed FPA were computed by aggregating the extra fuel burn from each flight.A year's worth of the ARTCC track data during 2011 for the three airports, DFW, JFK, and LAX, were used for the analysis.Due to occasional data feed issues, only 95% of the track data were available for the analysis.The flight plan information was used to distinguish the relevant arrival flights from those flying to other airports in the TRACON.Flights that originated in the freeze horizon were not considered.Table 4 lists the number of flights analyzed for each airport and the percentage of flights for each gate.Figure 10 shows the ten most frequent small jet aircraft types observed and used for the analysis.The top five most frequent types take up 87%, 93%, and 87% of the entire fleet for LAX, DFW, and JFK, respectively.Note that more large regional jets were observed for JFK (E190 and CRJ9), whereas more small jets such as E135 were observed for DFW. +VII. ResultsThis section is organized as follows: Sections VII.A, VII.B, and VII.C present the selected FPA and fuel burn comparison for the DFW, JFK, and LAX airports, respectively; Section VII.D compares the fuel burn results between the three airports and identifies the major factors that affect the relative fuel burn merits between adaptation types and strategies; Section VII.E discusses the effect of the speed brake usage condition on the extra fuel burn of the simple strategies; Section VII.F investigates the wind conditions that rendered fuel-efficient shallow descent FPAs; and finally, Section VII.G estimates the fuel efficiency of the two simple strategies adapted to the NAS level, using the analysis result for the three airports. +VII.A. Dallas/Fort Worth International AirportSection VII.A.1 shows the distribution of FPAs selected by Min-Fuel FPA and discusses their correlation with winds.Sections VII.A.2 and VII.A.3 show the FPA and FPA function selected by Universal FPA and Descent-Speed FPA, respectively.Section VII.A.4 compares the fuel-burn benefits of the three strategies. +VII.A.1. Min-Fuel FPAFigure 11 presents distributions of the FPAs selected by Min-Fuel FPA categorized by metering fixes.From November to April, steeper FPAs are selected for KARLA and HOWDY whereas shallower FPAs are selected for DEBBB and FEVER.This is expected since DEBBB and FEVER arrivals experienced mostly tailwinds in these months, while flights entering KARLA and HOWDY experienced mostly headwinds.For HOWDY (SE), the third quartile of the selected FPAs reached the steepest -3.5 • on Febuary 1, April 15, and April 27.During July and August, the FPAs selected for all four gates were close to -2.6 • and -2.7 • .This observation was consistent with the weaker winds in summer shown in Figure 4.The direction and magnitude of wind was the strongest discriminator, causing correlated fluctuations of the selected FPAs for all four gates.Although steeper FPAs are typically selected for strong headwinds, a few exceptions occurred, as shown for KARLA on January 31, Febuary 28, Deccember 21, and December 23, where the first quartile FPA was found to be -1.8 may be the most fuel-efficient in some wind conditions, and will be further investigated in Section VII.F. +VII.A.2. Universal FPAFigure 12 shows the values of γ univ selected for the airport-specific adaptations of Universal FPA to DFW.In general, the γ univ selected for an adaptation of shorter timespan fluctuated mostly at or above (steeper than) the γ univ selected for a longer timespan.This is because the average of the FPAs of shorter timespan, although possibly fuel-efficient, very often did not satisfy the criteria of the 99% feasibility rate.When the fluctuation of winds is large, the average value may be particularly unflyable on some days and therefore must be rejected.For the γ univ selected for the Airport-Day adaptation, the fluctuation was larger in winter and smaller in summer.The γ univ selected by the Airport-Season adaptation for summer (July, August, and September) appeared to be shallower than most of the Airport-Day values of γ univ during this period of time.This was because the γ univ for this season was bounded by September 5 and 6, which required a shallow FPA of -2.2 • for 80% of the flights entering DEBBB to have flyable trajectories.Figure 13 shows values of γ univ selected for the gate-specific adaptations of Universal FPA.Generally speaking, the selected values of γ univ for KARLA and HOWDY were very close, with HOWDY having slightly steeper of γ univ .The FPAs for DEBBB and FEVER were very close, with DEBBB having slightly shallower values of γ univ .The variation of Gate- Day γ univ between gates reached 1.2 • on April 15, when -3.1 • was selected for HOWDY and -1.9 • was selected for DEBBB.For a specific gate, the variation of Gate-Day γ univ between days reached 1.4 • for KARLA, when -1.8 • was selected on February 28 and -3.2 • was selected on December 4.-3.4 -3.0 -2.Note that the values of γ univ for the airport-specific adaptations shown in Figure 12 were very close to the values of γ univ selected for DEBBB shown in Figure 13.This is because the selected γ univ must ensure a feasibility rate of 80% for any given gate on any day.Therefore, the γ univ selected for airport-specific adaptations is mostly constrained by the shallowest FPAs selected for DEBBB. +VII.A.3. Descent-Speed FPARecall that the family of FPA functions described in Section III.A changes the selected FPA by 0.1 • for every 10 knots of the descent CAS.A selected FPA function is defined by the FPA it yields at 250 knots, denoted as γ 0 .Figure 14 shows γ 0 selected for the airport-specific adaptations of Descent-Speed FPA.Similar to the values of γ univ in Figure 12, values of γ 0 for an adaptation of shorter timespan fluctuate mostly at or above values of γ 0 selected for a longer timespan.For values of γ 0 selected for the Airport-Day adaptation, the fluctuation is larger in winter and smaller in summer.Figure 15 shows values of γ 0 selected for the gate specific adaptations of Descent-Speed FPA to DFW.Similar to Universal FPA, γ 0 selected for KARLA on Febuary 28 was -1.8 250 knots.Similar to Universal FPA, the selection of γ 0 in Figure 14 was constrained by the results for DEBBB in Figure 15.Therefore the values of γ 0 for airport-specific adaptations were very close to those in the gate-specific adaptations for DEBBB. +VII.A.4. Fuel Burn ComparisonThe fuel burn comparison was based on trajectories from the freeze horizon to the metering fix, and therefore had contributions from both the cruise and descent segments.Figure 16 shows the extra fuel-burn per flight computed for Universal FPA and Descent-Speed FPA relative to Min-Fuel FPA.The first observation made was that even the simplest strat- egy, a single static FPA adapted for DFW, has the potential to come within 25 lbs of the minimum-fuel solution for each flight on average.To put this into perspective, this represents approximately 5% of the fuel burned by a typical small-jet arrival transitioning over a 160 nmi segment from cruise to the TRACON boundary.To see the potential impact of directional and temporal adaptations on the fuel efficiency of Universal FPA and Descent-Speed FPA, consider the Universal FPA results first.Relative to the Airport-Static adaptation, the Gate-Static adaptation has the potential to reduce the "extra" 25 lbs of fuel per flight by 16%.In other words, publishing a universal FPA for each of the four gates will recover a little less than a fifth of the way to the minimum fuel solution.By comparison, adapting the universal "airport" FPA to season, month and day has the potential of reducing that extra 25 lbs of fuel per flight by 8%, 19%, and 26%, respectively.When combined, the directional and temporal adaptations together have the potential for reducing the 25 lbs of extra fuel burn by 52%.Essentially, the combined adaptation of a universally fixed FPA for each arrival gate, for each day of operations, recovers more than half of the fuel savings of the minimum-fuel solution for each flight.In considering the overall results for Descent-Speed FPA, both the directional and temporal adaptations yield similar improvements in fuel efficiency.By "adapting" to the descent speed, Descent-Speed FPA is essentially a surrogate adaptation for metering delay, one of the primary factors being considered in this analysis.As such, this strategy was anticipated to yield a fair amount of benefit under metering conditions.The results indicate that Descent-Speed FPA contributes a mere 3% reduction in the 25 lbs of extra fuel burn over Universal FPA for the Airport-Static case, and anywhere from 10% to 23% compared to their Universal FPA counterparts.The combined effect of Descent-Speed FPA with both directional and temporal adaptation has the potential to recover 63% of the fuel burn benefits of Min-Fuel FPA. +VII.B. John F. Kennedy International AirportThe selected FPAs for JFK correlated strongly with the winds aloft along the route.The following paragraph discusses the FPAs selected by the gate-specific adaptations of Universal FPA.Results of the other adaptations of Universal FPA and those of Descent-Speed FPA had similar distributions and are shown in the Appendix in Figs A.1, A.2, A.3, and A.4.Figure 17 shows values of γ univ selected for the gate-specific adaptations of Universal FPA.CCC clearly had the steepest FPAs selected on average compared to the other three gates.This is in accord with the fact that flights into CCC experienced mostly strong headwinds throughout the year.The other three gates had shallower FPAs selected.The Daily FPAs selected for HARTY were the shallowest of all gates.Figure 18 shows the extra fuel burn per flight computed for Universal FPA and Descent-Speed FPA.The Airport-Static adaptation has the potential to come within 22 lbs of the minimum-fuel solution for each flight on average.This amount is slightly less than that for DFW, and represents approximately 4% of the fuel burn for a typical small-jet arrival transitioning over a 160 nmi segment from cruise to the TRACON boundary.Adapting the universal airport FPA to season, month and day has the potential of reducing that extra 21.5 lbs of fuel per flight by 12%, 21%, and 38%, respectively.Relative to the Airport-Static adaptation, adaptation to the arrival gate only slightly reduces the extra 21.5 lbs of fuel per flight by 6%.Adaptation to gates becomes more effective when combined with finer granularity of timespan, as the Gate-Season, Gate-Month, and Gate-Day adaptations reduce the extra fuel burn from the corresponding airport-specific adaptations by 11%, 14%, and 38%, respectively.The combined adaptation of a universally fixed FPA for each arrival gate, for each day of operations, recovers about 61% of the fuel savings of the minimum-fuel solution for each flight.Descent-Speed FPA contributes a 21% reduction in the 21.5 lbs of extra fuel burn over Universal FPA for the Airport-Static case.In considering the overall results for Descent-Speed FPA, both the directional and temporal adaptations yield similar improvements in fuel efficiency ranging from 22% to 29% among the types of adaptation when compared to Universal FPA counterparts.The combined effect of Descent-Speed FPA with both directional and temporal adaptation has the potential to achieve 74% of the Min-Fuel FPA benefit. +VII.C. Los Angeles International AirportThe magnitude and variation of the winds aloft in LAX was much less than that of DFW and JFK, and the selected FPAs also showed less variation across gates and times.Figure 19 shows values of γ univ selected for the gate-specific adaptations of Universal FPA.The other adaptations of Universal FPA and those of Descent-Speed FPA had similar distributions and are shown in the Appendix in Figs A.1, A.2, A.3, and A.4.One interesting observation is that GRAMM had the shallow FPA of -1.8 • selected for many days.This will be further discussed in Section VII.F. Figure 20 shows the extra fuel burn per flight computed for Universal FPA and Descent-Speed FPA.The simplest strategy, a single static FPA adapted for LAX, comes within 14 lbs of the minimum-fuel solution for each flight on average.Adapting the universal airport FPA to season, month and day has the potential of reducing that extra 14 lbs of fuel per flight by 12%, 21%, and 38%, respectively.Adaptation to the arrival gate is even less effective for LAX than for JFK, only slightly reducing the extra 14 lbs of fuel per flight by 2%.Adaptation to gates becomes more effective when combined with finer granularity of timespan, as the Gate-Season, Gate-Month, and Gate-Day adaptations reduce the extra fuel burn from the corresponding airport-specific adaptations by 5%, 17%, and 34%, respectively.The combined adaptation of a universally fixed FPA for each arrival gate, for each day of operations, recovers about 44% of the fuel savings of the minimum-fuel solution for each flight.-3.4 -3.0 -2.6 -2.2 -1Interestingly, Descent-Speed FPA with Airport-Static adaptation performs slightly worse fuel burn of the Airport-Static adaptation.This is in contrast to JFK and LAX, for which the Gate-Static adaptation only recovered 6% and 2%, respectively.DFW's four-corner-post arrival configuration had two tailwind prevailing gates and two headwind prevailing gates.JFK had three tailwind prevailing gates that account for more than 80% of the small jet arrival traffic.LAX had small wind variations for all three gates.Therefore, the benefits of the Gate-Static adaptation relative to Airport-Static was much smaller for JFK and LAX.Figure 21 shows the FPA selected for each gate by the Gate-Static adaptation of Universal FPA.The gates are grouped by airports in the order of LAX, DFW, and JFK from top to bottom.These values were also shown as dotted lines in Figures 13, 17 that, as the variation of winds aloft increases, selected FPAs separate into "tailwind" and "headwind" groups.For LAX, the FPAs selected for the three gates SYMON, KONZL, and GRAMM, were close to one another and differ by only -0.1 • .For DFW, shallower FPAs were selected for the tailwind-prevailing gates of FEVER and DEBBB, whereas steeper FPAs were selected for the two headwind-prevailing gates of KARLA and HOWDY.The two sets of FPAs are separated by at -0.2 • .For JFK, an FPA of -2.4 was selected for the headwind-prevailing gate, CCC, while shallower FPAs were selected for the other three tailwind-prevailing gates.The three shallower FPAs were different than the other FPA by -0.4 • .Figure 21 also shows the extra fuel burn of the Gate-Static adaptation, associated with each gate, on the right side of the figure.Note that these values were not shown in Figures 16, 18, and 20, whose extra fuel burn for the Gate-Static were aggregated over all gates in each airport.In addition to the variation of winds with time, which increases the extra fuel burn, three other major factors impact the extra fuel burn of a Gate-Static adaptation:• The direction of wind along the route impacts the extra fuel burn in an interesting way.It was observed that many headwind distributions diminished the fuel burn difference among trajectories with varying FPAs.The fuel burn becomes a weak function of the FPA, and sometimes a very shallow FPA burns less fuel (to be discussed in Section VII.F).As a result, headwind-prevailing gates tended to have less extra fuel burn than tailwind-prevailing gates.KARLA and HOWDY had less extra fuel burn than FEVER and DEBBB mainly because the former two gates are headwind-prevailing.• High-altitude constraints reduce the length of the descent segment and diminish the difference between trajectories.As a result, the fuel burn difference between different FPAs is reduced.The gates of GRAMM and KONZL for LAX and LOLLY and HARTY for JFK had high-altitude constraints, and therefore the extra fuel burn for these gates were much reduced.• The aircraft fleet composition directly impacts the absolute values of the extra fuel burn.JFK has larger regional jets and therefore its gates can potentially have more extra fuel burn than LAX and DFW.Comparison of the extra fuel burn of Descent-Speed FPA to those of Universal FPA showed that, across the airport-specific and gate-specific adaptation types, Descent-Speed FPA was more effective for JFK and DFW than for LAX.A closer look reveals that Descent-Speed FPA was more beneficial for tailwind-prevailing gates than head-wind prevailing gates.Figure 22 shows the γ 0 for each gate by the Gate-Static adaptation of Descent-Speed FPA.Compared to Universal FPA, Descent-Speed FPA reduced the extra fuel of gates prevailed by strong tailwinds, such as CAMRN, DEBBB, and FEVER, by 27%, 24%, and 23%, respectively.Descent-Speed FPA did not reduce the extra fuel of the other gates more than 15%. +VII.E. Effect of Speed-Brake Usage on Fuel BurnRecall that two levels of planned speed-brake usage, SB0 and SBANY, were described in Section III.B.The SB0 condition ensured that planned-speed brake usage was not required; i.e., the aircraft always maintained a clean configuration during descent.This restriction had an effect on the selected FPA and the resulting extra fuel burn when compared with the SBANY condition.Figure 23 fuel burn of each adaptation type of the SB0 condition by 55% to 70%. Figure 24 shows the sizable effect that the planned-speed-brake usage had on the feasibility rate and, therefore, the selection of the parameters and the fuel burn for a group of flights.This figure sketches a notional average fuel burn per flight and feasibility rate as a function of γ univ or γ 0 .The γ univ or γ 0 that yields the least average fuel burn per flight while having a feasibility rate of at least 99% was selected.The SB0 condition ensured that planned speed brake usage was not required, which resulted in an early fall-off of the feasibility rate and therefore a steeper cutoff FPA that would restrict the acceptable γ univ or γ 0 to shallower values.The more forgiving SBANY condition very often allows the minimum-fuel FPA (aggregated over flights) to be selected, as the computed cutoff γ univ or γ 0 (using the criterion of 99% feasibility rate) for SBANY is frequently steeper than the minimum-fuel γ univ or γ 0 .The significant difference between the extra fuel burn of SB0 and SBANY suggested that, in actual operations, it may be worthwhile to trade the robustness of SB0 for the fuel efficiency of SBANY by a planned speed brake usage that allows certain fraction of the speed brake to be deployed.A way of quantifying the concept of robustness would be needed for such a analysis. +VII.F. Fuel-Efficient Shallow DescentsAnalysis of the fixed-FPA arrival trajectories showed that headwinds shift the minimumfuel FPA to a steeper value whereas tailwinds shift the minimum-fuel FPA to a shallower value.Some headwinds, however, also tend to reduce the fuel burn for shallow FPAs.In these wind conditions, the shallowest FPA analyzed, -1.8 • , can turn out to be the minimum-fuel FPA. Figure 25 demonstrates the fuel burn as a function of the FPA for two E145 flights from 160 nmi away to the NE gate KARLA at DFW. EGF2708 arrived at DFW on Feburay 1 and had a minimum-fuel FPA at -3.4 • .Such minimum-fuel FPA in the middle of the range of sampled FPAs is characteristic of the meet-time trajectories of most arrival flights.EGF3272 arrived at DFW on Febuary 28 and the fuel burn had an inversion in slope that resulted in -1.8 • (the shallowest FPA analyzed) being the minimum-fuel FPA.The Gate-Day adaptation of Universal FPA could have selected -1.8 • if a majority of the arriving flights had this inversion.Analysis of the winds aloft along the route revealed that such inversion of slope was highly correlated with the wind gradient with respect to the altitude.Instead of fighting against a strong headwind at a high altitude, if may be advantageous for the aircraft to descend early down to altitudes with a weaker headwind or even a tailwind.Take the two flights in Figure 25 as an example.Between FL360 and FL250 during the descent, the headwind experienced by EGF2708 decreased from 110 knots to 90 knots, while the headwind experienced by EGF3272 decreased from 120 knots to 60 knots.The faster decay of the headwind for EGF3272 contributed to favoring an early descent.Equation 6 considers an average wind gradient experienced by a flight:W = (W (x, y, z, t, Ψ) -W (x, y, z -∆z, t, Ψ)) ∆z ,(6)where W represents a measure of the change of the wind along the route with respect to altitude, x, y, z, t represent the flight's initial 4D position for the analysis, Ψ represents the flight's heading, and ∆z represents a characteristic change of altitude.A value of 8,000 ft was chosen for ∆z.Calculation of W for all flights arrving at DFW through KARLA showed that, on average, the flights on Feburay 28 experienced the greatest change of the along-theroute wind.The average value of W for the flights entering KARLA was 4.4 knots per 1,000 ft.This was correlated with the shallow FPA selected for Febuary 28 as shown in Figure 13.Interestingly, the Gate-Day adaptation for GRAMM of LAX also selected -1.8 • on many days of the year, despite the smaller wind magnitudes in LAX.Analysis of the wind gradient at LAX revealed that the average value of W also correlated strongly with the occurrence of the selected shallow FPA.Nonetheless, the correlation is not 100%, and other factors such as descent altitude ranges affect the fuel burn as well.In actual operations, a shallow FPA may extend the descent phase beyond the center boundary or cause more sector crossings, and implementation and execution of such shallow descents can be problematic.Therefore, airspace-related constraints must be taken into account when selecting the FPA for the simple strategies. +VII.G. Adaptation to the NASTo demonstrate the applicability of the methodology presented in this paper, analysis results of Universal FPA for the three airports of DFW, JFK, and LAX, were aggregated and used to select the FPA for the NAS adaptation types.While the three airports already covered a wide range of wind variation and disparate arrival traffic flows, additional analysis for other airports can be easily incorporated to improve the fuel burn estimates for the NAS adaptation types.Note that the aggregation averaged fuel burn over all the flights analyzed in the three airports, with each flight given the same weight.month adaptation has slightly less extra fuel burn than the Airport-Static adaptation.This suggests that publishing an FPA per month for the NAS may result in more fuel benefits than publishing a year-round FPA for each major airport.The finest Gate-Day adaptation has an extra fuel burn of 10 lbs and recovers about 73% of the extra fuel burn of the NAS-Static adaptation. +VIII. DiscussionThe motivation for a standard descent procedure for arriving small jet flights in the transition airspace is to improve trajectory predictability, especially near the top-of-descent region.The improved trajectory predictability leads to better airspace throughput and safety, which benefits not just the small jets but all flights.While Min-Fuel FPA provided the best fuel benefits, it is difficult to implement without datalink.Instead, two simpler strategies, Universal FPA and Descent-Speed FPA, were adapted to time, airspace, and direction to improve their fuel efficiency.The results showed that adaptation to gates and days could recover 50% to 70% of the extra fuel burn of the Airport-Static adaptation.Since regional jet operators consider fuel savings to the order of 10 lbs, the fuel-burn difference of 7 to 17 lbs per flight between the adaptation types may be important.Moreover, adaptation to shorter timespans may yield more fuel benefits in operation than estimated in the analysis, because high-fidelity weather data are unavailable for long look-ahead times such as a year, a season, or even a month.Variation of the performance envelope between aircraft, not modeled in the analysis, could compound with wind variation and raise the extra fuel burn, too.On the other hand, the speed brake usage, SB0, could be relaxed to reduce the extra fuel burn if it is determined that a certain level of relaxation would still maintain the required robustness of the procedure.Looking towards the operational implementation of continuous descents under metering conditions, the FAA will have to decide upon an approach for defining continuous descent FPAs for small jets.While this paper compares and contrasts the relative fuel efficiency of candidate strategies for defining descent FPAs, other factors will enter the implementation decision, not the least of which will be the complexity and cost of implementing and supporting a fixed-FPA procedure.Universal FPA has the advantage of defining a single descent FPA for an airport.Descent-Speed FPA, while slightly more complex in terms of defining FPA as a function of descent speed, has the potential to capture much of the fuel efficiency related to the descent speed.Both strategies lend themselves to fast-time analysis that can support the selection and publication of their parameters a day or more before flight.This would allow the FAA to adapt the appropriate parameters to each airport and disseminate those parameters through the aeronautical information network.Depending on the time horizon chosen for the adaptation (annual, seasonal, monthly, weekly or daily), this information may be made available in several ways.For longer periods of time, the parame-ters could be included as part of published arrival procedures or flight manual amendments.Alternatively, this information could be provided to pilots as part of their standard preflight planning and weather briefing.At the very least, this dissemination approach would be necessary for adaptations performed on a more frequent (e.g., daily) basis.Given the sensitivity of the selected FPA to prevailing winds and the lack of a precise wind forecast for longer time horizons, analysis based on historical data or climatological data would lend itself to parameter adaptation on an annual, seasonal or monthly basis.Historical data from previous years may be used for such prediction, assuming a similar pattern of weather and wind.For shorter adaptation horizons, on the order of daily updates, numerical weather prediction such as the National Oceanic and Atmospheric Administration's Rapid Refresh weather forecast 34 (succeeding RUC) would provide precise and relatively accurate forecasts. +IX. ConclusionPrediction of small jet descents has been challenging, as descent planning varies greatly from flight to flight and airline to airline.Lack of trajectory predictability makes it difficult for ATC to achieve Continuous Descent Arrivals (CDA) for these jets.This paper proposed a standard fixed-flight-path angle descent procedure for arriving small jets in the transition airspace in order to improve trajectory predictability and enhance airspace throughput.This standard descent procedure would help ground automation tools such as the Efficient Descent Advisor in delivering aircraft to the metering fix on time while maintaining separation.Three strategies for choosing the descent FPA were presented.While the three strategies vary in operational complexity, they are expected to achieve the same level of trajectory predictability.The selection of the FPA considered fuel burn and flyability, which was modeled by the planned speed-brake usage.Adaptation to time, airspace, and direction was proposed to improve the fuel benefits of the two simple strategies: Universal FPA and Descent-Speed FPA.The Min-Fuel FPA strategy served as a reference point for the fuel-burn metrics.Analysis of the winds aloft at twelve major US airports led to the selection of the JFK, DFW, and LAX airports for the application of the FPA selection methodology.The three airports had very different degrees of wind variation and disparate arrival traffic flows.Results showed that the selection methodology successfully selected fuel-efficient and flyable FPAs for all three strategies, and the selected FPA correlated strongly with winds along the route.The Airport-Static adaptation of the Universal FPA burned by 14 to 26 lbs extra fuel per flight compared to the Min-Fuel FPA solution.The finest adaptation could potentially recover 50% to 70% of the extra fuel.Various factors affecting the fuel benefits of the simple strategies were investigated.Planned speed brake usage, when allowed in descent, would decrease the extra fuel burn noticeably.The fuel burn comparison in this work provides information for the design of a standard descent procedure in the transition airspace to support the CDA.Although the comparison focused on small jets, the analysis could apply to any jets that perform fixed-FPA descents.The choice of an economically appropriate adaptation type of the simple strategies can be NAS-wide, airport-specific, or gate-specific.It is ultimately incumbent on the the FAA to perform the cost-benefit analysis of potential implementations. +AppendixThe results of the three strategies applied to JFK and LAX that are not shown in Section VII are shown here for completeness.Figure 1 .1Figure 1.A flight following a fixed-FPA procedure to descend and cross the metering fix at a STA of t MF . +Figure 2 .2Figure 2. Min-Fuel FPA selects the minimum-fuel FPA. +Figure 3 Figure 3 .33Figure 3. Variation of winds aloft at twelves major airports in the United States for 2011. +Figure 4 .4Figure 4. Winds aloft at 30,000 ft in summer and winter of year 2011 at LAX, DFW, and JFK.Red and green dots represent wind vectors in summer and winter, respectively. +Figure 5 .5Figure 5. Tracks of arival small jets into DFW on July 28th, 2011. +Figure 6 .6Figure 6.Tracks of arrival small jets to JFK on January 7, 2011. +Figure 7 .7Figure 7. Tracks of arrival small jets to LAX on August 5, 2011. +Figure 8 .8Figure 8.A modeled vertical profile of an arrival flight. +Figure 9 .9Figure 9.The three strategies select different FPAs in general. +Figure 10 .10Figure 10.Distribution of types of aircraft analyzed for the three airports. +Figure 12 .12Figure 12.Values of γ univ selected by the airport-specific adaptations of Universal FPA to DFW. +Figure 14 .14Figure 14.Values of γ 0 selected by the airport-specific adaptations of Descent-Speed FPA to DFW. +to Min-Fuel (lbs/flight) Extra Fuel Burn as a % of Baseline Universal FPA Descent Speed FPA +Figure 16 .16Figure 16.Extra fuel burn calculated for various adaptations of Universal FPAand Descent-Speed FPA to DFW. +Figure 18 .18Figure 18.Extra fuel burn calculated for eight adaptations of Universal FPA and Descent-Speed FPA to JFK. +Figure 21 .21Figure 21.Universal FPA: Gate-Static: The γ univ and extra fuel burn for each gate. +Figure 22 .22Figure 22.Descent-Speed FPA: Gate-Static: The γ 0 and extra fuel burn for each gate. +FuelSBANYFigure 23 .23Figure23.Impact of different levels of the planned speed brake usage on the extra fuel burn, when applied to DFW. +Figure 24 .24Figure 24.Notional representation of the average fuel burn and feasibility rate.The feasibility rate falls off at steeper values of γ univ or γ 0 . +Figure 25 .25Figure 25.Fuel burn as a function of the FPA for two flights.While EGF2708 had a minimum-fuel FPA of -3.4 • , EGF3272 had a minimum-fuel FPA of -1.8 • . +Figure 26 Figure 26 .2626Figure 26.Estimates of the extra fuel burn for NAS adaptation types. +Figure A. 1 :1distributions of the FPAs selected by Min-Fuel FPA for JFK. Figure A.2: γ univ selected by the airport-specific adaptations of Universal FPA to JFK. Figure A.3: γ 0 selected by the airport-specific adaptations of Descent-Speed FPA to JFK.Figure A.4: values of γ 0 selected by the gate specific adaptations of Descent-Speed FPA to JFK. +Figure A. 5 :5distributions of the FPAs selected by Min-Fuel FPA for LAX. +Figure A. 6 :6γ univ selected by the airport-specific adaptations of Universal FPA to LAX. +Figure A. 7 :7γ 0 selected by the airport-specific adaptations of Descent-Speed FPA to LAX. Figure A.8: γ 0 selected by the gate-specific adaptations of Descent-Speed FPA to LAX. +- +Table 1 . Adaptation of Universal FPA and Descent-Speed FPA to time and airspace/direction1The table illustrates the overall approach and potential scope.The Min-Fuel FPA is shown at the bottom-right corner to represent the finest "adaptation" of the simple strategies.The analysis in this paper will assess the adaptation types denoted with white cells in the table.The analysis of adaptations at the level of specific arrival routes and/or hours of the day is left for future work.AdaptationAirspace/DirectionNAS Airport Arrival Gate Arrival RouteStaticSeasonTimeMonthDayHourMin-Fuel FPAcolumn represents a progressively finer adaptation of the strategies to a specific airport,individual arrival gates (corner posts) feeding an airport, all the way down to specific arrivalroutes feeding each arrival gate. The rows represent a temporal scale starting at the top with +Table 2 .2The altitude and speed restrictions for arrival flightsAirportWindGateAltitude (ft) Speed (CAS in knots)Strength(Metering Fix)KARLA11,000250DFWMediumHOWDY FEVER11,000 11,000250 250DEBBB11,000250CCC12,000250JFKStrongCAMRN LOLLY12,000 FL200250 -HARTYFL210-KONZL17,000280LAXWeakGRAMMFL190280SYMON12,000280V.A. Dallas/Fort Worth International Airport +Table 3 .3Total number of 2011 flights analyzed for the airportsAirport Flights Gate Percentage (%)GRAMM18LAX48692KONZL33SYMON49KARLA39DFW81483HOWDY FEVER29 15DEBBB17CCC13JFK51576CAMRN LOLLY36 26HARTY25 +Figure 11.The first quartile, third quartile, and median FPAs selected by Min-Fuel FPA for DFW.DEBBB (NW)KARLA (NE)-3.4-3.4-3.0-3.0-2.6 γ (deg)-2.6-2.2-2.2-1.8-1.8JANFEBMARAPRMAYJUNJULAUGSEPNOVDECJANFEBMARAPRMAYJUNJULAUGSEPOCTNOVDEC1st QuartileFEVER (SW)HOWDY (SE)3rd Quartile Median-3.4-3.4-3.0-3.0-2.6 γ (deg)-2.6-2.2-2.2-1.8-1.8JANFEBMARAPRMAYJUNJULAUGSEPOCTNOVDECJANFEBMARAPRMAYJUNJULAUGSEPOCTNOVDEC•.This suggests that very shallow FPAs +Table 4 .4FPA and extra fuel burn for the Airport-Static and Gate-Static adaptation of Universal FPAAirportFPAExtra Fuel Burn (lbs)Extra Fuel Burn (lbs)% FuelAirport-StaticAirport-StaticGate-StaticRecoveredLAX-2.2 •13.813.52%DFW-2.1 •24.720.816%JFK-1.9 •21.620.26% +Figure A.1.The first quartile, third quartile, and median FPAs selected by Min-Fuel FPA for JFK.Figure A.2. Values of γ univ selected for the airport-specific adaptations of Universal FPA to JFK.Figure A.3.Values of γ 0 selected by the airport-specific adaptations of Descent-Speed FPA to JFK.Figure A.4.Values of γ 0 selected by the gate-specific adaptations of Descent-Speed FPA to JFK.The first quartile, third quartile, and median FPAs selected by Min-Fuel FPA for LAX.Figure A.6.Values of γ univ selected by the airport-specific adaptations of Universal FPA to LAX.Figure A.7. Values of γ 0 selected by the airport-specific adaptations of Descent-Speed FPA to LAX.Values of γ 0 selected for the gate-specific adaptations of Descent-Speed FPA to LAX.SYMON-3.4Airport-Static KONZL-3.4-3.4Airport-SeasonAirport-Month-3.0-3.0-3.0Airport-Day-2.6 γ (deg)γ univ (deg)-2.6-2.6-2.2-2.2-2.2-1.8-1.8JANFEBMARAPRMAYJUNJULAUGSEP-1.8 OCT NOVJAN DECFEBMARAPR JANMAY FEBJUN MARJUL APRAUG MAYSEP JUN JULOCT AUGNOV SEPDEC OCTNOVDEC1st QuartileGRAMM3rd Quartile Median-3.4-3.0-2.6-3.4SYMON-3.4Airport-Season Airport-Static -3.4 -2.2 GRAMMγ 0 (deg)-3.0 -2.6γ 0 (deg)-3.0 -2.6-1.8 -3.0 JAN -2.6Airport-Day Airport-Month FEB MAR APR MAY JUN JULAUGSEPOCTNOVDEC-2.2 -1.8 AUG SEP OCT Figure A.5. -3.4 -2.2 -1.8 JAN FEB MAR APR MAY JUN JULJAN NOVFEB DECMAR -2.2 APR -1.8 Airport-Season MAY JUN JUL AUG SEP JAN FEB MAR APR MAY JUN OCT JUL Airport-Month Airport-Static -3.4 KONZL NOV AUGDEC SEPOCTNOVDECGate-Season Gate-Static Gate-Day Gate-Month-3.0Airport-Day-3.0γ univ (deg)-2.6-2.6LOLLYCCC3.4LOLLY-2.2-3.4 -2.2CCC-3.4-3.4 -1.8-3.0 -2.6 -3.0 γ (deg) -2.6 γ 0 (deg) Figure A.8.-1.8JAN-3.0 -3.0 FEB MAR APR -2.6 -2.6MAY JANJUN FEBJUL MARAUG APRSEP MAYOCT JUN JULNOV AUGSEP DECOCTNOVDEC-2.2 -2.2-2.2 -2.2-1.8 -1.8 JAN -3.4 -3.4FEB JANMAR FEBAPR MARMAY APR HARTY JUN JUL MAY JUN HARTY AUG JULSEP AUG -3.4 OCT NOV SEP OCTDEC NOV-1.8 -1.8 JAN DEC -3.4 -3.4 Airport-Season FEB MAR APR MAY JUN JUL JAN FEB MAR APR MAY JUN CAMRN Airport-Month CAMRN AUG JUL Airport-StaticSEP AUGOCT SEPNOV OCTDEC NOVDECGate-Static 1st Quartile 3rd Quartile Median Gate-Day Gate-Month Gate-Season-3.0Airport-Day-3.0 -3.0-3.0 -3.0-2.6 γ (deg) -2.6 γ 0 (deg)γ 0 (deg)-2.6-2.6 -2.6-2.2 -2.2-2.2-2.2 -2.2-1.8 -1.8 JANFEB JANMAR FEBAPR MARMAY APRJUN MAYJUL JUNAUG JULSEP AUG -1.8 OCT NOV SEP OCTDEC NOV JAN-1.8 DEC -1.8 JAN FEB MAR APRFEB JAN MAYMAR FEB JUNAPR MAR JULMAY APR AUGJUN MAY SEP JUL JUN OCT AUG JULSEP AUG NOVOCT SEP DECNOV OCTDEC NOVDEC + + + +than the Universal FPA's counterpart by 1%.Other directional and temporal adaptations of Descent-Speed FPA do yield mild improvements in fuel efficiency ranging from 1% to 17% among the types of adaptation when compared to Universal FPA counterparts.The combined effect of Descent-Speed FPA with both directional and temporal adaptation has the potential to achieve 53% of the Min-Fuel FPA benefit. +VII.D. Comparison of the Three AirportsThis section compares results of the analysis for the DFW, JFK, and LAX airports in terms of the selected FPAs and fuel burn merits of the simple selection strategies.Such comparison gives insight to the design of a fixed-FPA descent procedure by identifiying a set of leading factors that can potentially affect the decision.The FPA selected by the Airport-Static adaptation of the Universal FPA showed correlation with the variation of winds aloft.Recall that LAX had the mildest variation of winds while JFK had the strongest variation of winds of all three airports.Table 4 shows that, going from top to bottom, shallower FPAs were selected for airports with stronger variation of winds.The extra fuel burn for the Airport-Static adaptation, however, does not show a clear trend.This is mainly due to the fact that the extra fuel burn for LAX and JFK was reduced by the shorter descent segments analyzed for gates that have high-altitude restrictions.A trend would have been seen for the extra fuel burn if the altitude restrictions were similar among the airports.Compared to the Airport-Static adaptation, the Gate-Static adaptation had the most benefits when the gates were dissimilar in terms of the winds along the route.This was the case for DFW, when the Gate-Static adaptation of Universal FPA recovered 16% of the extra + + + + + + + The Efficient Descent Advisor: Technology Validation and Transition + + RichardCoppenbarger + + + MiwaHayashi + + + GauravNagle + + + DouglasSweet + + + RenanSalcido + + 10.2514/6.2012-5611 + + + 12th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference and 14th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference + + American Institute of Aeronautics and Astronautics + 2012-5611, Sept. 2012. March 2011 + 2 + + + Federal Aviation Administration + Coppenbarger, R. A., Nagle, G., Sweet, D., and Hayashi, M., "The Efficient Descent Advisor: Technol- ogy Validation and Transition," Proceedings of the AIAA Aircraft Technology, Integration, and Operations Conference, AIAA-2012-5611, Sept. 2012. 2 "FAA's NextGen Implementation Plan," Federal Aviation Administration, March 2011. + + + + + The Next-Generation Air Transportation System's Joint Planning Environment: A Decision Support System + + EdgarWaggoner + + + ScottGoldsmith + + + JoshElliot + + 10.2514/6.2009-7011 + + + 9th AIAA Aviation Technology, Integration, and Operations Conference (ATIO) + + American Institute of Aeronautics and Astronautics + Oct. 2009 + + + "Concept of Operations for the Next Generation Air Transportation System," Joint Planning and Development Office (JPDO), Oct. 2009. + + + + + Optimized Profile Descent Arrivals at Los Angeles International Airport + + J-PClarke + + + JBrooks + + + GNagle + + + AScacchioli + + + WWhite + + + SRLiu + + 10.2514/1.c031529 + + + Journal of Aircraft + Journal of Aircraft + 0021-8669 + 1533-3868 + + 50 + 2 + + 2013 + American Institute of Aeronautics and Astronautics (AIAA) + + + Clarke, J.-P., Brooks, J., Nagle, G., Scacchioli, A., White, W., and Liu, S., "Optimized Profile Descent Arrivals at Los Angeles International Airport," Journal of Aircraft, Vol. 50, No. 2, 2013, pp. 360-369. + + + + + Field Evaluation of the Tailored Arrivals Concept for Datalink-Enabled Continuous Descent Approach + + RichardACoppenbarger + + + RobWMead + + + DouglasNSweet + + 10.2514/1.39795 + + + Journal of Aircraft + Journal of Aircraft + 0021-8669 + 1533-3868 + + 46 + 4 + + July-August 2009 + American Institute of Aeronautics and Astronautics (AIAA) + + + Coppenbarger, R. A., Mead, R. W., and Sweet, D. N., "Field Evaluation of the Tailored Arrivals Concept for Datalink-Enabled Continuous Descent Approach," Journal of Aircraft, Vol. 46, No. 4, July- August 2009, pp. 1200-1209. + + + + + Development, Design and Flight Test Evaluation of Continuous Descent Approach Procedure in FIR Zagreb + + DorisNovak + + + TinoBucak + + + TomislavRadišić + + 10.7307/ptt.v21i5.247 + + + PROMET - Traffic&Transportation + PROMET + 0353-5320 + 1848-4069 + + 21 + 5 + + Sept. 2009 + Faculty of Transport and Traffic Sciences + + + Novak, D., Bucak, T., and Radisić, T., "Development, Design and Flight Test Evaluation of Continuous Descent Approach Procedure in FIR Zagreb," PROMET -Traffic & Transportation, Vol. 21, No. 5, Sept. 2009, pp. 319-329. + + + + + Continuous Descent Approach: Design and Flight Test for Louisville International Airport + + John-Paul B.Clarke + + + NhutTHo + + + LilingRen + + + JohnABrown + + + KevinRElmer + + + Kwok-OnTong + + + JosephKWat + + 10.2514/1.5572 + + + Journal of Aircraft + Journal of Aircraft + 0021-8669 + 1533-3868 + + 41 + 5 + + 2004 + American Institute of Aeronautics and Astronautics (AIAA) + + + Clarke, J.-P. B., Ho, N. T., Ren, L., Brown, J. A., Elmer, K. R., Tong, K.-O., and Wat, J. K., "Continuous Descent Approach: Design and Flight Test for Louisville International Airport," Journal of Aircraft, Vol. 41, No. 5, 2004, pp. 1054-1066. + + + + + Benefits of Continuous Descent Operations in High-Density Terminal Airspace Considering Scheduling Constraints + + JohnRobinson Iii + + + MaryamKamgarpour + + 10.2514/6.2010-9115 + AIAA-2010-9115 + + + 10th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + + American Institute of Aeronautics and Astronautics + Sept. 2010 + + + Robinson, J. E. and Kamgarpour, M., "Benefits of Continuous Descent Operations in High-Density Terminal Airspace Under Scheduling Constraints," Proceedings of the 10th AIAA Aviation Technology, In- tegration, and Operations Conference, AIAA-2010-9115, Sept. 2010. + + + + + FAA Aviation Forecasts: Fiscal Years 1981-1992. Federal Aviation Administration, U.S. Department of Transportation, Washington, D.C. 20591. 1980. 69p + 10.1177/004728758102000159 + + + Journal of Travel Research + Journal of Travel Research + 0047-2875 + 1552-6763 + + 20 + 1 + + 2011 + SAGE Publications + + + "FAA Aerospace Forecast Fiscal Years 2012-2032," Federal Aviation Administration, 2011. + + + + + Design and Operational Evaluation of the Traffic Management Advisor at the Fort Worth Air Route Traffic Control Center + + HNSwenson + + + THoang + + + SEngelland + + + DVincent + + + TSanders + + + BSanford + + + KHeere + + + + Proceedings of the 1st USA + the 1st USA + + June 1997 + + + Swenson, H. N., Hoang, T., Engelland, S., Vincent, D., Sanders, T., Sanford, B., and Heere, K., "Design and Operational Evaluation of the Traffic Management Advisor at the Fort Worth Air Route Traffic Control Center," Proceedings of the 1st USA/Europe Air Traffic Management R&D Seminar , June 1997. + + + + + Experimental Approach for Efficient Arrival Procedures with Fixed-Flight Path Angle Descent + + MGWu + + + SMGreen + + 10.1109/icnsurv.2019.8735138 + + + 2019 Integrated Communications, Navigation and Surveillance Conference (ICNS) + + IEEE + 2011-215992, Nov. 2011 + + + Wu, M. G. and Green, S. M., "Analysis of Fixed Flight Path Angle Descents for the Efficient Descent Advisor," NASA/TM-2011-215992, Nov. 2011. + + + + + Study on Variations in Vertical Profile for CDA Descents + + GarySlater + + 10.2514/6.2009-6912 + AIAA-2009-6912 + + + 9th AIAA Aviation Technology, Integration, and Operations Conference (ATIO) + + American Institute of Aeronautics and Astronautics + Sept. 2009 + + + Slater, G., "Study on variations in vertical profile for CDA descents," Proceedings of the AIAA Aircraft Technology, Integration, and Operations Conference, AIAA-2009-6912, Sept. 2009. + + + + + Descent Profile Options for Continuous Descent Arrival Procedures within 3d Path Concept + + K.-OTong + + + ESchoemig + + + DBoyle + + + JScharl + + + AHaraldsdottir + + + + Proceedings of the IEEE/AIAA 26th Digital Avionics Systems Conference + the IEEE/AIAA 26th Digital Avionics Systems Conference + + Oct. 2007 + + + Tong, K.-O., Schoemig, E., Boyle, D., Scharl, J., and Haraldsdottir, A., "Descent Profile Options for Continuous Descent Arrival Procedures within 3d Path Concept," Proceedings of the IEEE/AIAA 26th Digital Avionics Systems Conference, Oct. 2007, pp. + + + + + Sensitivity Studies of 4D Descent Strategies in an Advanced Metering Environment + + KennethHIzumi + + 10.23919/acc.1986.4789026 + + + 1986 American Control Conference + + IEEE + 1986. June 1986 + + + + Izumi, K. H., "Sensitivity Studies of 4D Descent Strategies in an Advanced Metering Environment," Proceedings of the American Control Conference, 1986 , June 1986, pp. 687-692. + + + + + Sensitivity Studies of 4D Descent Strategies in an Advanced Metering Environment + + KennethHIzumi + + + RWSchwab + + + JLGroce + + + ACoote + + 10.23919/acc.1986.4789026 + + + 1986 American Control Conference + + IEEE + Dec. 1986 + + + Tech. Rep. NASA CR-178093 + Izumi, K. H., Schwab, R. W., Groce, J. L., and Coote, A., "An Evaluation of Descent Strategies for TNAV-Equipped Aircraft in an Advanced Metering Environment ATOPS," Tech. Rep. NASA CR-178093, Boeing Commercial Airplane Company, Dec. 1986. + + + + + 4 D fuel-optimal guidance in the presence of winds + + AChakravarty + + 10.2514/6.1983-2242 + AIAA-1983-2242 + + + Guidance and Control Conference + Gatlinburg, Tenn + + American Institute of Aeronautics and Astronautics + 1983 + + + Chakravarty, A., "Four-Dimensional Fuel-Optimal Guidance in the Presence of Winds," Proceedings of the AIAA Guidance and Control Conference, AIAA-1983-2242, Gatlinburg, Tenn., 1983. + + + + + Relative Significance of Trajectory Prediction Errors on an Automated Separation Assurance Algorithm + + TALauderdale + + + ACCone + + + ARBowe + + + + Proceedings of the 9th USA/Europe Air Traffic Management R&D Seminar + the 9th USA/Europe Air Traffic Management R&D Seminar + + June 2011 + + + Lauderdale, T. A., Cone, A. C., and Bowe, A. R., "Relative Significance of Trajectory Prediction Errors on an Automated Separation Assurance Algorithm," Proceedings of the 9th USA/Europe Air Traffic Management R&D Seminar , June 2011. + + + + + Predictability of Top of Descent Location for Operational Idle-Thrust Descents + + LaurelStell + + 10.2514/6.2010-9116 + + + 10th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + + American Institute of Aeronautics and Astronautics + June 2011 + + + Stell, L., "Prediction of Top of Descent Location for Idle-thrust Descents," Proceedings of the 9th USA/Europe Air Traffic Management R&D Seminar , June 2011. + + + + + Analysis of Top of Descent (TOD) uncertainty + + CraigMJohnson + + 10.1109/dasc.2011.6096041 + + + 2011 IEEE/AIAA 30th Digital Avionics Systems Conference + + IEEE + 2011 + + + + 2011 IEEE/AIAA 30th + Johnson, C., "Analysis of Top of Descent (TOD) uncertainty," Digital Avionics Systems Conference (DASC), 2011 IEEE/AIAA 30th, IEEE, 2011, pp. 2E3-1. + + + + + Distributed Scheduling Architecture for Multi-Center Time-Based Metering + + StevenLandry + + + ToddFarley + + + JohnFoster + + + StevenGreen + + + TyHoang + + + GregoryWong + + 10.2514/6.2003-6758 + AIAA-2003-6758 + + + AIAA's 3rd Annual Aviation Technology, Integration, and Operations (ATIO) Forum + + American Institute of Aeronautics and Astronautics + Nov. 2003 + + + Landry, S., Farley, T., Foster, J., Green, S., Hoang, T., and Wong, G. L., "Distributed Schedul- ing Architecture for Multi-Center Time-Based Metering," Proceedings of the AIAA Aviation Technology, Integration, and Operations Conference, AIAA-2003-6758, Nov. 2003. + + + + + Human-in-the-Loop Simulation of Three-Dimensional Path Arrival Management with Trajectory Error + + GauravNagle + + + DouglasSweet + + + GregoryCarr + + + ValentinoFelipe + + + AndrewTrapani + + + RichardCoppenbarger + + + MiwaHayashi + + 10.2514/6.2011-6877 + AIAA-2011-6877 + + + 11th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Virginia Beach, VA + + American Institute of Aeronautics and Astronautics + 2011 + + + Nagle, G., Sweet, D., Carr, G., Felipe, V., Trapani, A., Coppenbarger, R., and Hayashi, M., "Human- in-the-Loop Simulation of the Efficient Descent Advisor for 3D Path Arrival Management," Proceedings of the AIAA Aircraft Technology, Integration, and Operations Conference, AIAA-2011-6877, Virginia Beach, VA, 2011. + + + + + Development and Testing of Automation for Efficient Arrivals in Constrained Airspace + + RCoppenbarger + + + GDyer + + + MHayashi + + + RLanier + + + LStell + + + DSweet + + ICAS2010-11.11.3 + + + Proceedings of the 27th International Congress of the Aeronautical Sciences + the 27th International Congress of the Aeronautical SciencesNice, France + + 2010 + + + Coppenbarger, R., Dyer, G., Hayashi, M., Lanier, R., Stell, L., and Sweet, D., "Development and Testing of Automation for Efficient Arrivals in Constrained Airspace," Proceedings of the 27th International Congress of the Aeronautical Sciences, ICAS2010-11.11.3, Nice, France, 2010. + + + + + Design and Development of the En Route Descent Advisor (EDA) for Conflict-Free Arrival Metering + + RichardCoppenbarger + + + RichardLanier + + + DougSweet + + + SusanDorsky + + 10.2514/6.2004-4875 + AIAA-2004-4875 + + + AIAA Guidance, Navigation, and Control Conference and Exhibit + + American Institute of Aeronautics and Astronautics + Aug. 2004 + + + Coppenbarger, R. A., Lanier, R., Sweet, D., and Dorsky, S., "Design and Development of the En Route Descent Advisor (EDA) for Conflict-Free Arrival Metering," Proceedings of the AIAA Guidance, Navigation, and Control Conference, AIAA-2004-4875, Aug. 2004. + + + + + Improved Aircraft Path Stretch Algorithms for the En Route Descent Advisor (EDA) + + TysenMueller + + + DavidSchleicher + + + RichardCoppenbarger + + 10.2514/6.2003-5571 + AIAA-2003-5571 + + + AIAA Guidance, Navigation, and Control Conference and Exhibit + + American Institute of Aeronautics and Astronautics + Aug. 2003 + + + Mueller, K. T., Schleicher, D. R., and Coppenbarger, R. A., "Improved Aircraft Path Stretch Al- gorithms for the En Route Descent Advisor (EDA)," Proceedings of the AIAA Guidance, Navigation, and Control Conference, AIAA-2003-5571, Aug. 2003. + + + + + En route Descent Advisor concept for arrival metering + + StevenGreen + + + RobertVivona + + 10.2514/6.2001-4114 + AIAA-2001-4114 + + + AIAA Guidance, Navigation, and Control Conference and Exhibit + + American Institute of Aeronautics and Astronautics + Aug. 2001 + + + Green, S. M. and Vivona, R. A., "En Route Descent Advisor Concept for Arrival Metering," Proceed- ings of the AIAA Guidance, Navigation, and Control Conference, AIAA-2001-4114, Aug. 2001. + + + + + Trajectory Prediction Accuracy and Error Sources for Regional Jet Descents + + JeffreyMHenderson + + + RobertAVivona + + + StevenGreen + + 10.2514/6.2013-4535 + AIAA- 2013-4535 + + + AIAA Guidance, Navigation, and Control (GNC) Conference + + American Institute of Aeronautics and Astronautics + 2013 + + + Henderson, J., Vivona, R. A., and Green, S. M., "Trajectory Prediction Accuracy and Error Sources for Regional Jet Descents," Proceedings of the AIAA Guidance, Navigation, and Control Conference, AIAA- 2013-4535, 2013. + + + + + Analysis of flight management system predictions of idle-thrust descents + + LaurelStell + + 10.1109/dasc.2010.5655506 + + + 29th Digital Avionics Systems Conference + + IEEE + Oct. 2010 + + + Stell, L., "Analysis of Flight Management System Predictions of Idle-thrust Descents," Proceedings of the IEEE/AIAA 29th Digital Avionics Systems Conference, Oct. 2010. + + + + + Continuous Descent Approaches with Variable Flight-Path Angles under Time Constraints + + RichardSopjes + + + PaulDe Jong + + + ClarkBorst + + + RenéVan Paassen + + + MaxMulder + + 10.2514/6.2011-6219 + AIAA-2011-6219 + + + AIAA Guidance, Navigation, and Control Conference + + American Institute of Aeronautics and Astronautics + Aug. 2011 + + + Sopjes, R., De Jong, P., Borst, C., Van Paassen, M., and Mulder, M., "Continuous Descent Approaches with Variable Flight-Path Angles under Time Constraints," Proceedings of the AIAA Guidance, Navigation, and Control Conference, AIAA-2011-6219, Aug. 2011. + + + + + An Hourly Assimilation–Forecast Cycle: The RUC + + StanleyGBenjamin + + + DezsöDévényi + + + StephenSWeygandt + + + KevinJBrundage + + + JohnMBrown + + + GeorgAGrell + + + DongsooKim + + + BarryESchwartz + + + TatianaGSmirnova + + + TracyLorraineSmith + + + GeoffreySManikin + + 10.1175/1520-0493(2004)132<0495:ahactr>2.0.co;2 + + + Monthly Weather Review + Mon. Wea. Rev. + 0027-0644 + 1520-0493 + + 132 + 2 + + 1998 + American Meteorological Society + + + Benjamin, S. G., Brown, J. M., Brundage, K. J., Schwartz, B., Smirnova, T., Smith, T. L., Morone, L. L., and Dimego, G., "The Operational RUC-2. Preprints," Proceedings of the 16th Conference on Weather Analysis and Forecasting, Amer. Meteor. Soc., 1998, pp. 249-252. + + + + + The Trajectory Synthesizer Generalized Profile Interface + + AlanLee + + + XavierBouyssounouse + + + JamesMurphy + + 10.2514/6.2010-9138 + AIAA-2010-9138 + + + 10th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + + American Institute of Aeronautics and Astronautics + Sept. 2010 + + + Lee, A. G., Bouyssounouse, X., and Murphy, J. R., "The Trajectory Synthesizer Generalized Profile Interface," Proceedings of the 10th AIAA Aviation Technology, Integration, and Operations Conference, AIAA-2010-9138, Sept. 2010. + + + + + Trajectory Synthesis for Air Traffic Automation + + RhondaSlattery + + + YiyuanZhao + + 10.2514/2.4056 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 20 + 2 + + March 1997 + American Institute of Aeronautics and Astronautics (AIAA) + + + Slattery, R. and Zhao, Y., "Trajectory Synthesis for Air Traffic Automation," Journal of Guidance, Control and Dynamics, Vol. 20, No. 2, March 1997, pp. 232-238. + + + + + Design of Center-TRACON Automation System + + HErzberger + + + TJDavis + + + SMGreen + + + + Machine Intelligence in Air Traffic Management, Andre Benoit + + Oct. 1993 + + + + AGARD CP-538 + Erzberger, H., Davis, T. J., and Green, S. M., "Design of Center-TRACON Automation System," Machine Intelligence in Air Traffic Management, Andre Benoit, ed., AGARD CP-538 , Oct. 1993, pp. 11- 1-11-12. + + + + + Advanced Aircraft Performance Modeling for ATM: Enhancements to the Bada Model + + ANuic + + + CPoinsot + + + M.-GIagaru + + + EGallo + + + FANavarro + + + CQuerejeta + + 10.1109/dasc.2005.1563320 + + + 24th Digital Avionics Systems Conference + + IEEE + Nov. 2005 + + + Nuic, A., Poinsot, C., Iagaru, M.-G., Gallo, E., Navarro, F. A., and Querejeta, C., "Advanced Aircraft Performance Modeling for ATM: Enhancements to the BADA Model," Proceedings of the IEEE/AIAA 24th Digital Avionics Systems Conference, Nov. 2005. + + + + + An Hourly Assimilation–Forecast Cycle: The RUC + + StanleyGBenjamin + + + DezsöDévényi + + + StephenSWeygandt + + + KevinJBrundage + + + JohnMBrown + + + GeorgAGrell + + + DongsooKim + + + BarryESchwartz + + + TatianaGSmirnova + + + TracyLorraineSmith + + + GeoffreySManikin + + 10.1175/1520-0493(2004)132<0495:ahactr>2.0.co;2 + + + Monthly Weather Review + Mon. Wea. Rev. + 0027-0644 + 1520-0493 + + 132 + 2 + + 2006 + American Meteorological Society + + + Benjamin, S. G., Devenyi, D., Weygandt, T. S. S. S., Brown, J. M., Peckham, S., Brundage, K. J., Smith, T. L., Grell, G. A., and Schlatter, T. W., "From THE 13-KM RUC To the Rapid Refresh," 12th Conference on Aviation Range and Aerospace Meteorology, 2006, pp. 9.1.1-9.1. + + + + + + diff --git a/file798.txt b/file798.txt new file mode 100644 index 0000000000000000000000000000000000000000..9e443303d75d8385eab83473f8826071932eaff1 --- /dev/null +++ b/file798.txt @@ -0,0 +1,498 @@ + + + + +I. IntroductionSensor uncertainties can impact the ability of a detect-and-avoid (DAA) system to maintain separation between an unmanned aircraft system (UAS) and manned aircraft.A DAA system computes alerts and maneuver guidance based on the surveillance data to help UAS pilots maintain "well clear" with manned aircraft.For DAA, the "well clear" is defined by a quantitative separation standard called DAA well clear (DWC).Large uncertainties in surveillance data may create flickering alerts, causing UAS pilots and operators to lose confidence in the system.Uncertainties in the surveillance data may cause the maneuver guidance to be ineffective.As a result, the UAS may violate DWC even after following the DAA guidance.Even if pilots are able to maintain DWC, the increased pilot workload and stress level can make the DAA system operationally unsuitable.Characteristics of sensor uncertainties vary greatly across sensor types.Three surveillance components were specified in the Minimum Operational Performance Standards (MOPS) for DAA systems [1] published in 2017 by RTCA.This MOPS requires Automatic Dependent Surveillance-Broadcast (ADS-B) In, airborne active surveillance, and an air-to-air radar.A companion MOPS for the air-to-air radar for DAA was also published [2].These two MOPS are referred to as the Phase 1 MOPS.Additional work to extend these MOPS, the Phase 2 work, is underway.ADS-B can detect intruder aircraft equipped with ADS-B out systems.Active surveillance can detect aircraft that have a functioning mode S or mode C transponder.Air-to-air radar in theory can detect all intruder aircraft.For non-cooperative aircraft, i.e., aircraft without a functioning transponder, the air-to-air radar is the only sensor that can detect it.Both ADS-B and active surveillance provide highly accurate vertical state (altitude and vertical speed) information of the intruder aircraft.Current state-of-the-art air-to-air radar, on the other hand, is not as accurate in its vertical state estimate.For horizontal states, both ADS-B and air-to-air radar provide highly accurate measurements for DAA purposes.The DAA alerting and guidance algorithm needs to be tested with various combinations of sensor measurements to make sure it can adapt to varying degrees of sensor uncertainties.If sensor measurement accuracy information is available, the algorithm can take advantage of such information and adapt its behavior accordingly.This paper analyzes the impact of airborne radar sensor uncertainties on a few safety and operational suitability metrics of the DAA system.The safety metrics include the loss of DAA well clear (LoDWC) ratio and the near-midair-collision (NMAC) risk ratio.The operational suitability metrics include the alert ratio and the number of DAA maneuvers per LoDWC.Minimum accuracy parameters from the radar MOPS [2] will be applied as a worst-case sensor noise.The analysis focus on encounters between low-speed unmanned aircraft (UA) and non-cooperative aircraft.More than 72,000 encounters are simulated in both open and closed-loop configurations.Close-loop simulations involve the application of a pilot response model to select and execute DAA maneuvers.This paper is organized as follows.Section II presents additional operational assumptions for this simulation.Section III describes the simulation approach and various components in the simulation architecture.Section IV shows results and discuss trends.Conclusion is made in Section V. +II. Background on Detect-and-AvoidA DAA system consists of surveillance components, datalink capabilities, an alerting and guidance algorithm, and a display to the UAS pilot.UAS operations considered in the Phase 1 MOPS are those transitioning to and from Class A or special use airspace (higher than 500 ft Above Ground Level (AGL)), traversing Class D, E, and G airspace.The Phase 1 MOPS assumes the UAS operations follow instrument flight rules and involve a pilot in the loop of decision making.UAS pilots are expected to contact air traffic control to negotiate a DAA maneuver if time permits.The DAA system issues guidance maneuvers to help the UAS maintain DWC with other aircraft.The DWC is expected to be reasonably larger than the NMAC, a safety hazard the FAA defines as a separation less than 500 ft horizontally and less than 100 ft vertically [3].This study focuses on the non-cooperative DWC, which is for encounters between a UAS and a non-cooperative aircraft, i.e., an aircraft without a functioning transponder.The non-cooperative DWC is a cylinder that has a radius of 2,200 ft and a height of 900 ft, 450 ft above the UAS and 450 ft below.For more information about the selection of this DWC, see [4].The Phase 1 MOPS defines three types of alerts in increasing levels of severity: preventive, corrective, and warning.The first time an alert is issued by DAA is typically 1 to 2 minutes before the predicted time of closest point of approach (CPA).The lowest level, preventive, is primarily used to alert the pilot not to maneuver vertically when the aircraft are separated vertically by 450 to 700 feet.This alert should not be triggered by non-cooperative aircraft and is not modeled.The second level, corrective, indicates that a LoDWC is predicted to occur in the future, an avoidance maneuver is necessary, but there is still time for coordination with air traffic control (ATC).The highest level, warning, indicates that a LoDWC is imminent, an immediate avoidance maneuver is needed, and coordination with ATC before maneuvering is not a requirement, although ATC should be contacted about the deviation as soon as possible.The DAA system should present maneuver guidance to UAS pilots as DWC-based conflict-ensuing aircraft heading ranges or altitude ranges to avoid.In case a LoDWC is unavoidable, the DAA system presents regain-DWC guidance, a range of heading or altitude that can be executed to increase separation at CPA and regain DWC effectively.While the required surveillance volume in the Phase 1 MOPS of the radar has been derived from adequate pilot response times, and the DWC, sensor accuracies were derived from rudimentary engineering analysis.Assessment of the sensor accuracy requirements conducted at the end of the Phase 1 MOPS development [5] did not explore sensitivity of the DAA performance metrics to the magnitude of sensor uncertainty or pilots' selection.All but the simplest sensor uncertainty mitigation schemes were attempted.Another systematic study investigated the effect of a new scheme on the DAA performance and showed benefit to the safety metrics [6].However, the encounters analyzed there were limited to a few hundreds and were not enough to provide statistical trends.This gap in research motivated the study in this paper. +III. ApproachClosed-loop simulations refer to the utilization of a pilot response model to drive the maneuver of the UA upon maneuver guidance of an alerting and guidance algorithm. +A. MetricsThe safety and operational suitability metrics evaluate whether a DAA system provides adequate and effective alerts and guidance.These metrics and their formulation are described below.1) LoDWC Ratio (safety):P (LoDWC | encounter, with mitigation) P (LoDWC | encounter, without mitigation)(1)The denominator of Eq. 1 stands for the probability of a LoDWC during an encounter if the UAS pilot does nothing to mitigate the conflict.The numerator stands for the probability of an LoDWC during an encounter if the UAS pilot follows the DAA system's maneuver guidance to mitigate the conflict.If the LoDWC risk ratio is less than one, then the DAA system reduces the risk of LoDWC.For example, a LoDWC ratio of 0.1 indicates a 90% reduction in LoDWC rates.Smaller values are desirable.To compute the LoDWC ratio, a set of representative encounters are evaluated with and without engaging a DAA system in mitigating conflicts.2) NMAC Risk Ratio (safety):P (NMAC | encounter, with mitigation)P (NMAC | encounter, without mitigation)(2)The denominator of Eq. 2 stands for the probability of an NMAC during an encounter if the UAS pilot does nothing to mitigate the conflict (unmitigated NMAC).The numerator stands for the probability of an NMAC during an encounter if the UAS pilot follows the DAA system's guidance to mitigate the conflict.If the ratio is less than one, then the DAA system reduces the risk of NMAC.3) Alert Ratio (operational suitability):P (alert | encounter, without mitigation) P (LoDWC | encounter, without mitigation)(3)This metric computes the number of encounters that trigger alerts per encounter that progresses to an unmitigated LoDWC.Ideally, only encounters that progress to an unmitigated LoDWC should trigger alerts.In this ideal case, the alert ratio is 1.Due to sensor uncertainties and aircraft maneuvers, alerts may also trigger for those encounters that do not progress to an unmitigated LoDWC.This tends to increase this ratio.On the other hand, limited surveillance volume tends to decrease this ratio. +B. EncountersThe safety and operational suitability metrics are computed from simulation of more than 72,000 encounters.A suitable encounter set should provide a predicted statistical representation of the conflicts a UAS will come across during its mission.Since the encounters considered are between a UA and non-cooperative aircraft, no coordination of conflict resolution is expected to occur.The following paragraphs describe how the encounters are prepared.An entire day's worth of projected UAS flights in the near future are considered for this study.These flights consist of 19 different types of missions, including aerial imaging and mapping, law enforcement, and air quality monitoring.The demand and mission profiles were generated based on subject matter experts' opinions and socio-economical analysis [7].These missions cover the entire continental US and amount to a total of 20,000 hours of flight time.Details of the missions are described in a previous publication [8].For the intruder traffic, nation-wide visual flight rules (VFR) flight paths recorded in 2012 were extracted and processed from the historical Air Force 84th Radar Evaluation Squadron (RADES) radar data.Due to the limited number of non-cooperative trajectories available, 1200-code cooperative VFR aircraft (also extracted from the RADES data) are used as a surrogate for non-cooperative aircraft for this study.This is a reasonable approach since the flight characteristics of conventional non-cooperative aircraft are similar to those using cooperative VFR aircraft in terms of airspeed, acceleration, and turn rate [9].Encounters are identified when UAS trajectories are overlaid with the VFR traffic.A software suite was developed to detect and produce encounters from these overlaid trajectories [10].Filters were applied to the ownship (the UA) and intruder speeds and altitudes during the generation of encounters to ensure the dynamics of the sampled trajectories is within the bounds for the UAs and the intruders.The UA must have a speed between 40 and 110 knots true airspeed (KTAS) for the encounter to be selected.The upper bound of 110 KTAS was selected so as to focus on low-altitude UAS missions that usually have lower mission speeds.This speed range fits the assumption of UAS operations with low size, weight, and power (SWaP) sensors.The requirements of low SWaP sensors are a key objective of the Phase 2 work at RTCA special committee 228 (SC-228).The intruder speeds range from 0 to 170 KTAS, the 95 percentile speed for non-cooperative intruders [11].Only encounters occurring at altitudes between 500 ft above ground level (AGL) and 10,999 ft mean sea level (MSL) in airspace classes E and G are selected.Although non-cooperative aircraft will be present only below 10,000 ft MSL due to the ADS-B-mandate in the year 2020, altitudes up to 10,999 ft MSL were included to represent a few UAS missions that are flown slightly above 10,000 ft MSL. Figure 2 shows the speed and altitude distributions of UAS and intruder by their values at the closest point of approach (CPA). +C. Navigation Errors, Sensor, and Tracker ModelsThe alerting and guidance algorithm takes aircraft states from the UA and the intruder as input for computation of alerts and guidance.For this study, the UA's truth track is perturbed by small magnitudes of sensor errors before it is sent to the DAA algorithm.The intruder's truth states are processed by the sensor models that will be described below.The Phase 1 DAA MOPS requires ADS-B, active surveillance, and an air-to-air radar as onboard sensors.Only the radar can detect non-cooperative aircraft and is the only sensor modeled.This study applies a vertically cylindrical radar field of view (FOV).The radius of the cylinder varies from 8 NM to 4, 3, and 2 NM.Analysis with this type of FOV sheds light on the sensitivity of the DAA performance metrics to a finite surveillance range, a critical question for UAS operations with low SWaP sensors.Table 1 shows the key parameters in the radar model for measurement accuracies.The representative values, denoted as SN, are the values achieved by an airborne radar during a previous flight test.The Phase 1 MOPS values, denoted as LSN, are the required accuracy by the Phase 1 radar MOPS (DO-366). +Table 1 Radar measurement accuracy parameters in terms of bias and error. +ConfigurationRange (m) Azimuth ( • ) Elevation ( • ) Velocity (m/s) (Representative) Sensor Noise (SN) 5.5 ± 10 0 ± 0.4 0 ± 0.4 0 ± 2 Large Sensor Noise (LSN) 15 ± 21 0.5 ± 1.0 0.5 ± 1.0 0 ± 2The Phase 1 MOPS requires a tracker that fuses and correlates measurements from multiple sensors for a single intruder into tracks.The UA's state is also inputted to the tracker to produce the intruder's state in absolute coordinates.A fusion tracker developed by Honeywell [12] is used for this work.This tracker output the intruder's track position and velocity as well as estimated accuracies of the track. +D. Alerting and Guidance AlgorithmThe Detect and AvoID Alerting Logic for Unmanned Systems (DAIDALUS) [13], a reference algorithm developed to validate the Phase 1 MOPS requirements, computes the alerts and guidance during the simulation of encounters.A sensor uncertainty mitigation (SUM) feature has been added to DAIDALUS recently [14].The effect of SUM on the DAA performance metrics is investigated in this study.Upon alerting, DAIDALUS generates corresponding preventive, corrective, and warning guidance indicating a range of conflict-free headings and altitudes for a pilot to select from in order to maintain DWC separation.In the absence of an alert, DAIDALUS computes peripheral guidance by projecting candidate vertical and horizontal DAA maneuvers into future times to determine which would result in conflicts.This information aids UAS pilots in their situation awareness.It is worth noting that, for this study, no wind was assumed for the computation of the guidance or the flight model.DAIDALUS alerts guard against a buffered DWC.The buffer is meant to protect the UAS from sensor uncertainties and maneuvering intruders.This study investigates both static and dynamic buffers.For runs with a static buffer, the horizontal separation distance used by the DAA alerting and guidance algorithm is scaled by a factor of 1.52 to be consistent with the parameters referenced in the Phase 1 MOPS.Runs with a dynamic buffer utilize SUM that computes the buffer from track accuracy data supplied by the tracker and a set of configurable scaling factors.Generally speaking, larger measurement errors and scaling factors increase the dynamic buffer and cause the algorithm to be more conservative, resulting in earlier alerts and wider conflict bands.Table 2 shows the scaling factors assigned for SUM in this study.These factors are selected from results of a companion study [15].Note the horizontal velocity scaling factor decreases linearly from 2.0 to 0.4 between 0 and 3 NM. +Table 2 DAIDALUS SUM scaling factorsName Horizontal Position Horizontal Velocities at 0 NM and 3 NM Vertical Position Vertical Velocity value 2.0 2.0 and 0.4 1.0 1.0In addition to the buffering scheme around the DWC, a wrapper of DAIDALUS ensures the processed alert sent to the pilot response model (see Section III.E) remains on for at least 4 consecutive seconds.This wrapper also ensures the processed alert comes on only after at least 2 of 4 consecutive raw alerts are on.These basic alert stability schemes are required by the DAA MOPS and hence applied to all run configurations. +E. Pilot Response ModelThe following discussion in this and the next sections apply only to closed-loop simulations.In open-loop simulations, alerts and guidance are recorded as the encounter progresses, but no DAA maneuvers are executed.Therefore, the pilot response model does not apply.Upon alerting, DAIDALUS provides maneuver guidance indicating a range of conflict-free headings and altitudes.The SC-228 standard pilot model created by MIT Lincoln Laboratory [16] selects and executes an appropriate maneuver (see Figure 3).Model parameters were derived from human-in-the-loop experiment results 3. The model has the capability of stochastic sampling of response times and maneuvers.For this study, the model is executed in deterministic mode, in which all response times and maneuvers are deterministic.The response times are set to the expectation value of a distribution, which is either a Gamma distribution (for the ATC coordination time) or an exponential distribution.For corrective and warning guidance, only horizontal maneuvers were executed because vertical maneuvers against non-cooperative intruders are much less robust in most situations due to uncertainties in non-cooperative sensors' vertical measurements.The UA always maneuvers in the direction of the minimum heading change suggested by the guidance bands.In the event that the minimum suggestion is inconclusive, the ownship will turn left, as a preference for left turns was observed in human-in-the-loop experiments [16].After the first alert comes up, the model exerts a 5-second initial delay representing the time it takes the pilot to perceive the alert and devise a plan.For corrective alerts, an additional 11 second ATC coordination time elapses, representing the time the pilot spends to communicate the intended maneuver with ATC and receive approval.The model then follows with a 3-second execution delay representing the time it takes the pilot to enter a maneuver command into the control station and transmit this command to the UAS.The ownship may perform multiple maneuvers per encounter to resolve a conflict.The response time between decisions is determined by the alert state, as shown in Table 3.For example, if the pilot model selects a maneuver during a warning alert state, then the situation will be reevaluated after 6 seconds (the decision update period), and a different subsequent maneuver can be issued at that time, if needed.Human-in-the-loop simulations and flight tests have shown that UAS pilots' selection of a DAA heading maneuver tends to have a buffer away from the edge of the conflict heading range.For example, if the DAA guidance indicates that a 30-degree or more right turn can resolve the conflict, UAS pilots are likely to turn the UA by 35 degrees or more.This extra buffer, which appears to be unnecessary at the time the guidance is presented to the pilot, may be beneficial in guarding against intruder maneuvers or sensor uncertainties.This study attempts both zero and 5 • buffers to identify potential benefits of this pilot-initiated buffer. +F. UA Flight ModelOnce the pilot response model selects a DAA maneuver and the execution time delay elapses, the flight model takes control and deviates the UA's trajectory from its nominal trajectory by executing the selected DAA maneuver.The intruder's trajectory remains unchanged.Once the target heading is achieved by the flight model, the UA trajectory remains at that heading.If the ownship remains conflict-free for 15 consecutive seconds after the last maneuver and is at least 3 NM away from the intruder, simulation of this encounter terminates. +IV. Results +A. Experiment MatrixTable 4 shows the experiment matrix for the closed-loop simulation runs.The matrix consists of two parts: the first part (top) shows runs with an 8 NM horizontal surveillance range.The second part (bottom) shows runs with shorter horizontal surveillance ranges.In addition to these runs, a separate open-loop simulation, using a static buffer, as performed in order to provide the LoDWC and NMAC counts for the denominators of the metrics.The first part has a total of 10 runs.The two no-sensor-noise (NSN) runs, one with a turn buffer and one without, represents the baseline configuration.A static buffer around the DWC is applied to these two runs.The SN and LSN runs investigate the impact of the magnitude of sensor noise on DAA performance metrics.For SN and LSN configurations, The second part of the experiment matrix has three more runs with 4, 3, and 2 NM horizontal surveillance ranges.The sensor noise level is SN, and SUM is turned on for these runs.Results from these runs give insight to the sensitivity of the performance metrics to a limited surveillance range.Section IV.B presents results for the first part of the experiment matrix.Section IV.C presents results for the second part of the experiment matrix. +B. Surveillance Range of 8 NMFigure 4 shows the LoDWC ratio from various runs.The baseline run, shown on the left, as expected, achieves the lowest LoDWC ratio of 7.3%.The SN without SUM run (meaning a static buffer around the DWC) doubles the LoDWC ratio.The SN with SUM run is effective in reducing the LoDWC ratio to below 10%.The LSN run without SUM results in the highest LoDWC ratio of all.The LSN with SUM run effectively reduces the ratio to 11.2%.The 5 • turn buffer appears to provide limited benefit to the LoDWC ratio. +Fig. 6 Alert ratioFigure 6 shows the alert ratio for the five combinations of sensor noise and SUM configurations.Note that the turn buffer does not change this metric.Therefore, results from runs with a turn buffer are not shown.Compared to the baseline run on the leftmost, SN increases the alert ratio.SUM (middle bar) also increases the alert ratio by triggering alerts in more encounters.This shows the trade space between the safety metric such as the LoDWC ratio and an operational suitability metric such as the alert ratio.While the dynamic buffer improves the safety metric, it appears to create more nuisance alerts.This trade space is likely to be acceptable if such encounters occur with a low frequency.Interestingly, increase the sensor noise to LSN only increases the alert ratio slightly (<5%.)Figure 7 shows the number of maneuvers per LoDWC.Results without a turn buffer, shown in blue, are discussed first.The baseline run, NSN, indicates more than 4 maneuvers per LoDWC.SN increases this metric only marginally.SUM increases this metric from the sensor noise run (middle bar) by more than 20%.LSN yield similar results to the SN runs.With a turn buffer, the baseline run is much improved by 25%.With SUM on, the benefit of the turn buffer appears to diminish, reducing to 10% for the SN and LSN runs. +C. Surveillance Range Less than 8 NMFor the runs with limited surveillance range, only the safety metrics are presented.The operational suitability metrics are similar across runs and thus omitted.Figure 8a shows that the LoDWC ratio essentially remains the same as the radar's surveillance range reduces from 8 NM to 3 NM.Reducing the surveillance range to 2 NM increases the LoDWC ratio from 10% to 12%.This is consistent with results from a prior study [4] that performed closed-loop simulations without sensor noise.Results from that study showed lower LoDWC ratios that remain the same (8%) when the surveillance range is 3 NM and above and increases when the range reduces to 2 NM (10%.)Figure 8b shows a similar trend for the NMAC risk ratio, which remains the same at and above 3 NM.The NMAC risk ratio appears to be slightly higher at 2 NM.However, the large errors make these values statistically indistinguishable.Similar results were observed in the prior study [4], although the NMAC risk ratio was smaller (2.0%) mainly due to the lack of sensor noise. +V. ConclusionThis study analyzes the impact of sensor noise on the DAA system's performance using a few key safety and operational suitability metrics.More than 72,000 encounters, representative of UAS against non-cooperative manned aircraft, were analyzed using a reference DAA alerting and guidance algorithm for the DAA MOPS.Results show that, while sensor noise degrades safety metrics, suitably chosen buffers around the DWC, especially those computed by DAIDALUS SUM, can mitigate most of its impact.Larger buffers around the DWC tend to worsen operational suitability metrics such as the number of maneuvers per loss of DAA well clear.A 5 • buffer away from the heading range leading to a conflict selected for the pilot response model appears to reduce the pilot maneuver count.Results from this work will inform RTCA SC-228 directly in terms of validating sensor uncertainty requirements as well as selecting default parameters for DAIDALUS SUM.Trends observed from results of various configurations indicate trade space between safety and operational suitability, and can help manufacturers of DAA systems balance design and runtime parameters.The results may also provide supporting data to the FAA's system safety assessment.The trade space demonstrated by this study is likely to behave differently with higher UA speeds or a different type of sensor.For example, UAs flying with higher speeds need to start a maneuver away at a greater distance but with a lower turn rate.Also, the electro-optical/infrared sensor has very different error characteristics than an air-to-air radar.Additional work to expand this study to these areas is being planned.Figure 1 Fig. 111Fig. 1 Simulation architecture +Fig. 22Fig. 2 Speed and altitude distributions of UAS and VFR traffic. +Fig. 33Fig. 3 The SC-228 pilot response model +Fig. 55Fig. 5 NMAC risk ratios +Fig. 77Fig. 7 Maneuver count per LoDWC +Fig. 88Fig. 8 LoDWC ratio and NMAC risk ratio varying with surveillance range +This metric computes the average number of DAA maneuvers per unmitigated LoDWC.Lower values are desirable, since fewer maneuvers correspond to lower pilot workload.Note encounters that do not lead to unmitigated LoDWC also contribute to this metric if they trigger alerts that eventually lead to DAA maneuvers.Note that if on-and-off alerts are likely to lead to more DAA maneuvers.Unstable DAA guidance is also likely to lead to more DAA maneuvers.4) Number of Maneuvers per LoDWC (operational suitability):Total # of DAA Maneuvers Total # of LoDWCs(4) +Table 3 Pilot response model decision update times3Alert ConditionDecision Update Period (s)No Alert12Preventive Alert9Corrective Alert6Warning Alert6Regain-DWC Guidance0 +Table 4 Mitigated simulation test matrix. Configurations included in the experiment matrix are marked with a circle.4• pilot turn buffer from the edge of the heading conflict bands, 2) the dynamic buffer computed by DAIDALUS SUM.Surveillance Range = 8 NM +Figure5shows the NMAC risk ratio from various runs.Due to the large error estimates, no definitive trends can be ascertained except that the No SN runs result in lower NMAC risk ratios.0.070.060.05NMAC Risk Ratio0.03 0.040.020.010.00No SNSN without SUMSN with SUMLSN without SUM LSN with SUMWithout Turn BufferWith Turn Buffer0.250.20LoDWC Ratio0.10 0.150.050.00No SNSN without SUMSN with SUMLSN without SUM LSN with SUMWithout Turn BufferWith Turn BufferFig. 4 LoDWC ratios + + + + +AcknowledgmentsThe authors wish to thank Eric L. Wahl, Michael Abramson, and Mohamad Refai for their software support and feedback for this study. + + + + + + + + + Minimum Operational Performance Standard (MOPS) for Helicopter Hoist Systems + 10.4271/as6342 + + 2017 + SAE International + + + Minimum Operational Performance Standards (MOPS) for Detect and Avoid (DAA) Systems, DO-365, RTCA. Inc., 2017. + + + + + AeroMACS minimum operational performance standards (MOPS) compliance field trials for Hitachi prototype + + RafaelApaza + + 10.1109/icnsurv.2015.7121331 + + + 2015 Integrated Communication, Navigation and Surveillance Conference (ICNS) + + IEEE + 2017 + + + Minimum Operational Performance Standards (MOPS) for Air-to-Air Radar for Traffic Surveillance, RTCA. Inc., 2017. + + + + + Fixed-wing group petitions FAA for advisory circular + 10.1016/s0740-8315(88)80079-2 + + + + Hospital Aviation + Hospital Aviation + 0740-8315 + + 7 + 7 + + 2011 + Elsevier BV + + + 1%20Intro%20booklet.pdf + "Introduction to TCAS II Version 7.1," Federal Aviation Administration (FAA), 2011. URL http://www.faa.gov/ documentLibrary/media/Advisory_Circular/TCAS%20II%20V7.1%20Intro%20booklet.pdf. + + + + + Defining Well Clear Separation for Unmanned Aircraft Systems Operating with Noncooperative Aircraft + + ChristineChen + + + MatthewWEdwards + + + BilalGill + + + SamanthaSmearcheck + + + TonyAdami + + + SeanCalhoun + + + MinghongGWu + + + AndrewCone + + + SeungManLee + + 10.2514/6.2019-3512 + + + AIAA Aviation 2019 Forum + + American Institute of Aeronautics and Astronautics + 2019 Forum, 2019 + + + Chen, C., Edwards, M. W., Gill, B., Smearcheck, S., Adami, T., Calhoun, S., Wu, M. G., Cone, A., and Lee, S. M., "Defining Well Clear Separation for Unmanned Aircraft Systems Operating with Noncooperative Aircraft," AIAA Aviation 2019 Forum, 2019. + + + + + SC-228 DAA Phase 1 MOPS Verification Validation Simulation + + MHall + + + JKaznocha + + + ELester + + + SSherman + + + NReep + + MP170272 + + 2017 + Bedford, MA + + + MITRE Corporation + + + MITRE Product + Hall, M., Kaznocha, J., Lester, E., Sherman, S., and Reep, N., "SC-228 DAA Phase 1 MOPS Verification Validation Simulation," Tech. rep., MITRE Corporation, Bedford, MA, 2017. MITRE Product MP170272. + + + + + Unmanned Aircraft Systems Detect and Avoid System: End-to-End Verification and Validation Simulation Study of Minimum Operational Performance Standards for Integrating Unmanned Aircraft into the National Airspace System + + RaniaGhatas + + + DevinPJack + + + DimitriosTsakpinis + + + JamesLSturdy + + + MichaelJVincent + + + KeithDHoffler + + + RobertMyer + + + AnnaDehaven + + 10.2514/6.2017-3278 + + + 17th AIAA Aviation Technology, Integration, and Operations Conference + Kissimmee, FL + + American Institute of Aeronautics and Astronautics + 2018 + + + Ghatas, R. W., Tsakpinis, D., Sturdy, J. L., Jack, D. P., Myer, R. R., and DeHaven, A. M., "The Effects of Severity of Losses of Well Clear on Minimum Operations Performance Standards End-to-End Verification and Validation Simulation Study for Integrating Unmanned Aircraft Systems into the National Airspace System using Detect and Avoid Systems," Proceedings of the 2018 AIAA Information Systems-AIAA Infotech @ Aerospace, AIAA, Kissimmee, FL, 2018. + + + + + UAS Demand Generation Using Subject Matter Expert Interviews and Socio-economic Analysis + + SricharanKAyyalasomayajula + + + RohitSharma + + + FrederickWieland + + + AntonioTrani + + + NicolasHinze + + + ThomasSpencer + + 10.2514/6.2015-3405 + + + 15th AIAA Aviation Technology, Integration, and Operations Conference + + American Institute of Aeronautics and Astronautics + 2015 + + + Ayyalasomayajula, S., Sharma, R., Wieland, F., Trani, A., Hinze, N., and Spencer, S., "UAS Demand Generation Using Subject Matter Expert Interviews and Socio-Economic Analysis," Proceedings of the AIAA Aviation Conference, 2015. + + + + + Well Clear Trade Study for Unmanned Aircraft System Detect And Avoid with Non-Cooperative Aircraft + + MinghongGWu + + + AndrewCCone + + + SeungmanLee + + + ChristineChen + + + MatthewWEdwards + + + DevinPJack + + 10.2514/6.2018-2876 + + + 2018 Aviation Technology, Integration, and Operations Conference + + American Institute of Aeronautics and Astronautics + 2018 + + + Wu, M. G., Cone, A. C., Lee, S., Chen, C., Edwards, E. W. M., and Jack, D. P., "Well Clear Trade Study for Unmanned Aircraft System Detect And Avoid with Non-Cooperative Aircraft," 18th AIAA Aviation Technology, Integration, and Operations Conference, 2018. + + + + + Extended Airspace Encounter Models for Unmanned Aircraft Sense and Avoid Safety Evaluation + + AndrewJWeinert + + + EricHarkleroad + + + JohnGriffith + + + MatthewWEdwards + + 10.2514/6.2013-5049 + + + + AIAA Infotech@Aerospace (I@A) Conference + Lexington, Massachusetts + + American Institute of Aeronautics and Astronautics + Aug. 2013 + + + MIT Lincoln Laboratory + + + Tech. rep. + Weinert, A. J., Harkleroad, E. P., Griffith, J., Edwards, M. W., and Kochenderfer, M. J., "Uncorrelated Encounter Model of the National Airspace System, Version 2.0," Tech. rep., MIT Lincoln Laboratory, Lexington, Massachusetts, Aug. 2013. URL www.dtic.mil/cgi-bin/GetTRDoc?Location=U2&doc=GetTRDoc.pdf&AD=ADA589697. + + + + + Encounter-Based Simulation Architecture for Detect-And-Avoid Modeling + + MAbramson + + + MRefai + + + SLee + + + MGWu + + + + AIAA Science and Technology Forum and Exposition + + 2019. 2019 + + + Abramson, M., Refai, M., Lee, S., and Wu, M. G., "Encounter-Based Simulation Architecture for Detect-And-Avoid Modeling," AIAA Science and Technology Forum and Exposition 2019, 2019. + + + + + A Bayesian Approach to Aircraft Encounter Modeling + + MykelKochenderfer + + + JamesKuchar + + + JDGriffith + + + LeoEspindle + + 10.2514/6.2008-6629 + + + AIAA Guidance, Navigation and Control Conference and Exhibit + + American Institute of Aeronautics and Astronautics + 2008 + 17 + + + Kochenderfer, M. J., Espindle, L. P., Kuchar, J. K., and Griffith, J. D., "A Comprehensive Aircraft Encounter Model of the National Airspace System," Lincoln Laboratory Journal, Vol. 17, No. 2, 2008. + + + + + Multi-intruder aircraft, multi-sensor tracking system + + VibhorLBageshwar + + + EricAEuteneuer + + 10.1109/dasc.2015.7311425 + + + 2015 IEEE/AIAA 34th Digital Avionics Systems Conference (DASC) + + IEEE + 2015 + + + + Bageshwar, V. L., and Euteneuer, E. A., "Multi-Intruder Aircraft, Multi-Sensor Tracking System," 34th Digital Avionics Systems Conference (DASC), IEEE/AIAA, 2015, pp. 5A5-1-13. + + + + + DAIDALUS: Detect and Avoid Alerting Logic for Unmanned Systems + + CesarMunoz + + + AnthonyNarkawicz + + + GeorgeHagen + + + JasonUpchurch + + + AaronDutle + + + MariaConsiglio + + + JamesChamberlain + + 10.1109/dasc.2015.7311421 + + + 2015 IEEE/AIAA 34th Digital Avionics Systems Conference (DASC) + + IEEE + 2015 + + + + Muñoz, C., Narkawicz, A., Hagen, G., Upchurch, J., Dutle, A., Consiglio, M., and Chamberlain, J., "DAIDALUS: Detect and Avoid Alerting Logic for Unmanned Systems," 34th Digital Avionics Systems Conference (DASC), IEEE/AIAA, 2015, pp. 5A1-1. + + + + + Sensor Uncertainty Mitigation and Dynamic Well Clear Volumes in DAIDALUS + + AnthonyNarkawicz + + + CesarMunoz + + + AaronDutle + + 10.1109/dasc.2018.8569468 + + + 2018 IEEE/AIAA 37th Digital Avionics Systems Conference (DASC) + + IEEE + 2018 + + + Narkawicz, A., Muñoz, C., and Dutle, A., "Sensor Uncertainty Mitigation and Dynamic Well Clear Volumes in DAIDALUS," Proceedings of the 37th Digital Avionics Systems Conference, IEEE, 2018. + + + + + Applying Sensor Uncertainty Mitigation Schemes to Detect-and-Avoid Systems + + MichaelAbramson + + + MohamadRefai + + + MinghongGWu + + + SeungmanLee + + 10.2514/6.2020-3264 + + + AIAA AVIATION 2020 FORUM + + American Institute of Aeronautics and Astronautics + 2020 + + + To appear + Abramson, M., Refai, M., Wu, M. G., and Lee, S., "Applying Sensor Uncertainty Mitigation to Detect-and-Avoid Systems," AIAA Aviation 2020 Forum, 2020. To appear. + + + + + A Model of Unmanned Aircraft Pilot Detect and Avoid Maneuver Decisions + + RGuendel + + + MKuffner + + + DMaki + + + 2017 + Lexington, Massachusetts + + + MIT Lincoln Laboratory + + + Tech. rep + Guendel, R., Kuffner, M., and Maki, D., "A Model of Unmanned Aircraft Pilot Detect and Avoid Maneuver Decisions," Tech. rep., MIT Lincoln Laboratory, Lexington, Massachusetts, 2017. + + + + + + diff --git a/file799.txt b/file799.txt new file mode 100644 index 0000000000000000000000000000000000000000..835ea98d2e5a2d0376a486adc005d1e8c1371e91 --- /dev/null +++ b/file799.txt @@ -0,0 +1,871 @@ + + + + +I. IntroductionSuccessful integration of Unmanned Aircraft System (UAS) operations into airspaces populated with manned aircraft relies on an effective Detect and Avoid (DAA) System.A DAA system provides surveillance, alerts, and maneuver guidance to keep a UAS "well clear" of other aircraft [1,2].Despite significant advances in aircraft, surveillance, and communication technologies in the past three decades, lack of research-backed specific requirements for DAA systems continues to hinder the integration of UAS into the national airspace systems (NAS).In recent years, government, industry, and academia started working together on addressing this gap.In the United States, simulations as well as flight tests have provided supporting information for defining a DAA well clear (DWC) [1,3] and requirements for the alerting and maneuver guidance performance [4][5][6][7][8].Prototype alerting and maneuver guidance (referred to as guidance in this paper) algorithms have also been developed [9][10][11].These developments, along with other technical work, enabled the RTCA Special Committee 228 (SC-228) to publish the Minimum Operational Performance Standards (MOPS) for DAA systems [12] and air-to-air radar [13] in 2017.The corresponding Technical Standard Orders (TSO), TSO-C211 and TSO-C212, were published by the Federal Aviation Administration (FAA) in October 2017.These standards, referred to as the Phase 1 MOPS, target UAS operations transitioning to and from Class A or special use airspace (higher than 500 ft above ground level (AGL)), traversing Class D, E, and G airspace.The Phase 1 MOPS assumes the UAS operations follow instrument flight rules, involve a pilot in the decision loop, and fly with a speed between 40 and 200 KTAS.A Phase 1 MOPS compliant DAA system must include the following surveillance components: Automatic Dependent Surveillance-Broadcast (ADS-B) In, airborne active surveillance, and an air-to-air radar.Traffic Alert and Collision Avoidance System (TCAS) II [14] is an optional equipment.Phase 2 development for extending the MOPS to additional UAS categories and operations is underway.Ground-based surveillance technology [15] and DAA concept [16] developed in recent years is also to be included in the Phase 2 development.One objective of the Phase 2 development is to define an alternative DAA Well Clear (DWC) for UAS encountering non-cooperative aircraft, i.e., aircraft without a functioning transponder.Non-cooperative aircraft are estimated to comprise a small but non-negligible 15% of the flights in the airspaces outside the ADS-B mandate rule airspace [12].While aircraft with a functioning transponder can be detected by ADS-B and/or active surveillance, non-cooperative aircraft can only be detected by the air-to-air radar of the DAA surveillance.If the UAS is also equipped with TCAS II, both DAA and TCAS II alerts and guidance can be active during an encounter.The DWC in the Phase 1 MOPS was selected with considerations of interoperability of the DAA system with TCAS-II.DAA alerts have longer look-ahead time and should trigger before TCAS II alerts do.DAA guidance, when executed, should ideally be able to avoid TCAS II alerts in most situations.With this consideration, the DWC was defined to encompass a large portion of the TCAS II resolution advisories (RA) alerting volume.The resulting DWC is deemed very safe but is unnecessarily large for encounters of UAS with non-cooperative aircraft, which TCAS-II cannot detect and therefore need not be considered.In a previous study, four smaller candidate DWCs, specifically for non-cooperative aircraft, were proposed for additional evaluation [17].Selection of these candidate DWCs was based on unmitigated collision rates, maneuver initiation ranges, and a few other factors.Another objective of the Phase 2 development is to define requirements for operations of UAS equipped with low size, weight, and power (SWaP) sensors, or low SWaP UAS.While ADS-B and active surveillance can fit in the payload of many medium-sized UAS, the large, high-power radar required by the Phase 1 MOPS is physically infeasible and/or economically impractical for many UAS operations.Low SWaP sensors have favorable payloads but provide smaller surveillance volumes.For safety and operational suitability, UAS pilots need sufficient alerting times to evaluate and execute DAA maneuvers in order to maintain separation defined by the DWC.Compared to the Phase 1 radar surveillance volume requirements, two factors can potentially drive down the demand of the low SWaP sensors' surveillance volume, particularly its horizontal range.First, many UAS operations with low SWaP sensors are expected to be conducted at speeds much lower than 200 KTAS.Taking into account these lower mission speeds can reduce the demand of the low SWaP sensors' horizontal range.Second, an alternative DWC smaller than the Phase 1 definition gives UAS pilots more time from the maximum surveillance range to the DWC boundary to evaluate the situation and maneuver away from an intruder.Since UAS pilots do not need more time than what is already allocated in the set of Phase 1 requirement, this reduction of DWC will "free up" part of the originally required surveillance range.Still another objective of the Phase 2 development seeks to increase the upper bound of the unmanned aircraft (UA) speed assumed in the Phase 1 MOPS from 200 KTAS to 250 KTAS or higher.Note the term UA instead of UAS is used when referring to specifically the vehicle's performance parameters such as speed or turn rate.The ability of a DAA system to maintain separation effectively for such high-speed UAs during encounters needs to be investigated.The adequacy of the Phase 1 radar requirements for high-speed UAs should be assessed as well.This paper evaluates four candidate DAA Well Clear (DWC) definitions [17] by computing their safety and operational suitability metrics from closed-loop simulations.Execution of DAA maneuvers makes the simulations closed-loop.One million encounters covering a wide range of UA speeds are simulated.DAA maneuvers are computed using a reference alerting and guidance algorithm, and executed according to a research-backed pilot response model.Effects of finite surveillance volume as well as UA speeds on the metrics are also investigated.The main contributions of this paper are the following:1) Development of a simulation framework for the evaluation of DWCs 2) Providing safety and operational suitability metrics for informing RTCA SC-228 and the community in areas of the aforementioned three key objectives: 1. non-cooperative DWC, 2. low SWaP sensor requirements, and +high-speed UAS operationsThis paper is organized as follows: Section II presents additional background about the development of the DWC.Section II also reviews the considerations taken to propose the candidate DWCs.Section III describes the metrics, simulation architecture, and details of each simulation component.Results are presented in Section IV, additional discussion included in Section V, and conclusions are given in Section VI. +II. Background on Detect-and-Avoid Well ClearThe need for a DWC for UAS operations derives from the requirement for manned aircraft pilots stated under Title 14 of the Code of Federal Regulations (14 CFR).The intent of these regulations is to avoid collisions, remain well clear from other aircraft, and comply with right-of-way rules.For UAS, the pilots are not in the flight deck to "see and avoid," so pilots must rely on surveillance and algorithms for situation awareness and conflict avoidance.In this situation, the separation standard, or the DWC, must have a quantitative definition.The DWC is expected to be larger than the near-mid-air-collision (NMAC), a safety standard for the evaluation of collision avoidance systems.Two aircraft are in an NMAC if they are separated less than 500 ft horizontally and less than 100 ft vertically [14].The development of the DWC in the Phase 1 MOPS started in the Sense-and-Avoid Science and Research Panel (SaRP).The SaRP-recommended DWC was later modified by SC-228 based on recommendation from the FAA to improve its operational suitability [1,3].Note this DWC has been analyzed and researched only within the operational environment defined by the Phase 1 MOPS.It is likely to be also applicable to extended operations to be defined in the Phase 2 MOPS.For operational environments not covered by these MOPS, alternative DWCs may be more suitable [18].The DWC in the Phase 1 MOPS is defined by thresholds of three parameters.It does not have distinct physical boundaries because its definition depends on two aircraft's positions and velocities.Figure 1 The time parameter τ mod was introduced for the purpose of interoperability with TCAS II, because a similar time parameter for alerting is adopted by TCAS II.The definition of τ mod is [2]τ mod =              -r 2 -D mod 2 r r , r > D mod , 0, r ≤ D mod (1)where r and r are the horizontal range and range rate between the intruding aircraft and the UA, respectively.The range rate represents the rate of change of the two aircraft's horizontal distance, and is negative for closing geometries.The positive incremental distance modifier D mod defines the radius of a "protection" disk around the unmanned aircraft such that any manned aircraft close by, namely an intruder, with a horizontal range less than D mod is always considered "urgent."In the case of r < D mod , τ mod = 0.The value of D mod must be equal to HMD * to avoid undesirable on-and-off alerts during certain constant velocity encounters [19].The value of τ mod provides an estimate of the time for the UA to penetrate the protection disk defined by HMD * .The DWC in the Phase 1 MOPS was initially selected from three types of DWC definitions using eight performance metrics [1].The unmitigated collision risk, denoted as P, was used to tune the DWC threshold parameters such that all three DWCs yielded the same value of P [1].P represents the conditional probability of an NMAC given a LoDWC without executing DAA maneuvers:P = P(NMAC|LoDWC).(2)Values of P are usually computed from an encounter set representative of the UAS operations and intruder types considered.The target value of P was reduced from the recommended 5% [20] to 1.5% initially so as to expand the DWC volume to enclose most of the TCAS II RA alerting volume [1].The selected DWC, however, had an operationally unsuitable vertical separation threshold of h * = 700 ft, above the vertical separation of 500 ft permitted by visual flight rules (VFR) [3].The vertical separation threshold h * was hence reduced to 450 ft, and the final DWC resulted in a P of 2.2%.The Phase 1 DWC is expected to continue to be applied to cooperative aircraft in the Phase 2 MOPS development work.Previous work [17] proposed four candidate DWCs for non-cooperative aircraft.The candidates were selected based on P, maneuver initiation range (MIR), and a few other factors.P was evaluated using encounters representative of UAS against non-cooperative intruders (see [17] for more details).The UA speed for computing P and MIR was between 40 and 100 KTAS.MIR is the minimum horizontal range during a stressing case, head-on encounter at which a non-accelerating UA must start maneuvering away from a non-accelerating intruder to maintain separation defined by the DWC [21].Lower MIRs are preferred as a potential mean of reducing surveillance volume requirements.MIR is a function of both aircraft's airspeeds, encounter geometry, and the UA's maximum turn, climb, and descent rates.If the ranges of UAS and intruder airspeeds are specified, MIR is selected to be the value computed for the most stressing case, i.e., a combination of airspeeds that yield the highest MIR.Both P and MIR were computed based on assumptions for low SWaP UAS operations.The MIR computation considers the non-cooperative intruder to fly at a stressing-case speed of 170 KTAS (95 percentile [22]).The UA turns at a rate of 7 deg/sec, transitioning from a constant heading to aturn with a roll rate of 5 deg/sec, during a DAA maneuver.The four candidate DWCs' properties are given in Table 1.They are defined by thresholds of the same three parameters that define the Phase 1 DWC.Note that alternative forms of the DWC were explored in [17] but were not selected due to worse performance.These candidate DWCs vary in their HMD * and τ mod * values.The parameter D mod is set to be equal to the value of HMD * and therefore is not an independent parameter.Only one altitude separation threshold h * of 450 ft as that of the Phase 1 DWC was considered because 1) h * cannot exceed the legal separation of 500 ft for VFR flights and 2) decreasing h * from 450 ft is regarded as increasing collision risk unnecessarily.DWC1 and DWC2 are the two primary candidates because they both achieve a desirable P of 5%, a recommended value from a previous study [20].While various combinations of τ mod * and HMD * can achieve this value of P, DWC1 is selected for the minimum MIR it results in among these combinations.One the other hand, DWC2 is also selected for its simple form, representing a physical cylinder and not having a time component.DWC3 and DWC4 are backup candidates to be carried forward in case additional analyses reveal unfavorable metrics for DWC1 and DWC2.Their parameters were selected with certain level of subjectivity.DWC3 achieves a higher unmitigated collision risk of 7%.It was once proposed for terminal area UAS operations [23].DWC4 achieves an unmitigated collision risk of 3.6% and was considered a safer candidate.In addition to the four DWC candidates selected in [17], the Phase 1 DWC definition is also evaluated for the purpose of comparison. +III. ApproachThe four candidate DWCs are evaluated using safety and operational suitability metrics computed from results of open-loop and closed-loop simulations of a DAA system resolving conflicts arising from a large number of aircraft encounters.Closed-loop simulations refer to the utilization of a pilot response model to select and execute the maneuver of the UA, following the guidance from an alerting and guidance algorithm.The concept of UAS operations considered by RTCA SC-228 involves a UAS pilot, as well as air traffic control (ATC) if available, in the decision loop.An automated response is not considered.The following sections discuss the metrics, the simulation architecture, and each component in the simulation architecture in detail. +A. MetricsThe safety and operational suitability metrics evaluate a DAA system's ability to provide timely and effective alerts and guidance.Among the eight performance metrics considered for the Phase 1 DWC [1], the selection process for the four candidate non-cooperative DWCs already considered the MIR.Three of the criteria, mitigated risk ratio, controller acceptability, and well clear volume collision rate, are represented by precise or highly correlated metrics in this work.The closest point of approach (CPA) miss distance and cross track deviation were investigated but not presented due to the size limit of this paper.The remaining two criteria, TCAS II RA rate and vertical deviation, are deemed irrelevant by the authors to the consideration of non-cooperative DWCs.The metrics presented in this paper and their formulations are described below.1) NMAC Risk Ratio (Safety): P (NMAC | encounter, with mitigation) P (NMAC | encounter, without mitigation)The NMAC risk ratio indicates the system's effectiveness in reducing the occurrence of collision hazards and is a key safety metric for risk assessment.Small values are desirable.The denominator of Eq. 3 represents the probability of an NMAC during an encounter if the UAS pilot does nothing to mitigate the conflict.This is the probability of an unmitigated or nominal NMAC.The numerator represents the probability of an NMAC during an encounter if the UAS pilot follows the DAA system's guidance to mitigate the conflict, and is the probability of a mitigated NMAC.If the ratio is less than one, then the DAA system reduces the risk of NMAC.For example, a risk ratio of 0.1 indicates a 90% reduction in risk.To compute the NMAC risk ratio, a set of representative encounters are evaluated with (closed-loop) and without (open-loop) engaging a DAA system in mitigating conflicts.Each NMAC resulting from evaluation with a DAA system for conflict mitigation is put in one of two categories.Unresolved NMAC risk is comprised of encounters that lead to nominal NMACs (i.e., without a DAA system) and which still have NMACs with the DAA system.Induced NMAC is comprised of encounters that do not have nominal NMACs but develop into NMACs with the DAA maneuver in response to DAA guidance.2) LoDWC Ratio (Safety): P (LoDWC | encounter, with mitigation) P (LoDWC | encounter, without mitigation) (The denominator of Eq. 4 represents the probability of an LoDWC during an encounter if the UAS pilot does nothing to mitigate the conflict.The numerator stands for the probability of an LoDWC during an encounter if the UAS pilot follows the DAA system's guidance to mitigate the conflict.Similar to the NMAC risk ratio, if the LoDWC risk ratio is less than one, then the DAA system reduces the risk of LoDWC.Small values are desirable.Unresolved and induced LoDWCs are defined in a similar way to how unresolved and induced NMACs are defined.3) Alert Ratio (Operational Suitability):P (alert | encounter, with mitigation) P (NMAC | encounter, without mitigation) (5) This metric computes the number of encounters that issue actionable alerts (alerts that require pilots to start coordinating a DAA maneuver) per encounter that leads to an unmitigated NMAC.Lower alert ratios are desirable, since fewer alerts indicate fewer unnecessary maneuvers. +C. EncountersThe safety and operational suitability metrics are computed from simulations of one million encounters.Ideally, a suitable encounter set should provide a predicted statistical representation of the conflicts a UAS will come across during its mission.Encounters for this study are generated by sampling a UAS trajectory segment randomly from NASA's projected UAS mission flights [24] and creating around it an intruder trajectory from MIT Lincoln Laboratory's uncorrelated encounter model [25].The uncorrelated encounter model is appropriate for this purpose because no coordination of conflict resolution is expected to occur between the UAS and the non-cooperative aircraft. +D. Navigation Errors, Sensor, and Tracker ModelsThe alerting and guidance algorithm takes aircraft states from the UA and the intruder as input for computation of alerts and guidance.For this study, the UA's truth trajectory is sent to the DAA algorithm, and no navigation errors are added to the UA's truth states.The intruder's truth states are filtered by the sensor models in a way described below.The Phase 1 DAA MOPS requires ADS-B, active surveillance, and an air-to-air radar as onboard sensors.Only the radar can detect non-cooperative aircraft and is the only sensor modeled.This study sets the radar's field of regard (FOR) but does not model its measurement errors.This study applies two types of radar's FOR.The first type is a hypothetical vertical cylinder around the UA.The radius of the cylinder varies from infinity to 4, 3, and 2 NM.Analysis with this type of FOR sheds light on the sensitivity of the DAA performance metrics to a finite horizontal surveillance range, a key question for UAS operations with low SWaP sensors.A subset of the encounters that represent low SWaP UA operations, in which the maximum UA speed is between 40 and 100 KTAS, is simulated for this FOR.Results can potentially inform the requirements of low SWaP sensors.The second type of radar FOR is defined by a 8 NM slant range, ±15°elevation, and ±110°azimuth.The azimuth and elevation of an intruder are computed with respect to the UA's attitude.This FOR is comparable to the Phase 1 MOPS radar FOR and allows for comparisons to a previous study.This type of radar FOR is applied to all encounters.The Phase 1 MOPS also requires a tracker that fuses and correlates measurements from multiple sensors for a single intruder into tracks.The UA's state is also input to the tracker to produce the intruder's state in absolute coordinates.In this study, the tracker is modeled as a simple pass-through tracker, i.e., the radar measurements, including truth positions and velocities, are converted to tracks through coordinate transformations. +E. Alerting and Guidance AlgorithmThe Phase 1 MOPS defines three types of alerts in increasing levels of severity: preventive, corrective, and warning.The lowest level, preventive, alerts the pilot to not maneuver vertically when the aircraft are separated vertically by 450 to 700 feet.This alert should not be triggered by non-cooperative aircraft (for its lack of precise altitude information) and is not modeled.The second level, corrective, indicates that a LoDWC is predicted to occur in the future, an avoidance maneuver is necessary, but there is still time for coordination with ATC.The highest level, warning, indicates that a LoDWC is imminent, an avoidance maneuver is needed, and coordination with ATC before maneuvering is not a requirement.The DAA system must present maneuver guidance to UAS pilots about DWC-based, conflict-free aircraft headings or altitude ranges.In case a LoDWC is unavoidable, the DAA system should present regain-DWC guidance, a range of headings or altitudes that, if executed, can increase separation at the CPA and regain DWC effectively.Detect and AvoID Alerting Logic for Unmanned Systems (DAIDALUS) [10], a reference algorithm developed to validate the Phase 1 MOPS requirements, computes the alerts and guidance during the simulation of encounters.DAIDALUS computes UA trajectories resulting from executing vertical or horizontal DAA maneuvers to determine which maneuvers would result in conflicts and which are conflict-free.Upon alerting, DAIDALUS generates corresponding preventive, corrective, and warning guidance indicating a range of conflict-free headings and altitudes for a pilot to select from in order to maintain DWC separation.DAIDALUS generates regain-DWC guidance if a LoDWC is unavoidable.In this study, DAIDALUS alerts were issued based on a buffered DWC volume.Specifically, the HMD * used by the DAA alerting and guidance algorithm is scaled by a factor of 1.52 to be consistent with the parameters referenced in the Phase 1 MOPS.This buffer is meant to guard against maneuvering intruders and surveillance uncertainties (none in this work).A persistent corrective or warning alert would lead to the execution of a DAA maneuver.For maneuver guidance computation, DAIDALUS assumes an immediate transition of the UA from constant velocity to a constant turn rate for the trajectory.The turn rate is assumed to be 7 deg/sec for simulations using the cylindrical radar FOR (first type).This is a feasible turn rate for low UA speeds.For simulations using the Phase 1 comparable radar FOR (second type), the turn rate is set to 3 deg/sec, matching the turn rate assumption in the Phase 1 MOPS. +F. Pilot Response ModelFigure 4 shows the SC-228 standard response pilot model created by MIT Lincoln Laboratory [28].In this study, the model selects and executes appropriate maneuvers based on DAIDALUS's heading and altitude guidance.Model parameters were derived from human-in-the-loop experiment results [28].The model has the capability to sample response (or delay) times and maneuvers.For this study where the focus is assessment of the candidate DWCs rather than the total safety of the DAA system, the model is executed in deterministic mode, in which all response times are constant, and maneuvers are selected from the edge of the bands plus an optional buffer (none for the data shown in this study).The response times are set to the expectation value of a distribution, which is either a Gamma distribution (for the ATC coordination time) or an exponential distribution.For corrective and warning guidance, only horizontal maneuvers were executed.This is to reflect the fact that, in reality, vertical maneuvers against non-cooperative intruders are much less robust in these situations due to uncertainties in non-cooperative sensors' vertical measurements.For regain-DWC guidance, both horizontal and vertical maneuvers can be executed due to the improvement of vertical state accuracy in reality.For horizontal maneuvers, the UA always maneuvers in the direction of the minimum heading change suggested by the guidance bands.In the event that the minimum suggestion is inconclusive, the ownship will turn left, as a preference for left turns was observed in human-in-the-loop experiments [28].After the first alert comes up, the pilot response model implements a 5-second initial delay representing the time it takes the pilot to perceive the alert and devise a plan.For corrective alerts, an additional 11-second ATC coordination time elapses, representing the time it takes the pilot to communicate the intended maneuver with and receive approval from ATC.The model then implements a 3-second execution delay representing the time it takes the pilot to enter a maneuver command into the control station and transmit this command to the UA.The ownship may perform multiple maneuvers per encounter to resolve a conflict.The response time between decisions is determined by the alert state, as shown in Table 2.For example, if the pilot response model selects a maneuver during a warning alert state, the situation will be re-evaluated 6-seconds after the selection time, called the decision update period, and a different subsequent maneuver can be selected and executed, if needed.It takes 3 seconds after re-evaluation to select a maneuver for +G. UA Flight ModelOnce the pilot response model selects a DAA maneuver and the execution time delay elapses, the flight model takes control and deviates the UA's trajectory from its nominal trajectory by executing the selected DAA maneuver.Once the target heading or altitude is achieved by the flight model, the UA trajectory stays at that heading or altitude until the end of the encounter or until another maneuver is executed.The intruder's trajectory is not impacted by the UA's maneuvers, assuming a worst case in which the intruder either does not detect the UA or does not attempt to resolve the conflict. +IV. ResultsThe one million encounters are binned by the maximum UA speed during an encounter into four ranges listed in Results for the safety metrics are presented in Section IV.A, and results for the operational suitability metrics are presented in Section IV.B. +A. Safety Metrics +With Cylindrical Radar FOR and an Infinite Horizontal RangeOnly low-speed UA encounters in bin 1 are simulated with this radar FOR. Figure 6 shows the NMAC risk ratios (left) and LoDWC ratios (right) for the four candidate DWCs and the Phase 1 DWC.An important observation of the NMAC risk ratios is that there is no statistically significant difference among them, even when compared to the Phase 1 DWC.Interestingly, the Phase 1 DWC does not perform better.This suggests that all candidate DWCs are likely to be acceptable in terms of their resulting DAA performance to avoid NMACs, given sufficient surveillance volume (infinite for this simulation) and small surveillance uncertainties (none for this simulation.) +Fig. 6 Safety metrics for low-speed UA encountersIt should be pointed out that there is not a single value for the maximum acceptable NMAC risk ratio.The risk assessment makes use of the NMAC risk ratio in calculating the likelihood of a safety risk incident.Additional variables such as the encounter frequency, which varies greatly with location, must be estimated before the likelihood can be computed for a specific UAS mission.Regarding results of the LoDWC ratio, DWC2 yields the lowest value of 9%, and DWC1 yields a slightly worse 10%.DWC3, DWC4, and the Phase 1 DWC all have comparable values near 12%.Unresolved LoDWCs comprise the majority of the risk ratio.Adding a buffer of 5 degrees to the heading selected by the pilot response model to keep the selected heading further away from the edge of conflict band was attempted, but results showed no improvement.Intruder and ownship's nominal maneuvers appear to be the leading cause of unresolved NMACs and LoDWCs.Nominal maneuvers are maneuvers that are part of the unmitigated (original) trajectory.A close examination of encounters with NMACs showed that, in almost all these encounters, either the intruder or ownship had a late nominal maneuver near the nominal time of closest approach (TCA) during the encounter.The nominal TCA is based on unmitigated trajectories.Many LoDWCs are caused by late maneuvers as well.To confirm the impact of late maneuvers, NMAC risk ratios and LoDWC ratios were computed for a subset of encounters in which either the ownship or intruder has a nominal maneuver within 30 seconds of nominal TCA.This subset includes 57.4% of all the encounters.Results in Figure 7 show much higher NMAC risk ratios and LoDWC ratios across candidate DWCs compared to Figure 6.This confirms that late maneuvers (and hence, late alerts) are challenging for the DAA system, regardless of DWCs. +With Cylindrical Radar FOR and a Finite Horizontal RangeTo assess the impact of finite surveillance ranges on safety metrics, NMAC risk ratios and LoDWC ratios were compared in simulation runs of encounters in the low UA-speed bin 1 with 4 NM, 3 NM, and 2 NM horizontal ranges of the radar FOR imposed.Results are shown in Figure 8.The NMAC risk ratios for DWCs 1, 2, and 3 appear to +With Phase 1 Radar's FORThe Phase 1 comparable radar FOR has an 8 NM slant range, ±15°elevation range (up and down the nose of the UA), and ±110°azimuth range (left and right of the nose of the UA).With this radar's FOR, an intruder approaching the ownship from behind will not be detected by the DAA system.Fig. 9 shows the binned NMAC risk ratios.The NMAC risk ratios in bin 1 are the highest, mainly because the UA in this bin are more likely to be overtaken by aircraft from behind, coming from outside the radar's FOR.Compared to the NMAC risk ratio in Figure 6, the NMAC risk ratios in bin 1 are much higher due to the lack of surveillance of overtaking intruders.Low NMAC risk ratios are observed in higher speed bins.UAs in speed bins 2, 3, and 4 fly faster than most of the intruders and are therefore less likely to have undetected intruders approach from behind.There are no mitigated NMACs in the speed bin 4. indicating the NMAC risk ratio is likely to be extremely low. +B. Operational Suitability MetricsThese metrics are evaluated for only the low-speed UA bin 1 with a cylindrical radar FOR and an infinite horizontal range.The alert ratio measures the alert frequency relative to the unmitigated NMAC frequency.Ideally, the alert Alerting time and range are computed based on the first maneuver-triggering alert of any level that occurs in an encounter.For an alert to trigger maneuvers, the alert must remain on during the initial delay and ATC coordination (if it is corrective).Alerting time values are relative to the time of an unmitigated LoDWC.The solid lines in Figure 12 show the estimated cumulative distribution function for alerting time (left) and range (right) for all encounters that have an unmitigated LoDWC.The cumulative distribution function is the probability that alerting time or range will be less than or equal to the values on the x-axis.For example, the alerting range plot shows that 60% of encounters run with DWC4 have their first maneuver-triggering alert at a range of 3 NM or less.All encounters evaluated with DWC4 have their first maneuver-triggering alert within 6 NM.The solid lines in the left plot of Figure 12 show that the first alert time is driven more by τ mod * than by HMD * .the UA or the intruder.As a result, these encounters have on average shorter alert times and ranges.The left plot shows the shift of alerting times is more pronounced for candidate DWCs with a non-zero τ mod than for DWC2.The right plot show DWC3 has the lowest range.This can be explained by the small HMD * of DWC3 and the fact that many of these unresolved LoDWCs are from encounters with low closure rates, for which τ mod * has little effect in increasing the alerting range.The low alerting range of DWC3 does not make DWC3 preferable for reducing surveillance requirements, because these encounters are challenging cases that cannot be aided by surveillance volume. +V. DiscussionSection IV.A.1 shows the impact of ownship or intruder's late nominal maneuvers on the safety metrics.The causes of NMACs and LoDWCs in this study, even with DAA maneuvers, can be summarized as follows:1) Intruder and/or ownship's nominal maneuvers2) Surveillance volume limitations3) Guidance's ineffectiveness or instability 4) Pilot response unable to keep up with the situation For any specific encounter leading to an NMAC, all causes could have contributed.For example, analysis of a few select encounters leading to NMACs indicates an intruder maneuver near the TCA (cause 1), which causes the conflict-free guidance bands to shrink to zero width, leaving no conflict-free heading available.In this situation, the regain-DWC guidance comes up, is executed, but changes turn directions multiple times during the UAS's maneuver (cause 3).Combined with the pilot response delay (cause 4), this instability of guidance can cause a chase situation, resulting in the DAA system's failure to avoid an NMAC.The LoDWCs result from similar causes.Nominal ownship maneuvers are likely to bring the ownship into a immediate conflict that leaves little time for the DAA system to resolve.This may occur more frequently for UAS missions that fly loitering or grid patterns, accelerating often in horizontal and/or vertical directions.However, these conflicts occur in simulations more frequently than in reality, due to the limitation of the pilot response model.In reality, the DAA surveillance will provide UAS pilots situation awareness by showing peripheral conflict bands for intruders in the vicinity (if the intruder is within the DAA system's surveillance volume).UAS pilots can then make strategic nominal maneuvers different from the planned one to avoid conflicts effectively.The impact of finite surveillance range on the safety metrics, presented in Section IV.A.As shown in Section IV.A.3, NMAC and LoDWC risk ratios increase noticeably for the low-speed UA (bin 1) when the Phase 1 radar FOR is applied.This is due to a large number of undetected intruders approaching from the rear of the UAS.For encounters in which an intruder approaches the UA from an azimuth angle smaller than -110°or larger than 110°, SC-228 determined that the responsibility of maintaining separation in these situations falls on the intruder.Nonetheless, since UAs flying in the low-speed range are likely to be smaller than those flying at 200 KTAS, visual identification by pilots of a manned aircraft can be challenging.It may be beneficial in terms of safety for the sensor for the low-speed UA to cover a wider range of azimuth angles.Operational suitability metrics, presented in Section IV.B, show that DWC2 leads to more alerting times before LoDWC and less required horizontal surveillance range, giving it an edge over the other candidates.Although DWC3 achieves a lower alert ratio than the three other candidate DWCs, the importance of this metric is likely low for all but environments with very high traffic densities of non-cooperative aircraft.Encounters with non-cooperative aircraft are expected to occur with a low frequency everywhere in the National Airspace System [12].Therefore, the slight increase of alerts as a result of applying a DWC other than DWC3 is regarded as acceptable.The pilot response model is applied in its deterministic mode for this study, in which all response times take constant values.These response times are representative values, because they are expectation values of their corresponding distributions.Sampling response times from distributions, nonetheless, may have some effects on the metrics.Another aspect not modeled by the pilot response model is the effect of varying candidate DWCs on pilot response times.The pilot response model's parameters were derived from experiments using a Phase 1 DWC.For a smaller candidate DWC such as DWC2 that issues alerts at shorter ranges, pilots may feel a sense of urgency and act a little quicker.Sensor uncertainties are expected to increase the NMAC and LoDWC risk ratio.In a recent study by MITRE [29] that applied the same pilot response model, DAA algorithm, and the Phase 1 radar FOR to an encounter set consisting of UA flying between 50 and 100 kts, the NMAC risk ratio and LoDWC ratio for the Phase 1 DWC were computed to be 0.22 and 0.42, respectively.Comparing these values to 0.15 and 0.28 from the speed bin 1 results, without sensor uncertainties, the difference is believed to be the introduction of sensor uncertainties into the simulation. +VI. ConclusionA detect-and-avoid (DAA) system is a critical component for maintaining safety of UAS missions in the same airspace with manned aircraft.This study evaluates four candidate DAA Well Clear (DWC) definitions for UAS encountering non-cooperative aircraft, using safety and operational suitability metrics.One million encounters, covering a wide range of unmanned aircraft (UA) speeds from 40 to 250 KTAS, are simulated using a reference DAA algorithm, a pilot response model, and two types of radar field of regard (FOR).The findings from this study provide key supporting information for the requirements of DAA systems, and are summarized below.A general observation of safety metrics computed across candidate DWCs is that they are not improved by the introduction of the time parameter, τ mod * , into the DWC definition.In fact, the loss of DWC (LoDWC) frequency is worsened by τ mod * , particularly at high UA speeds.This suggests that τ mod * is not necessary for a DWC definition.Furthermore, operational suitability metrics show that τ mod * increases the required alerting range without providing additional safety benefit.Based on the findings from this work and a few related studies, SC-228 selected DWC2, the only candidate DWC without a τ mod * , for UAS encountering non-cooperative aircraft.This decision achieves a major milestone in SC-228's Phase 2 work.For surveillance requirements for UAS operations with low size, weight, and power (SWaP) sensors, this study also provides important supporting information.Simulation results show that reducing the horizontal surveillance range from infinity to 3 NM does not appear to impact safety metrics.Additional reduction from 3 NM is possible, but the value may not go below 2 NM, at which range the LoDWC ratio is impacted.Given the payload restrictions fromFig. 1 A1Fig. 1 A schematic representation of the DWC zone +4 )Fig. 242Figure 2 depicts the simulation architecture diagram.Data in both open-loop and closed-loop simulations follow the direction of solid arrows.Data in dashed arrows are for closed-loop simulations only.The box labeled "Encounter" provides the input data of ownship and intruder trajectories to the simulation.Each component in the diagram will be described in the following sections, including the encounter set, the UA navigation errors, the sensor and tracker models, the alerting and guidance algorithm, the pilot response model, and the UA flight model, respectively.Note that communication delays and failures are not modeled in the simulation. +Fig. 33Fig. 3 UAS Encounter Characteristics +Fig. 44Fig. 4 The SC-228 pilot response model +Figure 55Figure5shows the speed bins and the radar FOR types simulated in this study.A cylindrical radar FOR, defined by a maximum horizontal range around the UA, is applied to encounters in the low UA speed bin.This FOR is meant to shed light on the trade space between safety metrics and surveillance range requirements of low SWaP sensors, which are assumed to be carried by UAs with low mission speeds.The second radar FOR type, comparable to the Phase 1 radar's and defined by 8 NM slant range, ±110°azimuth, and ±15°elevation, is applied to encounters in all UA speed bins. +Fig. 55Fig. 5 Results coverage of UA speed bins and radar FOR types +Fig. 7 Fig. 878Fig. 7 Safety metrics for low-speed UA encounters with late ownship or intruder maneuvering +Fig. 9 NMAC9Fig. 9 NMAC Risk Ratios Binned by Speed +Fig. 1010Fig. 10 LoDWC ratios binned by speed +Fig. 11 Fig. 12 First1112Fig. 11 System operating characteristic for low UA speed bin 1 +UAs for low SWaP operations, it remains an open question whether this range is achievable by current surveillance technologies such as radar and EO/IR.For extending the Phase 1 DAA minimum operational performance standards (MOPS) to UA speeds greater than 200 KTAS, results for UAs flying between 200 and 250 KTAS show the DAA system achieves promising low NMAC risk ratios as well as low LoDWC ratios, given the Phase 1 radar FOR.This suggests the extension of the Phase 1 DAA MOPS to higher UA speeds such as 250 KTAS impacts neither safety nor the air-to-air radar's FOR requirements.This is the Accepted Manuscript.The final published version is available from the AIAA at https://arc.aiaa.org/doi/10.2514/1.D0199 This material is declared a work of the U.S. Government and is not subject to copyright protection in the United States. + + +This is the Accepted Manuscript. The final published version is available from the AIAA at https://arc.aiaa.org/doi/10.2514/1.D0199 This material is declared a work of the U.S. Government and is not subject to copyright protection in the United States. +Table 1 Candidate DWCs for non-cooperative aircraft, their definitions and properties1NameHMD * τ mod (ft) (sec) *h * (ft) (%) (NM) P MIRCommentDWC120001545051.8PrimaryDWC22200045052.0PrimaryDWC315001545071.7BackupDWC4250025450 3.62.3BackupPhase 1400035450 2.23.3Phase 1 +This is the Accepted Manuscript. The final published version is available from the AIAA at https://arc.aiaa.org/doi/10.2514/1.D0199 Thismaterial is declared a work of the U.S. Government and is not subject to copyright protection in the United States. +Table 2 Pilot Response Model Decision Update Times2Alert ConditionDecision Update Period (s)No Alert12Corrective Alert6Warning Alert6Regain-DWC Guidance3corrective and warning alerts.If regain-DWC guidance comes up, the selection of a maneuver occurs immediately.In all cases, another 3 seconds elapse before the maneuver is executed.If the alert goes away during the evaluation periods of initial delay or ATC coordination time, the pilot response model's state goes back to the No Alert state.However, if the alert goes away during the execution time delay period, the maneuver is still executed.If the severity of the alert level increases during the evaluation periods of time, update times for the more severe alert apply immediately. +Table 3 .3The low-speed bin of 40 to 100 KTAS aligns with the low SWaP UAS operations.The two medium speed bins +This is the Accepted Manuscript. The final published version is available from the AIAA at https://arc.aiaa.org/doi/10.2514/1.D0199 Thiscovered by the Phase 1 MOPS's UA range of 40 to 200 KTAS.The high-speed bin of 200 to 250 KTAS represents a potential extension of the MOPS.Table3also shows the percentage of encounters in each of these bins.material is declared a work of the U.S. Government and is not subject to copyright protection in the United States.are +Table 3 Ownship Speed Bins3Bin # Maximum Ownship Speed (KTAS) Percentage (%)140-10070.02100-1507.33150-20022.34200-2500.4 +2, suggests that a reasonable minimum surveillance range for low SWaP sensors is likely between 2 NM and 3 NM.While 3 NM is sufficient for any candidate DWC to maintain the same level of safety metrics as those with an infinite surveillance range, 2 NM increases the NMAC risk ratio only for DWC4 and LoDWC ratios for DWC1, DWC2, and DWC4.This indicates that, with 2 NM, in some situations UA pilots do not have enough time to evaluate and maneuver the UA in order to maintain separation.Although operationally undesirable, if the likelihood of an encounter with a non-cooperative aircraft is low, the amount of nuisance to pilots might be acceptable.On another note, whether the state-of-the-art low SWaP sensortechnologies, such as radar and electro-optical or infrared (EO/IR) sensors, can meet this level of DAA surveillance range requirements (to be defined) and payload requirements constrained by UA and mission types simultaneously remains an open question. + + + + +This is the Accepted Manuscript.The final published version is available from the AIAA at https://arc.aiaa.org/doi/10.2514/1.D0199 This material is declared a work of the U.S. Government and is not subject to copyright protection in the United States. + + + + +This is the Accepted Manuscript.The final published version is available from the AIAA at https://arc.aiaa.org/doi/10.2514/1.D0199 This material is declared a work of the U.S. Government and is not subject to copyright protection in the United States.Sensor uncertainties will be the subject of a follow-up study.Specific tasks include the impact of sensor uncertainties on the safety and operational suitability, pilots' performance and acceptance, and flight tests.All of this technical work will contribute to an updated DAA MOPS as well as related radar and EO/IR MOPS, scheduled to be published by early 2021.These MOPS documents will play a key role in reducing the barrier of enabling UAS operations in the national airspace systems. + + + + + + + + + + + Defining Well Clear for Unmanned Aircraft Systems + + StephenPCook + + + DallasBrooks + + + RodneyCole + + + DavisHackenberg + + + VincentRaska + + 10.2514/6.2015-0481 + + + + AIAA Infotech @ Aerospace + + American Institute of Aeronautics and Astronautics + 2015 + + + Cook, S. P., Brooks, D., Cole, R., Hackenberg, D., and Raska, V., "Defining Well Clear for Unmanned Aircraft Systems," Proceedings of AIAA Infotech@ Aerospace, AIAA, 2015. https://doi.org/10.2514/6.2015-0481, URL https://doi.org/10.2514/6. 2015-0481. + + + + + Characteristics of a Well Clear Definition and Alerting Criteria for Encounters between UAS and Manned Aircraft in Class E Airspace + + MJohnson + + + ERMueller + + + CSantiago + + + + + Eleventh UAS/Europe Air Traffic Management Research and Development Seminar + + 2015 + + + + Johnson, M., Mueller, E. R., and Santiago, C., "Characteristics of a Well Clear Definition and Alerting Criteria for Encounters between UAS and Manned Aircraft in Class E Airspace," Eleventh UAS/Europe Air Traffic Management Research and Development Seminar, 2015, pp. 23-26. URL https://ntrs.nasa.gov/search.jsp?R=20190027490. + + + + + FAA Position on Building Consensus Around the SARP Well-Clear Definition + + DWalker + + + + RTCA Special Committee + + 228 + 2014 + + + Walker, D., "FAA Position on Building Consensus Around the SARP Well-Clear Definition," RTCA Special Committee 228, 2014. + + + + + Flight Test Overview for UAS Integration in the NAS Project + + JamesRMurphy + + + PeggySWilliams-Hayes + + + SamKKim + + + WayneBridges + + + MichaelMarston + + 10.2514/6.2016-1756 + + + + AIAA Atmospheric Flight Mechanics Conference + + American Institute of Aeronautics and Astronautics + 2016 + + + Murphy, J. R., Hayes, P. S., Kim, S. K., Bridges, W., and Marston, M., "Flight Test Overview for UAS Integration in the NAS Project," AIAA Atmospheric Flight Mechanics Conference, AIAA SciTech, 2016. https://doi.org/10.2514/6.2016-1756, URL https://doi.org/10.2514/6.2016-1756. + + + + + Initial Characterization of the 30 kW Miniature Arc Jet (mARC II) at NASA Ames Research Center + + SMLee + + + CPark + + + DPThipphavong + + + DRIsaacson + + + CSantiago + + 10.2514/6.2020-3108.vid + + + + NASA/TM-2016-219067 + + American Institute of Aeronautics and Astronautics (AIAA) + 2016 + + + Tech. rep. + Lee, S. M., Park, C., Thipphavong, D. P., Isaacson, D. R., and Santiago, C., "Evaluating Alerting and Guidance Performance of a UAS Detect-And-Avoid System," Tech. rep., NASA Ames Research Center, NASA/TM-2016-219067, 2016. URL https://ntrs.nasa.gov/search.jsp?R=20160012543. + + + + + Analysis of alerting performance for detect and avoid of unmanned aircraft systems + + SamanthaSmearcheck + + + SeanCalhoun + + + WilliamAdams + + + JaredKresge + + + FabriceKunzi + + 10.1109/plans.2016.7479766 + + + + 2016 IEEE/ION Position, Location and Navigation Symposium (PLANS) + + IEEE + 2016 + + + + Smearcheck, S., Calhoun, S., Adams, W., Kresge, J., and Kunzi, F., "Analysis of Alerting Performance for Detect and Avoid of Unmanned Aircraft Systems," IEEE/ION Position, Location and Navigation Symposium (PLANS), 2016, pp. 710-730. URL https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=7479766. + + + + + The Impact of Suggestive Maneuver Guidance on UAS Pilot Performing the Detect and Avoid Function + + RobertCRorie + + + LisaFern + + + JayShively + + 10.2514/6.2016-1002 + + + + AIAA Infotech @ Aerospace + + American Institute of Aeronautics and Astronautics + 2016 + + + Rorie, R. C., Fern, L., and Shively, J., "The Impact of Suggestive Maneuver Guidance on UAS Pilot Performing the Detect and Avoid Function," AIAA InfoTech @ Aerospace Conference, AIAA SciTech, 2016. https://doi.org/10.2514/6.2016-1002, URL https://arc.aiaa.org/doi/pdf/10.2514/6.2016-1002. + + + + + An Evaluation of Detect and Avoid (DAA) Displays for Unmanned Aircraft Systems: The Effect of Information Level and Display Location on Pilot Performance + + LisaFern + + + RConradRorie + + + JessicaPack + + + JayShively + + + MarkDraper + + 10.2514/6.2015-3327 + + + + 15th AIAA Aviation Technology, Integration, and Operations Conference + + American Institute of Aeronautics and Astronautics + 2015 + + + Fern, L., Rorie, R. C., Pack, J. S., Shively, R. J., and Draper, M. H., "An Evaluation of Detect and Avoid (DAA) Displays for Unmanned Aircraft Systems: The Effect of Information Level and Display Location on Pilot Performance," Proceedings of 15th AIAA Aviation Technology, Integration, and Operations Conference, 2015. https://doi.org/10.2514/6.2015-3327, URL https://arc.aiaa.org/doi/pdf/10.2514/6.2015-3327. + + + + + The Generic Resolution Advisor and Conflict Evaluator (GRACE) for Detect-And-Avoid (DAA) Systems + + MichaelAbramson + + + MohamadRefai + + + ConfesorSantiago + + 10.2514/6.2017-4485 + + + + 17th AIAA Aviation Technology, Integration, and Operations Conference + + American Institute of Aeronautics and Astronautics + 2017 + + + Abramson, M., Refai, M., and Santiago, C., "The Generic Resolution Advisor and Conflict Evaluator (GRACE) in Applications to Detect-And-Avoid (DAA) Systems of Unmanned Aircraft," Proceedings of the 17th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, 2017. https://doi.org/10.2514/6.2017-4485, URL https://arc.aiaa.org/doi/pdf/10.2514/6.2017- 4485. + + + + + DAIDALUS: Detect and Avoid Alerting Logic for Unmanned Systems + + CesarMunoz + + + AnthonyNarkawicz + + + GeorgeHagen + + + JasonUpchurch + + + AaronDutle + + + MariaConsiglio + + + JamesChamberlain + + 10.1109/dasc.2015.7311421 + + + + 2015 IEEE/AIAA 34th Digital Avionics Systems Conference (DASC) + + IEEE + 2015 + + + + Muñoz, C., Narkawicz, A., Hagen, G., Upchurch, J., Dutle, A., Consiglio, M., and Chamberlain, J., "DAIDALUS: Detect and Avoid Alerting Logic for Unmanned Systems," 34th Digital Avionics Systems Conference (DASC), IEEE/AIAA, 2015, pp. 5A1-1. URL https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=7311421. + + + + + Development, Integration and Testing of a Stand-alone CDTI with Conflict Probing Support + + BrandonSuarez + + + KevinKirk + + + ErikTheunissen + + 10.2514/6.2012-2487 + + + + Infotech@Aerospace 2012 + + American Institute of Aeronautics and Astronautics + 2012. 2012 + 2487 + + + Suarez, B., Kirk, K., and Theunissen, E., "Development, Integration and Testing of a Stand-Alone CDTI with Conflict Probing Support," Infotech@ Aerospace 2012, 2012, p. 2487. https://doi.org/10.2514/6.2012-2487, URL https://arc.aiaa.org/doi/abs/10. 2514/6.2012-2487. + + + + + Minimum Operational Performance Standard (MOPS) for Helicopter Hoist Systems + 10.4271/as6342 + + 2017 + SAE International + + + Minimum Operational Performance Standards (MOPS) for Detect and Avoid (DAA) Systems, DO-365, RTCA. Inc., 2017. + + + + + AeroMACS minimum operational performance standards (MOPS) compliance field trials for Hitachi prototype + + RafaelApaza + + 10.1109/icnsurv.2015.7121331 + + + 2015 Integrated Communication, Navigation and Surveillance Conference (ICNS) + + IEEE + 2017 + + + Minimum Operational Performance Standards (MOPS) for Air-to-Air Radar for Traffic Surveillance, DO-366, RTCA. Inc., 2017. + + + + + Fixed-wing group petitions FAA for advisory circular + 10.1016/s0740-8315(88)80079-2 + + + + Hospital Aviation + Hospital Aviation + 0740-8315 + + 7 + 7 + + 2011 + Elsevier BV + + + 1%20Intro%20booklet.pdf + "Introduction to TCAS II Version 7.1," Federal Aviation Administration (FAA), 2011. URL http://www.faa.gov/documentLibrary/ media/Advisory_Circular/TCAS%20II%20V7.1%20Intro%20booklet.pdf. + + + + + Advances in UAS Ground-Based Detect and Avoid Capability + + RaymondYoung + + 10.1109/icnsurv.2019.8735385 + + + + 2019 Integrated Communications, Navigation and Surveillance Conference (ICNS) + + IEEE + 2019 + + + + Advances in UAS Ground-Based Detect and Avoid Capability + Young, R., "Advances in UAS Ground-Based Detect and Avoid Capability," 2019 Integrated Communications, Navigation and Surveillance Conference (ICNS), IEEE, 2019, pp. 1-14. https://doi.org/10.1109/ICNSURV.2019.8735385, URL https: //doi.org/10.1109/ICNSURV.2019.8735385. + + + + + Heavy Lift Multirotor Unmanned Aircraft System for Beyond-Visual-Line-of-Sight Operations in the National Airspace System + + TPSpriesterbach + + + KABruns + + + LIBaron + + + JESohlke + + 10.2514/6.2021-0043.vid + + + Johns Hopkins APL Technical Digest + + 32 + 3 + + 2013 + American Institute of Aeronautics and Astronautics (AIAA) + + + Spriesterbach, T. P., Bruns, K. A., Baron, L. I., and Sohlke, J. E., "Unmanned Aircraft System Airspace Integration in the National Airspace Using a Ground-Based Sense and Avoid System," Johns Hopkins APL Technical Digest, Vol. 32, No. 3, 2013, pp. 572-583. + + + + + Well Clear Trade Study for Unmanned Aircraft System Detect And Avoid with Non-Cooperative Aircraft + + MinghongGWu + + + AndrewCCone + + + SeungmanLee + + + ChristineChen + + + MatthewWEdwards + + + DevinPJack + + 10.2514/6.2018-2876 + + + + 2018 Aviation Technology, Integration, and Operations Conference + + American Institute of Aeronautics and Astronautics + 2018 + + + Wu, M. G., Cone, A. C., Lee, S., Chen, C., Edwards, E. W. M., and Jack, D. P., "Well Clear Trade Study for Unmanned Aircraft System Detect And Avoid with Non-Cooperative Aircraft," 18th AIAA Aviation Technology, Integration, and Operations Conference, 2018. https://doi.org/10.2514/6.2018-2876, URL https://arc.aiaa.org/doi/pdf/10.2514/6.2018-2876. + + + + + Well-Clear Recommendation for Small Unmanned Aircraft Systems Based on Unmitigated Collision Risk + + AndrewWeinert + + + ScotCampbell + + + AdanVela + + + DieterSchuldt + + + JoelKurucar + + 10.2514/1.d0091 + + + + Journal of Air Transportation + Journal of Air Transportation + 2380-9450 + + 26 + 3 + + 2018 + American Institute of Aeronautics and Astronautics (AIAA) + + + Weinert, A., Campbell, S., Vela, A., Schuldt, D., and Kurucar, J., "Well-Clear Recommendation for Small Unmanned Aircraft Systems Based on Unmitigated Collision Risk," Journal of Air Transportation, Vol. 26, No. 3, 2018, pp. 113-122. https://doi.org/10.2514/1.D0091, URL https://doi.org/10.2514/1.D0091. + + + + + Formal Analysis of Extended Well-Clear Boundaries for Unmanned Aircraft + + CésarMuñoz + + + AnthonyNarkawicz + + 10.1007/978-3-319-40648-0_17 + + + + Lecture Notes in Computer Science + + SRayadurgam + + + OTkachuk + + Cham + + Springer International Publishing + 2016 + + + + Muñoz, C., and Narkawicz, A., "Formal Analysis of Extended Well-Clear Boundaries for Unmanned Aircraft," NASA Formal Methods, edited by S. Rayadurgam and O. Tkachuk, Springer International Publishing, Cham, 2016, pp. 221-226. URL https://link.springer.com/chapter/10.1007/978-3-319-40648-0_17. + + + + + Establishing a Risk-Based Separation Standard for Unmanned Aircraft Self Separation + + RolandWeibel + + + MatthewEdwards + + + CarolineFernandes + + 10.2514/6.2011-6921 + + + + 11th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + + American Institute of Aeronautics and Astronautics + 2011 + + + Weibel, R. E., Edwards, M. W., and Fernandes, C. S., "Establishing a Risk-Based Separation Standard for Unmanned Aircraft Self Separation," Ninth USA/Europe Air Traffic Management Research & Development Seminar, 2011. https: //doi.org/10.2514/6.2011-6921, URL https://arc.aiaa.org/doi/pdf/10.2514/6.2011-6921. + + + + + Analysis of Influence of UAS Speed Range and Turn Performance on Detect and Avoid Sensor Requirements + + DevinPJack + + + JeremyHardy + + + KeithDHoffler + + 10.2514/6.2018-3507 + + + + 2018 Aviation Technology, Integration, and Operations Conference + + American Institute of Aeronautics and Astronautics + 2018 + + + Jack, D. P., Hardy, J., and Hoffler, K. D., "Analysis of Influence of UAS Speed Range and Turn Performance on Detect and Avoid Sensor Requirements," 18th AIAA Aviation Technology, Integration, and Operations Conference, 2018. https: //doi.org/10.2514/6.2018-3507, URL https://doi.org/10.2514/6.2018-3507. + + + + + A Bayesian Approach to Aircraft Encounter Modeling + + MykelKochenderfer + + + JamesKuchar + + + JGriffith + + + LeoEspindle + + 10.2514/6.2008-6629 + ATC-345 + + + AIAA Guidance, Navigation and Control Conference and Exhibit + Lexington, Massachusetts + + American Institute of Aeronautics and Astronautics + 2008 + + + MIT Lincoln Laboratory + + + Project Report + Kochenderfer, M. J., Kuchar, J. K., Espindle, L. P., and Griffith, J., "Uncorrelated Encounter Model of the National Airspace System, Version 1.0," Tech. rep., MIT Lincoln Laboratory, Lexington, Massachusetts, 2008. Project Report ATC-345. + + + + + A Recommended DAA Well-Clear Definition for the Terminal Environment + + MichaelJVincent + + + AnnaTrujillo + + + DevinPJack + + + KeithDHoffler + + + DimitriosTsakpinis + + 10.2514/6.2018-2873 + + + + 2018 Aviation Technology, Integration, and Operations Conference + + American Institute of Aeronautics and Astronautics + 2018 + + + + Vincent, M. J., Trujillo, A., Jack, D. P., Hoffler, K. D., and Tsakpinis, D., "A Recommended DAA Well-Clear Definition for the Terminal Environment," 2018 Aviation Technology, Integration, and Operations Conference, 2018. https://doi.org/10.2514/6. 2018-2873, URL https://doi.org/10.2514/6.2018-2873. + + + + + UAS Demand Generation Using Subject Matter Expert Interviews and Socio-economic Analysis + + SricharanKAyyalasomayajula + + + RohitSharma + + + FrederickWieland + + + AntonioTrani + + + NicolasHinze + + + ThomasSpencer + + 10.2514/6.2015-3405 + + + + 15th AIAA Aviation Technology, Integration, and Operations Conference + + American Institute of Aeronautics and Astronautics + 2015 + + + Ayyalasomayajula, S., Sharma, R., Wieland, F., Trani, A., Hinze, N., and Spencer, S., "UAS Demand Generation Using Subject Matter Expert Interviews and Socio-Economic Analysis," Proceedings of the AIAA Aviation Conference, 2015. https://doi.org/10.2514/6.2015-3405, URL https://doi.org/10.2514/6.2015-3405. + + + + + Extended Airspace Encounter Models for Unmanned Aircraft Sense and Avoid Safety Evaluation + + AndrewJWeinert + + + EricHarkleroad + + + JohnGriffith + + + MatthewWEdwards + + 10.2514/6.2013-5049 + + + + AIAA Infotech@Aerospace (I@A) Conference + Lexington, Massachusetts + + American Institute of Aeronautics and Astronautics + Aug. 2013 + + + MIT Lincoln Laboratory + + + Tech. rep. + Weinert, A. J., Harkleroad, E. P., Griffith, J., Edwards, M. W., and Kochenderfer, M. J., "Uncorrelated Encounter Model of the National Airspace System, Version 2.0," Tech. rep., MIT Lincoln Laboratory, Lexington, Massachusetts, Aug. 2013. URL https://apps.dtic.mil/dtic/tr/fulltext/u2/a589697.pdf. + + + + + Investigating the Impacts of a Separation Standard for UAS operations in Enroute and Transition Airspace + + MJohnson + + + ERMueller + + + CSantiago + + NASA/TR-2014-425425 + + + + NASA Ames Research Center + + 2014 + + + Tech. rep. + Johnson, M., Mueller, E. R., and Santiago, C., "Investigating the Impacts of a Separation Standard for UAS operations in Enroute and Transition Airspace," Tech. rep., NASA Ames Research Center, NASA/TR-2014-425425, 2014. URL https://ntrs.nasa.gov/search.jsp?R=20190025402. + + + + + A Bayesian Approach to Aircraft Encounter Modeling + + MykelKochenderfer + + + JamesKuchar + + + JDGriffith + + + LeoEspindle + + 10.2514/6.2008-6629 + + + AIAA Guidance, Navigation and Control Conference and Exhibit + + American Institute of Aeronautics and Astronautics + 2008 + 17 + + + + Kochenderfer, M. J., Espindle, L. P., Kuchar, J. K., and Griffith, J. D., "A Comprehensive Aircraft Encounter Model of the National Airspace System," Lincoln Laboratory Journal, Vol. 17, No. 2, 2008, pp. 41-53. + + + + + A Model of Unmanned Aircraft Pilot Detect and Avoid Maneuver Decisions + + RGuendel + + + MKuffner + + + DMaki + + + 2017 + Lexington, Massachusetts + + + MIT Lincoln Laboratory + + + ATC-434 + Guendel, R., Kuffner, M., and Maki, D., "A Model of Unmanned Aircraft Pilot Detect and Avoid Maneuver Decisions," Tech. rep., MIT Lincoln Laboratory, Lexington, Massachusetts, 2017. ATC-434. + + + + + SC-228 DAA Phase 1 MOPS Verification & Validation Simulation + + MHall + + + JKaznocha + + + ELester + + + SSherman + + + NReep + + MP170272 + + 2017 + Bedford, MA + + + MITRE Corporation + + + MITRE Product + Hall, M., Kaznocha, J., Lester, E., Sherman, S., and Reep, N., "SC-228 DAA Phase 1 MOPS Verification & Validation Simulation," Tech. rep., MITRE Corporation, Bedford, MA, 2017. MITRE Product MP170272. + + + + + + diff --git a/file800.txt b/file800.txt new file mode 100644 index 0000000000000000000000000000000000000000..88f884f58c8c46f92dfc97baf6ab9ed9b3e4272f --- /dev/null +++ b/file800.txt @@ -0,0 +1,980 @@ + + + + +Over the past thirty years, the civil aeronautical community has actively pursued research to enable efficient and environmentally friendly arrival operations.One outcome of this research is a collection of new ground automation tools [1][2][3], developed to assist air traffic controllers (ATC) in managing traffic while maintaining safety.Another outcome is a set of Optimal Profile Descent (OPD) and Continuous Descent Arrival (CDA) procedures [4,5], designed to reduce emissions and noise.Recent years have seen significant efforts to integrate these new tools and procedures to manage arrival flights even more efficiently [6,7].In one such research activity, each arrival flight is scheduled to pass certain points along a specific route defined from the transition airspace (i.e., en route airspace near the terminal area) all the way to the runway [8].This approach was initially called precision scheduling [9] and, in later literature, renamed to precision operations [10].With this approach, air traffic controllers would use speed clearances as the primary means of guiding flights, reserving vectoring clearances for exceptions only.Precision operations, with its speed control, is in contrast to the conventional approach, in which aircraft follow a routine of vectoring clearances from ATC upon entering the terminal airspace [10].In the United States, the technology developed for achieving precision operations is the Terminal Sequencing and Spacing (TSAS) system [11,12], tech-transferred from NASA to the Federal Aviation Administration (FAA) in 2015 and targeted for initial operating capability in 2018.TSAS requires a route-specific speed profile, referred to as nominal, as input for computation of each flight's predicted nominal trajectory.(The nominal speed profile is explained in detail in Section II.)Note that a route may not be completely defined (i.e., arrival procedures not connected to approach procedures) by the time TSAS is implemented.In this case, TSAS models the route by resorting to controllers' consistency in vectoring.The predicted nominal trajectories are used for ordering the aircraft at schedule points.Once the aircraft are ordered, TSAS computes for each flight a speed profile that would meet the schedule, perturbing the profile from the nominal one, if necessary, to meet separation restrictions at schedule points [13].During times of light arrival traffic in which flights rarely violate separation restrictions at the schedule points, TSAS advises all flights on the same route to fly by the same nominal speed profile.During times of medium to heavy traffic in which aircraft separation restrictions cannot be met at the schedule points, TSAS issues a perturbed speed profile to many of the flights.The current scheme of TSAS's speed perturbation considers only reduction of speed from the nominal speed profile.Nonetheless, perturbation that increases the speed from the nominal has been proposed [14].The selection of a nominal speed profile is expected to impact the efficiency of the TSAS schedule because this profile serves as a reference speed strategy and, in actual operations, will be flown by many flights.A nominal speed profile is most cost-efficient for a flight if it coincides with that flight's preferred speed profile.However, the preferred speed profile varies among airlines and aircraft types.Even flights from the same airline, with the same aircraft type, and on the same route can have different preferred speed profiles.The best that can be achieved by a nominal speed profile for TSAS is systemic efficiency; i.e., one that minimizes the total cost (cost to be defined in Section III A) of the arrival flights.Efficiency of a nominal speed profile for an individual flight, as defined specifically in terms of total cost, was not among the objects of study in the development of the TSAS system.(All the nominal speed profiles were validated for flyability only.)At the same time, the sensitivity of a speed profile's cost efficiency to individual flights' cost functions presents interest, because the degree of sensitivity tells how much consideration should be put into the selection of a nominal speed profile.This interest is the main motivation for this work.The main contributions of this paper are two.The first is a general methodology for computing a cost-optimal speed profile (described in Sec.III B) using a high-fidelity aircraft fuel model.This profile is computed using the exact formula derived in [15].The second is application of this methodology to an actual arrival route and its nominal speed profile, demonstrating sensitivity of the nominal speed profile's cost efficiency to a flight's cost function, parameterized by the aircraft type and the Cost Index [16].This work aims to answer questions of the following kinds:• Given a flight's cost function, how cost-efficient is a speed profile?• How does the cost efficiency of a speed profile vary with a flight's cost parameters?Although the application is to a specific route, the modeling methodology is applicable to any route that can be specified by waypoints, speed restrictions, and altitude restrictions (see Section II for more details).The rest of the paper is organized as follows.The arrival route as well as its nominal speed profile to be examined are described in Section II.The general exact formula for a cost-optimal speed profile is given in Section III.The aircraft fuel model for computing fuel consumption is described in Section IV.The aforementioned sensitivity analysis of the nominal speed profile's cost efficiency is carried out in Section V. Section V also compares a cost-optimal speed profile to a speed profile optimized by a flight simulator's Flight Management System (FMS).Discussion of the results is given in Section VI. +II. The Nominal Speed Profile +A. Arrival RoutePrecision operations [10] are based on specific arrival routes constructed from published procedures and, if necessary, a model of controllers' vectoring patterns.The resulting route consists of a sequence of waypoints all the way to the runway.Altitude restrictions are imposed at certain waypoints to provide vertical separation for different traffic flows, keep specific arrival flows in the same controller's sector, and account for terrain constraints [17].Although the upper and lower bounds of an altitude restriction can be different to accommodate different aircraft types' performance envelopes, ATC can rarely afford such flexibility in a congested terminal airspace with complex traffic flows.Therefore, altitude restrictions in the terminal airspace, and in some cases also the transition airspace, have either narrow windows or identical upper and lower bounds.When such altitude restrictions are closely spaced, arrival flights are observed to fly each descent segment by an approximately constant flight-path angle (FPA).An arrival route and its altitude restrictions are much constrained by considerations related to traffic flow, terrain, noise, and safety, and cannot be modified without considering the entire traffic pattern.Both the route and the altitude restrictions are fixed in this work.Only the speed profile is considered for optimization.The arrival route analyzed in this study, named here the GEELA route, was part of an actual arrival procedure and has been previously used in Human-In-The-Loop (HITL) simulations for demonstrating the TSAS technologies [9]. Figure 1 shows the scale of this route as well as the altitude restrictions along the route, starting from waypoint MOHAK and ending at JAMIL, a waypoint 8 nmi to the west of the runway.The short route between JAMIL and the runway is not considered in this work because operational safety and noise considerations leave little room for speed optimization.Table 1 specifies the speed restrictions along the GEELA route [18]; the speeds shown are the calibrated airspeeds (CAS).The path distance column in Table 1 estimates each waypoint's flight path distance from the end waypoint JAMIL, with a reference of 0 at the end of route and a negative sign to keep the path distance increasing along the route.down just in time to meet the speed restriction.This assumption derives from observation of a flight simulator's speed profile, which is to be described in Section II B. +TSAS constructs a nominal speed profile from speed restrictions by assuming the aircraft slowsAlthough this study analyzes only a specific arrival route, the methodology presented here is generally applicable to routes described by waypoints optionally furnished with restrictions of altitude and speed.By assuming realistic turn radii, these altitude and speed parameters can be fit with a 3-D curve, and its 2-D projection onto the approximately flat earth is parameterized by a variable called the path distance.The control problem is solved for the path distance and speed as functions of time (see Section III for more details). +B. Simulated FlightThe Aircraft Simulation for Traffic Operations Research (ASTOR) is a desktop-based aircraft simulator developed to support research of air traffic operations within future airspace environments [19,20].It has been used extensively in HITL simulations [12,21,22] and is capable of simulating realistic aircraft flight physics, pilot interface, automation, and communication infrastructures.ASTOR has high-fidelity components of aircraft performance, FMS, and autopilot.The nominal speed profile for the GEELA route was derived from an ASTOR simulation of a flight under standard atmospheric conditions and without winds.The aircraft type was set to Boeing 757-200.Altitude restrictions shown in Figure 1 and speed restrictions in Table 1 are applied to this flight.the altitude profile of the ASTOR flight (the modeled altitude profile for the GEELA route is to be described in Section V A).The closeness of the two altitude profiles (less than 100 ft apart anywhere) is important for the validity of the comparison work to be described in Section V B. The cost efficiency of the nominal speed profile is examined below for both B757-200 and B737-800 aircraft types and for a range of airlines' business objectives, parameterized by the ratio of a flight's fuel and operating time costs. +III. Cost-Optimal Speed Profile +A. Model FormulationThe cost-optimal speed profile defined here minimizes a modeled cost function, chosen for this work as the direct operating cost J (referred to simply as cost for the rest of the paper) of the trajectory.The cost J is the integral of a sum of two terms, corresponding to fuel consumption and flight duration, respectively:J = t f 0 p f [f (t) + r] dt. (1)Here r is related to the Cost Index [16] by a unit conversion factor, and the fuel consumption rate is modeled by this formula,f (t) = c f T + f 0 ,(2)where both c f and f 0 are functions of aircraft altitude and speed.Trajectories that minimize fuel or various forms of cost have received much attention in the literature, which is briefly reviewed in this paragraph.A fairly comprehensive review of fuel-and cost-optimal trajectories can be found in [23].Early work considered optimization of trajectories in climb, cruise, and descent phases combined [24][25][26][27].Such analysis can be formulated as a multi-phase optimization problem [28].Some recent work analyzed descent only or cruise/descent trajectories [29][30][31] and investigated trade-offs between fuel consumption and flight time [32].The optimal descent trajectory obtained from these efforts typically uses idle thrust engine control.However, none of these works consider altitude restrictions common in a congested terminal airspace.There is no guarantee such an optimal trajectory would meet the altitude restrictions and, in trajectory-based operations, may not be executable as is.The approach in this work incorporates altitude restrictions in the model and explores speed profile optimization within these restrictions.This work models an arrival route with a constrained altitude profile as a continuous curve in 3-dimensional airspace.(The method used here for constructing such an altitude profile from altitude restrictions is described in the second paragraph of Section V A.) The curve is parameterized by the aircraft's path distance coordinate, denoted by x, along its ground path.It follows that the altitude h and air density ρ can be written as functions of x:h = h(x), ρ = ρ(x),and therefore the tangent of the inertial FPA γ i is given bytan γ i = dh dx .The last equation is based on the sign convention that γ i is negative for descent.The wind along the route is assumed to be a function of x, i.e., w = w(x), but not of time.This is a good assumption if the duration of the flight segment considered is short.Crosswinds affect the aircraft's dynamics to a much lesser extent [33] than do along-track winds, hence are neglected in this model.Vertical winds are usually of smaller magnitudes than are head and tail winds, hence are also neglected.In the following state equations,ẋ = V + w,(3a)V = g T -D W -sin γ a -(V + w) dw dx ,(3b)the right-hand side of Eq. (3b) includes the thrust, drag, gravity, and the inertial force that results from the wind.The angles γ a and γ i are related byḣ = V sin γ a = (V + w) sin γ i .(4)The aircraft's total engine thrust is bounded by the engine's performance envelope,0 ≤ T ≤ T max .(5)The optimal control problem for the cost-optimal speed profile is defined by state variables x and V , control variable T , state equations in Eqs.(3), cost functional in (1), fuel rate in (2), initial time t = 0, final time t f free, boundary conditionsx(0) = x i , V (0) = V i , x(t f ) = x f , V (t f ) = V f ,                           (6)and control constraints by Eq. ( 5).For a cost-optimal speed profile to be feasible operationally, speed constraints may need to be imposed.One operational constraint imposed here is the FAA's speed restriction to keep the flight at or below 250 knots when at altitude 10,000 ft or below.Another constraint to consider is that the aircraft's speed in CAS should not increase in the terminal area.That constraint is not included in the model because it turns out to be inactive in the terminal area for the parameters considered (assuming terminal area includes the part of route below 10,000 ft). +B. SolutionThe optimal control problem described above has been solved analytically using Green's Theorem and observations of a typical aircraft's drag polar [15].In the solution, the aircraft aims to reach and keep a preferred speed, called the minimum-cost speed and denoted by V mc (x).In terms of the Hamiltonian formalism of optimal control theory, the minimum-cost speed curve is a singular arc [34].A typical cost-optimal speed profile starts from its initial speed and gets to V mc (x) as quickly as possible using an appropriate extremal value of thrust.It then stays on V mc (x) as long as possible, until it has to leave to reach the final speed.Figure 4 sketches typical cost-optimal speed profiles V (x).Such a profile consists of three segments.In Figure 4(a), the aircraft flies from its initial state, P i = (x i , V i ), to state Q 1 using a thrust of T max .It stays on the V mc curve (shown dashed) from Q 1 to Q 2 , using a non-extremal thrust.It then leaves the curve and arrives at the final state, P f = (x f , V f ), using idle thrust.Similarly, in Figure 4(b), the aircraft flies from its initial state, P i , to state Q 1 using idle thrust.It stays on the V mc curve (dashed line) from state Q 1 to state Q 2 , using a non-extremal thrust.It then leaves the curve and arrives at the final state, P f , using idle thrust.(a)xV P f Q 2 Q 1 V MC P i (b) x V P f P i Q 1 Q 2 V MCFig. 4 Notional sketch of a cost-optimal speed profile, V = V (x), in a bold curve.The plot ofVmc(x) is shown dashed. Note that V i > Vmc(x i ) in (a) and V i < Vmc(x i ) in (b).The minimum-cost speed V mc is the solution to0 = ∂c f ∂x - ∂R ∂V .(7)HereR ≡ c f D + r V + w ,(8)and D and r are defined by the following equations,D = g W (D + W sin γ a ) + (V + w) dw dx(9)andr = g W (f 0 + r) .(10)Eq. ( 7) is solved numerically for values of x to get V mc .In the special case of a level flight with r = 0 and a constant c f , the function V mc (x) becomes identical with the definition of the best range speed [33].Speed restrictions may prevent the cost-optimal speed profile from staying on V mc (x).The FAA's speed restriction to keep the flight at or below 250 knots CAS when at altitude 10,000 ft or below is applied.The restriction was applied in the following two steps: (i) solve for the cost-optimal speed profile without imposing the speed restriction, and (ii) check whether the computed solution violates the restriction.If it does, then replace the problem by a two-stage control problem [34],with the stages separated by the smallest (earliest) value of x at which the altitude h(x) reaches 10,000 ft.Stage 1 is solved using a final speed of 250 knots CAS, and stage 2 is solved using an initial speed of 250 knots CAS.The V mc (x) is replaced by 250 knots CAS for the portion of x where V mc (x) is greater than 250 knots CAS.If the flight-path angle changes rapidly, V mc may change rapidly too.As a result, the thrust required to follow V mc can fall outside bounds (5).This transient segment is usually short.In this case, the state (x, V ) of the aircraft leaves V mc momentarily with an extremal thrust and returns to V mc .The exact beginning point of this transient segment is sought numerically within a short segment of x [15].The computation of fuel in this work is based on a high-fidelity aircraft fuel model of B757-200 and B737-800 as part of the Center-TRACON Automation System (CTAS) [35].CTAS describes fuel rates with look-up tables.Values in these tables are approximated by fitting the coefficients c f and f 0 in Eq. ( 2).The slight variation of c f with thrust is neglected.These approximations are estimated to affect the fuel consumption by less than 3% for a descent trajectory.It will be shown in Section V E that the idle-thrust fuel rate, f 0 , contributes a significant percentage of the fuel consumption for a descending aircraft.That is, segments flown by idle thrust consume a percentage of the total fuel that is far from negligible.A nonzero value of f 0 captures the real fuel rate better than the zero fuel rate assumption for idle thrust descent (minimum-thrust) made in numerous previous works [25,28].The optimal speed profile was computed using MATLAB [36] code.The computation involves the V mc curve, forward integration from the initial state to the V mc curve, and backward integration from the final state to the V mc curve.With minimum attempt to optimize the performance of the code, each speed profile took about 5 minutes to complete.Additional computational details can be found in [15].V. Results +A. Route, Altitude, and Air DensityThe ground path of the route is modeled as a sequence of linear segments, pairwise connected by circular arcs, which correspond to turns.Turns are modeled with a 5 nmi turn radius each, a choice found to match closely those observed in the ASTOR trajectories.This assumption of speed-indepedent turn radii decouples the horizontal path from the speed.The total length of the GEELA route is about 101 nmi.Lift is assumed to be equal to the aircraft's weight at all times.These approximations have been used by trajectory generators of ground automation systems [37].Figure 3 shows the modeled altitude h(x) of the GEELA route.This change rate is derived from real flights' track data.The difference between the modeled altitude and the ASTOR altitude is everywhere less than 150 ft.The air density as a function of the altitude is computed using the standard atmospheric model.This density comes out 0.557 kg 3 /m at an altitude of 25,000 ft and 1.090 kg 3 /m at an altitude of 4000 ft.Wind effects have been investigated previously [15] and are not considered here. +B. Sensitivity Analysis of Cost EfficiencyThe sensitivity analysis of the nominal speed profile's cost efficiency to aircraft type and the Cost Index.Two aircraft types, B757-200 and B737-800, are investigated and compared.Values of the Cost Index investigated, denoted as CI, range from 0 to 70 (($/hr)/(cents/lb)).Typical values of CI range are 15 to 50 for B757-200 and 5 to 25 for B737-800 [16].The cost J is related to F and t f by integrating Eq. ( 1):J = p f [F + t f × CI/36],(12)where J is in $, p f in lbs/$, F in lbs, and t f in seconds.The fuel price p f for computing the cost is set to 0.45 $/lb.Table 2 juxtaposes the CTAS-estimated fuel consumption and flight time for the nominal speed profile for B757-200 and for B737-800.To illustrate variation of costs with CI, the cost of the nominal speed profile, J nominal , at three distinct values of CI are computed from Eq. ( 12) and shown.Cost efficiency is measured as the cost difference between the nominal speed profile and the cost-optimal one,∆J = J nominal -J optimal ,(13)where smaller ∆J indicates better cost efficiency. +C. B757-200Figure 6 shows a plot of ∆J against CI for B757-200.With CI varying from 0 to 70, ∆J varies between $18 and $41, reaching the minimum at CI=25 and maximum at CI=70. Figure 6 also shows ∆J as a percentage of J nominal , denoted as ∆J (%).∆J (%) varies from 7% for CI=0 to 2%-3% of J nominal for CI values between 15 and 70.For a pointwise (at each value of x) comparison between cost-optimal speed profiles corresponding to four different values of CI and the nominal one, all plotted as CAS vs x, see Figure 8.The speed profiles become identical slightly before the aircraft descends to 10,000 ft.The equality of the speed profiles results from the optimal control becoming and remaining idle-thrust (this follows from the model), which in this case happens slightly above 10,000 ft.The data points of format (total fuel burn, total flight time) for B737-800, plotted in Figure 10 (analogous to Figure 7 for B757-200), show that the data point corresponding to the nominal speed profile appears to lie near the envelope of the data points corresponding to the cost-optimal speed profiles obtained by varying CI.For a pointwise (at each value of x) comparison between cost-optimal speed profiles corresponding to four CI values and the nominal one, all plotted as CAS vs x, see Figure 11.Compared to the results of B757-200 in Figure 8, the cost-optimal speed profiles for B737-800 more sensitive to ranging from 220 knots at some part of the CI = 0 profile to 320 knots at some part of the CI = 70 profile.The speed profiles become identical after the aircraft descends to approximately 9,000 ft, in contrast to the case of B757-200. +E. Flight Management System Computed Speed Profile for B757-200In addition to the sensitivity analysis above, an FMS-optimized speed profile is compared to the corresponding cost-optimal speed profile for its cost efficiency.The FMS speed profile is constructed from ASTOR without waypoint-specific speed restrictions.ASTOR, however, imposed a hard-coded speed restriction of 240 knots restriction at or below 10,000 ft.A value of 59 for the Cost Index is applied to select the descent speeds in the ASTOR's FMS.The aircraft type is B757-800.A corresponding cost-optimal speed profile is computed using the same Cost Index of 59. shows the speed profiles of both.The authors could not conclusively determine why the FMS speed profile appears to follow another speed restriction of 265 knots at x ∼ 65, which corresponds to the waypoint HYDRR (see Table 1).Table 3 juxtaposes the fuel consumption, flight time, and cost resulting from the cost-optimal and the ASTOR FMS speed profiles.The FMS speed profile consumes less fuel but spends more flight time, resulting in a cost within $5 of the optimal cost.Table 3 also gives the "Idle Fuel" of 333 lb, the fuel consumption of the idle-thrust segment in the last 20 nmi.It turns out that the idle fuel takes up more than 30% of the total fuel.This is in contrast to the assumption of zero fuel consumption during an idle-thrust segment, a common assumption made in numerous previous works [25,28].Such an assumption can lead to noticeable errors in the computation of cost-optimal trajectories.For the range of CI studied, cost-optimal speed profile reduces up to $41 and $40 of the nominal speed profile's cost for B757-200 and B737-800, respectively.The maximum reduction of J ($) occurs at CI=70 for B757-200 and CI=0 for B737-800.Variation of ∆J (%) across values of CI is mild (2%-7% for B757-200 and 0%-10% for B737-800), indicating that cost efficiency of the nonimal speed profile is fairly insensitive to the Cost Index.Figure 12 +B. Cost Efficiency with Speed BrakesThe cost of the nominal speed profile for B737-800 gets very close to that of the corresponding cost-optimal speed profile, in particular, $3 at CI=40.The achievement of such apparent nearoptimality is likely a result of an additional control parameter, the speed brakes, feasible for flying the nominal and FMS speed profiles, but not included in the model for computing the cost-optimal speed profile.CTAS thrust and drag calculations show that, for keeping a B737-800 on the CAS deceleration speed and altitude profiles simultaneously, negative thrust is required.Negative thrust corresponds to the use of speed brakes, considered undesirable by many pilots, but common in arrival operations.Although speed brakes cause additional noise and vibrations, proper use of speed brakes can improve cost efficiency by attaining a favorable speed sooner.On the other hand, the computation of a cost-optimal speed profile in this work does not assume non-negative thrust defined in Eq. ( 5).Therefore, solutions that further reduce the cost via speed brake are not explored.A possible direction for future research would be to investigate additional gain of cost efficiency that can be achieved by extending the solution space to include speed brake usage. +C. Systemic Cost OptimalityATC's approach to managing arrival traffic will shift from routine vectoring to precision operations in upcoming years.STM Under precision operations, each arrival flight follows a specific route all the way to a runway with a speed profile issued by ATC clearances.This approach motivates a possible class of realistic optimization problems for finding cost-efficient speed profiles.For a convenient view of these optimization problems, consider for the moment the following two hypothetical approaches to selecting speed profiles:• each aircraft is assigned a speed profile customized to that aircrafts' and airlines' parameters and preferences;• all aircraft on the same route are to follow the same speed profile.The first of these, highly idealized, would achieve the minimal cost possible for the entire set of aircraft, but appears operationally infeasible.This is because, for instance, allowing each aircraft its preferred speed profile would make it difficult for ATC to maintain the required separation between pairs of flights, even in an arrival operation with relatively few flights.The second approach, less cost-optimal systematically than the first, mitigates ATC's burden in ordering and spacing aircraft, but can still lead to loss of separation near points where two routes merge.Nonetheless, it serves as the basis for the operationally feasible compromise achieved by TSAS: to start with a single nominal speed profile for all aircraft and then modify it for each aircraft as necessary to achieve the required separation.The speed modification may either increase or decrease the cost efficiency of a flight.For multi-aircraft arrival operations, one can pursue an optimization problem of finding a nominal speed profile for all arrival flights on the same route, over a period of time, so as to minimize the systemic cost.Anticipated traffic scenarios as well as TSAS's speed modification logic affect the outcome of such optimization.This problem and an extended problem for arrival operations with multiple arrival routes converging to the same runway (where one would seek a nominal speed profile for each route), appear good directions for future research.Xfe = flat-earth coordinate along a constant latitude Y fe = flat-earth coordinate along a constant longitude c f = thrust-specific fuel consumption f0 = fuel consumption rate at minimum thrust (idle thrust) f (t) = fuel consumption rate g = gravity of earth h = aircraft's altitude ḣ = aircraft's altitude rate p f = fuel price r = ratio of time cost and fuel cost r = ratio of time cost and fuel cost per unit mass per unit fuel price t = time t f = final time w = wind along the route (positive for tail wind) x = path distance coordinate ẋ = path distance rate x f = final ground path distance x i = initial ground path distance γa = aerodynamic flight-path angle γi = inertial flight-path angle ρ = air density I. Introduction +Fig. 11Fig. 1 Part of the GEELA arrival route to the Phoenix Airport. +Fig. 22Fig.2The nominal speed profile. +Fig. 33Fig.3 The flown ASTOR altitude profile and the modeled altitude profile. +While CTAS's clean configuration fuel models are high-fidelity, CTAS does not have a highfidelity flap model.An accurate flap model tends to make V mc change abruptly at the speeds when flaps are deployed.In this work, flaps were treated approximately by extending the clean configuration's drag polar below the minimum-drag speed.This approximation is a continuous function and tends to underestimate the drag and hence the fuel rate, but should affect only the last 10 nmi of the trajectory, where the flexibility of selecting the speeds diminishes. +Figure 5 Fig. 555Fig. 5 Test matrix for the sensitivity analysis. +Fig. 66Fig. 6 Cost efficiency for B757-200. +Figure 77Figure 7 is a scatter plot of the data points of format (total fuel consumption, total flight +Fig. 77Fig. 7 Fuel and time the cost-optimal speed profiles for B757-200. +Fig. 8 Figure 989Fig.8Selected cost-optimal speed profiles for B757-200. +Fig. 99Fig.9Cost efficiency for B737-800. +Fig. 10 Fuel10Fig. 10 Fuel and time of the cost-optimal speed profiles for B737-800. +Fig. 1111Fig.11Selected cost-optimal speed profiles for B737-800. +Fig. 1212Fig.12 Comparison of an FMS-computed speed profile to a cost-optimal speed profile (B757-200, CI=59). +Table 11Speed restrictions along the GEELA routeSpeedPath DistanceWaypoint(CAS knots)(nmi)MOHAK280-101RKDAM280-67HYDRR265-44GEELA250-31PUNNT230-24TEICH210-14ILIKE180-4JAMIL1800 +The altitude profile contains a sequence of fixed-FPA segments interconnected by transient segments that ensure continuity of the FPA.A fixed change rate of ±1 • per nmi is used to model these transient segments, i.e.,dγ i dx= ±1 degrees/nmi (for transient segments.) +Table 22Fuel consumption, time, and cost of the nominal speed profileTypeFuel (lbs)Time (sec)J nominal (CI=10) ($)J nominal (CI=40) ($)J nominal (CI=70) ($)B757-2001002595102614571150B737-8009015499801412 +Table 33Fuel and time of the speed profiles (B757-200, CI=59)Fuel Time Cost Idle FuelTrajectory(lb) (sec) ($)(lb)Cost-Optimal 1014 1082 1254333ASTOR FMS 979 1110 1259VI. DiscussionA. Sensitivity of Cost Efficiency + + + +This study evaluates the cost efficiency of an aircraft arrival speed profile by comparing it to an exactly computed cost-optimal speed profile.The cost of flying a speed profile is modeled as a simple direct operating cost.The computation is performed for an actual route with altitude restrictions along the route.A high-fidelity fuel model for two Boeing aircraft types, B757-200 and B737-800, is used in fuel computation.The results of a sensitivity analysis in Section V show the nominal speed profile is fairly cost-efficient for both aircraft types with Cost Index values between 0 and 70.The cost-optimal speed profile reduces by 2%-7% of the nominal speed profile's cost for B757-200 and by 0%-10% for B737-800.This indcates that the speed profile's cost efficiency is fairly insensitive to the Cost Index, especially for B757-200.The near-optimality for B737-800 is likely due to the use of speed brakes, a control parameter not considered in the solution to the cost-optimal speed profile.Comparison of a speed profile optimized by a Flight Management System (FMS) to a corresponding cost-optimal speed profile for a specific Cost Index of 59 show that the FMS speed profile is within $5 of the optimal cost and, therefore, is fairly cost-optimal.The methodology presented here can be used for assessing the design of a speed profile for an arrival route, the efficency of which is expected to add benefits to managing arrival traffic with tools such as the Terminal Sequencing and Spacing (TSAS) system. + + + + + + + The Efficient Descent Advisor: Technology Validation and Transition + + RichardCoppenbarger + + + MiwaHayashi + + + GauravNagle + + + DouglasSweet + + + RenanSalcido + + 10.2514/6.2012-5611 + + + 12th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference and 14th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference + + American Institute of Aeronautics and Astronautics + 2012-5611, Sept. 2012 + + + Coppenbarger, R. A., Nagle, G., Sweet, D., and Hayashi, M., "The Efficient Descent Advisor: Technol- ogy Validation and Transition," Proceedings of the AIAA Aircraft Technology, Integration, and Opera- tions Conference, AIAA-2012-5611, Sept. 2012. doi:10.2514/6.2012-5611 + + + + + Impact of Future Time-Based Operations on Situation Awareness of Air Traffic Controllers + + EOprins + + + DZwaaf + + + FEriksson + + + KVan De Merwe + + + RRoe + + + + 8th USA/Europe Air Traffic Management R&D Seminar + + 2009 + + + Oprins, E., Zwaaf, D., Eriksson, F., van de Merwe, K., and Roe, R., "Impact of Future Time-Based Operations on Situation Awareness of Air Traffic Controllers," 8th USA/Europe Air Traffic Management R&D Seminar , 2009. + + + + + Design and Operational Evaluation of the Traffic Management Advisor at the Fort Worth Air Route Traffic Control Center + + HNSwenson + + + THoang + + + SEngelland + + + DVincent + + + TSanders + + + BSanford + + + KHeere + + + + Proceedings of the 1st USA + the 1st USA + + June 1997 + + + Swenson, H. N., Hoang, T., Engelland, S., Vincent, D., Sanders, T., Sanford, B., and Heere, K., "Design and Operational Evaluation of the Traffic Management Advisor at the Fort Worth Air Route Traffic Control Center," Proceedings of the 1st USA/Europe Air Traffic Management R&D Seminar , June 1997. + + + + + Optimized Profile Descent Arrivals at Los Angeles International Airport + + J-PClarke + + + JBrooks + + + GNagle + + + AScacchioli + + + WWhite + + + SRLiu + + 10.2514/1.c031529 + + + Journal of Aircraft + Journal of Aircraft + 0021-8669 + 1533-3868 + + 50 + 2 + + 2013 + American Institute of Aeronautics and Astronautics (AIAA) + + + Clarke, J.-P., Brooks, J., Nagle, G., Scacchioli, A., White, W., and Liu, S., "Optimized Profile Descent Arrivals at Los Angeles International Airport," Journal of Aircraft, Vol. 50, No. 2, 2013, pp. 360-369. doi:10.2514/1.C031529 + + + + + Development, Design and Flight Test Evaluation of Continuous Descent Approach Procedure in FIR Zagreb + + DorisNovak + + + TinoBucak + + + TomislavRadišić + + 10.7307/ptt.v21i5.247 + + + PROMET - Traffic&Transportation + PROMET + 0353-5320 + 1848-4069 + + 21 + 5 + + Sept. 2009 + Faculty of Transport and Traffic Sciences + + + Novak, D., Bucak, T., and Radisić, T., "Development, Design and Flight Test Evaluation of Continuous Descent Approach Procedure in FIR Zagreb," PROMET -Traffic & Transportation, Vol. 21, No. 5, Sept. 2009, pp. 319-329. doi:10.7307/ptt.v21i5.247 + + + + + NASA's ATM Technology Demonstration-1: Moving NextGen Arrival Concepts from the Laboratory to the Operational NAS + + HSwenson + + + IiiRobinson + + + JE + + + + Journal of Air Traffic Control + + 55 + 2 + 2013 + + + Swenson, H., Robinson III, J. E., et al., "NASA's ATM Technology Demonstration-1: Moving NextGen Arrival Concepts from the Laboratory to the Operational NAS," Journal of Air Traffic Control , Vol. 55, No. 2, 2013. + + + + + AMAN Status Review + + NHasevoets + + + PConroy + + + + + Eurocontrol + + 1 + 2010. 2010 + + + Hasevoets, N. and Conroy, P., "AMAN Status Review 2010," Eurocontrol, Edition number 0.1 , Vol. 17, 2010, https://www.eurocontrol.int/sites/default/files/article/content/documents/ nm/fasti-aman-status-review-2010.pdf. + + + + + Figure 2. Measuring human sensitivity to image statistics. + 10.7554/elife.03722.007 + + + + Federal Aviation Administration (FAA) + + eLife Sciences Publications, Ltd + June 2005 + 52 + + + United States Standard for Required Navigation Performance (RNP) Approach Procedures with Special Aircraft and Aircrew Authorization Required (SAAAR), Order 8260.52, Federal Aviation Administration (FAA), June 2005, https://www.faa.gov/documentLibrary/media/Order/ND/8260_52.pdf. + + + + + Evaluation of the terminal area precision scheduling and spacing system for near-term NAS application + + JThipphavong + + + LMartin + + + HSwenson + + + PLin + + + JNguyen + + 10.1201/b12321-12 + + + Advances in Human Aspects of Aviation + + CRC Press + 2012 + + + + Thipphavong, J., Martin, L., Swenson, H., Lin, P., and Nguyen, J., "Evaluation of the Terminal Area Precision Scheduling and Spacing System for Near-Term NAS Application," 4th International Confer- ence on Applied Human Factors and Ergonomics (AHFE), 2012. + + + + + Tactical Scheduling for Precision Air Traffic Operations: Past Research and Current Problems + + DRIsaacson + + + AVSadovsky + + + DDavis + + 10.2514/1.i010119 + + + Journal of Aerospace Information Systems + Journal of Aerospace Information Systems + 2327-3097 + + 11 + 4 + + 2014 + American Institute of Aeronautics and Astronautics (AIAA) + + + Isaacson, D. R., Sadovsky, A. V., and Davis, D., "Tactical Scheduling for Precision Air Traffic Opera- tions: Past Research and Current Problems," Journal of Aerospace Information Systems, Vol. 11, 2014, pp. 234-257. doi:10.2514/1.I010119 + + + + + Enabling Performance-Based Navigation Arrivals: Development and Simulation Testing of the Terminal Sequencing and Spacing System + + JohnERobinson + + + JaneThipphavong + + + WilliamCJohnson + + 10.2514/atcq.23.1.5 + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 23 + 1 + + 2015 + American Institute of Aeronautics and Astronautics (AIAA) + + + Robinson, J. E., Thipphavong, J., and Johnson, W. C., "Enabling Performance-Based Navigation Ar- rivals: Development and Simulation Testing of the Terminal Sequencing and Spacing System," Air Traffic Control Quarterly, Vol. 23, No. 1, 2015, pp. 5-27. + + + + + Evaluation of the Terminal Sequencing and Spacing system for Performance-Based Navigation arrivals + + JaneThipphavong + + + JaewooJung + + + HarrySwenson + + + LynneMartin + + + MelodyLin + + + JimmyNguyen + + 10.1109/dasc.2013.6712503 + + + 2013 IEEE/AIAA 32nd Digital Avionics Systems Conference (DASC) + + IEEE + 2013. 2013 + + + + Thipphavong, J., Jung, J., Swenson, H., Martin, L., Lin, M., and Nguyen, J., "Evaluation of the terminal sequencing and spacing system for performance-based navigation arrivals," Digital Avionics Systems Conference (DASC), 2013 IEEE/AIAA 32nd , IEEE, 2013, pp. 1A2-1. doi:10.1109/DASC.2013.6712503 + + + + + The Dynamic Planner: the Sequencer, Scheduler, and Runway Allocator for Air Traffic Control Automation, NASA/TM + + GLWong + + + 2000-209586, 2000 + + + Wong, G. L., The Dynamic Planner: the Sequencer, Scheduler, and Runway Allocator for Air Traffic Control Automation, NASA/TM-2000-209586, 2000. + + + + + Enhancing the Traffic Management Advisor's Schedule by Time Advance + + MinghongGWu + + + HarrySwenson + + 10.2514/6.2015-1326 + + + AIAA Guidance, Navigation, and Control Conference + + American Institute of Aeronautics and Astronautics + Jan. 2015 + + + Wu, M. G. and Swenson, H. N., "Enhancing the Traffic Management Advisor's Schedule by Time Advance," AIAA Guidance, Navigation, and Control (GNC) Conference, Jan. 2015. doi:10.2514/6.2015-1326 + + + + + Cost Efficiency of an Aircraft Arrival Speed Profile + + MinghongGWu + + + AlexanderVSadovsky + + 10.2514/1.c034445 + NASA/TM-2015-218734 + + + Journal of Aircraft + Journal of Aircraft + 0021-8669 + 1533-3868 + + 56 + 1 + + April 2015 + American Institute of Aeronautics and Astronautics (AIAA) + + + Wu, M. G. and Sadovsky, A. M., "Minimum-Cost Aircraft Descent Trajectories with a Constrained Altitude Profile," NASA/TM-2015-218734, April 2015. + + + + + Fuel Conservation Strategies: Cost Index Explained + + BRoberson + + + RRoot + + + DAdam + + + + + Boeing Aero Quarterly + + 2 + + 2007, 2007 + + + Roberson, B., Root, R., and Adam, D., "Fuel Conservation Strategies: Cost Index Explained," Boeing Aero Quarterly, Vol. 2, No. 2007, 2007, pp. 26-45, http://www.boeing.com/commercial/ aeromagazine/articles/qtr_02_10/pdfs/AERO_FuelConsSeries.pdf. + + + + + Drew, Prof. George Charles, (10 Dec. 1911–19 Dec. 1997), London University Professor of Psychology, University College, 1958–79, then Professor Emeritus; Honorary Fellow, 1979 + + JMcelhatton + + + CDrew + + + PBuchanan + + 10.1093/ww/9780199540884.013.u178104 + + + NASA ASRS Directline + + 10 + + dec 1998 + Oxford University Press + + + McElhatton, J., Drew, C., and Buchanan, P., "Crossing-Restriction Altitude Deviations on SIDs and STARs," NASA ASRS Directline, Vol. 10, dec 1998, pp. 10-15. + + + + + ASTOR: An Avionics Concept Test Bed in a Distributed Networked Synthetic Airspace Environment + + Ding-JenLiu + + + WilliamChung + + 10.2514/6.2004-5259 + + + AIAA Modeling and Simulation Technologies Conference and Exhibit + + American Institute of Aeronautics and Astronautics + Aug. 2004 + + + Liu, D.-J. and Chung, W., "ASTOR: An Avionics Concept Test Bed in a Distributed Networked Syn- thetic Airspace Environment," AIAA Modeling and Simulation Technologies Conference and Exhibit, Aug. 2004. doi:10.2514/6.2004-5259 + + + + + A High-Performance Simulated On-Board Avionics Architecture to Support Traffic Operations Research + + MichaelPalmer + + + MarkBallin + + 10.2514/6.2003-5452 + + + AIAA Modeling and Simulation Technologies Conference and Exhibit + + American Institute of Aeronautics and Astronautics + Aug. 2003 + + + Palmer, M. and Ballin, M., "A High-Performance Simulated On-Board Avionics Architecture to Support Traffic Operations Research," AIAA Modeling and Simulation Technologies Conference and Exhibit, Aug. 2003. doi:10.2514/6.2003-5452 + + + + + A Traffic Information Service-Broadcast model for mixed-equipage Aircraft Simulation + + GoutamSatapathy + + + Jensan Chen + + + DevendraTolani + + + JamesLSturdy + + + JamesTHenion + + + GregoryKubat + + 10.1109/icnsurv.2010.5503264 + + + 2010 Integrated Communications, Navigation, and Surveillance Conference Proceedings + + IEEE + 2010 + + + + Satapathy, G., Chen, J., Tolani, D., Sturdy, J. L., Henion, J. T., and Kubat, G., "A Traffic Infor- mation Service-Broadcast Model for Mixed-Equipage Aircraft Simulation," Integrated Communications Navigation and Surveillance Conference (ICNS), IEEE, 2010, pp. E3-1. doi:10.1109/ICNSURV.2010.5503264 + + + + + Simulation Results for Airborne Precision Spacing along Continuous Descent Arrivals + + BryanBarmore + + + WilliamCapron + + + TerenceAbbott + + + BrianBaxley + + 10.2514/6.2008-8931 + + + The 26th Congress of ICAS and 8th AIAA ATIO + + American Institute of Aeronautics and Astronautics + Sept. 2008 + 22 + 27 + + + The 26th Congress of International Council of the Aeronautical Sciences + Barmore, B. E., Abbott, T. S., Capron, W. R., and Baxley, B. T., "Simulation Results for Airborne Precision Spacing along Continuous Descent Arrivals," The 26th Congress of International Council of the Aeronautical Sciences (ICAS), Vol. 22, Sept. 2008, pp. 27. doi:10.2514/6.2008-8931 + + + + + Analysis of optimal aircraft cruise with fixed arrival time including wind effects + + AntonioFranco + 0000-0001-7272-3388 + + + DamiánRivas + + 10.1016/j.ast.2013.10.005 + + + Aerospace Science and Technology + Aerospace Science and Technology + 1270-9638 + + 32 + 1 + + 2014 + Elsevier BV + + + Franco, A. and Rivas, D., "Analysis of Optimal Aircraft Cruise with Fixed Arrival Time Including Wind Effects," Aerospace Science and Technology, Vol. 32, No. 1, 2014, pp. 212-222. doi:10.1016/j.ast.2013.10.005 + + + + + Four-dimensional fuel-optimal guidance in the presence of winds + + AbhijitChakravarty + + 10.2514/3.19929 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 8 + 1 + + 1985 + American Institute of Aeronautics and Astronautics (AIAA) + + + Chakravarty, A., "Four-Dimensional Fuel-Optimal Guidance in the Presence of Winds," Journal of Guidance, Control, and Dynamics, Vol. 8, No. 1, 1985, pp. 16-22. doi:10.2514/3.19929 + + + + + 4-D Aircraft Flight Path Management in Real Time + + AbhijitChakravarty + + + JurisVagners + + 10.23919/acc.1983.4788220 + + + 1983 American Control Conference + + IEEE + 1983. 1983 + + + + American Control Conference + Chakravarty, A. and Vagners, J., "4-D Aircraft Flight Path Management in Real Time," American Control Conference, 1983 , IEEE, 1983, pp. 794-795. + + + + + Fuel-optimal aircraft trajectories with fixed arrival times + + JamesWBurrows + + 10.2514/3.19796 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 6 + 1 + + 1983 + American Institute of Aeronautics and Astronautics (AIAA) + + + Burrows, J. W., "Fuel-Optimal Aircraft Trajectories with Fixed Arrival Times," Journal of Guidance, Control, and Dynamics, Vol. 6, No. 1, 1983, pp. 14-19. doi:10.2514/3.19796 + + + + + Constrained Optimum Trajectories with Specified Range + + HeinzErzberger + + + HomerLee + + 10.2514/3.55950 + + + Journal of Guidance and Control + Journal of Guidance and Control + 0162-3192 + + 3 + 1 + + 1980 + American Institute of Aeronautics and Astronautics (AIAA) + + + Erzberger, H. and Lee, H., "Constrained Optimum Trajectories with Specified Range," Journal of Guidance, Control, and Dynamics, Vol. 3, No. 1, 1980, pp. 78-85. doi:10.2514/3.55950 + + + + + Optimization of Multiphase Aircraft Trajectories Using Hybrid Optimal Control + + AntonioFranco + + + DamiánRivas + + 10.2514/1.g000688 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 38 + 3 + + 2015 + American Institute of Aeronautics and Astronautics (AIAA) + + + Franco, A. and Rivas, D., "Optimization of Multiphase Aircraft Trajectories Using Hybrid Optimal Control," Journal of Guidance, Control, and Dynamics, 2015, pp. 1-16. doi:10.2514/1.G000688 + + + + + Optimal Control Based Vertical Trajectory Determination for Continuous Descent Arrival Procedures + + SangGyunPark + + + John-PaulClarke + + 10.2514/1.c032967 + + + Journal of Aircraft + Journal of Aircraft + 0021-8669 + 1533-3868 + + 52 + 5 + + 2015 + American Institute of Aeronautics and Astronautics (AIAA) + + + Park, S. G. and Clarke, J.-P., "Optimal Control Based Vertical Trajectory Determination for Continuous Descent Arrival Procedures," Journal of Aircraft, Vol. 52, No. 5, 2015, pp. 1469-1480. doi:10.2514/1.C032967 + + + + + Aircraft Shortest and Fastest Continuous Descent Approach Development + + SKhardi + + 10.2514/1.c031775 + + + Journal of Aircraft + Journal of Aircraft + 0021-8669 + 1533-3868 + + 49 + 6 + + 2012 + American Institute of Aeronautics and Astronautics (AIAA) + + + Khardi, S., "Aircraft Shortest and Fastest Continuous Descent Approach Development," Journal of Aircraft, Vol. 49, No. 6, 2012, pp. 1931-1939. doi:10.2514/1.C031775 + + + + + Fixed RTA Fuel Optimal Profile Descent Based on Analysis of Trajectory Performance Bound + + SGPark + + + J.-PClarke + + 10.1109/DASC.2012.6382316 + + + Digital Avionics Systems Conference (DASC) + + IEEE + 2012 + + + + 2012 IEEE/AIAA 31st + Park, S. G. and Clarke, J.-P., "Fixed RTA Fuel Optimal Profile Descent Based on Analysis of Trajectory Performance Bound," Digital Avionics Systems Conference (DASC), 2012 IEEE/AIAA 31st, IEEE, 2012, pp. 3D3-1. doi:10.1109/DASC.2012.6382316 + + + + + Rule derivation for arrival aircraft sequencing + + AdrianaAndreeva-Mori + + + ShinjiSuzuki + + + EriItoh + + 10.1016/j.ast.2013.08.004 + + + Aerospace Science and Technology + Aerospace Science and Technology + 1270-9638 + + 30 + 1 + + 2011 + Elsevier BV + + + Andreeva-Mori, A., Suzuki, S., and Itoh, E., "Scheduling of Arrival Aircraft Based on Minimum Fuel Burn Descents," ASEAN Eng J , Vol. 1, No. 1, 2011. + + + + + Introduction to Aircraft Performance, Selection, and Design + + FJHale + + + 1984 + Wiley + New York + + + Hale, F. J., Introduction to Aircraft Performance, Selection, and Design, Wiley New York, 1984. + + + + + Applied Optimal Control + + ArthurEBryson + + + Yu-ChiHo + + 10.1201/9781315137667 + + + Applied Optimal Control: Optimization, Estimation and Control + + 1975 + Routledge + + + Bryson, Jr., A. E. and Ho, Y.-C., Applied Optimal Control: Optimization, Estimation and Control , CRC Press, 1975. + + + + + Design of Center-TRACON Automation System + + HErzberger + + + TJDavis + + + SMGreen + + + + Machine Intelligence in Air Traffic Management + + ABenoit + + + AgardCp-538 + + + Oct. 1993 + 12 + + + + Erzberger, H., Davis, T. J., and Green, S. M., "Design of Center-TRACON Automation System," Machine Intelligence in Air Traffic Management, edited by A. Benoit, AGARD CP-538, Oct. 1993, pp. 11.1-11.12. + + + + + + + MatlabThe + + + Software + + + 2008 + The MathWorks Inc + Natick + + + Version 7.7.0.471 (R2008b + The MATLAB Software, Version 7.7.0.471 (R2008b), The MathWorks Inc., Natick, 2008. + + + + + Trajectory Synthesis for Air Traffic Automation + + RhondaSlattery + + + YiyuanZhao + + 10.2514/2.4056 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 20 + 2 + + March 1997 + American Institute of Aeronautics and Astronautics (AIAA) + + + Slattery, R. and Zhao, Y., "Trajectory Synthesis for Air Traffic Automation," Journal of Guidance, Control and Dynamics, Vol. 20, No. 2, March 1997, pp. 232-238. doi:10.2514/2.4056 + + + + + + diff --git a/file801.txt b/file801.txt new file mode 100644 index 0000000000000000000000000000000000000000..f7115138deb2f64b7e463292cd9729de14e964cb --- /dev/null +++ b/file801.txt @@ -0,0 +1,837 @@ + + + + +IntroductionEfficient arrival operations in the terminal airspace have been the focus of air traffic management research for years.In recent years, significant efforts have been made to integrate existing and new technologies as well as new procedures to manage arrival flights in a trajectory-based, precision-scheduling fashion [1,2].Arrival flights are scheduled based on specific arrival routes defined all the way from the top-of-descent to the runway threshold [3,4].With this trajectory-based, precision-scheduling approach, air traffic controllers use speed clearances to guide flights to meet the schedule and maintain separation while reserving vectoring clearances for exceptions only.This speed-centric approach is in contrast to the conventional approach, in which aircraft routinely follow air traffic controllers' vectoring clearances as soon as they enter the terminal airspace.An arrival procedure that supports trajectory-based, precision-scheduling requires a flight to traverse a specific route consisting of a ordered sequence of waypoints 1 .Such a procedure also imposes altitude restrictions at certain waypoints.The altitude restrictions provide vertical separation for different traffic flows, keep specific arrival flows in the same controller's sector, and account for terrain constraints [6].An altitude restriction can have an upper bound, a lower bound, or both.While the upper and lower bounds can be made different to accommodate aircraft types' varying performance envelopes in the en route airspace, ATC can rarely afford such flexibility in a congested terminal airspace with complex traffic flows.Therefore, altitude restrictions in the terminal airspace as well as some transition airspace always have identical upper and lower bounds.When such altitude restrictions are closely spaced, arrival flights are observed to fly a sequence of constant flight-path angle (FPA) segments.While an arrival flight's altitude profile is much constrained by considerations related to traffic flow, terrain, and safety, its speed profile leaves room for optimization.Cost-effective speed strategies are of common interest to airlines and ATC, and insight of such can benefit both parties.Specifically, the insight can be used to improve scheduling efficiencies in the ground automation tools for ATC.For example, the Terminal Sequencing and Spacing (TSS) system in the United States requires route-specific nominal speed profiles as input parameters for computing the arrival flights' schedule.Each aircraft is scheduled by TSS's scheduler at route merge points such as the meter fix2 and the runway threshold [1,3].Route-specific nominal speed profiles are used for ordering aircraft on the schedule.TSS then computes for each aircraft a speed profile that would meet the schedule, perturbing the speed profile from the nominal one to meet separation restrictions at the merge points if necessary [8].While the nominal speed profiles are critical to the efficiency of the schedule, little attention has been paid to their selection.They have been selected only to ensure flyability without any check for systemic fuel or cost efficiency.This lack of consideration motivated the study in this work.Trajectories that minimize fuel or cost have been studied extensively.A fairly comprehensive review of work in this area can be found in [9].Early work considered optimization of trajectories in climb, cruise, and descent phases combined [10][11][12][13].Such analysis can be formulated as a multi-phase optimization problem [14].Some recent work analyzed descent only or cruise/descent trajectories [15][16][17][18] and investigated trade-off between fuel and flight time [19].The optimal descent trajectory obtained from these work typically utilizes idle thrust engine control.However, none of these work considered altitude restrictions during the descent.Therefore, the optimal trajectory may violate the altitude restrictions in a trajectory-based arrival procedure and cannot be executed in operations without modification.This paper derives the minimum-cost speed profile for an arrival flight following a specific route with a constrained altitude profile.The direct operating cost [20] serves as the objective functional to be minimized.The total engine thrust, referred to as thrust, is the control variable.The central result consists of an analytical formula for the speed profile.It reveals that an optimal speed profile generally gets to a minimum-cost speed as quickly as possible.It then stays on the minimum-cost speed as long as possible, until it has to leave to reach the final speed.The minimum-cost speed profile is computed for Boeing 737-800 along an actual route, and its sensitivity to wind and airlines' business objectives is investigated.The rest of the paper is organized as follows: Section 2 gives additional background information and discusses the modeling assumptions; Section 3 formulates the optimal control model; Section 4 derives the analytical formula for the speed profile; Section 5 applies the formula to an actual arrival route and compares the minimum-cost trajectory to that of a simulated flight; Section 6 summarizes the findings and discusses the potential applications of the minimum-cost speed profile. +Background +Minimum-Cost DescentAn arrival flight transitions from cruise to descent at a distance of 100 to 150 nmi from the destination airport.In the absence of wind and altitude restrictions, the most fuelefficient operation for an arrival flight is to maintain at the best-range speed in cruise until the top-of-descent point for an idle-thrust descent at the aircraft's minimum-drag speed [20].However, airlines can save time-related costs by flying an aircraft faster to the destination airport.If time saving is the sole objective, the ideal cruise and descent speeds would be the maximum speeds allowable by the aircraft performance envelope.In actual operations, airlines consider both fuel and time costs and the minimum-cost speeds lie between these two extremes.The relative importance of expediency to fuel is modeled by the Cost Index, a parameter selected by airlines during flight planning.The Cost Index varies from route to route and from aircraft type to aircraft type [21].A zero Cost Index represents the case in which flight time does not contribute to the cost.A large Cost Index represents that expediency is more important than fuel. +Arrival ProcedureIn the United States, arrival flights to major airports follow published arrival procedures called the Standard Terminal Arrival Route (STAR).While current STARs do not connect the route all the way to the runway threshold, they are expected to be extended to the runway in the near future to support trajectory-based operations [1].For the rest of the paper, it is assumed that a specific route is defined for an arrival flight all the way to the runway threshold.The aircraft descent trajectory considered in this work begins from the first altitude restriction along the arrival route and ends at the initial approach fix that is usually 5 to 10 nmi from the airport.The total path distance of the trajectory is between 50 and 100 nmi.3 An Optimal Control Model for Aircraft Descent +Modeling AssumptionsThe following assumptions are made throughout this section, grouped here by the different aspects of the operation.1. Route geometry: The descent trajectory is modeled as a continuous curve in 3dimensional airspace, parameterized by the ground path distance. +Aircraft kinematics and kinetics:(a) The aircraft is modeled as a point mass.(b) The weight of the aircraft is treated as a constant.(c) The thrust acts along the direction of the flight path.(d) Fuel flow is linear with thrust. +Flight conditions:(a) The wind velocity is constant in time.Crosswind and vertical wind are small compared to the airspeed and, therefore, is neglected.(b) Both the inertial and aerodynamic flight path angles are small enough that their cosines are assumed to equal 1. +Model FormulationLet x denote the ground path distance of the aircraft's position along the coordinate defined by the horizontal path.By assumption (1), the altitude h of the aircraft, the air density ρ, and inertial flight path angle, γ i , of the aircraft can each be represented as a function of x alone,h = h(x),(1)ρ = ρ(x), tan γ i = dh dx .The last equation is based on the sign convention that γ i is negative for descent.Let V = V (t) denote the speed of the aircraft with respect to the air, i.e., the true airspeed.In the following sections, speed refers to the true airspeed unless noted otherwise.The wind along the route is denoted byw = w(x)(2)and, by assumption (3a), is independent of time.By assumption (3b), the difference between the horizontal component of the true airspeed and the true airspeed itself is ignored:V horiz ∼ V w horiz ∼ wIn the following state equations, T denotes thrust, D drag, γ a the aerodynamic flight path angle, W the aircraft's weight, g the gravity of Earth, and the dot over a symbol denotes differentiation with respect to time:ẋ = V + w,(3a)V = g T -D W -sin γ a -(V + w) dw dx ,(3b)where D = D(x, V ) and x = x(t).The right-hand side of Eq. (3b) consists of forces from the thrust, drag, gravity, and the inertial force as a result of wind.Note the use of assumptions (2b) and (2c) in Eq. (3b).The FPAs γ a and γ i are related byḣ = V sin γ a = (V + w) sin γ i .(4)The use of γ a in the state equations has been preferred to that of γ i for brevity.The aircraft's total engine thrust is bounded,T min ≤ T ≤ T max ,with the bounds determined by the engine's performance envelope.The objective to be minimized is the direct operating cost, which is a sum of two terms corresponding to fuel and flight duration, respectively:J = t f 0      Fuel p f × c f × [T (t) -T min ] + f 0 + Time p f × r      dt.(5)Here, t f stands for the arrival time at the end of the route, p f the fuel price, c f the thrust-specific fuel consumption, f 0 the minimum-thrust fuel rate, and r a constant Cost Index [20].In what follows, it will be convenient to combine the drag, gravity, and wind gradient into the "effective drag per unit mass",D = g W (D + W sin γ a ) + (V + w) dw dx ,(6)and to normalize thrust to a "thrust per unit mass",T = g W T.(7)These definitions simplify the state equations toẋ = V + w,(8a)V = T -D(8b)The optimal control problem central to this paper is the problem with state variables x and V , control variable T , state equations ( 8), initial time t = 0, final time t f free, boundary conditionsx(0) = x i , V (0) = V i , x(t f ) = x f , V (t f ) = V f ,            (9) control constraints T min ≤ T ≤ T max ,(10)and the objective to minimize being the normalized cost per unit aircraft mass:J = t f 0 c f T (t) + r dt, where r = g W (-c f T min + f 0 + r) .(11)4 An Analytical Optimal Solution in Feedback Form +Comparison of Two Feasible State TrajectoriesThe optimal control problem ( 8), ( 9), ( 10), (11) has two state variables and one control variable which, furthermore, enters the state equations and the cost linearly.This special form of the problem allows one to find optimal control strategies using a technique based on Green's Theorem; see, e.g., [22].This technique consists of comparing two feasible (i.e., compliant with ( 8), ( 9), ( 10), but possibly suboptimal) state trajectories for performance.Let Γ 1 = (x 1 (t), V 1 (t)), 0 ≤ t ≤ t 1 f , and Γ 2 = (x 2 (t), V 2 (t)), 0 ≤ t ≤ t 2 f ,be two feasible state trajectories corresponding to control strategies T 1 (t) and T 2 (t) with generally different final times.The respective operating costs (11) of the two trajectories can be briefly writtenJ k = t f k 0 c f T k (t) + r dt, k = 1, 2.The direct operating cost can be written as a path integral, along the state trajectory, of a differential form P (x, V ) dx + M (x, V ) dV (derived explicitly in Appendix A), i.e.,J k = Γ k [P (x, V ) dx + M (x, V ) dV ] , k = 1, 2. (12)At constant airspeed, i.e. dV = 0, the quantity P has the interpretation of steady-state cost per unit ground path distance.The differenceJ 1 -J 2(13)is the difference in performance between the two state trajectories.Henceforth, assume that the two trajectories share no other states besides (x i , V i ) and (x f , V f ).This assumption, which entails no loss of generality but simplifies analysis, implies that Γ 1 and reverseoriented Γ 2 together constitute a closed contour [23], denoted here by Γ 1 ∪ -Γ 2 and, for definiteness, assumed oriented counter-clockwise.See Figure 1 for a notional illustration.Expression ( 13) can be written as a path integral along the aforementioned closed contour:Γ 1 ∪-Γ 2 (P dx + M dV ) .By Green's Theorem, the latter integral can be written as an integral over the area A enclosed and oriented by the contour:A ∂M ∂x - ∂P ∂V dx dV.(14)The two state trajectories can now be compared for performance by using the knowledgex V P f P i Г 2 Г 1 I = I(x, V) A Γ Γ Figure 1.Green's Theorem predicts J 1 < J 2 if I < 0 everywhere in the enclosed area; and J 1 > J 2 if I > 0 everywhere in the enclosed area.of the integrandI ≡ ∂M ∂x - ∂P∂V to examine the sign of the difference (13).In particular, if I is nonnegative in and on the boundary of A, then Γ 2 performs as well as does Γ 1 or better.Consequently, a comparison of two feasible state trajectories reduces to an examination of the sign of I in the appropriate regions of the state space.This examination will now be carried out.By the derivation in Appendix A,P = c f D + r V + w , M = c f ,(15)and, consequently,I = ∂c f ∂x - ∂ ∂V c f D + r V + w .For typical commercial aircraft and descent FPAs, c f does not vary substantially with the ground path distance.Therefore, the first term on the right hand side can be ignored without noticeable effects on the results, i.e.,I - ∂ ∂V c f D + r V + w . (16)For aircraft with typical drag formulas (described in detail below) and, within the range of airspeeds considered, one finds that the zero-level set [I = 0] is the graph of a function V mc (x) called the minimum-cost speed.The subscript mc stands for minimum cost because∂P ∂V V =Vmc = 0and∂ 2 P ∂V 2 > 0,(17)which will be shown at the end of this section.It follows thatI(x, V )    < 0 if V > V mc (x), = 0 if V = V mc (x), > 0 if V < V mc (x). (18)To conclude the section, we justify stipulation (17).A typical drag polar [24] for modeling drag isD = 1 2 ρV 2 SC D = 1 2 ρV 2 S C d0 + KC L 2 ,where C L denotes the lift coefficient, C D the drag coefficient, C d0 the zero-lift drag coefficient, K the lift drag constant, S the aircraft's wing area, and ρ = ρ(x) the air density.Replacing C L with its definition [24],L = W 1 + V 4 g 2 R 2 = 1 2 ρV 2 SC L ,where R is the instantaneous turn radius, the drag can be writtenD = 1 2 ρV 2 SC d0 + KW 2 1 2 ρV 2 S 1 + V 4 g 2 R 2 .(19)Equations ( 6), ( 16), (19) and the dependency of D on V together imply thatI ∼        - g 2W ρSC d0 - 2KW ρSgR 2 < 0, for a large V, 6KW gρSV 4 > 0, for a small V.(20)It can also be shown that∂ 2 P ∂V 2 = - ∂I ∂V = 24KW gρV 5 S > 0, if w = r = 0. (21)Results ( 20) and ( 21) together imply (18).Numerical evidence obtained for the case when w and r have realistic nonzero values suggests that the inequality in Eq. ( 21) always holds for the range of V considered, hence confirming (17) and (18). +Optimal State Trajectories and Control StrategiesThe following observation will be instrumental throughout this section: the curve in the state space traversed by a feasible state trajectory (x(t), V (t)) can be parameterized by x instead of by t; i.e., can be viewed as the graph of a function V = V (x).This follows from the fact that, by dividing Eq. (8b) by Eq. (8a), one obtainsdV dx = T -D V + w is finite.Therefore, in what follows, whenever convenient, a state trajectory (x(t), V (t)) will be treated as a function V = V (x) without further explanation.The following lemma is a direct consequence of the results of Section 4.1.The five parts of the lemma are illustrated, respectively, by the five panels of Figure 2. It will be convenient to use the control-theoretic terms of admissibility [25] and attainability: A control value T satisfying ( 10) is said to be admissible.This term will be applied as well to a control strategy with the value T (t) admissible at every t.A state trajectory corresponding to an admissible control strategy will be referred to as attainable.The last part of the following lemma addresses the situation when not all segments of [I = 0] are attainable trajectories.Lemma 1 Consider an initial state a = (x a , V a ) and a final state b = (x b , V b ), with x a < x b .For (a) through (d), consider two attainable state trajectories (x 1 (t), V 1 (t)) and (x 2 (t), V 2 (t)) that have the same initial state a and the same final state b: The optimal trajectory described in the last part, (e), of the Lemma is completely determined by the point of intersection (the gray dot in Fig. 2(e)) between a b and trajectory a 1 b 1 ; i.e., by the choice of that state on a b which minimizes the cost of getting from a to b.Since segment a b is closed and bounded, such a state necessarily exists.(a) If a is above [I = 0], b is on [I = 0],andV mc (x) ≤ V 1 (x) ≤ V 2 (x) for x a ≤ x ≤ x b , then V 1 (x) outperforms V 2 (x). (This is because the integrand I in the area A enclosed by Γ 1 ∪ -Γ 2 is negative.) (b) If a is below [I = 0], b is on [I = 0],andV 2 (x) ≤ V 1 (x) ≤ V mc (x) for x a ≤ x ≤ x b , then V 1 (x) outperforms V 2 (x). (c) If a is on [I = 0], b is above [I = 0],andV mc (x) ≤ V 1 (x) ≤ V 2 (x) for x a ≤ x ≤ x b , then V 1 (x) outperforms V 2 (x). (d) If a is on [I = 0], b is below [I = 0],andV 2 (x) ≤ V 1 (x) ≤ V mc (x) for x a ≤ x ≤ x b , then V 1 (x) outperforms V 2 (x).In summary, the optimal trajectory goes towards [I = 0] as rapidly as possible, as in Figure 2(a) and 2(b), and leaves [I = 0] as late as possible, as in Figure 2(c) and 2(d).This observation along with Eq. (8b) suggests that extreme values of T must be used in getting the optimal trajectory to or from [I = 0].That the optimal trajectory must, in fact, use T min or T max when it is not on the [I = 0] curve, can be proved using the Hamiltonian [26] of the problem,H = c f T + c i + λ x (V + w) + λ s T -D ,(22)where λ x and λ s are the respective costate variables of x and V .The optimal control minimizes (per the sign convention in [26]) the Hamiltonian at every point of the trajectory.The Hamiltonian H is linear in T and is minimized by:T = T min if ∂H ∂T > 0 T max if ∂H ∂T < 0. (23)If ∂H/∂T = 0, then the optimal control is on a singular arc [26].The curve [I = 0] turns out to be the singular arc of the Hamiltonian.In particular, for an optimal trajectory passing through a state in I < 0 and later intersecting [I = 0], as in Lemma 1(a), the optimal control at that state is T min .Similarly, for a state on an optimal trajectory passing through a state in I > 0 and later intersecting [I = 0], as in Lemma 1(b), the optimal control at that state is T max .This classification of optimal control strategies can also be carried out by more elementary means, without involving the Hamiltonian formalism, but expending more verbiage.This simpler approach to optimal control classification is presented in Appendix B. +Control on the Minimum-Cost Speed CurveThe control that keeps the aircraft on the minimum-cost speed curve, T mc , is computed here by setting the right-hand side of Eq. ( 16) to zero and using the state equations (8).For convenience, define Ĩ ≡ (V + w) 2 I.Since (V + w) is never zero in the range of airspeed and wind considered, the curves [I = 0] and [ Ĩ = 0] coincide.The following equality holds on [ Ĩ = 0]:d V d x Ĩ=0 = - ∂ Ĩ ∂V -1 ∂ Ĩ ∂x .(25)DefiningF ≡ c f D,(a)V mc (x)I < 0 I > 0 x V a b V 1 (x)(b)V mc (x) 24) and ( 16) into the right-hand side of Eq. ( 25), one obtainsI < 0 I > 0 x V a b V 1 (x) (c) V mc (x) I < 0 I > 0 x V a b V 1 (x) (d) V mc (x) I < 0 I > 0 x V a b V 1 (x) (e) V mc (x) I < 0 I > 0 x Vd V d x Ĩ=0 = (V + w) ∂ 2 F ∂V 2 -1 - ∂F ∂V dw dx + ∂F ∂x -(V + w) ∂ 2 F ∂x ∂V ,(26)From Eqs. ( 8) and ( 26), one obtainsT mc = D + ∂ 2 F ∂V 2 -1 - ∂F ∂V dw dx + ∂F ∂x -(V + w) ∂ 2 F ∂x ∂V . (27)With analytic expressions for the drag and the wind, the right-hand side of Eq. ( 27) can be evaluated for specific values of x and V .The thrust T mc on [I = 0] is given byT mc = W g T mc .The thrust T mc computed using Eq. ( 27) can turn out to violate the admissibility condition (10) on an interval of the x-axis, leading to the situation of Lemma 1(e).In this case, the optimal trajectory leaves the minimum-cost speed curve with an extremal value of thrust until it either rejoins the minimum-cost speed curve or reaches the final state. +V mc with Constant c f and without WindThe content of this section is a number of consequences derived from the above results for the special case when c f is assumed constant, and wind speed assumed is zero at all altitudes.Under the assumption of constant c f , Eq. ( 24) simplifies toĨ = -(V + w) 2 c f ∂ ∂V D + r/c f V + w = 0. (28)Furthermore, if the wind effect is negligible, thenĨ = -V 2 c f ∂ ∂V D + r/c f V = 0. (29)By using Eqs.( 6) and ( 19) to model the drag, Eq. ( 29) is solved analytically for V .The minimum-cost speed is found to beV mc (x) 2 = (W/S) C d0 ρ(x) α + α 2 + 12KC d0 . (30)Here α = sin γ i + r c f g captures the combined effect of the FPA and of the Cost Index on the minimum-cost speed, In particular, α increases with r and with γ i in a neighborhood of γ i = 0. Equality (30) has the following interpretations and implications * The minimum-cost speed V mc can be regarded as a modified best-range speed modulated by the factor α.For a flight segment which is level (i.e., with FPA zero) and has a zero Cost Index, the resulting special case α = 0 corresponds to an V mc equal to the best-range speed at an altitude specified by h(x) [24].* The minimum-cost speed V mc (x) decreases with altitude due to air density change.* The minimum-cost speed V mc (x) increases with α, which means V mc increases with γ i and with the Cost Index.5 Case Study: GEELA Arrival Route to the Phoenix Airport +Route Definition and Test ConditionsThis section investigates the sensitivity of the optimal speed profile to the wind condition and Cost Index, explores the effects of certain modeling assumptions on the fuel computation, and compares the computed optimal trajectory to a simulated flight trajectory.The specific example of an arrival procedure used as a basis for this section is a portion of the GEELA route from waypoint HYDRR to waypoint PHX08 (runway 8) of the Phoenix Airport. 3igure 3 shows the waypoints and altitude restrictions specified in this procedure.The coordinates X fe and Y fe in Fig. 3 increase Eastward and Northward, respectively.Note that the origin of the coordinates is arbitrary, since this figure is mainly for demonstrating the scale of an arrival route.This procedure is closely based on current operations but has additional waypoints to precisely guide the arrival flights all the way to the runway.This procedure has been used in simulations conducted to test arrival management technologies [3].The ground path distance of the entire route was constructed by connecting consecutive waypoints' positions and span roughly 54 nmi.The coordinate x, representing the ground path distance, starts negative and increases to 0 at the runway threshold.Turns were modeled as having a five nautical mile turn radius each.This approximate treatment of the turns decoupled the horizontal path from the speed.Lift was treated approximately as being equal to the aircraft weight all the time.These approximations have been used by trajectory generators of the ground automation systems [27].Figure 4 shows the modeled altitude h(x) of the GEELA route.The FPA is modeled as a constant on each segment between two consecutive waypoints, except varying near the waypoint so as to remain continuous.This change of FPA takes place after the aircraft has crossed the waypoint.A fixed change rate of ±1 • per nmi was used to model these transient segments.The air density was a function of the altitude computed from the standard atmosphere model.It was 0.778 kg 3 /m at the beginning of the descent and 1.187 kg 3 /m at the end.The initial airspeed at the waypoint HYDRR, V i , for the aircraft was chosen to be the calibrated airspeed (CAS) [24] of 260 knots (174.7 m/sec), required at HYDRR by the Boeing 737-800 aircraft type was modeled using the Base of Aircraft Database (BADA) 3.8 [28].Flap schedule was not considered and only the clean configuration was used.The minimum-cost speed curve, V mc (x) (see Fig. 2), was computed using Eq. ( 28) for the entire range of x.Substituting Eq. ( 19) into Eq.( 28), a 5th-order polynomial in V was obtained whose coefficients depended on x.The polynomial reduces to 4th order when w = 0, in which case V mc can be computed from Eq. (30).In all test conditions, one and only one real root between V i and V f was found for each x.The optimal control strategy were applied to four benchmark conditions, labeled as Min-Fuel, Headwind, Tailwind, and Fuel-and-Time.These conditions differ only in the choice of wind function and Cost Index, listed in Table 1 and justified in the following sections.Table 1.Parameters that define the four benchmark conditions. +Condition +Wind GradientCost Index (see Eq. ( 2)) (see Eq. ( 5)) w(x) (knots; x in nmi) r ($/hr / (cents / lb))Min-Fuel 0 0 Headwind -3.79 + 0.226x 0 Tailwind 1.03 -0.301x 0 Fuel-and-Time 0 30 In all conditions, the initial airspeed is above the minimum-cost speed and the final airspeed is below the minimum-cost speed, corresponding to Lemma 1(a) and 1(c), respectively.The discussion in Section 4.2 indicates that an extreme value of thrust, T min , must be used in the initial and final phases of the state trajectory.The speeds for those phases were found by integrating the state equations with that value of thrust forward from the first two boundary conditions of (9) and backward from the last two boundary conditions of (9).The integration was carried out numerically using MATLAB's [29] ode45 function.The thrust T mc was computed as a function of x using Eq. ( 27).Table 2 summarizes the flight times and fuel burn obtained for the four conditions.Details of the control strategy and speed profile will be discussed for each condition in the following sections. +Min-Fuel Condition: No Wind, No Time CostFigure 5(a) shows the optimal speed profile.The thin line represents the minimum-cost speed, V mc .Q 1 and Q 2 stand for points that join a T min speed profile with a T mc speed profile.The value V mc was computed essentially from Eq. (30).The rapid changes of V mc at some values of x were due to change of the FPA before and after a waypoint.The general trend was that V mc increased with increasing (shallower) γ i .For example, the increase of V mc near x = -34 nmi corresponded to the waypoint of PUNNT.Between PUNNT and TEICH, the FPA was shallower than before PUNNT (see Fig. 4), resulting in higher values of V mc .shows T as a function of x, which has non-zero thrust during the period of time when the trajectory follows the minimum-cost speed curve.The thrust for this period of time was equal to T mc .Shallower FPAs result in higher values of T mc (e.g., on the segment from PUNNT to TEICH; ).This corresponds to the physical intuition that a shallower descent does not convert as much potential energy to kinetic energy to compensate for the drag and therefore must use more thrust instead.The speed profile shown in Fig. 5 turns out to be unattainable near x = -24 nmi and near x = -20 nmi.This situation corresponds to Lemma 1(e).In physical flight control, the effect of negative thrust could have been attained by using speed brakes.This use, however, would not result in the reduction of cost (11) as would negative thrust.In this model, the trajectory segment is computed, following the notation in Lemma 1(e), by iterating through possible segments of a 1 b 1 until the minimum cost between ab is obtained.As expected, the actual optimal speed has a flatter |dV /dx| than that of |dV mc /dx| on these segments. +Headwind and Tailwind ConditionsThe main purpose of studying these two wind conditions is to understand the effect of wind variation on the optimal control strategy and the resulting speed profile.To select representative winds conditions, winds along the GEELA route were computed using the two-hour look-ahead time, 40-km Rapid Update Cycle weather forecast [30] for year 2011.Twenty-four forecast winds were generated for each day on an hourly basis.Daily averages were computed for every day of year 2011 at the waypoints PUNNT and PHX08, two of the waypoints on the GEELA route (see Fig. 3).Figure 6 shows the daily averages of the along-the-route components of wind at PUNNT and PHX08.The wind components for PUNNT and PHX08 were computed at 10,000 ft and 1,111 ft, respectively.The data indicated predominantly tailwinds for PUNNT all year long.Wind components at PHX08 were smaller in magnitude and varied somewhat by the season.The Headwind and Tailwind conditions were selected from the two days of year 2011 that had the most negative and positive wind gradients, dw/dx, respectively.The Headwind condition was defined based on February 3's average wind, which had the most negative wind gradient.Both PUNNT and PHX08 experienced headwinds along the route on February 3.The Tailwind condition was defined based on February 19's average wind, which had the most positive wind gradient.Both PUNNT and PHX08 experienced tailwinds along the route on February 19.The wind gradient was assumed to be constant throughout the path, and was computed by dividing the difference of the wind components at PHX08 and PUNNT by the ground path distance between these two waypoints.The resulting wind functions are listed in Table 1.The optimal speed profile and control strategy for the Headwind condition are shown in Fig. 7.The wind perturbed the minimum-cost speed curve noticeably, increasing V mc by 16 knots at the beginning (x i ) of the trajectory and by only 4 knots at the end of the trajectory.The initial and final arcs descended faster in speed (more negative dV mc /dx), causing the aircraft to stay longer on the minimum-cost speed curve than in the Min-Fuel condition.The thrust on the minimum-cost speed curve increased slightly when compared to the Min-Fuel condition.For transient segments on which the FPA increases rapidly around x = -34 nmi, the thrust takes relatively high values.The optimal speed profile for the Tailwind condition is shown in Fig. 8(a).The wind perturbed the minimum-cost speed curve in the opposition direction, decreasing V mc by 15 knots of V mc at the beginning of the trajectory and by only 1 knot at the end of the trajectory.Fig. 8(b) shows the T min arcs descended slower in speed, causing the aircraft to stay only a short distance on the minimum-cost speed curve.The thrust on the minimumcost speed curve decreased slightly compared to the Min-Fuel condition.The effect of wind on the minimum-cost speed is consistent with the dependency of the best-range speed on winds [24]: higher for headwind and lower for tailwind.Comparing the fuel burn to the Min-Fuel condition, the trajectory for the Headwind condition burns 36 lb more fuel while the trajectory for the Tailwind condition burns 31 lb less fuel (see Table 2). +Fuel-and-Time ConditionThe Fuel-and-Time condition considered both fuel and time costs.The Cost Index selected by airlines varies greatly from route to route and among aircraft types [20].For B737, a typical value [20,31] of r = 30 $/hr cents $/lb was used for the analysis.The fuel price, p f , was chosen to be 0.43 ($/lb).The optimal speed profile for the Fuel-and-Time condition is shown in Fig. 9(a).Compared to the Min-Fuel condition, the minimum-cost speed curve moves toward higher speeds significantly.This is because the aircraft must fly faster in order to reduce the time cost at the expense of the fuel cost.The minimum-cost speed curve was reached here much earlier than in the Min-Fuel condition.The speed profile for the Fuel-and-Time condition saved 45 seconds compared to that for the Min-Fuel condition (see Table 2), at the cost of 23 lb of fuel.Fig. 9(b) shows the required thrust T as a function of x.A large value of thrust was required to accelerate the aircraft around x = -34 nmi due to the increase of the minimum-cost speed. +ConclusionsThe key contribution of this paper is an exact feedback-control formula for the speed strategy that accrue minimum cost during a descent under altitude restrictions.The sensitivity of the optimal speed strategy to winds and airlines' business objectives was analyzed by applying the formula to an actual arrival route.This analytical formula has the following potential applications:Evaluation of arrival procedures: While most altitude restrictions are set by terrain, noise, and separation of traffic flows, speed restrictions can be compared to the minimum-cost speed profile to see whether the procedure can be refined to reduce aircrafts' direct operational costs.Since weather, airlines' business objectives, and aircraft types all affect the minimum-cost speed profile, a static arrival procedure should at best achieve systemic fuel/cost savings on an aggregated basis.Scheduling efficiency of ground automation tools: Ground automation tools that assist ATC in scheduling arrival aircraft can take advantage of the minimum-cost speed profile and use this information to improve its scheduling efficiency.Specifically, the TSS system in the United States can use this information to improve the nominal speed profiles used as reference points for scheduling aircraft.Guidance of small jets: While large commercial jets equipped with performance-based Flight Management Systems (FMS) have similar capabilities of computing company-preferred speed profile based on the airlines' business objective, aircraft performance model, and limited wind information, small jets such as regional and business jets are equipped with simpler FMS that cannot compute minimum-cost speed profiles.These jets accounts for 25% to 30% of the operations in the United States [32].The methodology developed in this paper can be used to aid the airlines in the selection of speed profiles for these jets.The selection can presumably be made during the flight-planning phase.A reference point for multi-aircraft operations: During periods of medium to heavy traffic, separation constraints must be taken into account and minimum-cost speed profiles may not be always feasible for individual aircraft.Nonetheless, the minimum-cost speed profiles can still be used as a reference point for evaluating the level of optimality of the system and for computing systemic cost-effective strategies.Among the directions for future research, the following seem reasonable: * Compare the minimum-cost speed profile to actual trajectories and determine how far-off is the current-day operations from optimality.* Predict a systemic minimum-cost speed profile for a group of aircraft and a period of future time, using airlines' flight schedules and weather forecast.( e )eSuppose a and b are on the curve [I = 0], and let ab denote the segment of [I = 0] connecting these two states.If a segment a b of [I = 0] is contained in ab and is not an attainable state trajectory (i.e., state b cannot be reached from a using admissible controls), then the optimal state trajectory from a to b consists of three segments: the segment of [I = 0] from a to a state a 1 on aa , one (attained by a suitable extreme value of control) from a 1 to a state b 1 on b b, and the segment of [I = 0] from b 1 to b. +1 Figure 2 .12Figure 2. A notional illustration of Lemma 1.In panels (a) through (d), the trajectories from state a to state b shown in solid outperform those shown dashed. +Figure 3 .3Figure 3.The GEELA arrival route to the Phoenix Airport. +Figure 4 .4Figure 4.The altitude profile along the GEELA arrival route. +Figure 5 .5Figure 5.The minimum-cost speed profile and control strategy for the Min-Fuel condition. +Figure 6 .6Figure 6.Daily averages of the along-the-route wind components at PUNNT and PHX08. +Figure 5 (5Figure 5(b) shows T as a function of x, which has non-zero thrust during the period of time when the trajectory follows the minimum-cost speed curve.The thrust for this period of time was equal to T mc .Shallower FPAs result in higher values of T mc (e.g., on the segment from PUNNT to TEICH; ).This corresponds to the physical intuition that a shallower descent does not convert as much potential energy to kinetic energy to compensate for the drag and therefore must use more thrust instead. +Figure 7 .7Figure 7.The speed profile and control strategy for the Headwind condition. +Figure 8 .8Figure 8.The speed profile and control strategy for the Tailwind condition. +Figure 9 .9Figure 9.The minimum-cost speed curve and control strategy for the Fuel-and-Time Condition. +(iii) If an optimal trajectory passes through two states on [I = 0], then the portion of the trajectory between these two states lies entirely on [I = 0].(In other words, an optimal trajectory cannot leave [I = 0] and then come back to [I = 0].) +Table 2 .2Flight times and fuel burn for the four conditions.ConditionFlight Time (sec) Fuel Burn (lb)Min-Fuel846127Headwind830166Tailwind85482Fuel-and-Time801150300V (knots)220Q 1Q 2140 + A waypoint is a geographical position defined by its latitude and longitude coordinates[5]. + A meter fix is a waypoint at the boundary of the terminal area that can have scheduled times of arrival for flights during high-density traffic[7]. + The GEELA route has two upstream branches that merge at HYDRR.Only the portion from HYDRR to the airport is analyzed in this work. + + + +Appendix A Direct Operating Cost as a Path IntegralThe direct operating cost in Eq. ( 11) can be written as a path integral along the state trajectory [22].First, rewrite the equations of motion in Eqs.(8) in a matrix form:Invert the two-by-two matrix on the right hand side to write the control T as a linear combination of ẋ and V [22],Substituting Eq. (A2) into Eq.( 11),Multiplying the last term by a factor of unity, ẋ/(V + w),and changing the variables of integration,Here Γ stands for the trajectory, and dx and dV are related by the equations of motion and the control T .Eq. (A3) expresses the cost function as a line integral of a differential form P (x, V )dx + M (x, V )dV , where +Appendix B Optimal Control Strategy Connecting A State to the [I = 0] CurveThe following lemma proves that the optimal control strategy for the problem considered in this work must use either T min or T max when connecting a state to the [I = 0] curve.Lemma 2 A state trajectory that has no part in I > 0, has the final state on [I = 0], and passes through a state (x(t), V (t)) in I < 0 at some instant t * , with the control strategy T (t) strictly greater than T min in some neighborhood of t * , is not optimal.Proof.Construct a new control strategy, denoted here by T * (t), as follows.Replace T (t) by T min on the interval [t * , t 1 ], where t 1 is the earliest instant at which the newly obtained state trajectory, (x * (t), V * (t)), intersects either (x(t), V (t)) (Case (a)) or the curve [I = 0] (Case (b)), whichever happens first (if the two cases coincide, the situation is treated as as Case (a)).In Case (a), let T * (t) = T (t) for t ≥ t 1 , to obtainIn Case (b), let T * (t) = T mc (t) from t = t 1 until such earliest instant t 2 when (x * (t), V * (t)) intersects the (x(t), V (t)) (possibly at the final state).Assume this portion of arc on [I = 0] is attainable.From t = t 2 on, let T * (t) = T (t).This yieldsIn both cases, V * (x) by construction satisfies both of the following conditions: V * (x) ≤ V (x) for every x, and V * (x) < V (x) in some right-sided neighborhood of x(t * ).By Lemma 1a at Section 4.2, (x * (t), V * (t)) outperforms (x(t), V (t)).This completes the proof.Immediate consequences of Lemma 2, and of results analogous to it, are as follows:(i) An optimal trajectory passing through a state (x, V ) in I < 0 and later reaching [I = 0] has in state (x, V ) control T min .(ii) An optimal trajectory passing through a state (x, V ) in I > 0 and later reaching [I = 0] has in state (x, V ) control T max . +REPORT DOCUMENTATION PAGEForm Approved OMB No. 0704-0188The public reporting burden for this collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering and maintaining the data needed, and completing and reviewing the collection of information.Send comments regarding this burden estimate or any other aspect of this collection of information, including suggestions for reducing this burden, to Department of Defense, Washington Headquarters Services, Directorate for Information Operations and Reports (0704-0188), 1215 Jefferson Davis Highway, Suite 1204, Arlington, VA 22202-4302.Respondents should be aware that notwithstanding any other provision of law, no person shall be subject to any penalty for failing to comply with a collection of information if it does not display a currently valid OMB control number. +PLEASE DO NOT RETURN YOUR FORM TO THE ABOVE ADDRESS. +Standard Form 298 (Rev. 8/98)Prescribed by ANSI Std.Z39.18 +REPORT DATE (DD-MM-YYYY)01-04-2015 An electronic version can be found at http://ntrs.nasa.gov. +REPORT TYPETechnical Memorandum +ABSTRACTAn analytical formula for solving the speed profile that accrues minimum cost during an aircraft descent with a constrained altitude profile is derived.The optimal speed profile first reaches a certain speed, called the minimum-cost speed, as quickly as possible using an appropriate extreme value of thrust.The speed profile then stays on the minimum-cost speed as long as possible, before switching to an extreme value of thrust for the rest of the descent.The formula is applied to an actual arrival route and its sensitivity to winds and airlines' business objectives is analyzed. + + + + + + + NASAs ATM Technology Demonstration-1: Moving NextGen Arrival Concepts from the Laboratory to the Operational NAS + + HSwenson + + + IiiRobinson + + + JE + + + + Journal of Air Traffic Control + + 55 + 2 + 2013 + + + Swenson, H.; Robinson III, J. E.; et al.: NASAs ATM Technology Demonstration- 1: Moving NextGen Arrival Concepts from the Laboratory to the Operational NAS. Journal of Air Traffic Control , vol. 55, no. 2, 2013. + + + + + + NHasevoets + + + PConroy + + + + AMAN Status Review + + 2010. 2010 + 1 + + + Hasevoets, N.; and Conroy, P.: AMAN Status Review 2010. Eurocontrol, Edition num- ber 0.1 , vol. 17, 2010. + + + + + Evaluation of the terminal area precision scheduling and spacing system for near-term NAS application + + JThipphavong + + + LMartin + + + HSwenson + + + PLin + + + JNguyen + + 10.1201/b12321-12 + + + Advances in Human Aspects of Aviation + + CRC Press + 2012 + + + + Thipphavong, J.; Martin, L.; Swenson, H.; Lin, P.; and Nguyen, J.: Evaluation of the Terminal Area Precision Scheduling and Spacing System for near-term NAS application. 4th International Conference on Applied Human Factors and Ergonomics (AHFE), 2012. + + + + + Determining operational benefits of required navigation performance (RNP) authorization required (AR) approaches + + DavidGouldey + + 10.1109/icnsurv.2014.6820048 + + + 2014 Integrated Communications, Navigation and Surveillance Conference (ICNS) Conference Proceedings + + IEEE + June 2005 + 52 + + + United States Standard for Required Navigation Performance (RNP) Approach Pro- cedures with Special Aircraft and Aircrew Authorization Required (SAAAR). Order 8260.52, Federal Aviation Administration (FAA), June 2005. + + + + + Tactical Scheduling for Precision Air Traffic Operations: Past Research and Current Problems + + DRIsaacson + + + AVSadovsky + + + DDavis + + 10.2514/1.i010119 + + + Journal of Aerospace Information Systems + Journal of Aerospace Information Systems + 2327-3097 + + 11 + 4 + + 2014 + American Institute of Aeronautics and Astronautics (AIAA) + + + Isaacson, D. R.; Sadovsky, A. V.; and Davis, D.: Tactical Scheduling for Precision Air Traffic Operations: Past Research and Current Problems. Journal of Aerospace Information Systems, vol. 11, 2014, pp. 234-257. + + + + + Drew, Prof. George Charles, (10 Dec. 1911–19 Dec. 1997), London University Professor of Psychology, University College, 1958–79, then Professor Emeritus; Honorary Fellow, 1979 + + JMcelhatton + + + CDrew + + + PBuchanan + + 10.1093/ww/9780199540884.013.u178104 + + + NASA ASRS Directline + + 10 + + dec 1998 + Oxford University Press + + + McElhatton, J.; Drew, C.; and Buchanan, P.: Crossing-Restriction Altitude Deviations on SIDs and STARs. NASA ASRS Directline, vol. 10, dec 1998, pp. 10-15. + + + + + Design and Evaluation of a Stochastic Time-Based Arrival Scheduling Simulation System + + DanielMulfinger + + + AlexanderSadovsky + + 10.2514/6.2011-6874 + AIAA-2011-6874 + + + 11th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + + American Institute of Aeronautics and Astronautics + 2011 + + + Mulfinger, D.; and Sadovsky, A.: Design and Evaluation of a Stochastic Time-Based Arrival Scheduling Simulation System. 11th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, AIAA-2011-6874, 2011. + + + + + The Dynamic Planner: the Sequencer, Scheduler, and Runway Allocator for Air Traffic Control Automation. NASA/TM + + GLWong + + + 2000-209586, 2000 + + + Wong, G. L.: The Dynamic Planner: the Sequencer, Scheduler, and Runway Allocator for Air Traffic Control Automation. NASA/TM-2000-209586, 2000. + + + + + Analysis of optimal aircraft cruise with fixed arrival time including wind effects + + AntonioFranco + 0000-0001-7272-3388 + + + DamiánRivas + + 10.1016/j.ast.2013.10.005 + + + Aerospace Science and Technology + Aerospace Science and Technology + 1270-9638 + + 32 + 1 + + 2014 + Elsevier BV + + + Franco, A.; and Rivas, D.: Analysis of Optimal Aircraft Cruise with Fixed Arrival Time Including Wind Effects. Aerospace Science and Technology, vol. 32, no. 1, 2014, pp. 212-222. + + + + + Four-dimensional fuel-optimal guidance in the presence of winds + + AbhijitChakravarty + + 10.2514/3.19929 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 8 + 1 + + 1985 + American Institute of Aeronautics and Astronautics (AIAA) + + + Chakravarty, A.: Four-Dimensional Fuel-Optimal Guidance in the Presence of Winds. Journal of Guidance, Control, and Dynamics, vol. 8, no. 1, 1985, pp. 16-22. + + + + + 4-D Aircraft Flight Path Management in Real Time + + AbhijitChakravarty + + + JurisVagners + + 10.23919/acc.1983.4788220 + + + 1983 American Control Conference + + IEEE + 1983. 1983 + + + + Chakravarty, A.; and Vagners, J.: 4-D Aircraft Flight Path Management in Real Time. American Control Conference, 1983 , IEEE, 1983, pp. 794-795. + + + + + Fuel-optimal aircraft trajectories with fixed arrival times + + JamesWBurrows + + 10.2514/3.19796 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 6 + 1 + + 1983 + American Institute of Aeronautics and Astronautics (AIAA) + + + Burrows, J. W.: Fuel-optimal Aircraft Trajectories with Fixed Arrival Times. Journal of Guidance, Control, and Dynamics, vol. 6, no. 1, 1983, pp. 14-19. + + + + + Constrained Optimum Trajectories with Specified Range + + HeinzErzberger + + + HomerLee + + 10.2514/3.55950 + + + Journal of Guidance and Control + Journal of Guidance and Control + 0162-3192 + + 3 + 1 + + 1980 + American Institute of Aeronautics and Astronautics (AIAA) + + + Erzberger, H.; and Lee, H.: Constrained Optimum Trajectories with Specified Range. Journal of Guidance, Control, and Dynamics, vol. 3, no. 1, 1980, pp. 78-85. + + + + + Optimization of Multiphase Aircraft Trajectories Using Hybrid Optimal Control + + AntonioFranco + + + DamiánRivas + + 10.2514/1.g000688 + + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 38 + 3 + + 2015 + American Institute of Aeronautics and Astronautics (AIAA) + + + Franco, A.; and Rivas, D.: Optimization of Multiphase Aircraft Trajectories Using Hybrid Optimal Control. Journal of Guidance, Control, and Dynamics, 2015, pp. 1-16. Doi:http://arc.aiaa.org/doi/pdf/10.2514/1.G000688. + + + + + Optimal Control Based Vertical Trajectory Determination for Continuous Descent Arrival Procedures + + SangGyunPark + + + John-PaulClarke + + 10.2514/1.c032967 + + + + Journal of Aircraft + Journal of Aircraft + 0021-8669 + 1533-3868 + + 52 + 5 + + 2015 + American Institute of Aeronautics and Astronautics (AIAA) + + + Park, S. G.; and Clarke, J.-P.: Optimal Control Based Vertical Trajectory De- termination for Continuous Descent Arrival Procedures. Journal of Aircraft, 2015. Doi:http://arc.aiaa.org/doi/abs/10.2514/1.C032967. + + + + + Aircraft Shortest and Fastest Continuous Descent Approach Development + + SKhardi + + 10.2514/1.c031775 + + + Journal of Aircraft + Journal of Aircraft + 0021-8669 + 1533-3868 + + 49 + 6 + + 2012 + American Institute of Aeronautics and Astronautics (AIAA) + + + Khardi, S.: Aircraft Shortest and Fastest Continuous Descent Approach Development. Journal of Aircraft, vol. 49, no. 6, 2012, pp. 1931-1939. + + + + + Fixed RTA Fuel Optimal Profile Descent Based on Analysis of Trajectory Performance Bound. Digital Avionics Systems Conference (DASC), 2012 IEEE/AIAA 31st + + SGPark + + + JClarke + + + 2012 + IEEE + + + + Park, S. G.; and Clarke, J.: Fixed RTA Fuel Optimal Profile Descent Based on Analysis of Trajectory Performance Bound. Digital Avionics Systems Conference (DASC), 2012 IEEE/AIAA 31st, IEEE, 2012, pp. 3D3-1. + + + + + Fixed-Range Optimal Trajectories of Supersonic Aircraft by First-Order Expansions + + RobertWindhorst + + + MarkArdema + + + DavidKinney + + 10.2514/2.4797 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 24 + 4 + + 2001 + American Institute of Aeronautics and Astronautics (AIAA) + + + Windhorst, R.; Ardema, M.; and Kinney, D.: Fixed-Range Optimal Trajectories of Supersonic Aircraft by First-Order Expansions. Journal of Guidance, Control, and Dy- namics, vol. 24, no. 4, 2001, pp. 700-709. + + + + + Rule derivation for arrival aircraft sequencing + + AdrianaAndreeva-Mori + + + ShinjiSuzuki + + + EriItoh + + 10.1016/j.ast.2013.08.004 + + + Aerospace Science and Technology + Aerospace Science and Technology + 1270-9638 + + 30 + 1 + + 2011 + Elsevier BV + + + Andreeva-Mori, A.; Suzuki, S.; and Itoh, E.: Scheduling of Arrival Aircraft Based on Minimum Fuel Burn Descents. ASEAN Eng J , vol. 1, no. 1, 2011. + + + + + Index + + BRoberson + + + RRoot + + + DAdam + + 10.1680/cdmre.34969.bm07 + + + CDM Regulations 2007 Explained + + Thomas Telford Publishing + 2007, 2007 + 2 + + + + Roberson, B.; Root, R.; and Adam, D.: Fuel Conservation Strategies: Cost Index Explained. Boeing Aero Quarterly, vol. 2, no. 2007, 2007, pp. 26-45. + + + + + Flight Profile Variations due to the Spreading Practice of Cost Index Based Flight Planning + + WRumler + + + TGünther + + + UWeißhaar + + + HFricke + + + + Transportation + + 2010 + Budapest + + + Rumler, W.; Günther, T.; Weißhaar, U.; and Fricke, H.: Flight Profile Variations due to the Spreading Practice of Cost Index Based Flight Planning. 4th International Conference on Research in Air Transportation, Budapest, 2010. + + + + + Green's Theorem Method for a Control System with Outstanding Control + + SYosida + + + + Funkcialaj Ekvacioj + + 16 + + 1973 + + + Yosida, S.: Green's Theorem Method for a Control System with Outstanding Control. Funkcialaj Ekvacioj , vol. 16, 1973, pp. 255-263. + + + + + Mathematical Handbook for Scientists and Engineers: Definitions, Theorems, and Formulas for Reference and Review + + GAKorn + + + TMKorn + + + 2000 + Courier Dover Publications + + + Korn, G. A.; and Korn, T. M.: Mathematical Handbook for Scientists and Engineers: Definitions, Theorems, and Formulas for Reference and Review . Courier Dover Publi- cations, 2000. + + + + + Introduction to Aircraft Performance, Selection, and Design + + FJHale + + + 1984 + Wiley + New York + + + Hale, F. J.: Introduction to Aircraft Performance, Selection, and Design. Wiley New York, 1984. + + + + + Mathematical Theory of Optimal Processes + + LSPontryagin + + 10.1201/9780203749319 + + + + Classics of Soviet Mathematics, Taylor & Francis + + Routledge + 1987 + + + Pontryagin, L.: Mathematical Theory of Optimal Processes. Clas- sics of Soviet Mathematics, Taylor & Francis, 1987. URL http://books.google.com/books?id=kwzq0F4cBVAC. + + + + + Applied Optimal Control + + ArthurEBryson + + + Yu-ChiHo + + 10.1201/9781315137667 + + + Applied Optimal Control: Optimization, Estimation and Control + + Routledge + 1975 + + + Bryson, Jr., A. E.; and Ho, Y.-C.: Applied Optimal Control: Optimization, Estimation and Control . CRC Press, 1975. + + + + + Trajectory Synthesis for Air Traffic Automation + + RhondaSlattery + + + YiyuanZhao + + 10.2514/2.4056 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 20 + 2 + + Mar. 1997 + American Institute of Aeronautics and Astronautics (AIAA) + + + Slattery, R.; and Zhao, Y.: Trajectory Synthesis for Air Traffic Automation. Journal of Guidance, Control and Dynamics, vol. 20, no. 2, Mar. 1997, pp. 232-238. + + + + + Advanced Aircraft Performance Modeling for ATM: Enhancements to the Bada Model + + ANuic + + + CPoinsot + + + M.-GIagaru + + + EGallo + + + FANavarro + + + CQuerejeta + + 10.1109/dasc.2005.1563320 + + + 24th Digital Avionics Systems Conference + + IEEE + Nov. 2005 + + + Nuic, A.; Poinsot, C.; Iagaru, M.-G.; Gallo, E.; Navarro, F. A.; and Querejeta, C.: Advanced Aircraft Performance Modeling for ATM: Enhancements to the BADA Model. Proceedings of the IEEE/AIAA 24th Digital Avionics Systems Conference, Nov. 2005. + + + + + + + MatlabThe + + + Software + + + 2008 + The MathWorks Inc + Natick + + + Version 7.7.0.471 (R2008b + The MATLAB Software: Version 7.7.0.471 (R2008b). The MathWorks Inc., Natick, 2008. + + + + + An Hourly Assimilation–Forecast Cycle: The RUC + + StanleyGBenjamin + + + DezsöDévényi + + + StephenSWeygandt + + + KevinJBrundage + + + JohnMBrown + + + GeorgAGrell + + + DongsooKim + + + BarryESchwartz + + + TatianaGSmirnova + + + TracyLorraineSmith + + + GeoffreySManikin + + 10.1175/1520-0493(2004)132<0495:ahactr>2.0.co;2 + + + Monthly Weather Review + Mon. Wea. Rev. + 0027-0644 + 1520-0493 + + 132 + 2 + + 1998 + American Meteorological Society + + + Benjamin, S. G.; Brown, J. M.; Brundage, K. J.; Schwartz, B.; Smirnova, T.; Smith, T. L.; Morone, L. L.; and Dimego, G.: The Operational RUC-2. Preprints. Proceedings of the 16th Conference on Weather Analysis and Forecasting, Amer. Meteor. Soc., 1998, pp. 249-252. + + + + + An Approach for Balancing Delay and Fuel Burn in Separation Assurance Automation + + AishaBowe + + + ConfesorSantiago + + 10.2514/6.2012-5416 + + + 12th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference and 14th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference + + American Institute of Aeronautics and Astronautics + 2012 + + + Bowe, A.; and Santiago, C.: An Approach for Balancing Delay and Fuel Burn in Sepa- ration Assurance Automation. AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, 2012. + + + + + Benefits of Continuous Descent Operations in High-Density Terminal Airspace Considering Scheduling Constraints + + JohnRobinson Iii + + + MaryamKamgarpour + + 10.2514/6.2010-9115 + AIAA-2010-9115 + + + 10th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + + American Institute of Aeronautics and Astronautics + Sept. 2010 + + + Robinson, J. E.; and Kamgarpour, M.: Benefits of Continuous Descent Operations in High-Density Terminal Airspace Under Scheduling Constraints. Proceedings of the 10th AIAA Aviation Technology, Integration, and Operations Conference, AIAA-2010-9115, Sept. 2010. + + + + + + diff --git a/file802.txt b/file802.txt new file mode 100644 index 0000000000000000000000000000000000000000..5b68601604809346a612abf09f667d0cd6ccf5e1 --- /dev/null +++ b/file802.txt @@ -0,0 +1,448 @@ + + + + +Harry N. Swenson † NASA Ames Research Center, Moffett Field, California, 94035-1000A time advance algorithm associated with the scheduling functionalities of the Traffic Management Advisor (TMA) for arrival flights is presented and evaluated.The algorithm enhances TMA's meter fix schedule by advancing the flights' Scheduled Time of Arrival (STA) by an amount that minimizes their systemic operating cost.The systemic operating cost leverages the inherent trade-off of time and fuel efficiency resident in the cost index of modern flight management systems.The resulting STAs are achievable by speeding up the leading flights from their desired nominal speed profiles.A key advantage of this approach is that it reduces systemic delay to tight groupings of arriving aircraft as well as increases sustained throughput of the operation.A fast-time, Monte Carlo simulation that emulates TMA's scheduling functionalities is performed for arrival flights to the Phoenix Airport to quantify the benefit of the time advance algorithm.Results show consistent time saving benefits, ranging from 3 to 50 minutes for 112 flights with varying levels of traffic congestion. +Nomenclature +I. Introduction2][3] In the United States, the operationally deployed Traffic Management Advisor (TMA) has served as an important component in achieving this objective. 4TMA provides air traffic controllers and Traffic Management Coordinators (TMCs) with a time-based metering function, arrival flow visualization and statistics, and runway allocation to increase capacity and reduce delay.The currently fielded TMA is referred to as the Time-Based Flow Management (TBFM) system within the Federal Aviation Administration's (FAA) portfolio of automation systems and programs.TMA is now deployed in all 20 en-route Air Route Traffic Control Centers (ARTCCs, or Centers) and many major Terminal Radar Approach Control (TRACON) facilities in the US.TMA sequences the arrival flights on a constrained first-come-first-serve basis.To determine the sequence of flights for scheduling, TMA computes the Estimated Time of Arrival (ETA) to the meter fix a and the runway threshold for each aircraft.TMA then uses the ETAs to sequence aircraft and determine the scheduled times of arrival (STAs) to the meter fix and runway threshold for each aircraft.The STAs are spaced according to scheduling constraints based on airport/runway configuration, separation requirements, and flow rates entered by the TMC.In the absence of scheduling constraints, TMA sets the computed STA for each aircraft equal to its ETA.When the traffic is heavy enough, TMA will begin to delay aircraft to accommodate the scheduling constraints.In this case, the STAs for these aircraft will be delayed from their ETAs.TMA in its current form, however, does not consider a time advance, which sets an STA of flight earlier than its ETA.The choice of using only delays to maintain separation constraints on TMA's schedule, although straightforward and robust, does not consider the opportunity of reducing the systemic operating cost of the flight by time advance.The systemic cost refers to the collective direct operating cost of the flights.Each airline seeks to minimize the direct operating costs of its own operations by considering both fuel and time costs in flight planing. 5This leads to the selection of preferred cruise and descent speeds that vary among aircraft types and from route to route.TMA models airlines' preferred speeds and uses them in computing the nominal trajectories and resulting ETAs for all flights.With proper selection of the speeds, the ETA represents the arrival time that would minimize a flight's direct operating cost.A delay or a time advance deviates from the minimum-cost ETA and increases the cost.Whereas not all flights can fly their minimum-cost speed profiles during heavy traffic due to scheduling constraints, it can be shown that a time advance combined with delay can reduce the projected systemic cost of the schedule when compared to the cost of using delays alone.This paper proposes an algorithm to improve the projected systemic cost of achieving TMA's meter fix schedule, as a step towards more efficient arrival operations.This algorithm identifies "packs" of arrival flights and, for each pack, shifts the computed STAs ahead for some flights by an amount that minimizes the pack's systemic cost.Packs are defined as groups of aircraft whose ETAs will require controller action to achieve proper separation.The algorithm is fast, straightforward, and can be easily integrated with TMA's scheduler without changes to its infrastructure.Further background information about the TMA's scheduler is given in Section II.Section III derives the amount of time advance for a group of flights and describes the algorithm for identifying packs of flights on TMA's schedule.Section IV describes the Monte Carlo simulations designed to quantify the amount of cost savings as well as assert the robustness of the algorithm.Section V presents the simulation results of the Phoenix Airport arrival traffic scheduling.The envisioned integration of the proposed algorithm with TMA's scheduler is described in Section VI. +II. TMA's SchedulerThis section provides a brief, simplified description of TMA's scheduler component, the Dynamic Planner (DP), with a focus on the meter fix/runway scheduling and runway allocation.A comprehensive description of the DP can be found elsewhere. 6Figure 1 shows the key features of the airspace relevant to the scheduling functionalities of DP.The airspace is divided into Center airspace and TRACON airspace.The Center and TRACON facilities provide TMA with static information such as airspace constraints, arrival procedures, separation requirements, and dynamic information such as flight plans, radar track data, and weather forecast.When running TMA in real time, DP receives periodic updates of arrival flights' meter fix and runway threshold ETAs from another TMA component.DP computes the STAs for the meter fix and the runway threshhold according to the ETAs and scheduling constraints.Flights are scheduled one by one, on a firstcome-first-serve basis, in an order defined roughly by their runway ETAs.DP also allocates a runway for each flight and uses the ETA for that runway for its order consideration.When the traffic is heavy enough, DP will begin to delay aircraft to accommodate scheduling constraints.The scheduling constraints for the meter fix come from in-trail separation requirements, the airport flow rate imposed by the TMCs, and other considerations.The STA of a flight is frozen once it crosses the freeze To demonstrate DP's scheduling logic with a simple scenario, consider the scheduling of four arriving aircraft that are to cross the same meter fix. Figure 2 shows the ETAs and STAs for these four arrival flights.AC1 has the earliest ETA and is expected to cross the meter fix earlier than the other three flights.AC1 is followed by AC2 and AC3, whose ETAs are very close to each other.AC4 follows AC3 and has the latest ETA.DP assigns each flight's STA, starting with the flight with the earliest ETA, which is AC1.In the absence of scheduling constraints, DP sets the STA for AC1 equal to AC1's ETA.AC2, AC3, and AC4 must be delayed so that their STAs satisfy the in-trial separation requirements, usually input as a distance and converted to a time in DP.Various scheduling events trigger partial or complete updates of the schedule.Radar track and flight plan updates only trigger recalculation of the STAs of flights that are on the schedule but outside the freeze horizon.In general, the following steps are performed by DP to update the STAs (for all flights, for some flights, or only for flights outside the freeze horizon): +ETA STA1. DP computes preliminary meter fix STAs for all flights that have yet to cross the meter fix.2. DP computes the runway STAs for flights one by one.In the absence of any scheduling constraints, a runway STA is equal to the Proposed Time of Arrival (PTA), which is defined as the flight's preliminary meter fix STA plus a nominal TRACON transit time between the meter fix and the runway considered.In the presence of scheduling constaints, DP assigns a runway STA later than the PTA.The delay (STA -PTA) needs to be absorbed in the TRACON.3. If the delay to be absorbed in the TRACON is greater than the maximum TRACON delay (an input parameter to DP), DP distributes the excessive delay to the Center by pushing the preliminary meter fix STA for the flight to a later time, which is the final meter fix STA.Otherwise, the preliminary meter fix STA stands as the final meter fix STA.DP also allocates runways for flights upon specific event triggers.This allocation affects the runway schedule and therefore can affect the meter fix schedule too, due to the runway schedule feedback.The runway allocation calculation is intensive, and DP performs it twice for each arrival flight: once when its first STA is calculated and again when it is about to be frozen.For each flight to be considered, DP first tries every eligible runway for the flight and computes the schedule of all the trailing flights that may be affected by the flight.A runway that minimizes the System Schedule Time, which is the sum of the STAs of the flights that are affected, will be selected for the flight being considered.Currently, only the meter fix schedule is directly used by controllers in guiding the arrival traffic.Research to extend the scheduling functions in the terminal area is underway, as the extended TMA is one of the critical components of the Air Traffic Management Technology Demonstration -1 (ATD-1), 7 which showcases an integrated set of technologies that provide an efficient arrival solution for managing aircraft beginning from just prior to top-of-descent and continuing down to the runway.The FAA has renamed the extended scheduling functions as the Terminal Sequencing and Spacing (TSS) technologies and is planning to adopt this technology as an enhancement of the TBFM program. +III. Time Advance AlgorithmEarly work on time advance by Neuman and Erzberger 8 proposed two versions of a Time-Advance (TA) algorithm: the idealized fuel-saving TA algorithm and the pure TA algorithm.Both versions attempted to minimize the systemic cost of merging arrival flights by using a simple cost model that does not distinguish aircraft types.A heuristic maximum allowable time advance was proposed as a function of traffic demand estimated by the number of aircraft per hour.Both versions first shift the STAs of all flights on the schedule by a fixed amount of time.The idealized fuel-saving TA algorithm then examined the resulting STAs one by one to remove unnecessary time advances for flights, while the pure TA algorithm made no further adjustment.Another independent study 9 also advanced the STAs of all flights by a parametric amount of time, and investigated the fuel benefits as a function of the amount of time advance using arrival schedules at one runway of the Dallas/Fort Worth International Airport.Although systemic benefits were observed, both approaches used simple cost models and did not clearly relate the amount of minimum-cost time advance to local traffic information.Therefore, neither of these two approaches were operationally acceptable without further research.A recent study computed separation-compliant trajectories for arrival flights, using time advance to reduce the timespan of the landing flights. 10Compared to TMA, of which the schedule ensures separation compliance only at the scheduling points such as the meter fix and runway, this separation-compliant approach ensures separation compliance at every point on an arrival flight's trajectory.However, computation of such trajectories is time-consuming and can be a performance bottleneck for real-time ground automation tools.The time advance algorithm proposed in this work utilizes information already computed by TMA or already accessible to TMA's scheduler, the DP.Instead of advancing all flights' STAs with the same amount of time, it identifies packs of flights according to local traffic information, such as the flights' ETAs, originally computed STAs, speed performance envelopes, and the scheduling constraints.For each pack of flights, the algorithm advances each flight's STA by a computed amount, called the Minimum-Cost Time Advance, that minimizes the systemic direct operating cost of achieving the STAs for the pack. +III.A. Direct Operating CostThe systemic direct operating cost is the sum of each flight's direct operating cost,C(t) = Fuel Cost P f × f (t) + Time Cost P f × CI × (t -t ref ) . (1)Here t is the time of crossing a reference point, which can be a meter fix or a runway.P f is the fuel price, f is the fuel consumption, t ref is a reference time, and CI is the Cost Index, which is the ratio of the time-related cost to the fuel cost.The time-related cost accounts for items such as flight crew wages, airplane lease cost, maintenance costs, connection flight constraints, and others that are potentially a function of the flight time.The Cost Index is determined by individual airlines, and usually varies among aircraft types and from route to route.Before a flight takes off, the flight crew enters the Cost Index into the flight management computer for it to compute the preferred speeds for climb, cruise, and descent. 5When CI is zero, the direct operating cost accounts for solely the fuel cost and the computed preferred speeds are lower.When CI is large, the time cost far outweighs the fuel cost and the preferred speeds are higher.A direct operating cost reaches a minimum at a specific crossing time.As CI increases, the time cost becomes important and the minimumcost time decreases.Figure 3 shows the estimated direct operating cost of a B757 flight from 35,000 ft and 180 nmi away from the airport to about 10 nmi away from the airport.Details of the computation of these curves are described in the Appendix.For illustration's purpose, the reference time was arbitrarily chosen to be the arrival time of a minimum-fuel trajectory so as to keep both curves within the same range.The choice of the reference time does not affect the time advance algorithm.The time advance algorithm is designed based on two critical assumptions:• The direct operating cost can be approximated by a quadratic function in the vicinity of the minimumcost arrival time.• The ETAs correspond to the flights' minimum-cost arrival times.The following sections derive the minimum-cost time advance for a group of flights and describe how the algorithm identifies packs of flights on the schedule.Although the derivation is for the meter fix schedule, it can be applied to the runway schedule as well. +III.B. Minimum-Cost Time AdvanceConsider a group of N arrival flights scheduled to arrive at a meter fix.The flights are assumed to be outside the freeze horizon, although the derivation applies to arrival flights anywhere in the Center.The STAs computed by the DP are named the original STAs to be distinguished from the STAs shifted later by the Time Advance algorithm.Let t i,ETA denote the ETA of the ith flight and t i,STA denote the original STA of the ith flight.They are ordered by the index such that b t i,ETA ≤ t i+1,ETA , i = 1, . . ., N -1.and t i,STA < t i+1,STA , i = 1, . . ., N -1.The original delay for each flight isD i = t i,STA -t i,ETA .(2)Without loss of generality, assume that the DP must delay all but the first flight,D i = 0, i = 1 > 0, i = 2, 3, . . . , NThe Minimum-Cost Time Advance for this group of flights is derived as follows.Using the quadratic approximation, each flight's cost function, C i , is expressed asC i (t) C i (t i,ETA ) + C i (t i,ETA ) (t -t i,ETA ) 2 , (3)where C is the second-order derivative of C andC i (t) > 0in the vicinity of t i,ETA .The systemic direct operating cost of achieving the STAs, i.e., flying the group of flights to the meter fix according to the STAs, isC tot = N i=1 C i (t i,STA ) = N i=1 C i (t i,ETA ) + C i (t i,ETA )D i 2 .(4)Now, shift the STAs of these flights by an amount of ∆t, thenC tot (∆t) = N i=1 C i (t i,STA -∆t) = N i=1 C i (t i,ETA ) + C i (t i,ETA )(D i -∆t) 2(5)The Minimum-Cost Time Advance is the time, ∆t a , that minimizes C tot (∆t).Differentiating the right-hand side of Equation 5 and equating it to zero,dC tot d∆t ∆t=∆ta = -2 N i=1 C i (t i,ETA )(D i -∆t a ) = 0. (6)After rearranging,∆t a (N ) = N i=1 C i (t i,ETA )D i N i=1 C i (t i,ETA ) ,(7)where the superscript (N ) emphasizes the fact the the Minimum-Cost Time Advance is computed for a group of N flights.The following remarks are made regarding Eq. 7, a critical formula for the time advance algorithm:• The Minimum-Cost Time Advance, ∆t a , is non-negative and can be regarded as a weighted average of the original delays, D i .• Other constraints impose an upper bound on ∆t a .One possible constraint is the frozen STA of another flight in front of the first flight.Another key constraint is the maximum time advance achievable by each flight using speed changes, which depends on aircraft type and how far each flight's preferred cruise and descent speeds are from its maximum cruise and descent speeds.• The weight C i is a function of the aircraft type and the aircraft's distance from (or nominal flight time to) the meter fix.The latter dependency affects ∆t a much less in many situations because the effect of distance cancels out in the ratio of the weights.As a first-order approximation, C i can be modeled as an aircraft-type specific constant.b DP actually schedules the STAs in an order that is based on flights' runway ETAs, which can be different from the order of meter fix ETAs in some special cases.The derivation still holds even if these cases are considered. +III.C. Identifying PacksThe computed ∆t a can be too aggressive for the trailing flights in the group if there are gaps in the ETAs of the flights.In this case, it is better to allow gaps in the STAs by splitting the group of flights into multiple packs.As an example, Figure 4 shows a group of six flights that have delays for all but the leading flight.Since there is a gap between the ETAs of AC4 and AC5, applying the computed ∆t a for all six flights will time-advance both AC5 and AC6 uncessarily.Instead, STAs of AC1 to AC4 should be shifted using computed ∆t a (4) computed from the first four flights' original STAs and ETAs, and STAs of AC5 and AC6 should be shifted using ∆t a (2) computed from the original STAs and ETAs for just these two flights.The rest of this section describes how to identify packs in the group of flights in a algorithmic way.The time advance algorithm identifies packs of flights by ensuring that the last flight in the pack is not time-advanced, i.e, its STA is not earlier than its ETA after applying ∆t a .A pack can have only as few as one flight, although a one-flight pack does not result in a time advance.Consider a meter fix schedule of N flights, some having delays.The algorithm takes the following steps to identify packs:1. Look for the first flight that can be time-advanced, starting from the one with the earliest STA.A flight cannot be time-advanced if its STA is minimally-spaced with another flight's STA ahead of it on the schedule.If such a flight is found, label it as flight 1 and the subsequent flights on the schedule as 2, 3,. .., etc.2. Identify the delays D i , for flights 1, 2, 3,. . .,etc.D 1 should be zero, because otherwise flight 1 must have been delayed due to the scheduling constraints and cannot be time-advanced.3. Consider a preliminary pack of flights 1 and 2. Compute ∆t a (2) for the pack, using Eq. 7. If ∆t a (2) is smaller than D 2 , extend the preliminary pack to include flight 3. Repeat the procedure until one of the following two conditions is met:• The last aircraft N on the meter fix schedule is reached: Identify flights 1 to N as a pack, and advance the STAs of flights in this pack by ∆t a (N ) .• ∆t a (i) ≥ D i for flight i: Exclude flight i.Identify flights 1 to i -1 as a pack, and advance the STAs of flights 1 to i -1 by ∆t a (i-1) .Search for the next pack by repeating step 1, relabeling flight i as the flight 1.The following remarks are made regarding this algorithm:• It works from earlier packs to later ones, using a short look-ahead horizon of one flight.• Once a pack is identified and their STAs advanced, the algorithm makes no additional adjustment to this pack.• Although it computes a ∆t a that minimizes the systemic operating cost for each pack, it does not result in the minimum systemic operating cost for all the flights on the schedule.One extreme example is a single flight followed by a large pack.The one-flight pack does not result in a time advance, and its STA constrains the available time advance of the following large pack.The algorithm is expected to most beneficial for medium to heavy traffic with clustered ETAs.If the traffic is light, there is no delay to reduce and the time advance algorithm has no effect.If the traffic is extremely heavy, the time advance algorithm does not find room on the meter fix schedule to advance the flights' STAs.In both extreme cases no benefit will be found. +IV. Monte Carlo SimulationA fast-time Monte Carlo simulation was designed to assess the benefits and robustness of the time advance algorithm.While the algorithm was intended to be ultimately implemented in DP, the architecture of DP was designed for real-time usage and did not easily facilitate fast-time simulation with various test traffic scenarios.Instead, The Stochastic Terminal Arrival Scheduling Software (STASS) 11 was used.STASS models all the essential scheduling functionalities of DP with the following distinctions:• STASS computes the schedule for all flights in one batch without periodic schedule updates.• STASS models primitive runway allocation functionalities which are less efficient than the DP's runway allocation.Specifically, STASS takes a traffic scenario, Center and TRACON nominal transit times, and in-trail and runway separation requirements as its input.A traffic scenario contains a list of flights, each characterized by aircraft type, weight class (which depends on the aircraft type), meter fix, and a freeze horizon ETA.The freeze horizon ETA models the time a flight is expected to reach the freeze horizon.STASS computes the meter fix ETA by adding a freeze horizon transit time, set to 19 minutes in this analysis, to the freeze horizon ETA.Each flight is assigned a nominal runway, and its runway ETA is computed by adding to the meter fix ETA a nominal TRACON transit time.The nominal TRACON transit time is a function of the runway and meter fix pair.The nominal runway ETA was used for the order of consideration, which is the order by which STASS follows to assign meter fix and runway STAs.STASS first computes the preliminary meter fix STAs for all the flights, then assigns the runway STAs for all the flights one by one.An assigned runway STA may require delay to be absorbed.A maximum TRACON delay time is imposed to ensure flights can absorb the TRACON delay with simple speed reduction.If the maximum TRACON delay time is not enough for absorbing all the delay, the preliminary meter fix STA for that flight must be pushed to a later time to absorb the extra delay.The Monte Carlo simulation generates random arrival scenarios spanning light, medium, and heavy traffic, from a template traffic scenario, by varying the flights' freeze horizon ETAs.The template traffic scenario's aircraft composition and arrival routes are are applied to all scenarios.A time window for constraining the variation of freeze horizon ETAs for each scenario was varied to achieve different levels of traffic demand.A large time window represents light traffic and a narrow time window represents heavy traffic.To capture the effect of traffic demand variation during the same time window, the expected separation violation at the runway is computed from the nominal runway ETAs.The runway ETAs represents the time of arrival at the runway threshold if an aircraft follows its desired descent procedures.If two aircraft have close ETAs at the same runway, they will violate the runway separation constraints.Thus the expected separation violation at the runway characterizes the overall level of traffic demand challenge.The performance metric for the scheduler was chosen to be the System Schedule Time, 6 which was the sum of the flights' runway STAs.The reason to use the runway STAs instead of the meter fix STAs as a performance metric is to include the effect of the time advance algorithm on the TRACON flight times.An alternative performance metric would be the systemic direct operating cost, which was expected to be highly correlated with the System Schedule Time.However, the systemic direct operating cost involves fuel burn comparison, and cannot be used unless all the TRACON transit routes are characterized by fuel burn.This is due to the fact that aircraft may be assigned different runways in different runs.The template traffic scenario was taken from real traffic data of flights that landed in the Phoenix Airport between 5pm and 7:30pm on September 10, 2014.It contained 101 jet flights and 11 turboprop flights arriving from all directions.A total of 600 traffic scenarios with varying levels of traffic demand were generated from the template traffic scenario.Figure 5 shows the four meter fixes whose schedules were computed for the jet flights.Turboprop flights were scheduled at different meter fixes (or the same meter fixes with different crossing altitudes) which are not shown here.The maximum time advance achievable by speeding was modeled as 100 seconds for all flights.Whenever the minimum-cost time advance computed by Eq. 7 exceeded 100 seconds, a value of 100 seconds was used instead.Modeling of the runways and TRACON transit times assumed the East Flow airport configuration, with which aircraft land towards the east on two runways, 07R and 08.Table 1 shows the TRACON transit times from each meter fix to each arrival runway.The transit times were estimated using constructed routes for the ATD-1 simulations 12 and the Trajectory Synthesizer 13 component of TMA. Figure 6 shows the modeled arrival routes from each meter fix to each runway.Each route is identified by a meter-fix specific color, while gray routes are shared by flights from different meter fixes.Two modes of STASS's runway selection were explored:• Nominal Runway: always schedule a flight to its nominal runway• Earliest Runway: always schedule a flight to the runway that has the earliest time to the Phoenix Airport corresponds to an expected runway separation violation of about 40 (among 112 flights).Time advance has time saving benefits between 3 and 50 minutes for scenarios in which the number of expected runway separation violations was between 20 and 60.For heavier traffic scenarios in which the number of expected runway separation violations was above 60, Time advance's benefit starts to vary greatly, with savings of around 100 minutes for some scenarios and zero for some others.Also shown in Fig. 7 was the average delay per flight, computed by +V. ResultsD avg = N i=1 (t i,STA -t i,ETA ) N ,(8)where t i,STA and t i,ETA were the times at the runway.Note that traffic demand resulting in more than 10 minutes' average delay is very unlikely to occur on a regular basis, for the FAA will often institute other traffic management initiatives such as ground stops and ground holds to prevent it from occuring.The 10 minute average delay benchmark corresponds roughly to 85 expected runway separation violations in Fig. 7.Figure 8 shows the time saved by the Time-Advance algorithm with the Earliest Runway selection mode.While the time saved per scenario is on average positive, individual savings are more scattered than those for the Nominal Runway mode, and can be negative for some scenarios.This was because the Earliest Runway mode only attempted to optimize the landing time for the flight considered without regarding landing times of the trailing flights.Time advance changed the selected runway for certain flights and such change can have negative impact on the schedule of trailing flights.Figure 9 shows an example in which the Earliest Runway Selection mode leads to excessive meter fix traffic delays of its trailing flights.Consider five aircraft AC1, AC2, AC3, AC4, and AC5, and two runways Rwy1 and Rwy2.STASS already allocates Rwy1 to AC1 and AC2.Rwy2's schedule is empty.AC3, AC4, and AC5 come through the same meter fix.The TRACON transit time to Rwy1 is shorter than that to Rwy2.AC3 is being considered for runway allocation between its nominal runway, Rwy1, and Rwy2.In the absence of scheduling constraints, AC3 can always land on Rwy1 earlier than on Rwy2 and no delay is to be absorbed.With the presence of AC2, AC3 can still land on Rwy1 earlier than Rwy2, and STASS's Earliest Runway selection mode selects Rwy1.However, the schedule requires delay for AC3 to be absorbed, and the maximum TRACON delay is not enough for absorbing all the delay.Therefore, the feedback from the runway schedule pushes the meter fix STA of AC3 to later times, thus pushing the STAs of the trailing flights AC4 and AC5 to later times as well.As a result, traffic is delayed at the meter fix and AC3, AC4, and AC5 end up with later STAs at both the meter fix and the runway.Had Rwy2 been selected for AC3, AC3 could maintain its preliminary meter fix STA and fly to Rwy2 without delay.The trailing flights AC4 and AC5 will not suffer from delays at the meter fix and will have a better chance of landing earlier, therefore achieving a better System Schedule Time.In summary, the simulation results showed consistent time saving benefits with the time advance algorithm across all levels of traffic demand, especially at medium to high traffic demand.Although running STASS in the Earliest Runway mode can result in negative time saving for certain traffic scenarios, the time saving average over traffic scenarios is still positive.While the variation of runway allocation can counter the benefits of the time advance algorithm under certain traffic conditions, this effect is attibuted to STASS's primitive runway allocation model.Better modeling of TMA's runway allocation logic is expected to reduce the scattering of the time saving benefits and will be investigated in the future. +MF Prelim +VI. Integration with TMA's SchedulerThe current time advance algorithm can be implemented as a part of DP, TMA's scheduler, in a straightforward way.Recall the steps performed by DP to update the STAs described in Section II.With the time advance algorithm implementation, the steps are modified as follows:• Compute the preliminary meter fix STAs for flights that are in the Center.• Apply the time advance algorithm to shift the meter fix STAs of these flights.• Compute the runway STAs for flights in the Center and in the TRACON.Delay the preliminary meter fix STAs, if necessary, to avoid absorbing excessive delays in the TRACON.The weights in Equation 7, C , can be approximated by a function of the weight class or the aircraft type, both of which are available to the DP.The maximal allowable time advance for individual flights can be estimated by the difference in the fast and nominal trajectories already computed by TMA.The time advance algorithm can potentially be applied to the schedules at additional route merge points between the meter fix and the runway.Scheduling flights at these merge points has an approach taken by the Terminal Area Precision Scheduling and Spacing System to improve airport throughput and reduce controllers' workload. 12 +VII. Conclusions and Future WorkThis paper presents a time advance algorithm that can enhance the efficiency and reduce delay of the arrival schedule computed by the Traffic Management Advisor (TMA).Monte Carlo simulation results showed time-saving benefits across a wide range of traffic demand.For a set of representative 112 flights into the Phoenix Airport, the time advance algorithm reduced flights' delay by 3 to 50 minutes in traffic scenarios with varying levels of traffic congestion.The algorithm is easily implementable on TMA, using information already computed by TMA's scheduler and requires minimal changes to TMA's infrastructure.Future work being considered include the following:• Evalulation of the time saving and fuel benefits with better modeling of TMA's runway allocation logic.• Exploration of the benefits of time advance on the schedule of runways and the merge points between meter fixes and runways.• Better modeling of the coefficients C . +Appendix: Fuel Consumption as a Function of Arrival TimeFigure 3 presents the direct operating cost of a B757 flight at 35,000 ft and 180 nmi away from the destination airport as a function of the arrival time.The fuel price, P f , was set to $0.43.The fuel consumption part of the direct operating cost was computed in the following way.For a specific arrival time, a minimum-fuel trajectory from the aircraft's position to a point on the arrival route that is 3000 ft in altitude and 10 nmi from the airport was constructed.This minimum-fuel trajectory was computed using a set of 5-state aircraft equations of motion and the GPOPS-II optimal control software. 14The computation was repeated for a range of arrival times to obtain the fuel consumption as a function of the arrival time and thus the curve for CI = 0. Speeds were bounded by the minimum-drag speed 15 and a Mach number of 0.85.Later arrival times required the trajectory to have path stretch along with a minimum-drag speed.The CI = 27 curve was derived from the CI = 0 curve using Equation 1.When the fuel consumption was at a global minimum, i.e., without the arrival time constraint, the minimum-fuel trajectory has a cruise segment with the maximum-range speed and a descent segment with the minimum-drag speed. 5Figure 1 .1Figure 1.The Center/TRACON diagram (Courtesy of Larry Meyn, NASA Ames Research Center). +horizon (see Fig. 1), a boundary inside which arrival flights' STAs are no longer updated.The freeze horizon can be time-based or distance-based.A time-based freeze horizon freezes the STAs of flights whose ETAs at the meter fix are less than or equal to 19 minutes, for example, into the future.A distance-based freeze horizon freezes the STAs of flights who are within 130 nmi, for example, of the meter fix.The STAs are frozen so that the Center air traffic controllers can guide each arrival flight towards realizing a fixed schedule. +Figure 2 .2Figure 2. Scheduling four arrival flights at a meter fix. +Figure 3 .3Figure 3.The direct operating cost of an arriving B757 flight. +Figure 4 .4Figure 4. Two packs of flights are identified.Blue STAs are modified by the Time Advance Algorithm. +Figure 5 .5Figure 5.The Phoenix Airport arrival routes. +Figure 7 Figure 6 .76Figure7shows the time saved by the Time-Advance algorithm with the Nominal Runway selection mode.Each data point corresponds to one traffic scenario.The abscissa represents the number of expected runway separation violation.As a reference point, the original template scenario created from the evening traffic +Figure 7 .7Figure 7. Time-Advance benefits with nominal runway assignment. +Figure 8 .8Figure 8. Time-Advance benefits with earliest runway assignment. +Figure 9 .9Figure 9. Optimal runway selection of AC3 delays trialing flights AC4 and AC5. +Table 1 .1Phoenix Airport TRACON transit times in secondsEngineType: JetRunway/Fix HOMRR BRUSR GEELA SQUEZ07R10971028682967089617746831087EngineType: TurbopropRunway/Fix HOMRR BRUSR PAYNT SUNSS07R11951167713974081076870711960 + Downloaded by NASA AMES RESEARCH CENTER on January 28, 2015 | http://arc.aiaa.org| DOI: 10.2514/6.2015-1326 + of 13 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on January 28, 2015 | http://arc.aiaa.org| DOI: 10.2514/6.2015-1326 + of 13 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on January 28, 2015 | http://arc.aiaa.org| DOI: 10.2514/6.2015-1326 + American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on January 28, 2015 | http://arc.aiaa.org| DOI: 10.2514/6.2015-1326 + of 13 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on January 28, 2015 | http://arc.aiaa.org| DOI: 10.2514/6.2015-1326 + of 13 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on January 28, 2015 | http://arc.aiaa.org| DOI: 10.2514/6.2015-1326 + + + +For the minimum-fuel trajectory calculation, the aircraft performance model parameters were taken from the Base of Aircraft Database (BADA) 16 3.8 with slight simplification.The airspeed-dependent thrustspecific fuel consumption in BADA was replaced with an average, constant value for simplification. + + + + + + + High Throughput Arrival Operations from Simulation to Reality + + TPrevot + + + BBaxley + + + TCallantine + + + WJohnson + + + LQuon + + + JRobinson + + + HNSwenson + + + + Proceedings of the International Conference on Human-Computer Interaction in Aerospace + the International Conference on Human-Computer Interaction in Aerospace + + 2012. 2012 + + + NASAs ATM Technology Demonstration-1: Transitioning Fuel Efficient + Prevot, T., Baxley, B., Callantine, T., Johnson, W., Quon, L., Robinson, J., and Swenson, H. N., "NASAs ATM Technology Demonstration-1: Transitioning Fuel Efficient, High Throughput Arrival Operations from Simulation to Reality," Proceedings of the International Conference on Human-Computer Interaction in Aerospace (HCI-Aero 2012), 2012. + + + + + The Efficient Descent Advisor: Technology Validation and Transition + + RichardCoppenbarger + + + MiwaHayashi + + + GauravNagle + + + DouglasSweet + + + RenanSalcido + + 10.2514/6.2012-5611 + AIAA-2012-5611 + + + 12th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference and 14th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference + + American Institute of Aeronautics and Astronautics + Sept. 2012 + + + Coppenbarger, R. A., Nagle, G., Sweet, D., and Hayashi, M., "The Efficient Descent Advisor: Technology Validation and Transition," Proceedings of the AIAA Aircraft Technology, Integration, and Operations Conference, AIAA-2012-5611, Sept. 2012. + + + + + Impact of Future Time-Based Operations on Situation Awareness of Air Traffic Controllers + + EOprins + + + DZwaaf + + + FEriksson + + + KVan De Merwe + + + RRoe + + + + 8th USA/Europe Air Traffic Management R&D Seminar + + 2009 + + + Oprins, E., Zwaaf, D., Eriksson, F., van de Merwe, K., and Roe, R., "Impact of Future Time-Based Operations on Situation Awareness of Air Traffic Controllers," 8th USA/Europe Air Traffic Management R&D Seminar , 2009. + + + + + Design and Operational Evaluation of the Traffic Management Advisor at the Fort Worth Air Route Traffic Control Center + + HNSwenson + + + THoang + + + SEngelland + + + DVincent + + + TSanders + + + BSanford + + + KHeere + + + + Proceedings of the 1st USA + the 1st USA + + June 1997 + + + Swenson, H. N., Hoang, T., Engelland, S., Vincent, D., Sanders, T., Sanford, B., and Heere, K., "Design and Operational Evaluation of the Traffic Management Advisor at the Fort Worth Air Route Traffic Control Center," Proceedings of the 1st USA/Europe Air Traffic Management R&D Seminar , June 1997. + + + + + Index + + BRoberson + + + RRoot + + + DAdam + + 10.1680/cdmre.34969.bm07 + + + CDM Regulations 2007 Explained + + Thomas Telford Publishing + 2007. 2007 + 2 + + + + Roberson, B., Root, R., and Adam, D., "Fuel Conservation Strategies: Cost Index Explained," Boeing Aero Quarterly, Vol. 2, No. 2007, 2007, pp. 26-45. + + + + + The Dynamic Planner: the Sequencer, Scheduler, and Runway Allocator for Air Traffic Control Automation, NASA/TM + + GLWong + + + 2000-209586, 2000 + + + Wong, G. L., The Dynamic Planner: the Sequencer, Scheduler, and Runway Allocator for Air Traffic Control Automa- tion, NASA/TM-2000-209586, 2000. + + + + + NASAs ATM Technology Demonstration-1: Moving NextGen Arrival Concepts from the Laboratory to the Operational NAS + + HSwenson + + + IiiRobinson + + + JE + + + + Journal of Air Traffic Control + + 55 + 2 + 2013 + + + Swenson, H., Robinson III, J. E., et al., "NASAs ATM Technology Demonstration-1: Moving NextGen Arrival Concepts from the Laboratory to the Operational NAS," Journal of Air Traffic Control, Vol. 55, No. 2, 2013. + + + + + NATIONAL SYMPOSIUM ON DEVELOPMENTS IN IRRADIATION TESTING TECHNOLOGY, SEPTEMBER 9--11, 1969, NASA LEWIS RESEARCH CENTER, SANDUSKY, OHIO. + + FNeuman + + + HErzberger + + 10.2172/4165376 + + + NASA/TM-103880, National Aeronautics and Space Administration + + Office of Scientific and Technical Information (OSTI) + 1991 + + + Neuman, F. and Erzberger, H., Analysis of Delay Reducing and Fuel Saving Sequencing and Spacing Algorithms for Arrival Traffic, NASA/TM-103880, National Aeronautics and Space Administration, Ames Research Center, 1991. + + + + + Fuel Cost, Delay and Throughput Tradeoffs in Runway Scheduling + + HLee + + + HBalakrishnan + + + 2008. 2008 + IEEE + + + + American Control Conference + Lee, H. and Balakrishnan, H., "Fuel Cost, Delay and Throughput Tradeoffs in Runway Scheduling," American Control Conference, 2008 , IEEE, 2008, pp. 2449-2454. + + + + + Separation-Compliant Time Advance in Terminal Area Arrivals: Tradeoff between Makespan and Fuel Burn + + AlexanderVSadovsky + + + BrianFabien + + 10.2514/6.2013-4783 + AIAA-2013-4783 + + + AIAA Guidance, Navigation, and Control (GNC) Conference + + American Institute of Aeronautics and Astronautics + 2013 + + + Sadovsky, A. and Fabien, B., "Separation-Compliant Time Advance in Terminal Area Arrivals: Tradeoff between Makespan and Fuel Burn," AIAA Guidance, Navigation, and Control (GNC) Conference, AIAA-2013-4783, 2013. 11 + + + + + Airport Arrival Capacity Benefits Due to Improved Scheduling Accuracy + + LarryMeyn + + + HeinzErzberger + + 10.2514/6.2005-7376 + + + AIAA 5th ATIO and16th Lighter-Than-Air Sys Tech. and Balloon Systems Conferences + + American Institute of Aeronautics and Astronautics + 2005 + + + Meyn, L. A. and Erzberger, H., "Airport Arrival Capacity Benefits Due to Improved Scheduling Accuracy," AIAA Aviation, Technology, Integration and Operations (ATIO) Conference, 2005. + + + + + Evaluation of the terminal area precision scheduling and spacing system for near-term NAS application + + JThipphavong + + + LMartin + + + HSwenson + + + PLin + + + JNguyen + + 10.1201/b12321-12 + + + Advances in Human Aspects of Aviation + + CRC Press + 2012 + + + + Thipphavong, J., Martin, L., Swenson, H., Lin, P., and Nguyen, J., "Evaluation of the Terminal Area Precision Scheduling and Spacing System for near-term NAS application," 4th International Conference on Applied Human Factors and Ergonomics (AHFE), 2012. + + + + + Trajectory Synthesis for Air Traffic Automation + + RhondaSlattery + + + YiyuanZhao + + 10.2514/2.4056 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 20 + 2 + + March 1997 + American Institute of Aeronautics and Astronautics (AIAA) + + + Slattery, R. and Zhao, Y., "Trajectory Synthesis for Air Traffic Automation," Journal of Guidance, Control and Dynam- ics, Vol. 20, No. 2, March 1997, pp. 232-238. + + + + + GPOPS-II Version 1.0: A General-Purpose MATLAB Toolbox for Solving Optimal Control Problems Using the Radau Pseudospectral Method + + MAPatterson + + + AVRao + + + 2013 + + + University of Florida + + + Patterson, M. A. and Rao, A. V., "GPOPS-II Version 1.0: A General-Purpose MATLAB Toolbox for Solving Optimal Control Problems Using the Radau Pseudospectral Method," University of Florida, 2013. + + + + + Introduction to Aircraft Performance, Selection, and Design + + FJHale + + + 1984 + Wiley + New York + + + Hale, F. J., Introduction to Aircraft Performance, Selection, and Design, Wiley New York, 1984. + + + + + Advanced Aircraft Performance Modeling for ATM: Enhancements to the Bada Model + + ANuic + + + CPoinsot + + + M.-GIagaru + + + EGallo + + + FANavarro + + + CQuerejeta + + 10.1109/dasc.2005.1563320 + + + 24th Digital Avionics Systems Conference + + IEEE + Nov. 2005 + + + Nuic, A., Poinsot, C., Iagaru, M.-G., Gallo, E., Navarro, F. A., and Querejeta, C., "Advanced Aircraft Performance Modeling for ATM: Enhancements to the BADA Model," Proceedings of the IEEE/AIAA 24th Digital Avionics Systems Conference, Nov. 2005. + + + + + + diff --git a/file803.txt b/file803.txt new file mode 100644 index 0000000000000000000000000000000000000000..577e7429507e7c185ff6a178a7a4776b96c79bc6 --- /dev/null +++ b/file803.txt @@ -0,0 +1,419 @@ + + + + +I. IntroductionIn the national airspace system, high traffic demand and inefficient airspace designs cause overloaded sectors.Because some overloads can be eliminated by redesigning sector boundaries, the congested sectors impose unnecessary delays and traffic rerouting.These problems become worse when traffic patterns and demands fluctuate.An automatic algorithm is needed to design airspace boundaries such that overloads can be reduced.Then, to accommodate the fluctuating traffic patterns and demand, the automatic algorithm can be applied as needed to address the current airspace constraints.2][3][4][5][6][7][8][9] However, these methods are still facing some challenges.For instance, the GA-based algorithms [1][2][3] struggle with handling a large volume of real traffic.The cell-based algorithms [4][5][6][7] produce jagged boundaries or unacceptable sector shapes.The sector boundaries generated by the computational geometry algorithm 8 so far cannot conform to major flows.Also, as shown by Zelinski, 10 the flight clustering based algorithm 9 produces too many sectors.In previous work, 11 an approach that combines Voronoi diagrams, a genetic algorithm, and an iterative deepening algorithm was proposed.The method guarantees that the sector boundaries are convex and accepts nonlinear objective functions other than aircraft count.Designs produced by the method also conform to the dominant flows.However, the number of sectors was predefined and fixed, and only twodimensional partitions were allowed.These restrictions constrained the Voronoi-based method's ability to achieve efficient sector designs.The current work extends the Voronoi-based method by automatically selecting the number of sectors and allowing three dimensional partitions.Automatic selection of the number of sectors enhances the method's ability to create sector boundaries that conform with traffic patterns.The new method includes three costs: primary, secondary, and design cost.The primary and secondary costs were empirically designed for traffic pattern conformance.The design cost was set up to judge if a sector is overloaded or not.The method was applied to sector designs at Fort-Worth and Indianapolis centers with different traffic scenarios.The new designs have self-defined numbers of sectors and conform to traffic patterns.Compared with the previous method, the designs generated by the enhanced method can eliminate or reduce overloads without changing the traffic.In this paper, a preliminary study of Voronoi and the iterative deepening algorithms is described in Section II.Section III presents the new design strategy.Section IV describes cost functions.Section V presents the results and comparisons.Conclusions are discussed in Section VI. +II. Preliminary StudyIn order to enhance the previous Voronoi-based method, studies must be conducted to examine the strength and weakness of included algorithms.In the following sections, the limitations of Vornoi diagrams are investigated and the way the iterative deepening algorithm compensates for these limitations is studied.Finally, the benefits of vertical partitions are discussed. +A. Voronoi Diagrams and Iterative Deepening AlgorithmGiven a set of generating points, Voronoi diagrams can partition space into proximal regions such that any point in the region is associated with the closest generating point. 12One of the properties of Voronoi cells is that they are always convex. 12There are a great variety of Voronoi diagrams 12 depending on the way of defining generators (points, lines, areas, or 3D objects), the way of defining distance (Euclidian distance, Hausdorff distance, etc), and the way of weighting distance to generators.The Voronoi adopted in this work is the ordinary Voronoi diagrams that apply point generators, Euclidian distance, and uniform weights on distances.The property of partitioning space based on generating points reduces the number of parameters for optimization.An optimizer only needs to adjust the locations of the generating points to find a good sector design.However, using the ordinary Voronoi diagrams alone might over-restrict the design space, especially when the number of partitions becomes high.In experiments, when a center was divided into a large number of sectors in one step, the resulting performance cost was not improved much, even after many generations of GA.It was found that, as the number of partitions increases, the Voronoi design space could be more limited.This implies that the number of partitions should not be too high when applying Voronoi diagrams directly at a single step.In order to cope with this restriction, the iterative deepening algorithm is used to break up a partition with a large number of sectors into multiple steps.To illustrate the principle, a theoretical case where the traffic is not real will be used.Suppose there exists a dominant traffic flow surrounded by several intersection areas, as shown in Fig. 1(a).Assuming seven partitions is the best choice for this structure, if Voronoi diagrams are applied directly with seven partitions, the dominant flow cannot be contained in a single sector.In fact, as shown by the blue lines in Fig. 1(a) that were generated by Voronoi directly, Voronoi segregates the flow into several parts.However, if the iterative deepening algorithm is exploited, this airspace can first be separated into three parts, as shown by blue lines in Fig. 1(b).Then two of the parts can be further partitioned in accordance with the traffic pattern.The resulting partition, shown by the blue and green lines in Fig. 1(b), is more desirable because it does not break the main flow.Thus, the iterative deepening algorithm can compensate for the limitations of Voronoi diagrams. +IntersectionIn this work, the number of divisions at a single step is selected to be less than six.Although, five or six divisions may still be high, however, they offer flexibility.For example, Voronoi diagrams can easily form a good partition for a typical hub-like traffic pattern. +B. Vertical PartitionsIf only horizontal partitions are allowed, sectorization is constrained.Fig. 2(a) and 2(b) show an example of how vertical segregation can help create a good partition.Suppose, as in the scenario shown in the figure, there are two groups of aircraft flying at two different levels.(For simplicity, instead of symbols of aircraft, only time-elapsed trajectories are shown in this figure .)Assuming the capacity is approximated based on average dwell time and the demand is based on peak aircraft count (e.g. using the criterion of "P eakAircraf tCount > 5/3 × AverageDwellT ime" a ), this airspace is overloaded because the capacity estimation is 25 and peak count is 30.Further division may be needed.If the airspace is segregated with the green dash lines in the figures, assuming aircraft are evenly distributed at the peak time, both sector average dwell time and peak aircraft count will be halved to 12.5 and 15, respectively.Thus the resulting two sectors are still overloaded.Whereas if the airspace is vertically stratified using the blue dash line, only the peak count is halved and sector dwell is kept the same.The resulting two sectors would not be overloaded anymore.Therefore, in this case a vertical partition is desired, both for maintaining good controllers' situation awareness and meeting the design requirement.These costs are discussed further in subsection E. It is preferable not to use too many vertical partitions because entering a sector vertically causes more workload than horizontal boundary crossings.In this work, a threshold (e.g. 1, 500ft) was set up to limit the smallest vertical range for a single sector.Additionally, in terms of when to stratify vertically and when to stop, three types of rules are tested in this work: conservative, moderate, and aggressive.In the "conservative" regime, stratification is applied only when, right after the application, the design requirement can be satisfied.In the "moderate" and "aggressive" regimes, even though the immediate consequences of vertical partitions do not satisfy the design cost, stratifications will be performed if horizontal partitions cannot produce a design that is closer to the design requirement.The difference between "moderate" and "aggressive" is that, in "moderate" regime, although the requirements cannot be satisfied after stratification, the resulting divisions will not be further divided, whereas in "aggressive" cases, the resulting divisions are still open for further partitioning.a The five thirds is calculated from the table in FAA Order 7210.3 using the curve-fitting method.This formula has been used as a simple capacity estimation in many airspace redesign studies +III. ApproachBased on the above discussions, a new tri-cost strategy that uses three costs -a primary cost, a secondary cost, and a design cost -is established.Optimization with the primary cost is repeated for different numbers of sectors to create a set of eligible designs.The secondary cost is used to judge the preference among designs.The design cost is utilized as a criterion to determine the ultimate number of sectors.When the design cost reaches a defined threshold, the partitioning process stops.Step 2 +IsReset N = 0, Clear "sub-database" design cost > thresh ?A flow chart is shown in Fig. 3 to illustrate the detailed procedure.In this chart, H represents the number of horizontal partitions, and V denotes the number of vertical partitions.H 0 is the limit of the number of horizontal partitions, and V 0 is the limit of the number of vertical partitions.In this work, H 0 and V 0 are set to be six and three, respectively.N is a temporary count of sectors stored in the sub-database.A "divisible" airspace can gain capacity by further division.Whereas an "indivisible" airspace implies that dividing the airspace will reduce capacity, thus it is good not to be divided.A "satisfied" airspace is an airspace that meets design cost or requirement.The "conservative" regime is shown in the flow chart.The following are detailed descriptions:Yes NoStep 1 The algorithm first checks if partitioning is needed for a given airspace according to the design requirement.If no, the airspace is marked with "satisfied" and sent to the main database, and the process goes to Step 3. If yes, optimal sectorization is performed based on the primary cost function.Different numbers of partitions including vertical stratifications are generated.In this work, the exploration range is set at between two and six partitions for horizontal segregation and between two and three for vertical division.If any resulting partitions meet the design requirement, they are marked with "satisfied".In this step, if any partition improves the design cost, it will be stored in "sub-database" for Step 2. If no partition that provides better design cost than upper-iteration sub-airspace, the initial sub-airspace is marked "indivisible".Step 2 A secondary cost is utilized to judge the partitions in the "sub-database" to see which partition among them may lead to a good design.The winner's partitions are marked with "divisible" and sent to the main database.Meanwhile, the "sub-database" gets cleared, and N is reset to zero.Step 3 All sub-airspaces stored in the main database are checked.If there is any "divisible" sub-airspace or sub-sectors, the algorithm will pick one and go to Step one.If all sub-airspace are either "satisfied" or "indivisible", they will be output as final sectors and the entire process will be terminated. +IV. ObjectivesIn this work, efforts were dedicated to developing new costs on the basis of previous work.The primary and secondary costs were identified based on numerous experiments.Before describing these costs, the next few sections present key metrics used in costs, including the sector average dwell time, the intersection proximity, the dominant flow proximity, the aircraft count of short dwell flights, and the variance of peak aircraft count.Straightforward metrics such as the sector boundary crossings are skipped.Later sections describe how the costs are constructed. +A. Average Sector Dwell Time (M )As discussed in previous work, the average sector dwell time M j is defined as Eqn. 1, where T ij is the dwell time for flight i at Sector j, and n j is the total number of aircraft in Sector j.Experiments show the dwell time is a critical metric for a good sector design.As demonstrated in Fig. 4(a), maximizing dwell time will lead to the sector boundary with the blue edges rather than the red one, since the blue one has higher dwell time than the red one.Similarly, in a region which includes a dominant flow, as in Fig. 4(b), the cost of dwell time metric will prefer the blue design instead of the red one.A similar result is obtained when handling 3D traffic.As displayed in Fig. 4(c), the entire major ascent or descent flow will be included in the blue boundaries in order to maintain a higher dwell time than would be obtained with the red boundaries.M j = i T ij n j(1) +B. Intersection Proximity (P I )However, using merely the dwell time is not enough.Adjustment should be made by incorporating an intersection-related cost.To set up such a cost, intersection areas should be identified first.Fig. 5 shows the time-elapsed traffic above FL240 of Fort-Worth (ZFW) center for an entire day.Each pixel corresponds to a grid point with a resolution of 0.01 degree in both latitude and longitude.In order to find the intersection grid points, a window is first defined for each grid point, and the orientation of all traffic entering or leaving the window is computed, as shown in Fig. 6(a).The slope of a trajectory as it crosses the horizontal axis is used to represent the orientation.As a result, a trajectory going in and out of the window will have only one orientation.A histogram of orientations at each pixel is calculated based on the surrounding window.The range of orientations is then divided into 18 bins with 10 degrees for each bin.These bins act as accumulators to count the orientations in a given range.A sample histogram at the pixel of longitude -97.1 • E and latitude 32.9 • N is displayed in Fig. 6 Next a threshold is set to be the average pixel value of the entire center.For example, the threshold for the ZFW center is 51.37 and is shown as a red horizontal line in Fig. 6(b).The bars that are higher than the threshold are kept and the others are discarded.Consecutive bars among those that remain are treated as one.At the end of this process, the number of bars that remain is counted as the "intersection index" for that pixel.Any pixel that has "intersection index" greater than or equal to two will be treated as an intersection point.The remaining bars for the sample point are labeled No.1 to 5 in Fig. 6(b), thus the "intersection index" is five, which roughly corresponds to the five distinct flow directions in Fig. 7(a).The An image composed of intersection indexes can be created, as presented in Fig. 7(b).Pixel values in this figure are set equal to corresponding intersection indices.Dark blue means no intersection while a "hot" color denotes a major intersection.To construct a scalar cost, a Gaussian-like penalty (Eqn.2) is defined according to the distances between sector boundaries and these intersection points.d i is the minimum distance between pixel i and any boundary.In order to simplify the computation, only the penalty from the closest boundary is taken into account.Γ is a distance threshold that is defined to trigger the penalty.I i is the index value of pixel i, and c 1 and σ 1 are defined constants.P I = i (d i < Γ) • I i • e - d i •c 1 2 2•σ 2 1(2)C. Dominant Flow Proximity (P D )A boundary may lie on a major flow when pursuing long dwell time.Although the number of sector boundary crossing may not be significant, it definitely lowers the situation awareness of controllers because the boundary is too close to the trajectories.This should be avoided in sector designs, and thus a penalty is needed to prevent this.The penalty should push the boundaries away from the flows or at least let the boundaries cross the flows with near-perpendicular angles.Although there exist other ways to identify dominant flows, the method used here is the same as the one for identifying intersection areas, but with the index threshold changed from two to one.Any pixel with a nonzero index will be recognized as a dominant flow point.The identified dominant flows in ZFW center are shown in Fig. 8.Note that the dominant flows match well with the time elapsed traffic in Fig. 5. Similarly, a Gaussian-like penalty function is shown in Eqn. 3, where c 1 and σ 2 are defined constants, and others have the same meanings as in the intersection penalty.P D = i e - d i •c 1 2 2•σ 2 2 i I i > 0 (3) D. Aircraft Count of Short Dwell (N s )Because the primary costs try to increase the "average" of sector flight dwell time, some undesired extremely short dwell flights may remain.In experiments, according to observations, most cases can be ameliorated by a minor adjustment.Thus, a cost based on the aircraft count of short dwell is built in.In this work, the flights that have a sector dwell time less than three minutes are counted and used for setting the penalty. +E. The Variance of Sector Peak Aircraft Count (V p )If there is no constraint for balancing of sector peak aircraft counts, the rule of "long dwell for high peak" may lead to a partition with tiny pieces containing low traffic areas.The process would be like pie-biting, which is not preferred.Thus the variances of the peak aircraft counts (V p ) for the partitions at the same level are applied to penalize such extreme cases.High costs will cancel out the benefits of long dwell if the partitions are extremely-unbalanced. +F. Primary CostA scalar primary cost is computed based on the fundamental metrics discussed above as well as the results of numerous experiments.The optimization is to maximize the primary cost described as:F p = min j c0•Mj (Np)j a 1 • a 2 • a 3 • a 4(4)where:a 1 = e P I + c 2 • P D(5)a 2 = c 3 + N c N • (N p -1)(6)a 3 = c 4 + N s N(7)a 4 =    c 5 + V p , for N a > Θ , or N a ≤ Θ and V p < N a • c 6 , ∞, for N a ≤ Θ and V p ≥ N a • c 6 .(8)The basic principle is to allocate long dwell time for regions with high peak aircraft, which is shown as the numerator on the right side of Eqn. 4. M j is proportional to the dwell time in sector j, and (N p ) j is the peak aircraft count in sector j.The remaining "a"s are costs for tuning designs.The traffic pattern related penalty is composed of the proximities to intersections and major flows.This is shown in Eqn. 5, where the P I is the penalty for the intersection proximity and P D is for major flows.All "c"s are constants defined based on experimental observations.The second adjustment is a function of sector boundary crossings as in Eqn.6, where N c is the count of sector boundary crossings, N is total number of aircraft, and N p is the number of partitions.Then the adjustment for extremely short dwell time is established in Eqn. 7. N s is the count of flights that have a sector dwell time less than three minutes.The final adjustment shown in Eqn. 8 is related to the variance of sector peak counts, where V p denotes the variance of peak aircraft counts, N a is the average peak aircraft count over sectors, and Θ is a threshhold defined to strengthen the variance constraint when partitions approach the end of the process or sectors become small. +G. Secondary CostAt each level of the iterative deepening algorithm, a secondary cost is needed to judge which partition among the set of partitions may be good and would lead to a good final partition.The purpose of the secondary cost is to make sure the division follows the nature of the traffic pattern.In this work, based on the observations of experiments, the crossing angles between trajectories and sector boundaries are identified as the key component.The crossing angles for an ideal partitioning should be close to a right angle, which means that in the ideal case the flows either don't cross the sector boundaries or cross them with near-perpendicular angles.The number of crossings should also matter.Therefore the sum of the shallow crossing angles is used as a secondary cost as shown in Eqn. 9.Here θ k is the crossing angle between a trajectory and a sector boundary for the kth crossing.Of course, one flight can have multiple crossings.Only shallow-angle crossings are counted and Φ is a define threshold for a shallow angle.F s = k [(θ k -90 • ) -Φ > 0] • [(θ k -90 • ) -Φ](9) +H. Design CostDesign cost serves as a stop criterion.In following experiments, two design costs are used.The first one is the cost for increasing capacity shown in Eqn.10.The average sector dwell time is used to estimate sector capacity, and the sector peak aircraft count is applied as an estimation of complexity.Then, the residual capacity based on these two estimations are maximized to increase the capacity.Additionally, the high bound of the calculated capacity is limited to be 18, and the lower bound of the demand (the peak aircraft count) is limited to be five.Another design cost requires the peak count of each sector not to exceed a threshold as in Eqn.11.M odel I :      Capacity ≥ Demand, Capacity = 5 3 • average sector dwell time, Demand = sector peak aircraf t count. +V. ResultsIn this work, simulated traffic data were used for sector designs.The traffic was generated from the flight schedule for August 20, 2005, and simulated using the Airspace Concept Evaluation System (ACES) 13 based on filed flight plans without capacity constraints in both airports and en-route airspaces.Detailed description can be found in Zelinski's work. 10Both design requirements mentioned in the previous section were tested as stop criteria.The results were obtained on a Mac-Pro computer with a dual-core CPU at 2.8MHz and with the GA code multi-threaded.In the experiments, unless otherwise specified, the design cost of Model I in Eqn 10 was applied and the "conservative" rule was used for vertical stratifications.To demonstrate the benefits, several cases are studied and compared.The results for high altitude Fort-Worth(ZFW) center are presented.Then the results for combined airspace with FL240 and above are presented and compared.In the comparisons, different design costs are also applied.Finally, design results for Indianapolis(ZID) center, which is typically heavy-loaded and intersection-dominated, are compared to show the adaptability of designs to the flow patterns. +A. Sector Design Comparison for High Altitude Fort-Worth CenterIn order to demonstrate the benefit of the new tri-cost strategy over the original method, several cases were studied for high altitude using the simulated unconstrained traffic data between FL240 and FL350 for Fort Worth center (ZFW).First, the results from the original method are presented.Then, the results from the tri-cost method with and without 3D are discussed, respectively.Finally, comparisons are made. +Case I: Predefined Number of SectorsIn this case, the original Voronoi-based method from previous work was applied.The design cost of increasing capacity was directly used as the optimization cost.The overall number of partitions and the number of divisions at each level were fixed and defined before the experiments.For example, given a 15-sector design, the first level of partition segregated the airspace into three sectors, and the second level divided each of those into five.Fig. 9(a) shows a 15-sector design, whereas Fig. 9(b) shows an 18-sector design.In the figures, the "C" means the calculated capacities, and the "P" means the peak aircraft counts of the sectors.If a sector is overloaded, which means the "P" is higher than the "C", these symbols and values are red, otherwise they are yellow.Because the number of partitions at each level is blindly defined and fixed in advance, parts of the airspace have excessive sectors while others need more sectors.As shown in Fig. 9(a), sector C and D are still overloaded in this 15-sector design.Although an 18-sector design eliminates the overloaded sectors, it introduces unnecessary sectors in region B as shown in Fig. 9(b).Because the design cost is used directly in the optimization, there exist several prohibitive boundaries in both designs.For instance, in Fig. 9(a), sector boundaries are too close to major flows in the A and B regions.And in Fig. 9(b), sector boundaries cross the intersection area in region A and the major flows in regions B, C, and D. +Case II: Tri-cost Strategy Without Vertical StratificationsIn this case the new tri-cost method was applied, but vertical stratification was not allowed.The design results are presented in Fig. 10(a).At the first level, a four-sector partition (shown as green dash lines) was picked by the secondary cost.It appears that the secondary cost performs well, because the four-sector partition follows the hub-like traffic pattern and provides a good beginning for subsequent partitions.In the final design, the regions surrounding the center are well sectorized.Sectors such as A and B conform to the major flows.The only undesired part is that the center area containing major flows has been divided into five sectors, C, D, E, F, andG.In order to meet the design requirement, this kind of partitioning sacrifices controllers' situation awareness because of the proximity between boundaries and flows.Counting these five sectors, the number of sectors required becomes 16. +Case III: Tri-cost Strategy With Vertical StratificationsThe previous case demonstrates that when a region contains only one integral traffic pattern, further horizontal divisions may damage the integrity of the traffic pattern and lower controllers' situation awareness.In this situation, appropriate vertical partitions may resolve the dilemma of either keeping the traffic pattern or meeting design requirements.Fig. 10(b) shows the design produced by the complete version of the tri-cost strategy, which allows "conservative" 3D sectorization.The total number of sectors in this design is 14.Similar to the previous case, a four-sector partition was picked at the first level from the explorations of different partitions.In the final design, the surrounding sectors, especially A and B, follow the traffic patterns well.The center region C is vertically partitioned into two layers.One is from FL240 to FL290 and another from FL290 to FL350.Both layers satisfy the design requirement.With the vertical division, the pattern in region C has been well preserved and the total number of sectors is also lowered.In fact, the number of boundary crossings between the upper and lower layers is 244, which is lower than the 254 crossings among five sectors in the previous case. +ComparisonsWith the original Voronoi-based method, Case I could not allocate the sectors efficiently.Unnecessary sectors increased the total number of partitions and could damage controllers' situation awareness.Without the pattern-related cost, sector boundaries were allowed to cross the intersection areas and encroach major flows.Case II utilized the tri-cost method without vertical partitions.Although some problems in Case I were resolved, it created a trade-off situation between reaching the design requirements and keeping integrated traffic patterns within the sectors.Case III resolved these issues using the complete tri-cost method.Appropriate vertical partitions were applied to remove unnecessary partitions and both the integrity of traffic patterns and design requirements were satisfied.In Table 1 several metrics are presented for these cases in order to show the differences quantitatively.The last column shows the measurements for current ZFW sectors.Although the altitude ranges of current ZFW sectors are FL240 and above, to be comparable, the same range of traffic data (FL240 to FL350) was used to calculate those metrics.The "Dominant Flow Proximity" and "Intersection Proximity" are normalized cost values.A low value means that boundaries are away from flows and intersections.The lower values in the table are preferred and shown in bold.The comparisons show that Case III has the most preferred values and the new tri-cost method is a substantial improvement over the original method. +B. Sector Design for Combined High and Super-high Fort-Worth CenterIn the above studies, the sectorization was applied to the airspace between FL240 and FL350, with FL350 chosen as an artificial split between high and superhigh airspace.The purpose of the next set of experiments was to design for combined high and super-high airspace (FL240 and above) without such artifacts.The combined airspace in ZFW center had 4, 210 flights on Apr.21 compared with 3, 493 flights in the high sectors alone.Three experiments will be presented in the following sections.Increasing capacity was used as design costs in the first two, denoted as "CAP", and the peak aircraft count was used in the third one, denoted as "PEAK". +CAP2D: Capacity Increasing Design Cost without 3D PartitionIn the first experiment, only horizontal partitions were allowed with the tri-cost strategy.The results are shown in Fig. 11(a).In this design most of the sectors capture the flow patterns.However, without vertical partitions, the algorithms have been pushed to the limit.Even though there are 22 sectors, the design still has two saturated sectors, A and B. The sizes of many sectors are small as well, especially sectors C and D. +CAP3D: Capacity Increasing Design Cost with 3D PartitionThe next experiment allowed 3D partitions, and the results are shown in Fig. 11(b).With 3D partitions allowed, the integrity of the hub has been preserved.The 3D design has no overloaded sector, and the sector count is only 18.This shows that three dimensional partitions not only produce preferred sector shapes but also dramatically enhance the capability of meeting design requirements.The comparison further testifies that appropriate 3D partitions have a significant advantage over 2D partitions. +PEAK3D: Peak Count Capping Design Cost with StratificationThe third case explored a different requirement to show the adaptation of the method.The second design requirement in Section IV.H, which requires the sector peak count not exceed 15, was utilized.Fig. 12 presents the final design.It shows that major traffic patterns have been captured by pursuing the peak aircraft count.The total number of sectors 17 is similar to previous ones. +ComparisonComparisons of the above designs are shown in Table 2. "Requirement 1" denotes the first design cost of increasing airspace capacity, while "Requirement 2" refers to the second cost of capping peak aircraft counts.The optimal values for metrics are in bold.Both "CAP3D" and "PEAK3D" achieved their respective design goals.In the table, all optimal values occured in "CAP3D" and "PEAK3D"."PEAK3D" has the lowest intersection proximity while "CAP3D" has the rest of the optimal values. +C. Sector Design for Heavy-loaded AirspaceTo explore how the method behaves when sectorizing a heavily-loaded airspace, the new method was applied to ZID center.On the same day, ZID had 6, 401 flights based on simulated data.That is 50% more than the traffic volume of ZFW and even more in terms of traffic density because ZID is smaller than ZFW.The other reason for picking the ZID center is that ZID has traffic patterns that are dominated by intersections, which is different from the hub-like patterns in ZFW.Because of the high traffic volume, the "moderate" or "aggressive" stratification was also used for vertical partitions.The first two designs used the requirement of increasing capacity.The third design applied the requirement of capping peak count. +CAP3DMOD: Capacity Increasing Design Cost with "Moderate" StratificationThe first "CAP3DMOD" uses the "moderate" rule for vertical partitions, and the result is shown in Fig. 13.The result again demonstrates that when the new method is used, the intersection areas and major patterns are well preserved.Due to the high volume of traffic, even though 3D partitions were used, 11 out of 49 sectors are still overloaded.Regions A and B are stratified because the resulting sectors lower the severity of saturation, although the resulting sectors are still overloaded.The process stopped there because no improvement could be achieved with further divisions of the airspace.Region C is relatively better than A and B because one of the vertical stratifications is not overloaded, and the overloaded sectors exceed capacity by only one flight.To examine the function of different rules in vertical stratification, the "aggressive" stratification was applied.The result in Fig. 14 shows that the airspace volume of overloaded sectors was less, but the number of sectors increased.13 out of 62 sectors are saturated.By carefully examining the region surrounded by the green polygon and comparing it to Region A in the "moderate" case, one can see that the "aggressive" rule stratified the region into six layers.Five of them are not overloaded anymore, while the bottom layer from FL240 to FL305 is laterally segregated into three sectors, and one becomes unsaturated.So with a different rule of vertical partition, the exact same region has been divided into eight sectors with two overloaded, in stead of three saturated sectors in the previous experiment.Thus, an aggressive use of vertical partitions may increase the overall capacity and may cause fewer delays, but more stratifications can happen.Overall, these two experiments imply that if the calculated capacity is used as the criterion, then there exists a limit of capacity that a design can provide with a given rule of vertical partition, and further partitions will not increase the capacity anymore. +CAP3DAGR: Capacity Increasing Design Cost with "Aggressive" Stratification +PEAK3DCON: Peak Count Capping Design Cost with "Conservative" StratificationThe third design applied the requirement of capping peak count.The results, shown in Fig. 15, capture the intersection areas and major flows as well.Note that in order to meet the requirement, only 36 sectors were needed and all sectors except one have 15 or fewer aircraft at any one time.The exception is due to the constraint on sector size that was imposed in the design.Because many aircraft were lumped together in that small area, the peak aircraft count could only be lowered by generating small sectors.Theoretically, any peak count requirement could be achieved by chopping the sector into smaller and smaller pieces.From the above experiments, it can be seen that it is easier to fulfill the requirement of capping peak count compared to increasing sector dwell time.Table 3 presents a comparison of these two designs with current sectors.The bold values are the lowest compared with others.Obviously, the current system has a high number of violations on both requirements.It is noted that current sectors have the lowest boundary crossings and dominant flow proximity cost.The reason is that the current sectors are much fewer in number, which causes unfair comparisons in these two metrics.As an extreme example, if the center contains only one sector, these two values would be zero.Taking this fact into account, the new designs perform much better on these metrics.The "CAP3DMOD" and "PEAK3DCON" have the fewest number of violations on the capacity and peak aircraft count limit, respectively.Both follow the traffic pattern well, based on the pattern-related metrics. +VI. ConclusionsThis paper presents an enhanced Voronoi-based sector design method, which automatically selects the number of sectors and allows 3D partitions.New costs enforcing traffic pattern conformance have been designed based on numerous experiments.It was found that Voronoi diagrams have the advantage of simplifying the graph partitioning problem but they have limitations when the number of partitions becomes high.Using the iterative deepening algorithm was found to compensate for the limitations caused by Voronoi diagrams, instead of simply speeding up the computation.It is also demonstrated that vertical partitions can enhance sector designs.The experiments show that the new traffic pattern related costs have the capability of avoiding intersection areas, and enhance the ability of aligning with major flows.In the experiments these new costs lead to designs that capture trajectory patterns, lower sector hands-off, and meet the user's requirements.Meanwhile, the results also show that the secondary cost can pick out the preferred partitions among the options with different numbers of partitions.It was also found that vertical partitions can be used to preserve integrated traffic patterns while increasing the ability to achieve the user's design requirements.Since new method enlarges design space by using iterative deepening and vertical partition, it results in less necessary sectors while meeting the capacity constraints.Overall, the new methodology, along with the new costs, substantially improves the original one and makes the sector designs more acceptable.Figure 1 .1Figure 1.A sample of airspace partition.(a) Voronoi diagrams only.(b) Voronoi diagrams and the iterative deepening algorithm +Figure 2 .2Figure 2. A sample of vertical partition (a) top view (b) side view +Figure 3 .3Figure 3. Procedure of the tri-cost strategy +Figure 4 .4Figure 4. Function of dwell Time.(a) intersection area.(b) 2D dominant flow.(c) 3D dominant flow. +Figure 5 .5Figure 5.Time elapsed traffic of Fort-Worth (ZFW) center (FL240-999) +(b). +Figure 6 .6Figure 6.Computation of intersection index.(a) computation of flow orientations.(b) histogram of flow orientations. +(a) Computation of Flow Orientation (b) Distribution of Flow Orientation +Figure 7 .7Figure 7. Intersection index map of ZFW center (FL240-999) +Figure 8 .8Figure 8. Dominant flow in ZFW center (FL240 and above) +( 10 )10M odel II : sector peak aircraf t count <= a threshold(11) +Figure 9 .9Figure 9. Designs for ZFW center using original method (FL240-350) (a) a 15-sector design.(b) a 18-sector design. +Figure 10 .10Figure 10.Designs for ZFW center using new method (FL240-350) (a) horizontal partition only.(b) threedimensional partition. +Figure 11 .11Figure 11.Designs to increase capacity for ZFW center (FL240 and above) (a) CAP2D: horizontal partition only.(b) CAP3D: three-dimensional partition. +Figure 12 .12Figure 12.PEAK3D: sector design to cap peak counts with 3D partition for ZFW center (FL240 and Above) +Figure 13 .13Figure 13.CAP3DMOD: Sector design to increase capacity for ZID center ("moderate" stratification) +Figure 14 .14Figure 14.CAP3DAGR: Sector design to increase capacity for ZID center ("aggressive" stratification) +Figure 15 .15Figure 15.PEAK3DCON: Sector design to cap peak count for ZID center (FL240 and above) +Table 1 .1Comparison of Designs for ZFW from FL240 to FL350Case IA Case IB Case II Case III CurrentNumber of sectors1518161419# of overloaded sectors (Model I)20001Boundary crossings2,6982,8222,3682,4712,851Variance of peak counts97%94%69%47%69%Dominant flow proximity cost2.893.702.392.142.96Intersection proximity cost253.7271.74.550.9744.6 +Table 2 .2Comparison of Designs for ZFW center FL240 and aboveCAP2D CAP3D PEAK3D CurrentNumber of sectors22182219# of overloaded sectors (Model I)2046# of overloaded sectors (Model II)6708Boundary crossings6,5834,8896,1605,579Variance of peak counts60.3%43.8%53.5%78.5%Dominant flow proximity cost3.552.963.303.16Intersection proximity cost74.173.860.8288.5 +Table 3 .3Comparison of Designs for ZID center FL240 and aboveCAP3DMOD CAP3DAGR PEAK3DCON CurrentNumber of sectors49623628# of overloaded sectors (Model I)11131920# of overloaded sectors (Model II)64115Boundary crossings15,24116,77513,53912,341Variance of peak counts97.1%142%55.5%66.9%Dominant flow proximity cost5.05.24.13.9Intersection proximity cost37.9107.131.1391.0 + + + + + + + + + Genetic Algorithms for Automatic Regrouping of Air Traffic Control Sectors + + DDelahaye + + + JAlliot + + + MSchoenauer + + + JFarges + + 10.7551/mitpress/2887.003.0057 + + + Evolutionary Programming IV + + The MIT Press + 1995 + + + Delahaye, D., Alliot, J., Schoenauer, M., and Farges, J., "Genetic Algorithms for Automatic Regraupement of Air Traffic Control Sectors," Proceedings of the International Conference on Evolutionary Programming, 1995. + + + + + Airspace sectoring by evolutionary computation + + DDelahaye + + + MSchoenauer + + + J-MAlliot + + 10.1109/icec.1998.699504 + + + 1998 IEEE International Conference on Evolutionary Computation Proceedings. IEEE World Congress on Computational Intelligence (Cat. No.98TH8360) + + IEEE + 1998 + + + Delahaye, D., Schoenauer, M., and Alliot, J., "Airspace Sectoring by Evolutionary Computation," Proceedings of IEEE International Congress on Evolutionary Computation, 1998. + + + + + Airspace sectorization with constraints + + HuyTrandac + + + PhilippeBaptiste + + + VuDuong + + 10.1051/ro:2005005 + + + RAIRO - Operations Research + RAIRO-Oper. Res. + 0399-0559 + 1290-3868 + + 39 + 2 + + June 23-27 2003 + EDP Sciences + Budapest, Hungary + + + Trandac, H. and Duong, V., "Optimized Sectorization of Airpace with Constraints," 5th Erocontrol/FAA ATM R&D Seminar , Budapest, Hungary, June 23-27 2003. + + + + + Temporal and Spatial Distribution of Airspace Complexity for Air Traffic Controller Workload-Based Sectorization + + ArashYousefi + + + GeorgeDonohue + + 10.2514/6.2004-6455 + + + AIAA 4th Aviation Technology, Integration and Operations (ATIO) Forum + Chicago, IL + + American Institute of Aeronautics and Astronautics + September 2004 + + + Yousefi, A. and Donohue, G., "Temporal and Spatial Distribution of Airspace Complexity for New Methodologies in Airpace Design," 4th AIAA Aviation Technology, Integration, and Operations conference (ATIO),, Chicago, IL, September 2004. + + + + + Airspace availability estimation for traffic flow management using the scanning method + + AlexanderKlein + + + LaraCook + + + BryanWood + + 10.1109/dasc.2008.4702802 + + + 2008 IEEE/AIAA 27th Digital Avionics Systems Conference + + IEEE + + + + th + Klein, A., "An Efficient Method for Airspace Analysis and Partitioning Based on Equalized Traffic Mass," 6th + + + + + + + Usa/ + + + + Europe Seminar on Air Traffic Management Research and Development + + June 27-30 2005 + Baltimore, MD + + + USA/Europe Seminar on Air Traffic Management Research and Development, Baltimore, MD, June 27-30 2005. + + + + + Analysis of an optimal sector design method + + MichaelDrew + + 10.1109/dasc.2008.4702801 + + + 2008 IEEE/AIAA 27th Digital Avionics Systems Conference + + IEEE + October 26-30 2008 + + + Drew, M., "Analysis of an Optimal Sector Design," 27th Digital Avionics Systems Conference, October 26-30 2008. + + + + + A Weighted-Graph Approach for Dynamic Airspace Configuration + + StephaneMartinez + + + GanoChatterji + + + DengfengSun + + + AlexandreBayen + + 10.2514/6.2007-6448 + + + AIAA Guidance, Navigation and Control Conference and Exhibit + Hilton Head, South Carolina + + American Institute of Aeronautics and Astronautics + August 20-23 2007 + + + Martinez, S., Chatterji, G., Sun, D., and Bayen, A., "A Weighted-Graph Approach for Dynamic Airspace Configuration," Proceedings of AIAA Guidance, Navigation, and Control Conference, Hilton Head, South Carolina, August 20-23 2007. + + + + + Geometric Algorithms for Optimal Airspace Design and Air Traffic Controller Workload Balancing + + AmitabhBasu + + + JosephS BMitchell + + + GirishkumarSabhnani + + 10.1137/1.9781611972887.8 + + + 2008 Proceedings of the Tenth Workshop on Algorithm Engineering and Experiments (ALENEX) + San Francisco, California + + Society for Industrial and Applied Mathematics + January 19 2008 + + + + Basu, A., Mitchell, J., and Sabhnani, G., "Geometric Algorithms for Optimal Airspace Design and Air Traffic Controller Workload Balancing," Proceedings of the 9th Workshop on Algorithm Engineering and Experiments(ALENEX), San Francisco, California, January 19 2008. + + + + + Airspace partitioning using flight clustering and computational geometry + + ChrisRBrinton + + + StephenPledgie + + 10.1109/dasc.2008.4702800 + + + 2008 IEEE/AIAA 27th Digital Avionics Systems Conference + + IEEE + October 26-30 2008 + + + Brinton, C. and Pledgie, S., "Airspace Partitioning using Flight Clustering and Computational Geometry," 27th Digital Avionics Systems Conference, October 26-30 2008. + + + + + A Comparison of Algorithm Generated Sectorizations + + ShannonZelinski + + 10.2514/atcq.18.3.279 + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 18 + 3 + + June 29 -July 2 2009 + American Institute of Aeronautics and Astronautics (AIAA) + Napa, CA + + + 8th USA + Zelinski, S., "A Comparison of Algorithm Generated Sectorizations," 8th USA/Europe Air Traffic Management Research and Development Seminar , Napa, CA, June 29 -July 2 2009. + + + + + Airspace Sector Redesign Based on Voronoi Diagrams + + MinXue + + 10.2514/6.2008-7223 + + + AIAA Guidance, Navigation and Control Conference and Exhibit + Honolulu, Hawaii + + American Institute of Aeronautics and Astronautics + August 18-21 2008 + + + Xue, M., "Airspace Sector Redesign Based on Voronoi Diagrams," AIAA Guidance, Navigation, and Contorl Conference and Exhibit, Honolulu, Hawaii, August 18-21 2008. + + + + + + AOkabe + + + BBoots + + + KSugihara + + + SNChiu + + Spatial Tessellations: Concepts and Applications of Voronoi Diagrams + Chichester, UK + + John Wiley & Sons + 1999 + + + nd edition + Okabe, A., Boots, B., Sugihara, K., and Chiu, S. N., Spatial Tessellations: Concepts and Applications of Voronoi Diagrams (2nd edition), John Wiley & Sons, Chichester, UK, 1999. + + + + + Build 4 of the Airspace Concept Evaluation System + + LarryMeyn + + + RobertWindhorst + + + KarlinRoth + + + DonaldVan Drei + + + GregKubat + + + VikramManikonda + + + SharleneRoney + + + GeorgeHunter + + + AlexHuang + + + GeorgeCouluris + + 10.2514/6.2006-6110 + + + AIAA Modeling and Simulation Technologies Conference and Exhibit + Keyston, CO + + American Institute of Aeronautics and Astronautics + 2006 + + + Meyn, L., Windhorst, R., Roth, K., Van Drei, D., Kubat, G., Manikonda, V., Roney, S., Hunter, G., Huang, A., and Couluris, G., "Build 4 of the Airspace Concept Evaluation System," AIAA Modeling and Simulation Technologies Conference an Exhibit, Keyston, CO, 2006. + + + + + + diff --git a/file804.txt b/file804.txt new file mode 100644 index 0000000000000000000000000000000000000000..2db4fc83eb7c0d0656e9325225b3aeb244007a57 --- /dev/null +++ b/file804.txt @@ -0,0 +1,403 @@ + + + + +I. IntroductionIn Dynamic Airspace Configuration (DAC) 1 research, dynamic resectorization has shown initial promise for restructuring sectors to accommodate fluctuating demand.Dynamically changing sectors are expected to balance demand and capacity and eventually reduce controller workload.Sectorization is a constrained multi-objective optimization problem.It mixes graph partitioning and optimization.Besides balancing and minimizing workload, the sectorization should also meet additional constraints or preferences.For example, the shapes of sectors are preferred to be convex, which not only assists the controllers but also minimizes the possibility of the same aircraft entering the same sector multiple times.Long flight dwelling times within each sector are also preferred to increase the time available to controllers for resolving conflicts.Many approaches to this problem have been proposed.Delahaye et al. 2,3 applied Genetic Algorithms for regroupment of sectors.Trandac et al. 4 proposed a constrained programming approach to optimize the sectorization while satisfying specific constraints.In the above research, the authors set up artificial scenarios, simplified the air traffic into networks with major routes and intersections, and finally grouped the modeled networks.There is still no evidence that such algorithms would work in practice because no one has applied these approaches to real traffic.Yousefi et al. 5 developed an approach that discretized the airspace into a hexagonal grid and clustered hexagonal cells using network flow algorithms.Each hexagonal grid cell contains the local traffic flow directionality and ATC workload information.Klein 6 suggested a new fast algorithm for sectorization based on hexagonal cells.Martinez et al. 7 proposed an algorithm based on graph theory.These algorithms [5][6][7] use network flow algorithms to approximately capture the flow pattern.However, there are some unresolved issues with these approaches.First, the convex or approximately convex shapes of sectors cannot be guaranteed.Some boundaries of sectors are "jagged", and some of the sectors are enclosed within others or have "C" shapes, which should be avoided in sector designs.Second, computing workload factors other than aircraft count might be prohibitive since all cost computations are required to be calculated in cells, and they have to be additive to form the final costs of the sectors.This also forbids the use of advanced complexity measurements like dynamic density. 8To overcome the disadvantages caused by grid-based methods, Basu et al. 9 developed geometric algorithms based on binary space partitions, pie cuts and dynamic programming for sectorization problems.The binary space partition and pie cut algorithm meets the convexity requirement, but in the available literature final sectors still have undesired shapes.It is necessary to have a method that has no limitation on choosing costs and is able to simultaneously balance the workload, maintain the preferred shape, and optimize given costs.Therefore, a strategy that is composed of a graph partition method (Voronoi diagrams), and an optimization scheme that combines Genetic Algorithms (GA) with the iterative deepening algorithm is developed to perform sectorization on real traffic data.Furthermore, conforming to dominant flow is one of the major factors contributed to the sector shape preference, thus examinations are needed to check if the optimization can take it into account without clustering algorithms.This paper applies the Voronoi diagram, GA and iterative deepening algorithm directly to the real flight track data.Experiments are performed based on different objectives.The effects of different fundamental metrics in the costs has been analyzed. +II. Algorithm for Airspace ResectorizationThis section briefly describes the algorithm used for sectorization.The Voronoi Diagram is applied to divide the airspace.Therefore, the sectorization problem is simplified to finding generating points.Then, bundled optimization algorithms will solve the optimization problem.The Iterative Deepening Algorithm provides a way to achieve this optimality with feasible computational efforts. +A. Partition -Voronoi DiagramThe Voronoi Diagram 10 decomposes a space into subdivisions around given sites.Each subdivision corresponds to one site, and all points in the region around the site are closer to the site than other sites.This algorithm has been used for solving numerous, and surprisingly different, geometric problems, e.g., nearest neighbors, minimum spanning trees, shortest paths, geometric clustering, and motion planning.Figure 1 shows a typical Voronoi Diagram in a rectangle.Given some generating points, the Voronoi Diagram divides a 2D space into a group of convex polygons with no overlap.It is noticed that all the points on the common edges have equal distances to their neighbor generating points.To compute the vertices of Voronoi Diagrsm, the Fortune's sweep-line algorithm 11 was applied in this work.Besides of the convex sector shapes, the major advantage of using the Voronoi Diagrams is detaching the graph partition and optimization.The goal of the problem essentially becomes finding optimal generating sites that minimize given costs.Additionally, since no small grids are involved in the process, computing workload does not rely on small cells.Thus, there is no limitation on the choice of costs, and advanced density measurement metrics can be easily incorporated, although it is left for the future work.Of course, this conversion confines the final sectors to the space of Voronoi Diagrams.However, this problem is likely has many optima, so finding a global optimal point is not primary goal.Existing methods do not guarantee global optimality. +B. Optimization -Genetic AlgorithmThe Genetic Algorithm 12,13 is a guided random search based on the mechanics of biological evolution.GA models two natural phenomena: genetic inheritance and Darwinian evolution.It provides efficient and effective techniques for optimization and machine learning applications, and has been widely used in scientific and engineering applications.It first creates a population of potential solutions or "chromosomes".After evaluating the fitness of each solution, it goes through a natural selection process loop if the termination criteria is not satisfied.While in the loop, the main operators -mutation and crossover -are executed among selected parents with respect to a predefined probability distribution based on fitness values.Then a new generation is produced.This recursive process stops when the termination condition is met.The roulette wheel method is used as a selection criteria.The crossover probability is 0.8 and mutation probability is 0.2.The population size is set to 500, and the process is stopped after 200 generations.The generating points for the Voronoi diagram are the optimization parameters.Given a number of sectors N , considering the latitude and longitude of each point, there will be 2N parameters to be optimized. +C. Efficiency -Iterative deepening algorithmDue to the time-consuming cost evaluations, the optimization might not finish in a feasible time period if GA is simply applied.The iterative deepening algorithm 14 is a state space search strategy in which a depth-limited search is run repeatedly.This search is applied to divide the single problem into sub-problems with the designated depth level.For this work, the depth limit is set to 1. Figure 2 describes an example of such a strategy when optimizing 16 sectors for an airspace using two levels.First, it optimally divides the airspace into 4 sectors.Then each sector will be further decomposed to 4 leaf sectors.The total 16 sectors will be the final sectors.This strategy leads final solutions from the global optima but expedites the optimization process.The experiments show that on a MAC 2.8GHZ platform with single processer, typical running time for a center is about 10 to 20 minutes. +D. ApproachThe approach can be described as follows.Figure 3 shows the procedure (Step 2 to 6) +III. DataTraffic data were obtained from the FAA's Aircraft Situation Display to Industry (ASDI) files or data for the entire day of August 24, 2005.The ASDI file for that day contains over 50, 000 flights.Without loss of generality, the Fort Worth Center (ZFW) was studied.The track data were generated using the Future Air traffic management Concepts Evaluation Tool (FACET). 15The flight track data are not confined to current flight plan.Any other track data e.g.wind-optimal trajectory, can be fed into this algorithm.The tracks with latitudes and longitudes were then rounded to the tenths place to expedite the cost evaluations.It was noted that during the whole day: 1) there were 4, 372 flights that overflew ZFW center, 2) a total of 1886 waypoints inside ZFW were involved, and 3) 10, 357 links within ZFW were flown.In preprocessing, each link was assigned with a weight, which corresponded to the number of flights that flew that link. +IV. ObjectivesIn this section, fundamental cost metrics will be discussed first.Then the objective functions will be set up for the case studies in next section.Advanced dynamic density metrics can be used for the calculation of monitoring workload variance, but in this initial study, aircraft counts were used as a rough estimate of monitoring workload.To expedite the evaluation of aircraft counts during optimization, flight data were preprocessed first using FACET to divide the ZFW center into small rectangular grids and to count the number of aircraft inside of the grids within a given time period.Figure 4 presents the workload distribution during the whole day based on the grids whose sizes were defined to be [0.1 • × 0.1 • ] in latitude and longitude, respectively.Bright colors mean high volume of traffic, from which the locations of the dominant traffic flows can be told.The total number of aircraft in a given sector was approximated by summing the counts of the grids associated with that sector.A grid was associated with a given sector if the grid center point was within the sector boundary.Total workload in the jth sector was approximated as: +A. Average and Peak Aircraft CountW j = i∈Sectorj a i(1)where a i is was the aircraft count in the ith grid.The cost of monitoring workload variance will be expressed as:f b1 = max j abs(W j -W avg ) W avg(2)where the average workload W avg = i a i N and N is was the desired number of sectors.In the cases where the workload variance is was based on the peak aircraft counts, the cost is:f b2 = max j abs[max t ( i∈Sectorj (a i,t )) -W avg ] W avg(3) +B. Sector Boundary CrossingsGiven the sectors, the number of flights that cross their boundaries was used as the estimate of the coordinating workload.For each link which connects two way points/fixes, the number of common boundary between two sectors were counted if they intersect with the link.The crossings over the boundaries of ZFW center were neglected.Assuming the number of common edges that the ith link crosses was M i , the total sector crossings can be obtained by summing them up.To facilitate the optimization, the coordination workload cost f c was defined as the normalization of the total crossings with respect to the total monitoring workload:f c = i∈Sectorj M i • w i i∈Sectorj a i(4)where w i was the traffic volume for the ith link. +C. Average Sector Flight TimeCalculating the average sector flight time or dwelling time is was straightforward.First, the total flight time of a sector was calculated by summing the durations for all flights that flew over the sector.Then it was divided by the total flight counts in this sector.The cost of dwelling time was the minimum of the flight sector time:f t = min j i T i i n i(5)where T i was the flight time. +D. Objective Function for BalancingThe fitness function of GA was then defined as the combination of the above costs:f = c 1 •f bi + c 2 •f c -c 3 •f t(6)where c i were the coefficients, with which the costs are guaranteed to stay in comparable magnitude and can be easily turned on and off.f bi represents f b1 and f b2 . +E. Objective Function for Increasing CapacityTo study the ability of new design to increase sector capacity, a cost based on the sector residual capacity is introduced.The sector capacity is calculate based on MAP from FAA Order 7210.3, which is roughly 5/3 of average sector flight time.Although the usefulness or accuracy of this formula is debatable, it enables a study of how sector design can increase the capacity.f r = min j { 5 3 × average sector f light time -peak ac count}(7)In the objective function, the peak aircraft count is used in the sector as the measurement of complexity, and the goal is maximizing the gap between the capacity and peak aircraft counts for a given number of sectors. +V. ResultsCurrently ZFW center is divided into 17 sectors.In this work, for simplicity, the center was decomposed into 18 sectors.Using the Iterative Deepening Algorithm, at the first level 6 sectors were optimized then, at the next level, each of them was optimally divided into 3 sectors to have a total of 18 sectors.In a MacOS platform with Intel Core 2 Duo Processor with 2.8 GHz and 8GB RAM, this process takes approximately 20 minutes with no parallel scheme involved.In the following sections, several cases will be explored to examine the effects of different costs discussed above, and a preliminary benefit analysis of dynamic sectorization will also be conducted.In the first case, the effect of the cost of balancing workload or average aircraft counts was studied.The coefficients c 2 and c 3 are equal to zero in Eqn. 6. Figure 5(a) shows the final configuration of sectors.Figure 5(b) presents the variance of average aircraft counts for each sector, where the solid line is the variance of current sectors and the dashed line is the variance for the new ones.With the maximum variance around 2.6%, the new solution has a more balanced workload than the current sectorization.As discussed above, sectorization is complicated due to the multi-objective cost.If the results are examined carefully, it is noticed that although the variances of aircraft counts are low, the new sectors do not satisfy other preferences.The number of sector crossings is 14, 213, which means the flights in ZFW will cross the sector boundaries 14, 213 times.This high volume will yield high workload of coordination among sectors.Additionally, the minimum of the average sector flight time in the sectors is relatively low -7.2 minutes. +B. Case II: Balancing Workload and Minimizing Sector CrossingsTo incorporate both monitoring workload balancing and minimizing sector crossings, c 1 , c 2 , and c 3 were set to 1.0, 1.0, and 0 respectively.Figure 6 displays the resulting sectorization.The solution gives a variance of 2.35% on average aircraft counts, which is approximately equal to the previous case.In addition, the number of sector crossings has been decreased from 14, 213 to 8, 047.The 43.4% less crossings will lower the coordination workload dramatically.Furthermore, the minimum average sector flight time is 7.8 minutes, which is longer than before.The longer sector flight time or dwelling time will yield a larger capacity associated with the new design.These results also indicate a correlation between the flight sector time and sector crossings.Visually examining the new configuration, it is noted that some dominant traffic flow have less interactions with the sector boundaries.For instance, the one at the bottom-right corner is kept in one sector, and in the middle part the major top-down traffic passes less sectors in the new design than the previous one.To get insight of the effect of sector flight time, in the third case, the cost of transition time was added into the objective function.Additionally, for workload balancing, peak aircraft counts were used instead of average counts to show the flexibility.In this case, the maximum aircraft counts are kept around 15 in each sector.The solution is shown in Figure 7.With this configuration, the peak aircraft counts are balanced.The number of sector crossings is 8, 947, which is similar to previous case.The average sector flight time is increased to 8.3 minutes, which is 7% better than before.From Figure 7, one can tell the dominant flows have been well-considered in this design.Several major flows from the top-right corner only need to pass through two to three sectors in ZFW center.8(a) presents the final solution using this method.The dominant flow has been captured even better than in previous cases.For instance, the major traffic flows from the upper-right corner are kept in one sector to increase the sector flight time, thus increasing the capacity as defined above.In Figure 8(b), the upper plot shows that all peak aircraft counts are capped by the defined capacities.This shows that the new design increases throughput by placing capacity where needed.However, both the new sectors and current sectors have unevenly distributed average aircraft counts. +D. Case IV: Maximizing Sector Residual CapacityThis result implies that balancing the workload and increasing throughput are conflicting goals. +E. Experiments for Different CentersIn this section, different centers were explored.Meanwhile, since the method is reactive to the input traffic data, to show flexibility, in this experiment, unconstrained traffic data were used.They were generated based on the scheduled flight plan from April 20, 2007 using the Airspace Concept Evaluation System (ACES).The costs in Case III and Case IV were used, respectively, for these centers to show the difference between increased capacity and comprehensive balancing.As an example, Figure 9 shows the final designs for ZAU center and ZDC center.The designs in Figure 9(b) and Figure 9(d) were based on balancing aircraft count which is the same as Case III.Although sector flight time and boundary crossings have been taken into account, they did not capture the major flow as well as the design in Figure 9(a) and Figure 9(c) since the balancing the aircraft count is the dominant objective.As noticed, the designs with increased capacity for ZDC center aligned with major flows.So does the new sectors of ZAU center, especially the eastern part.These experiments further demonstrate that without cell clustering, dominant flow can be taken into account by setting proper costs in the optimization. +F. Preliminary Benefit Analysis of Dynamic SectorizationWhile many practical and operational concerns need to be addressed before implementing dynamic sectorization, it is still valuable to examine the benefits under ideal assumptions.As a preliminary examination, a peak aircraft count per sector of 15 is used as a criteria for sector utilization.It is assumed that the sector configuration can be changed every two hours.The goal is to determine how many sectors are needed for different time periods given the fluctuating demand.Figure 10 gives the experimental results.The blue line is the time history of instant aircraft counts and the green line denotes the number of sectors needed during different periods.In the highest traffic period, 14 sectors are needed.During the lowest traffic period, only a few sectors are necessary to keep the sectors fully utilized.The number of sectors required is strongly correlated with the instantaneous aircraft counts. +VI. ConclusionsIn this work, a methodology based on the Voronoi Diagram and Genetic Algorithm is investigated and applied to the resectorization problem.With the Voronoi Diagram, the convexity requirement is automatically satisfied and the choice of costs is flexible.The sectorization can be encoded as the generating points.Genetic Algorithm is used to perform the multi-objective optimization.The Iterative Deepening Algorithm is applied to expedite the process.Initial results in 2D showed that this strategy is promising for sectorization.This method balanced the workload satisfactorily with a small deviation from average workload, and maintained convex shapes for sectors by the nature of the Voronoi diagram.By lowering the crossing volume and increasing sector flight time, the method captured the flow structure in some extent.The case study on maximizing sector residual capacity shows that increasing capacity, which is based on 5/3 sector flight time, has conflicts with the objective of balancing aircraft counts.Experiments show the cost of increased capacity can lead a design that aligned with major flow in the frame of optimization.In future work, advanced complexity measurements such as dynamic density metrics will be investigated as a cost.It will also be interesting to incorporate a new formula for capacity measurements.Additionally, its application to 3D airspace sectorization will be developed and examined.Figure 1 :1Figure 1: Voronoi Diagram in a planner space +(a) Division at the first level (b) Division at the second level +Figure 2 :2Figure 2: Description of using iterative deepening algorithm in sector design +Figure 3 :3Figure 3: Procedure of the optimization at one level sectorization +Figure 4 :4Figure 4: Grid based Workload Distribution for ZFW +A. Case I: Balancing Workload Only (a) New Sectors for Balancing AC Counts Only +Variance of Average AC Counts in Each Sector +Figure 5 :5Figure 5: Balance Average Aircraft Counts Only (variance of aircraft counts = 2.61%, total sector crossings = 14,213, minimum of average sector flight time = 7.2 min) +Figure 6 :6Figure 6: Balance Average Aircraft Counts and Minimize Sector Crossings (variance of aircraft counts = 2.35%, total sector crossings = 8,047, minimum of average sector flight time = 7.8 min) +Figure 7 :7Figure 7: Balance Average Aircraft Counts, Minimize Sector Crossings, and Maximize the Sector Flight Time (peak aircraft counts ≈ 15, total sector crossings = 8,947, minimum of average sector flight time = 8.3 min) +sectors Peak aircraft counts of current sectors Capacity of new sectors Peak aircraft counts of new sectors Workload variance with current sectors Workload variance with new sectors (b) Analysis of New Sectors +Figure 8 :8Figure 8: Maximize Sector Residual Capacity +Figure +(a) ZAU with cost of increased capacity (b) ZAU with cost of comprehensive balancing (c) ZDC with cost of increased capacity (d) ZDC with cost of comprehensive balancing +Figure 9 :9Figure 9: Experiments for ZAU center and ZDC center +Figure 10 :10Figure 10: Preliminary Benefits of Dynamic Sectorization for ZFW + + + + + + + + + Initial Concepts for Dynamic Airspace Configuration + + ParimalKopardekar + + + KarlBilimoria + + + BanavarSridhar + + 10.2514/6.2007-7763 + + + 7th AIAA ATIO Conf, 2nd CEIAT Int'l Conf on Innov and Integr in Aero Sciences,17th LTA Systems Tech Conf; followed by 2nd TEOS Forum + Belfast, Northern Ireland + + American Institute of Aeronautics and Astronautics + 18-20 September 2007 + + + P. Kopardekar, K. Bilimoria, and B. Sridhar. Initial concepts for dynamic airspace configuration. In 7th AIAA Aviation Technology, Integration and Operations Conference (ATIO), Belfast, Northern Ireland, 18-20 September 2007. + + + + + Genetic Algorithms for Automatic Regrouping of Air Traffic Control Sectors + + DDelahaye + + + JMAlliot + + + MSchoenauer + + + JLFarges + + 10.7551/mitpress/2887.003.0057 + + + Evolutionary Programming IV + + The MIT Press + 1995 + + + D. Delahaye, J.M. Alliot, M. Schoenauer, and J.L. Farges. Genetic algorithms for automatic re- graupement of air traffic control sectors. In Proceedings of the International Conference on Evolutionary Programming, 1995. + + + + + Airspace sectoring by evolutionary computation + + DDelahaye + + + MSchoenauer + + + J-MAlliot + + 10.1109/icec.1998.699504 + + + 1998 IEEE International Conference on Evolutionary Computation Proceedings. IEEE World Congress on Computational Intelligence (Cat. No.98TH8360) + + IEEE + 1998 + + + D. Delahaye, M. Schoenauer, and J.M. Alliot. Airspace sectoring by evolutionary computation. In Proceedings of IEEE International Congress on Evolutionary Computation, 1998. + + + + + Airspace sectorization with constraints + + HuyTrandac + + + PhilippeBaptiste + + + VuDuong + + 10.1051/ro:2005005 + + + RAIRO - Operations Research + RAIRO-Oper. Res. + 0399-0559 + 1290-3868 + + 39 + 2 + + June 2003 + EDP Sciences + Budapest, Hungary + + + H. Trandac and V. Duong. Optimized sectorization of airpace with constraints. In 5th Erocontrol/FAA ATM R&D Seminar, Budapest, Hungary, 23-27 June 2003. + + + + + Temporal and Spatial Distribution of Airspace Complexity for Air Traffic Controller Workload-Based Sectorization + + ArashYousefi + + + GeorgeDonohue + + 10.2514/6.2004-6455 + + + AIAA 4th Aviation Technology, Integration and Operations (ATIO) Forum + Chicago, IL + + American Institute of Aeronautics and Astronautics + September 2004 + + + A Yousefi and G.L. Donohue. Temporal and spatial distribution of airspace complexity for new method- ologies in airpace design. In 4th AIAA Aviation Technology, Integration, and Operations conference (ATIO),, Chicago, IL, September 2004. + + + + + An efficient method for airspace analysis and partitioning based on equalized traffic mass. In 6th USA + + AKlein + + + + Europe Seminar on Air Traffic Management Research and Development + + + June 2005 + Baltimore, MD + + + A. Klein. An efficient method for airspace analysis and partitioning based on equalized traffic mass. In 6th USA/Europe Seminar on Air Traffic Management Research and Development, Baltimore, MD, 27-30 June 2005. + + + + + A Weighted-Graph Approach for Dynamic Airspace Configuration + + StephaneMartinez + + + GanoChatterji + + + DengfengSun + + + AlexandreBayen + + 10.2514/6.2007-6448 + + + AIAA Guidance, Navigation and Control Conference and Exhibit + Hilton Head, South Carolina + + American Institute of Aeronautics and Astronautics + August 2007 + + + + S. Martinez, G. Chatterji, D. Sun, and A. Bayen. A weighted-graph approach for dynamic airspace configuration. In Proceedings of AIAA Guidance, Navigation, and Control Conference, Hilton Head, South Carolina, 20-23 August 2007. + + + + + Dynamic density: measuring and predicting sector complexity [ATC] + + PKopardekar + + + SMagyarits + + 10.1109/dasc.2002.1067920 + + + Proceedings. The 21st Digital Avionics Systems Conference + The 21st Digital Avionics Systems Conference + + IEEE + June 2003 + + + P. Kopardekar and S. Magyarits. Measurement and prediction of dynamic density. In 5th FAA & Eurocontrol ATM Conference, June 2003. + + + + + Geometric Algorithms for Optimal Airspace Design and Air Traffic Controller Workload Balancing + + AmitabhBasu + + + JosephS BMitchell + + + GirishkumarSabhnani + + 10.1137/1.9781611972887.8 + + + 2008 Proceedings of the Tenth Workshop on Algorithm Engineering and Experiments (ALENEX) + San Francisco, California + + Society for Industrial and Applied Mathematics + 19 January 2008 + + + + A. Basu, J.S.B. Mitchell, and G. Sabhnani. Geometric algorithms for optimal airspace design and air traffic controller workload balancing. In Proceedings of the 9th Workshop on Algorithm Engineering and Experiments(ALENEX), San Francisco, California, 19 January 2008. + + + + + + AOkabe + + + BBoots + + + KSugihara + + Spatial Tessellations: Concepts and Applications of Voronoi Diagrams + Chichester, UK + + John Wiley & Sons + 1993 + + + A. Okabe, B. Boots, and K. Sugihara. Spatial Tessellations: Concepts and Applications of Voronoi Diagrams. John Wiley & Sons, Chichester, UK, 1993. + + + + + A sweepline algorithm for Voronoi diagrams + + StevenJFortune + + 10.1007/bf01840357 + + + Algorithmica + Algorithmica + 0178-4617 + 1432-0541 + + 2 + 1-4 + + 1987 + Springer Science and Business Media LLC + + + S.J. Fortune. A sweepline algorithm for Voronoi Diagrams. Algorithmica, 2:153-174, 1987. + + + + + Adaptation in Natural and Artificial Systems + + JHHolland + + + 1975 + University of Mchigan Press + + + J.H. Holland. Adaptation in Natural and Artificial Systems. University of Mchigan Press, 1975. + + + + + Genetic Algorithms + Data Structures = Evolution Programs + + ZMichaelwicz + + 10.1007/978-3-662-03315-9 + + 1996 + Springer Berlin Heidelberg + Berlin Heidelberg + + + Third Edition + Z. Michaelwicz. Genetic Algorithms + Data Structures = Evolution Programs, Third Edition. Springer- Verlag Berlin Heidelberg, 1996. + + + + + Artificial Intelligence: A Modern Approach + + SJRussell + + + PNorvig + + + 2003 + Prentice Hall + Upper Saddle Rive, NJ + + + Second Edition + S.J. Russell and P. Norvig. Artificial Intelligence: A Modern Approach, Second Edition. Prentice Hall, Upper Saddle Rive, NJ, 2003. + + + + + FACET: Future ATM Concepts Evaluation Tool + + KarlDBilimoria + + + BanavarSridhar + + + ShonRGrabbe + + + GanoBChatterji + + + KapilSSheth + + 10.2514/atcq.9.1.1 + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 9 + 1 + + 2001 + American Institute of Aeronautics and Astronautics (AIAA) + + + K.D. Bilimoria, B. Sridhar, G. Chatterji, K.S. Sheth, and S. Grabbe. FACET: Future ATM concepts evaluation tool. Air Traffic Control Quarterly, 9(1):1-20, 2001. + + + + + + diff --git a/file805.txt b/file805.txt new file mode 100644 index 0000000000000000000000000000000000000000..062330510ce7ae00a7286c07d4751b284ed5a866 --- /dev/null +++ b/file805.txt @@ -0,0 +1,407 @@ + + + + +I. IntroductionIn Dynamic Airspace Configuration (DAC) 1 research, dynamic resectorization has shown initial promise for restructuring sectors to accommodate fluctuating demand.Dynamically changing sectors are expected to balance demand and capacity and eventually reduce controller workload.Sectorization is a constrained multi-objective optimization problem.It mixes graph partitioning and optimization.Besides balancing and minimizing workload, the sectorization should also meet additional constraints or preferences.For example, the shapes of sectors are preferred to be convex, which not only assists the controllers but also minimizes the possibility of the same aircraft entering the same sector multiple times.Long flight dwelling times within each sector are also preferred to increase the time available to controllers for resolving conflicts.Many approaches to this problem have been proposed.Delahaye et al. 2,3 applied Genetic Algorithms for regroupment of sectors.Trandac et al. 4 proposed a constrained programming approach to optimize the sectorization while satisfying specific constraints.In the above research, the authors set up artificial scenarios, simplified the air traffic into networks with major routes and intersections, and finally grouped the modeled networks.There is still no evidence that such algorithms would work in practice because no one has applied these approaches to real traffic.Yousefi et al. 5 developed an approach that discretized the airspace into a hexagonal grid and clustered hexagonal cells using network flow algorithms.Each hexagonal grid cell contains the local traffic flow directionality and ATC workload information.Klein 6 suggested a new fast algorithm for sectorization based on hexagonal cells.Martinez et al. 7 proposed an algorithm based on graph theory.These algorithms [5][6][7] use network flow algorithms to approximately capture the flow pattern.However, there are some unresolved issues with these approaches.First, the convex or approximately convex shapes of sectors cannot be guaranteed.Some boundaries of sectors are "jagged", and some of the sectors are enclosed within others or have "C" shapes.Second, computing workload factors other than aircraft count might be prohibitive since all cost computations are required to be calculated in cells, and they have to be additive to form the final costs of the sectors.This also forbids the use of advanced complexity measurements like dynamic density. 8To overcome the disadvantages caused by grid-based methods, Basu et al. 9 developed geometric algorithms based on binary space partitions, pie cuts and dynamic programming for sectorization problems.The binary space partition and pie cut algorithm meets the convexity requirement, but in the available literature final sectors still have undesired shapes.It is necessary to have a method that has no limitation on choosing costs and is able to simultaneously balance the workload, maintain the preferred shape, and optimized given costs.Therefore, a strategy that is composed of a graph partition method (Voronoi diagrams), and an optimization scheme that combines Genetic Algorithms (GA) with the iterative deepening algorithm is developed to perform sectorization on real traffic data.This paper applies the Voronoi diagram, GA and Iterative Deepening Algorithm.The whole strategy is applied directly to the real flight track data.With the Voronoi diagram and a multi-objective cost function, the sectorization is converted to an optimization problem.Thereafter, GA, which is bundled with the Iterative Deepening Algorithm will perform the multi-objective optimization. +II. Algorithm for Airspace ResectorizationThis section briefly describes the algorithm used for sectorization.The Voronoi Diagram is applied to divide the airspace.Therefore, the sectorization problem is simplified to finding generating points.Then, bundled optimization algorithms will solve the optimization problem.The Iterative Deepening Algorithm provides a way to achieve this optimality with feasible computational efforts. +A. Voronoi DiagramThe Voronoi Diagram 10 decomposes a space into subdivisions around given sites.Each subdivision corresponds to one site, and all points in the region around the site are closer to the site than other sites.Figure 1, which is reproduced from Ref, 11 shows a typical Voronoi Diagram in a rectangle.This algorithm has been used for solving numerous, and surprisingly different, geometric problems, e.g., nearest neighbors, minimum spanning trees, shortest paths, geometric clustering, and motion planning.Given some generating points, the Voronoi Diagram divides a 2D space into a group of convex polygons with no overlap.Additionally, since no grid decomposition is involved in the process, computing workload does not rely on small cells.Thus, there is no limitation on the choice of costs, and advanced density measurement metrics can be easily incorporated, although it is left for the future work.Using the Voronoi Diagram, the goal of the problem essentially becomes finding optimal generating sites that minimize given costs.This conversion confines the final sectors to the space of Voronoi Diagrams.However, this problem is likely has many optima, so finding a global optimal point is difficult.Existing methods do not guarantee global optimality. +B. Genetic AlgorithmThe Genetic Algorithm 12,13 is a guided random search based on the mechanics of biological evolution.GA models two natural phenomena: genetic inheritance and Darwinian evolution.It provides efficient and effective techniques for optimization and machine learning applications, and has been widely used in scientific and engineering applications.It first creates a population of potential solutions or "chromosomes".After evaluating the fitness of each solution, it goes through a natural selection process loop if the termination criteria is not satisfied.While in the loop, the main operators -mutation and crossover -are executed among selected parents with respect to a predefined probability distribution based on fitness values.Then a new generation is produced.This recursive process stops when the termination condition is met.The roulette wheel method is used as a selection criteria.The crossover probability is 0.8 and mutation probability is 0.2.The population size is set to 500, and the process is stopped after 200 generations.The generating points for the Voronoi diagram are the optimization parameters.Given a number of sectors N , considering the latitude and longitude of each point, there will be 2N parameters to be optimized. +C. Iterative deepening algorithmDue to the time-consuming cost evaluations, the optimization might not finish in a feasible time period if GA is simply applied.The Iterative Deepening Algorithm 14 is a state space search strategy in which a depth-limited search is run repeatedly.This search is applied to divide the single problem into sub-problems with the designated depth level.For this work, the depth limit is set to 1.The leaf nodes will eventually compose the final sectors for the airspace.Figure 2 describes an example of such a strategy when optimizing 18 sectors using two levels.First, it optimally divides the airspace into 6 sectors.Then each sector will be further decomposed to 3 leaf sectors.The total 18 sectors will be the final sectors.This strategy leads final solutions from the global optima but expedites the optimization process. +D. ApproachThe approach can be described as follows:1. Define the objectives, the airspace region S needed to be sectorized, and the initial number of subdivisions N at the first level.2. Randomly generate N points in S.3. Using the N generating points, apply Voronoi Diagrams to generate boundaries of the N sub-divisions.4. Evaluate the total cost for new divisions.5. Move the generating points using GA optimization algorithms. +Go back toStep 3 if terminal condition is not satisfied.7. If the defined depth is not reached, for each sub-division at current depth level, reassign the airspace S and the number N , and go through Step 2 -6. +III. ObjectivesTraffic data were obtained from the FAA's Aircraft Situation Display to Industry (ASDI) files or data for the entire day of August 24, 2005.The ASDI file for that day contains over 50, 000 flights.Without loss of generality, the Fort Worth Center (ZFW) was studied.The track data were generated using the Future Air traffic management Concepts Evaluation Tool (FACET). 15The tracks with latitudes and longitudes were then rounded to the tenths place to expedite the cost evaluations.It was noted that during the whole day: 1) there were 4, 372 flights that overflew ZFW center, 2) a total of 1886 waypoints inside ZFW were involved, and 3) 10, 357 links within ZFW were flown.In preprocessing, each link was assigned with a weight, which corresponded to the number of flights that flew that link.With Voronoi diagrams, the sectors were guaranteed to be convex.To compute the vertices of each sector in the optimization progress, the Fortune's sweep-line algorithm 16 was applied.Other costs or constraints then will be put into objective function.In this work, the objectives include minimizing monitoring workload variance, minimizing coordinating workload, and maximizing dwelling time in a sector.Advanced dynamic density metrics can be used for the calculation of monitoring workload variance, but in this initial study, aircraft counts were used as a rough estimate of monitoring workload.To expedite the evaluation of aircraft counts during optimization, flight data were preprocessed first using FACET to divide the ZFW center into small rectangular grids and to count the number of aircraft inside of the grids within a given time period.Figure 3 presents the workload distribution during the whole day based on the grids whose sizes were defined to be [0.1 • × 0.1 • ] in latitude and longitude, respectively.Bright colors mean high volume of traffic, from which the locations of the dominant traffic flows can be told.The total number of aircraft in a given sector was approximated by summing the counts of the grids associated with that sector.A grid was associated with a given sector if the grid center point was within the sector boundary.Total workload in the jth sector was approximated as:W j = i∈Sectorj a i(1)where a i is was the aircraft count in the ith grid.The cost of monitoring workload variance will be expressed as:f b = max j abs(W j -W avg ) W avg(2)where the average workload W avg = i ai N and N is was the desired number of sectors.In the cases where the workload variance is was based on the peak aircraft counts, the cost is:f b = max j abs[max t ( i∈Sectorj (a i,t )) -W avg ] W avg(3)Given the sectors, the number of flights that cross their boundaries was used as the estimate of the coordinating workload.For each link which connects two way points/fixes, the number of common boundary between two sectors were counted if they intersect with the link.The crossings over the boundaries of ZFW center were neglected.Assuming the number of common edges that the ith link crosses was M i , the total sector crossings can be obtained by summing them up.To facilitate the optimization, the coordination workload cost f c was defined as the normalization of the total crossings with respect to the total monitoring workload:f c = i∈Sectorj M i • w i i∈Sectorj a i(4)where w i was the traffic volume for the ith link.Calculating the average sector flight time or dwelling time is was straightforward.First, the total flight time of a sector was calculated by summing the durations for all flights that flew over the sector.Then it was divided by the total flight counts in this sector.The cost of dwelling time was the minimum of the flight sector time:f t = min j i T i i a i (5)where T i was the flight time.The fitness function of GA was then defined as the combination of the above costs:f = c 1 •f b + c 2 •f c -c 3 •f t(6)where c i were the coefficients, with which the costs are guaranteed to stay in comparable magnitude and can be easily turned on and off. +IV. ResultsCurrently ZFW center is divided into 17 sectors.In this work, for simplicity, the center was decomposed into 18 sectors.Using the Iterative Deepening Algorithm, at the first level 6 sectors were optimized then, at the next level, each of them was optimally divided into 3 sectors to have a total of 18 sectors.In a MacOS platform with Intel Core 2 Duo Processor with 3.0 GHz and 8GB RAM, this process takes approximately 20 minutes with no parallel scheme involved.In the following sections, several cases will be explored to examine the effects of different costs discussed above, and a preliminary benefit analysis of dynamic sectorization will also be conducted.In the first case, the effect of the cost of balancing workload or average aircraft counts was studied.The coefficients c 2 and c 3 are equal to zero in Eqn. 6. Figure 4(a) shows the final configuration of sectors.Figure 4(b) presents the variance of average aircraft counts for each sector, where the solid line is the variance of current sectors and the dashed line is the variance for the new ones.With the maximum variance around 2.6%, the new solution has a more balanced workload than the current sectorization.As discussed above, sectorization is complicated due to the multi-objective cost.If the results are examined carefully, it is noticed that although the variances of aircraft counts are low, the new sectors do not satisfy other preferences.The number of sector crossings is 14, 213, which means the flights in ZFW will cross the sector boundaries 14, 213 times.This high volume will yield high workload of coordination among sectors.Additionally, the minimum of the average sector flight time in the sectors is relatively low -7.2 minutes. +A. Balancing Workload Only +B. Balancing Workload and Minimizing Sector CrossingsTo incorporate both monitoring workload balancing and minimizing sector crossings, c 1 , c 2 , and c 3 were set to 1.0, 1.0, and 0 respectively.Figure 5 displays the resulting sectorization.The solution gives a variance of 2.35% on average aircraft counts, which is approximately equal to the previous case.In addition, the number of sector crossings has been decreased from 14, 213 to 8, 047.The 43.4% less crossings will lower the coordination workload dramatically.Furthermore, the minimum average sector flight time is 7.8 minutes, which is longer than before.The longer sector flight time or dwelling time will yield a larger capacity associated with the new design.These results also indicate a correlation between the flight sector time and sector crossings.Visually examining the new configuration, it is noted that some dominant traffic flow have less interactions with the sector boundaries.For instance, the one at the bottom-right corner is kept in one sector, and in the middle part the major top-down traffic passes less sectors in the new design than the previous one.Sector design is expected to increase the capacity of the NAS.In this experiment, the sector capacity is calculate based on FAA Order 7210.3, which is roughly 5/3 of average sector flight time.Although the usefulness or accuracy of this formula is debatable, it enables a study of how sector design can increase the capacity.In the objective function, the peak aircraft count is used in the sector as the measurement of complexity, and the goal is maximizing the gap between the capacity and peak aircraft counts for a given number of sectors.Figure 7(a) presents the final solution using this method.The dominant flow has been captured even better than in previous cases.For instance, the major traffic flows from the upper-right corner are kept in one sector to increase the sector flight time, thus increasing the capacity as defined above.In Figure 7(b), the upper plot shows that all peak aircraft counts are capped by the defined capacities.Current sectors have some violations as the green lines show.This shows that the new design increases throughput by placing capacity where needed.However, both the new sectors and current sectors have unevenly distributed average aircraft counts.This result implies that balancing the workload and increasing throughput are conflicting goals. +E. Preliminary Benefit Analysis of Dynamic SectorizationWhile many practical and operational concerns need to be addressed before implementing dynamic sectorization, it is still valuable to examine the benefits under ideal assumptions.As a preliminary examination, a peak aircraft count per sector of 15 is used as a criteria for sector utilization.It is assumed that the sector configuration can be changed every two hours.The goal is to determine how many sectors are needed for different time periods given the fluctuating demand.Figure 8 gives the experimental results.The blue line is the time history of instant aircraft counts and the green line denotes the number of sectors needed during different periods.In the highest traffic period, 14 sectors are needed.During the lowest traffic period, only a few sectors are necessary to keep the sectors fully utilized.The number of sectors required is strongly correlated with the instantaneous aircraft counts. +V. ConclusionsIn this work, a methodology based on the Voronoi Diagram and Genetic Algorithm is investigated and applied to the resectorization problem.With the Voronoi Diagram, the convexity requirement is automatically satisfied and the choice of costs is flexible.The sectorization can be encoded as the generating points.Genetic Algorithm is used to perform the multi-objective optimization.The Iterative Deepening Algorithm is applied to expedite the process.Initial results in 2D showed that this strategy is promising for sectorization.This method balanced the workload satisfactorily with a small deviation from average workload, and maintained convex shapes for sectors by the nature of the Voronoi diagram.By lowering the crossing volume and increasing sector flight time, the method captured the flow structure.The case study on maximizing sector residual capacity shows that increasing capacity, which is based on 5/3 sector flight time, has conflicts with the objective of balancing aircraft counts.In future work, advanced complexity measurements such as dynamic density metrics will be investigated as a cost.It will also be interesting to incorporate a new formula for capacity measurements.Additionally, its application to 3D airspace sectorization will be developed and examined.Figure 1 :1Figure 1: Voronoi Diagram in a planner space +Figure 2 :2Figure 2: Application of Iterative Deepening Algorithm +Figure 3 :3Figure 3: Grid based Workload Distribution for ZFW +(a) New Sectors for Balancing AC Counts Only +Variance of Average AC Counts in Each Sector +Figure 4 :4Figure 4: Balance Average Aircraft Counts Only (variance of aircraft counts = 2.61%, total sector crossings = 14,213, minimum of average sector flight time = 7.2 min) +Figure 5 :5Figure 5: Balance Average Aircraft Counts and Minimize Sector Crossings (variance of aircraft counts = 2.35%, total sector crossings = 8,047, minimum of average sector flight time = 7.8 min) +Figure 6 :6Figure 6: Balance Average Aircraft Counts, Minimize Sector Crossings, and Maximize the Sector Flight Time (peak aircraft counts ≈ 15, total sector crossings = 8,947, minimum of average sector flight time = 8.3 min) +Figure 7 :7Figure 7: Maximize Sector Residual Capacity +Figure 8 :8Figure 8: Preliminary Benefits of Dynamic Sectorization for ZFW + + + + + + + + + Initial Concepts for Dynamic Airspace Configuration + + ParimalKopardekar + + + KarlBilimoria + + + BanavarSridhar + + 10.2514/6.2007-7763 + + + 7th AIAA ATIO Conf, 2nd CEIAT Int'l Conf on Innov and Integr in Aero Sciences,17th LTA Systems Tech Conf; followed by 2nd TEOS Forum + Belfast, Northern Ireland + + American Institute of Aeronautics and Astronautics + 18-20 September 2007 + + + P. Kopardekar, K. Bilimoria, and B. Sridhar. Initial concepts for dynamic airspace configuration. In 7th AIAA Aviation Technology, Integration and Operations Conference (ATIO), Belfast, Northern Ireland, 18-20 September 2007. + + + + + Genetic Algorithms for Automatic Regrouping of Air Traffic Control Sectors + + DDelahaye + + + JMAlliot + + + MSchoenauer + + + JLFarges + + 10.7551/mitpress/2887.003.0057 + + + Evolutionary Programming IV + + The MIT Press + 1995 + + + D. Delahaye, J.M. Alliot, M. Schoenauer, and J.L. Farges. Genetic algorithms for automatic regraupement of air traffic control sectors. In Proceedings of the International Conference on Evolutionary Programming, 1995. + + + + + Airspace sectoring by evolutionary computation + + DDelahaye + + + MSchoenauer + + + J-MAlliot + + 10.1109/icec.1998.699504 + + + 1998 IEEE International Conference on Evolutionary Computation Proceedings. IEEE World Congress on Computational Intelligence (Cat. No.98TH8360) + + IEEE + 1998 + + + D. Delahaye, M. Schoenauer, and J.M. Alliot. Airspace sectoring by evolutionary computation. In Proceedings of IEEE International Congress on Evolutionary Computation, 1998. + + + + + Airspace sectorization with constraints + + HuyTrandac + + + PhilippeBaptiste + + + VuDuong + + 10.1051/ro:2005005 + + + RAIRO - Operations Research + RAIRO-Oper. Res. + 0399-0559 + 1290-3868 + + 39 + 2 + + June 2003 + EDP Sciences + Budapest, Hungary + + + H. Trandac and V. Duong. Optimized sectorization of airpace with constraints. In 5th Erocontrol/FAA ATM R&D Seminar, Budapest, Hungary, 23-27 June 2003. + + + + + Temporal and Spatial Distribution of Airspace Complexity for Air Traffic Controller Workload-Based Sectorization + + ArashYousefi + + + GeorgeDonohue + + 10.2514/6.2004-6455 + + + AIAA 4th Aviation Technology, Integration and Operations (ATIO) Forum + Chicago, IL + + American Institute of Aeronautics and Astronautics + September 2004 + + + A Yousefi and G.L. Donohue. Temporal and spatial distribution of airspace complexity for new methodologies in airpace design. In 4th AIAA Aviation Technology, Integration, and Operations conference (ATIO),, Chicago, IL, September 2004. + + + + + An efficient method for airspace analysis and partitioning based on equalized traffic mass. In 6th USA + + AKlein + + + + Europe Seminar on Air Traffic Management Research and Development + + + June 2005 + Baltimore, MD + + + A. Klein. An efficient method for airspace analysis and partitioning based on equalized traffic mass. In 6th USA/Europe Seminar on Air Traffic Management Research and Development, Baltimore, MD, 27-30 June 2005. + + + + + A Weighted-Graph Approach for Dynamic Airspace Configuration + + StephaneMartinez + + + GanoChatterji + + + DengfengSun + + + AlexandreBayen + + 10.2514/6.2007-6448 + + + AIAA Guidance, Navigation and Control Conference and Exhibit + Hilton Head, South Carolina + + American Institute of Aeronautics and Astronautics + August 2007 + + + + S. Martinez, G. Chatterji, D. Sun, and A. Bayen. A weighted-graph approach for dynamic airspace configuration. In Proceedings of AIAA Guidance, Navigation, and Control Conference, Hilton Head, South Carolina, 20-23 August 2007. + + + + + Dynamic density: measuring and predicting sector complexity [ATC] + + PKopardekar + + + SMagyarits + + 10.1109/dasc.2002.1067920 + + + Proceedings. The 21st Digital Avionics Systems Conference + The 21st Digital Avionics Systems Conference + + IEEE + June 2003 + + + P. Kopardekar and S. Magyarits. Measurement and prediction of dynamic density. In 5th FAA & Eurocontrol ATM Conference, June 2003. + + + + + Geometric Algorithms for Optimal Airspace Design and Air Traffic Controller Workload Balancing + + AmitabhBasu + + + JosephS BMitchell + + + GirishkumarSabhnani + + 10.1137/1.9781611972887.8 + + + 2008 Proceedings of the Tenth Workshop on Algorithm Engineering and Experiments (ALENEX) + San Francisco, California + + Society for Industrial and Applied Mathematics + 19 January 2008 + + + + A. Basu, J.S.B. Mitchell, and G. Sabhnani. Geometric algorithms for optimal airspace design and air traffic controller workload balancing. In Proceedings of the 9th Workshop on Algorithm Engineering and Experiments(ALENEX), San Francisco, California, 19 January 2008. + + + + + + AOkabe + + + BBoots + + + KSugihara + + Spatial Tessellations: Concepts and Applications of Voronoi Diagrams + Chichester, UK + + John Wiley & Sons + 1993 + + + A. Okabe, B. Boots, and K. Sugihara. Spatial Tessellations: Concepts and Applications of Voronoi Diagrams. John Wiley & Sons, Chichester, UK, 1993. + + + + + HTML Colors + + WolframMathworld + + 10.24097/wolfram.03358.data + + + null + Wolfram Research, Inc. + + + Wolfram Mathworld. http://mathworld.wolfram.com/VoronoiDiagram.html. + + + + + Adaptation in Natural and Artificial Systems + + JHHolland + + + 1975 + University of Mchigan Press + + + J.H. Holland. Adaptation in Natural and Artificial Systems. University of Mchigan Press, 1975. + + + + + Genetic Algorithms + Data Structures = Evolution Programs + + ZMichaelwicz + + 10.1007/978-3-662-03315-9 + + 1996 + Springer Berlin Heidelberg + Berlin Heidelberg + + + Third Edition + Z. Michaelwicz. Genetic Algorithms + Data Structures = Evolution Programs, Third Edition. Springer-Verlag Berlin Heidelberg, 1996. + + + + + Artificial Intelligence: A Modern Approach + + SJRussell + + + PNorvig + + + 2003 + Prentice Hall + Upper Saddle Rive, NJ + + + Second Edition + S.J. Russell and P. Norvig. Artificial Intelligence: A Modern Approach, Second Edition. Prentice Hall, Upper Saddle Rive, NJ, 2003. + + + + + FACET: Future ATM Concepts Evaluation Tool + + KarlDBilimoria + + + BanavarSridhar + + + ShonRGrabbe + + + GanoBChatterji + + + KapilSSheth + + 10.2514/atcq.9.1.1 + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 9 + 1 + + 2001 + American Institute of Aeronautics and Astronautics (AIAA) + + + K.D. Bilimoria, B. Sridhar, G. Chatterji, K.S. Sheth, and S. Grabbe. FACET: Future ATM concepts evaluation tool. Air Traffic Control Quarterly, 9(1):1-20, 2001. + + + + + A sweepline algorithm for Voronoi diagrams + + StevenJFortune + + 10.1007/bf01840357 + + + Algorithmica + Algorithmica + 0178-4617 + 1432-0541 + + 2 + 1-4 + + 1987 + Springer Science and Business Media LLC + + + S.J. Fortune. A sweepline algorithm for Voronoi Diagrams. Algorithmica, 2:153-174, 1987. + + + + + + diff --git a/file806.txt b/file806.txt new file mode 100644 index 0000000000000000000000000000000000000000..7d0594d1552444af75da37d1921a705cc3612533 --- /dev/null +++ b/file806.txt @@ -0,0 +1,333 @@ + + + + +I. IntroductionCorridors-in-the-sky, also referred to as tubes, have been proposed 1 to accommodate increasing and fluctuating traffic volume with designs based on traffic demand.They are expected to increase the airspace capacity and reduce delays.Corridors can absorb high-density traffic flows that have similar trajectories.Less controllers' workload may be demanded with corridors than with classic airspace structures due to the organized traffic flow.If automation is involved, such as advanced equipment for self-merging and self-spacing, the associated controllers' workload may be negligible.Many studies have been conducted to design/evaluate corridors or corridor networks.Alipio et al. 2 and Yousefi et al. 3 initially proposed to construct corridors between city-pairs.Sridhar et al. 4 developed a corridor network interconnecting airports in clusters seeded by major airports to impact a significant amount of traffic.Gupta et al 5 refined a corridor network based on airport clusters using optimization based on Mixed Integer Programming (MIP) according to the cost of deviations.In previous work, 6 a method that combines the Hough transform and Genetic Algorithm was proposed to find corridor candidates based on great circle trajectories.Hering 7 proposed the "Freeways" because of the different structure of Europe.The "Freeways" are corridors that try to pass through, or as close as possible to, enlarged major airport areas.Furthermore, Hoffman et al. 8 analyzed a list of design and operational issues for corridors.Sheth et.al. 9 compared different corridor designs based on several developed metrics.All of above the research focuses on finding/analyzing 2D corridors, but more realistic concerns need to be addressed to move the design and evaluation of corridors forward.This work discusses some missing issues, such as number of lanes, altitude allocations of corridors, and how these will affect the benefits/costs of corridors.Capacity constraints at airports, separation assurance for the flights in the corridors, and the impact of weather conditions are still neglected in this work.This study chooses a single corridor from previous work 6 and presents analysis of its spatial and temporal utilization through which the number of lanes can be derived.To construct a 3D corridor, a metric measuring the number of crossings between remaining flights and the corridor is used as performance.A cell-decomposition method is proposed to make crossing detection a hundred times faster than the bruteforce method.This paper starts with an introduction of the corridor model.Next, it presents the analysis of operations in the corridor, from that the number of lanes can be derived.Then the impact on remaining traffic is discussed, meanwhile the effect of the allocation of altitudes is described.The final section analyzes the possible benefits in underlying classic sector by deploying corridors based on the reduction of the peak aircraft count. +II. Corridor ModelAlthough the analysis presented in this paper can be applied to other corridor models, without loss of generality, a single corridor from previous work 6 is studied.In this model, a method that combines the Hough transform and Genetic Algorithm was developed to identify 2D locations of corridors.Flights were assumed to fly great circle flight trajectories.The schedule was based on flight plans from April 20, 2007.A flight was allowed to fly on a corridor if it had less than 5% extra flight distance compared with its shortest path distance.The method was applied to optimize the corridor location such that the number of corridor-attendees was maximized under the constraint of 5% extra flight distance.The output corridors were ranked according to the number of attendees.Figure 1 shows the top 60 candidates that can accommodate 44% or 13, 015 flights.The colors indicate the number of attendees in different corridors and vary with its geographical location.The candidate (see the right figure in Fig. 2) with the most attendees (774 flights during entire day) was chosen for analysis.All generated corridors, including this one, follow great circle trajectories.A corridor attendee was assumed to join in and exit from a corridor perpendicularly, as shown in the left figure in Fig. 2.This is the worst case, since other intersection angles will generate less extra distance.As an initial set-up, this model will be used for calculations through the paper. +III. Operations in CorridorsThere could be many ways for measuring utilization of a corridor, such as the number of attended flights.But, they may not answer how crowded are the flights and what is the minimum number of lanes needed.The utilization has to be analyzed in a temporal and spatial manner.Therefore, a space-time map based on studies in ground transportation 10,11 is developed. +A. Space-time MapTo generate a space-time map, a spatial scale is needed.An origin is first defined at one end of the corridor.Then points on the great circle corridor are located according to their distances from the origin.As shown in Fig. 3(a), the upper end is the origin, and the scale along the corridor corresponds to distances (in nautical miles) from the origin.Next, a space-time map is set up.If the schedule, entry and exit points, and speed of a flight is known, a flight can be represented as a curve on it.If the speed of the flight is assumed to be constant, the curve becomes a straight line whose slope is the speed.Figure 3(b) shows the space-time map and a straight line corresponding to a flight.From the line, it can be seen that the flight joined the corridor at the coordinate of 1, 200 nmi (close to Atlanta) at 8 AM and exited at the corrdiante of 500 nmi (close to New York) at 10 AM.Then, this space is descretized into grids.Each grid is 10 nautical miles in width and 2 minutes in height.A grid can be looked on as a safety zone in which only one flight is allowed.Since a flight can be anywhere in the occupied grid, the size of grids is doubled based on classic safety requirement.If the line intersects a grid, the grid will be incremented by one.Figure 3(b) shows the grids that are intersected by the sample curve.Although the speed of a flight in the corridor is assumed to be constant, it is applicable to have flights with varied speeds when necessary.In that case, a straight line will be replaced by a curve which corresponds to the varied speeds.Furthermore, uncertainty of the schedule, speed, and entry location can be taken into account to generate a probabilistic space-time map for corridor operations. +B. Analysis of OperationsUsing the method described above, the space-time map of 774 flights in corridor No.1 is generated.Typical speeds of different aircraft types are chosen.Entry and exit locations are computed following the rule of "entering and exiting perpendicularly".The flight schedules are used as the basis for time.The resulting map is shown in Fig. 4. As described in the last section, each grid accumulates the times occupied by flights.High pixel value of a grid means high occupancy, which is represented by hot color.Cooler colors represent lower occupancy.The hottest grids have 7 flights showing up simultaneously.Considering the separation rule, at least 7 lanes might be needed for these spots.In Figure 5(a), there are only a few flights in the dark blue region R, whereas heavy traffic occurs in the brightest regions Q and P between 500 nmi and 1, 250 nmi during the daytime.It may not be necessary to construct a corridor for the region of R. A dynamic corridor may be opened to follow the first flight of heavy traffic and ended with the last flight of heavy traffic.By setting up a threshold for triggering a corridor, regions in the space-time map can be defined, where that corridor is active.Fig. 5(a) shows a sample region included in orange frames specifying a dynamic corridor.Starting at 3 : 20 UTC (time stamp A), the full length corridor will be shrunk gradually from the location at 500 nmi towards the two ends.At 6 : 00 UTC (time stamp B), the corridor will be totally deactivated.At 10 : 00 UTC (time stamp C), the corridor will be gradually activated again.It will start from the location at 500 nmi.At 14 : 00 UTC, the entire corridor will be fully activated.Fig. 6 shows the space-time locations of the entries and exits.The blue points represent entries and the red points represent exits.This figure visualizes when and where a flight would enter or exit the corridor in terms of its schedule.For example, before 09 : 00 UTC, many flights will join the corridor from the location of 2, 500 nmi, which is somewhere above the Gulf of Mexico.This suggests that only one entry ramp is needed at that location at that time period.While after 12 : 40 UTC, because most flights would exit through this position, one exit ramp should be enough.To further check the entry and exit clustering natures, the spatial coordinates histogram is presented in Fig. 7. Along the corridor, several locations accommodate the majority of the entries and exits.This indicates that only a few ramps should meet the demand and the assumption of entering and exiting corridors anywhere doesn't necessarily introduce large number of ramps.The distribution and histogram can be used to guide construction of ramps in the future. +IV. Impact on Remaining TrafficAnother important operational issue that needs to be investigated is the impact on non-corridor users.If a significant number of flights are absorbed by corridors, the underlying sectors will have more capacity.But the traffic in underlying sectors will be forced to cap or tunnel if the corridor airspace is assumed to be impenetrable.Therefore, a corridor lane option that has less impact on crossing traffic will be preferred.This section will analyze the impact and provide suggestions on lane options for the sample corridor.In the analysis, the safety zone is defined as 5 nmi horizontally and 1000 ft vertically. +A. Cell-Decomposition Method for Detecting CrossingsOn April 20, 2007, there were more than 50, 000 flights.In the track data, the trajectories are composed of one-minute flight segments.Therefore, the number of flight segments are many times more than the number of flights in the daily track data.Because the crossing detection is not a one-time task, significant checks might be expected.For instance, an optimal altitude needs to be determined based on an exhaustive search, or the crossing detection may be integrated into optimization for designing corridors.Thus, although brute-force can detect the number of crossings in feasible time, a fast or real-time method to detect crossings is desired.The quad-tree cell decomposition method has been used in robot path planning problems. 12It constructs an obstacle-free solution space for building a shortest path in the presence of obstacles.Its theory can be applied in the crossing detection problem to rule out most unrelated flight segments.To apply quad-tree cell-decomposition, a rectangular region, which includes the entire US continent, is defined.The rectangular region is called the root cell.The rule is to decompose the cell into 4 quadrants if it contains designated "obstacles".This decomposition process is recursively executed until a defined depth level is reached or there are no quadrants with obstacles.The pseudo code of the cell decomposition is shown in Figure 8.Procedure CellDecomp(T reeDepth, CurrDepth, Cell, Obstacles) ; ; T reeDepth is predef ined and CurrDepth is initialized as 0; begin j ← CurrDepth; while j < T reeDepth and Cell is M ixed do ; W hen a cell intersects Obstacles, it is M ixed .Otherwise Empty ; Decompose Cell into f our sons {N W, N E, SW, SE}; j ← j -1; for each c ∈ {N W, N E, SW, SE} do CellDecomp(T reeDepth, j, c, Obstacles); end The following is the procedure for using quad-tree cell-decomposition to detect crossings:1. Define a depth level and location of a corridor.2. Preprocessing : Perform cell-decomposition with flight segments as obstacles in the entire NAS up to the given depth level.Then every cell at the deepest level is associated with a list of flight segments.3. Decomposing : Perform cell-decomposition with the corridor as an obstacle up to the given depth level so that the corridor is associated with a list of leaf cells (cells with the smallest size).4. Checking : For each leaf cell associated with the corridor, check its associated flights to see if the flights intersect with the corridor.If yes, count the number of distinct flights.Fig. 9(a) displays the decomposed cells for National Airspace System (NAS) traffic with depth level 6 as in step 2. Because the flights are all over the US continental region, most of the cells reach the deepest level.Only a few cells at the left-bottom part are big because there were no flight records.Fig. 9(b) shows the decomposed cells for given corridor as in Step 3. Cells around the corridor are small.Since cells in step 2 are mapped with the cells in step 3, only flight segments associated with these small cells in step 3 will be considered for checking.This is expected to significantly lower the number of flights that needs to be examined.!!"# !!$# !!!# !!## !%# !&# !'# $# $( "# "( )# )( (#(b) +B. Crossing Detection PerformanceTo check the performance of the crossing detection, the track data on April 20, 2007 is utilized, and the brute-force method is used as baseline.All flights above FL290 will be checked.According to the data, there are 3, 454, 185 flight segments above FL290.The sample case will be, given a corridor at FL310, check the number of crossings between non-corridor traffic and the corridor.The brute-force method can take advantage of the altitude range by only considering the flight segments between FL300 and FL320, But the number of flight segments in that range is as high as 205, 222.The experiments were run on a Mac machine with an Intel dual core CPU at 3.0 GHz.By applying the brute-force method, it takes 8.6 s to find the solution.While using quad-tree cell decomposition method, it can take as little as 0.048 s, which is 180 times faster than the the brute-force method.The reason is that brute-force wastes a large amount of time on checking flight segments far from the corridor.Figure 10 presents the relative performance when the number of tree depth is increased to 8. Depth 7 and 8 have the same computational time.It is noted that depth level 8 can not gain any more benefit over depth 7.That is because the cell sizes at depth 7 are already comparable to the size of one-minute flight segments.Further decomposing cells can not rule out more flight segments.Figure 11 shows the preprocessing time for different depth levels.Since the tree only needs to be built once, the long preprocessing time should be negligible when the detection needs to be performed many times.Based on these two figures, depth level 7 is recommended when performing cell-decomposition for detecting crossings.With a computation time of 0.048s, this method can be incorporated into the optimization to find good corridor candidates that also minimize crossings of non-corridor traffic. +of 11American Institute of Aeronautics and Astronautics +C. Lane OptionsSection III.B indicated that at least 7 lanes might be needed for the given corridor.If corridors are expected to be impenetrable, it is desired to find a lane option that minimizes the number of crossing flights.The cell-decomposition method described above is used.The middle part of the corridor from the location of 500 nmi to the location of 1250 nmi is studied using three lane options.Fig. 12(a) shows vertically stacked lanes.Fig. 12(b) displays side-by-side lanes.While Fig. 12(c) shows a combined option, with a 3-lanes at one altitude and a 4-lanes at another altitude.Figure 13 presents the comparison of these three lane options.Three sets of stacked lanes are shown as horizontal lines that cover 7 altitudes range, respectively.To simplify the figure, "combined" lanes are shown for four different altitudes chosen for the 3-lane part.The 4-lane part is allowed to vary between FL290 and FL410.The 3-lane altitudes for combined lanes 1, 2, 3, and 4 are FL410, FL350, FL340, and FL290, respectively.Contradicting one's intuition, except at super-high altitudes, the side-by-side lane option is the worst according to the number of crossing flights.Although it has low crossings at super high altitudes, these altitudes may not be feasible for all corridor users.The stacked lane options have relatively low crossings, but since they block seven consecutive altitudes, this option may cause high cost for non-corridor traffic to cap and tunnel.The combined options that have one part at high altitude and another at low altitude seem attractive.They have relatively low numbers of crossings and provide corridor users the flexibility of choosing two different altitudes.Although, these don't serve as final analysis, they provide insights of the impact on remaining traffic.There are big differences in the impacts if different lane options are chosen. +V. Initial Benefit AnalysisBased on previous research, 1 one possible advantage of having corridor airspace is to increase overall airspace capacity.As a preliminary study, the benefit and cost will be discussed based on the reduction of the peak aircraft count and the number of crossings, respectively.From the analysis in above sections, if the corridor is enabled, the number of crossings would be around 2, 000 (see Fig. 13), which is dependent on the lane profile.The peak aircraft count will be used as an approximate measure of complexity.The corridor and its underlying sectors are shown in Fig. 14.The peak aircraft count is first examined for the sectors without the corridor.Next, to study the peak aircraft count of sectors when the corridor is enabled, the corridor traffic is simply removed from the sectors.The time history of the peak aircraft count of Washington Center high sector ZDC72 with and without the corridor are shown in Fig. 15 with a blue and red curve, respectively.The difference of these two is shown with a green curve.The difference of traffic has coincided peaks with overall traffic (the blue curve) in ZDC72.Thus, the average reduction of the peak aircraft count due to enabling the corridor is as high as 26.1%.In fact, similar situations happen in other sectors with this corridor enabled.The results show that Sectors ZDC04, ZTL28, and ZNY09 have 17.4%, 27.3%, and 24.1% reduction of the peak aircraft count, respectively.Thus, the corridor concept seems to be a positive solution to increase the NAS capacity. +VI. ConclusionA single corridor design is analyzed in this work.To examine and visualize the utilization of this corridor, a space-time map was developed.It can be used to guide construction of dynamic corridors and to suggest the minimum number of lanes.A quad-tree cell-decomposition was developed to speed up the crossing checks by a factor over one hundred.This method can be used to minimize crossings in the initial corridor design.In the discussion of lane options in terms of minimizing crossings, it was found that combining multiple lanes at multiple altitudes reduces the number of crossings and provides flexibility for corridor users.The analyses of corridor utilization and impact on the remaining traffic is important for understanding or evaluating corridors.They actually can be integrated into the initial corridor design.Preliminary benefit analysis was discussed based on the peak aircraft count and the number of crossings.Results show around 25% reduction of the peak aircraft count in underlying sectors if only one corridor is enabled.Crossings can be low if a combined lane option was used.This benefit analysis provides new metrics for corridor airspace.In the future, research will focus on how flights would be operated in corridors.Feasible simulations will be carried out based on these design information of a corridor, such as 2D locations, preferred altitudes, lane option, and ramp locations.Figure 1 .1Figure 1.Top 60 corridors ranked by number of attendees +Figure 2 .2Figure 2. Corridor Model.(a) An example of how a user joins and exits a corridor.(b) the corridor picked for analysis +Figure 3 .3Figure 3. Space Time Map.(a) A corridor with scale.(b) Sketch of Space-time Map +Figure 4 .Figure 5 .45Figure 4. Space Time Map for Corridor No. 1 +Figure 6 .Figure 7 .67Figure 6.Space-Time Distribution of Entries and Exits +Figure 8 .8Figure 8. CellDecomp Algorithm +Figure 9 .9Figure 9. Cell-Decomposition for Detecting Crossings.(a) Preprocessing in Step 2. (b) Decomposition of the corridor in Step 3 +Figure 10.Relative Performances of Crossing Detection +vertical +Figure 12 .12Figure 12.Corridor Cross Sections of Lane Options.(a) Stacked Lanes.(b) Side-by-side Lanes.Right: Combined Lanes. +500 1000 1500 2000 2500 3000 3500 4000 Flight Level Number of Crossing Flights Stacked Lane 1 Stacked Lane 2 Stacked Lane 3 Side!by!side Lane Combined Lane 1 Combined Lane 2 Combined Lane 3 Combined Lane 4Figure 13.Comparison Among Different Lane Options280 0300320340360380400420 + of 11 American Institute of Aeronautics and Astronautics + of 11 American Institute of Aeronautics and Astronautics + of 11 American Institute of Aeronautics and Astronautics + + + + + + + + + Initial Concepts for Dynamic Airspace Configuration + + ParimalKopardekar + + + KarlBilimoria + + + BanavarSridhar + + 10.2514/6.2007-7763 + + + 7th AIAA ATIO Conf, 2nd CEIAT Int'l Conf on Innov and Integr in Aero Sciences,17th LTA Systems Tech Conf; followed by 2nd TEOS Forum + Belfast, Northern Ireland + + American Institute of Aeronautics and Astronautics + 18-20 September 2007 + + + Kopardekar, P., Bilimoria, K., and Sridhar, B., "Initial Concepts for Dynamic Airspace Configuration," 7th AIAA Aviation Technology, Integration and Operations Conference (ATIO), Belfast, Northern Ireland, 18-20 September 2007. + + + + + Dynamic airspace super sectors (DASS) as high-density highways in the sky for a new US air traffic management system + + JAlipio + + + PCastro + + + HKaing + + + NShahid + + + OSherzai + + + GLDonohue + + + KGrundmann + + 10.1109/sieds.2003.158005 + + + IEEE Systems and Information Engineering Design Symposium, 2003 + + IEEE + October 2003 + + + + Alipio, J., Castro, P., Kaing, H., Shahd, N., Sheizai, O., Donohue, G., and Grundmann, K., "Dynamic Airspace Super Sectors (DASS) As High-Density Highways in The Sky for A New US Air Traffic Management System," AIAA/IEEE Digital Avionics Systems Conference, 12-16 October 2003. + + + + + High-volume tube-shape sectors (HTS): a network of high capacity ribbons connecting congested city pairs + + AYousefi + + + GLDonohue + + + LSherry + + 10.1109/dasc.2004.1391296 + + + The 23rd Digital Avionics Systems Conference (IEEE Cat. No.04CH37576) + Salt Lake City, CT + + IEEE + 2004 + + + Yousefi, A., Donohue, G., and Sherry, L., "High-Volume Tube-Shape Sectors(HTS): A Network of High Capacity Ribbons Connecting Congested City Pairs," Proceedings of the 23rd Digital Avionics Systems Conference, Salt Lake City, CT, 2004. + + + + + Initial Study of Tube Networks for Flexible Airspace Utilization + + BanavarSridhar + + + ShonGrabbe + + + KapilSheth + + + KarlBilimoria + + 10.2514/6.2006-6768 + + + AIAA Guidance, Navigation, and Control Conference and Exhibit + Keystone, Colorado + + American Institute of Aeronautics and Astronautics + August 2006 + + + + Sridhar, B., Grabbe, S., Sheth, K., and Bilimoria, K., "Initial Study of Tube Networks for Flexible Airspace Utilization," AIAA Guidance, Navigation, and Control Conference and Exhibit, Keystone, Colorado, 21-24 August 2006. + + + + + Freeways in the Sky: Exploring Tube Airspace Design Through Mixed Integer Programming + + GautamGupta + + + BanavarSridhar + + + AvijitMukherjee + + 10.2514/6.2008-6824 + + + AIAA Guidance, Navigation and Control Conference and Exhibit + Washington, D.C. + + American Institute of Aeronautics and Astronautics + October 2008 + + + Gupta, G., Sridhar, B., and Mukherjee, A., "Freeways in the Sky: Exploring Tube Airspace design through Mixed Integer Programming," INFORMS Annual Meeting, Washington, D.C., October 2008. + + + + + High-Capacity Tube Network Design using the Hough Transform + + MinXue + + + ParimalKopardekar + + 10.2514/6.2008-7396 + + + AIAA Guidance, Navigation and Control Conference and Exhibit + Honolulu, Hawaii + + American Institute of Aeronautics and Astronautics + August 18-21 2008 + + + Xue, M. and Kopardekar, P., "High-Capacity Tube Network Design using the Hough Transform," AIAA Guidance, Navigation and Control Conference and Exhibit, Honolulu, Hawaii, August 18-21 2008. + + + + + Air Traffic Freeway System for Europe + + HHering + + No. 20/05 + + 2005 + Eurocontrol Experimental Centre + + + Tech. Rep. EEC Nore + Hering, H., "Air Traffic Freeway System for Europe," Tech. Rep. EEC Nore No. 20/05, Eurocontrol Experimental Centre, 2005. + + + + + Principles of Airspace Tube Design for Dynamic Airspace Configuration + + RobertHoffman + + + JosephPrete + + 10.2514/6.2008-8939 + + + The 26th Congress of ICAS and 8th AIAA ATIO + Anchorage, Alaska + + American Institute of Aeronautics and Astronautics + September 14-19 2008 + + + Hoffman, R. and Prete, J., "Principles of Airspace Tube Design for Dynamic Airspace Configuration," The 8th AIAA Aircraft Technology, Integration, and Operation Conference, Anchorage, Alaska, September 14-19 2008. + + + + + Analysis of airspace tube structures + + KapilSSheth + + + TanimSIslam + + + ParimalHKopardekar + + 10.1109/dasc.2008.4702805 + + + 2008 IEEE/AIAA 27th Digital Avionics Systems Conference + St. Paul, Minnesota + + IEEE + October 26-30 2008 + + + Sheth, K. S., Islam, T. S., and Kopardekar, P. H., "Analysis of Airspace Tube Structures," The 27th Digital Avionics Systems Conference, St. Paul, Minnesota, October 26-30 2008. + + + + + On kinematic waves II. A theory of traffic flow on long crowded roads + + MJLighthill + + + GBWhitham + + 10.1098/rspa.1955.0089 + + + Proceedings of the Royal Society of London. Series A. Mathematical and Physical Sciences + Proc. R. Soc. Lond. A + 0080-4630 + 2053-9169 + + 229 + 1178 + + 1956 + The Royal Society + + + Lighthill, M. J. and Whitham, G. B., "On Kinemetic Waves. II. A Theory of Traffic Flow on Long Crowded Roads," Proceedings of the Royal Society of London, Vol. 229, No. 1178, 1956, pp. 317-345. + + + + + A finite difference approximation of the kinematic wave model of traffic flow + + CarlosFDaganzo + + 10.1016/0191-2615(95)00004-w + + + Transportation Research Part B: Methodological + Transportation Research Part B: Methodological + 0191-2615 + + 29 + 4 + + 1995 + Elsevier BV + + + Daganzo, C. F., "A Finite Difference Approximation of The Kinematic Wave Model of Traffic Flow," Transportation Research B , Vol. 29B, No. 4, 1995, pp. 261-276. + + + + + Robot Motion Planning + + J.-CLatombe + + 10.1007/978-1-4615-4022-9 + + + Robot Motion Planning + + Springer US + 1991 + + + Latombe, J.-C., Robot Motion Planning, Kluwer Academic Press, 1991. + + + + + + diff --git a/file807.txt b/file807.txt new file mode 100644 index 0000000000000000000000000000000000000000..5f97f7f53f16aa5204e68264c225200e2e39c023 --- /dev/null +++ b/file807.txt @@ -0,0 +1,655 @@ + + + + +I. IntroductionIn recent years, driven by the rapidly growing interest in air passenger and cargo transportation within an urban area, the concepts of unmanned aircraft system traffic management (UTM) [1] and urban air mobility (UAM) [2] have been proposed to enable safe and efficient aerial vehicle operations.These new concepts change the paradigm of existing air traffic systems because of a variety of factors, including various vehicle characteristics, increased density of operations, urban terrain environments, and complex atmospheric condition in low-altitude airspace.How to safely and efficiently handle high-density operations in low-altitude airspace faces great challenges.Because of the above characteristics, many minor factors in conventional aviation and large Unmanned Aerial System (UAS) operations may play important roles in the new operational environments.For instance, the communication update rate becomes critical as vehicles fly much closer to each other than in conventional aviation.When vehicle density gets high, the communication signal interference becomes severe due to the high communication channel load [3,4].Moreover, while new vehicle types (e.g., multicopters) can execute agile operations in urban areas, they also impose complexity on analyzing airspace operation capacity.As vehicles operate closely, the effects of position accuracy and wind on a UAM/UTM traffic system need to be investigated.Although there were many studies on complexity or capacity for conventional aviation [5][6][7][8][9] and some initial studies on small UAS (sUAS) operations [10,11], the understanding of key factors that affect the high-density UAM and UTM operations is still limited.In-depth studies must be conducted to help stakeholders understand and derive requirements for these factors, including equipages, conflict management models, and wind conditions.These studies can be used to help enable safe and efficient high-density air traffic operations.This work analyzed five factors that can impact the traffic system: position accuracy, communication latency, wind uncertainty, traffic density, and separation buffers.To analyze the impact of these factors on safety and efficiency in high density operations, hundreds of experiments were conducted using the Fe 3 simulation tool [12].Section II introduces the experimental setup in this work, including the conflict resolution model and test scenarios.Section III presents sensitivity study results and analyses.Section IV summarizes the findings in this work. +II. Experiment SetupThe Fe 3 simulation tool incorporates many models that are involved in low-altitude air traffic operations, such as six degrees of freedom (DOF) vehicle aerodynamic and control models, conflict resolution models, wind, communication, navigation, and energy consumption.A description of these models can be found in previous research [12].To understand experimental results and analyses in this work, the experiment setup is presented in this section to provide necessary context, including the conflict management model and test scenarios. +A. Conflict Management ModelAccording to the UTM conflict management model published by the NASA UTM research transition team (RTT) sense and avoid (SAA) working group [13], a three-phase confliction management model was proposed to resolve predicted conflicts between two sUASs: strategic conflict management, separation provision, and collision avoidance.The strategic conflict management changes flight plans to resolve conflicts and normally occurs prior to departure.The separation provision is a tactical process for conflicts predicted while airborne to prevent well-clear violations.The collision avoidance is triggered when the well-clear definition is violated and serves as the last layer of conflict resolution.Because the strategic separation phase occurs prior to departure and will be handled by scheduling and re-routing flight plans through the UAS Service Supplier (USS) and/or Supplemental Data Service Provider (SDSP), the conflict management model in this work only covers separation provision and collision avoidance phases when sUASs are aloft after departure. +Fig. 1 Structure used in the conflict resolution algorithmsSince the RTT work package didn't specify the complete structure for the separation provision, a detailed framework is proposed to implement the conflict resolution algorithm for this work.Figure 1 shows the structure (not to scale) used in the simulator.The well-clear (WC) threshold d WC provides minimal space for collision avoidance for which both ownship and intruder can take actions as the last effort before colliding with each other and serves as the boundary between the collision avoidance and separation provision phases.The well-clear buffer d WC B defines the extra buffer used to improve error tolerance.The look-ahead time (LAT) t L AT defines the time horizon for trajectory prediction.The conflict warning threshold (CWT) defines the time t CWT when a conflict warning becomes effective.The conflict resolution threshold (CRT) defines the time t C RT when the ownship can start to maneuver to resolve the conflict.Since developing the well-clear definition is still an open research question [11,14] The model used in this work utilizes discrete heading changes as the only resolution maneuver option.Vehicles are assumed to fly cooperatively following a set of predefined conflict resolution logics to decide the right of the way and the preferred resolution maneuver when there are multiple solutions.While defining a safe and efficient set of conflict resolution logic remains as a challenging problem [15,16], this work adopts the following simple logics to decide the right of way and the resolution maneuver preference:• If the intruder is on the right side of the ownship, the intruder has the right of way and the ownship shall take a conflict resolution maneuver.• If it's a head-on conflict, both the ownship and intruder shall take right turns.• If the ownship and intruder are in-trail, the front vehicle has the right of the way and the trailing vehicle shall take an avoidance maneuver.• If the well-clear definition is violated, both ownship and intruder need to take maneuvers to avoid collision.• When there are multiple resolutions, the maneuver with minimum deviation from its planned trajectory is chosen.The conflict resolution model uses the constant-velocity ("dead-reckoning") trajectory projection with a maximum turn rate when producing trial trajectories corresponding to various heading change options.The update rate of the conflict resolution model is set to 500 milliseconds, which is assumed to be the approximated computational time needed for a conflict resolution algorithm [17].For reference, the update rate of ACAS-X is expected at 1Hz [18] and the same is expected for the Detect And Avoid (DAA) algorithms in the Java Architecture for Detect and Avoid Extensibility and Modeling (JADEM) [19]. +B. Test ScenariosPairwise encounter scenarios are typically used in analyzing the performance of DAA algorithms.Despite their popularity, pairwise encounter scenarios are not enough for systematic analyses in medium and high density operational environments (e.g., >1 aircraft/nmi 2 ), where aircraft operate much closer to each other and multi-aircraft encounters are common.Previous work [12] showed that even though a conflict resolution algorithm resolved all conflicts in thousands of pairwise encounter scenarios with varied relative speeds and encounter angles, it still could not prevent losses of separation in multi-aircraft encounter scenarios, where a loss of separation happens when the well-clear threshold is violated.Apparently, multi-aircraft encounter scenarios serve the purpose better in systematic analyses.The traffic density is a simple and intuitive metric commonly used to describe the airspace operation complexity.It's usually defined by the number of aircraft divided by the operating area over a given period (e.g., number of aircraft per nmi 2 ).However, this definition is not an accurate complexity measurement, especially when comparing a scenario of multiple aircraft flying in parallel against a scenario of multiple aircraft flying towards each other with potential conflicts.In this work, six multiple-aircraft scenarios were created, not only with increased traffic density but also with increased number of potential conflicts (when no conflict resolution maneuver is used) to ensure increased complexity.Because quantifying the complexity is still an ongoing research topic, it's not clear if the complexity of these scenarios increases proportionally with the aircraft count or with the number of potential conflicts.Flights in these scenarios are planned to fly direct routes, which means that each flight flies directly from its own origin to destination without any planned turns as shown in Fig. 2(a).The target ground speeds of all sUASs are 22 fps (or 15 mph) and approximated flight times are around 450 seconds.The operation area is about 3 nmi 2 .In this study, a quad-rotor vehicle model is used for all vehicles.Table 1 lists some attributes of these six scenarios.As presented in the table, the aircraft count varies from 5 to 30 aircraft and the number of potential conflicts (without any conflict resolution maneuver) increases from 1 to 10. Convex hull was used to create a boundary around all sUASs (with each sUAS wrapped with separation buffers).This boundary represents the current operating area.The history of instantaneous traffic density (over one second) is calculated and shown in Fig. 2(b) to provide some insight on the complexity of these test scenarios.The averages of these instantaneous traffic densities are presented in Table 1. +A. Position accuracyAn intruder's position can be obtained through sensors (on-board or ground-based) or shared by the intruder via various communication approaches.This section examines the effect of intruder position error despite the means of obtaining the information.It assumes that each vehicle's own position is known accurately, whereas intruder positions include errors that follow a normal distribution with a zero mean.Position errors are introduced purely by the standard deviation.With a specified standard deviation, the Fe 3 simulator can generate random errors to simulate intruder positions received by the ownship and feed them into the conflict management algorithm.About 1,000 Monte Carlo simulations are performed for each position accuracy level defined by standard deviation to obtain the probability of loss of separation and the percentage of extra flight distance.Figure 3(a) presents the probability of loss of separation under various density levels and intruder position accuracy levels.The cold colors denote low probability of loss of separation whereas the warm colors represent high possibility of loss of separation.It is noted that the dark blue grids are safe operational regions where no losses of separation are expected.According to this plot, one may derive requirements for safe operations given the default setup in this work: At the density level with 5 vehicles, the standard deviation of intruder position received by the ownship must be less than 80 feet; when the number of operating vehicles increases to 20, the intruder position error must be less than 40 feet; when the aircraft count is between 20 and 30 vehicles, no position error should be allowed for any safe operation.Figure 3(b) shows the percentage of extra flight distance, which can be used to help define requirements for energy reserves.This metric is directly related to requirements for battery life.It is shown that within the region where there is a zero percent chance of loss of separation (the dark blue area in Fig. 3(a)), when the aircraft count is less than 15 vehicles, the percentage of extra flight distance is typically less than 5%, which implies that 5% energy reserve should be adequate.When the aircraft count is 20 and the position standard deviation is over 60 feet, the percentage of extra flight distance is over 10%, which indicates that the energy reserve has to be greater than 10% to deal with position uncertainties.It's clear that position uncertainty affects both safety and efficiency.Improving position accuracy not only increases the safety level but also helps improve operational efficiency.see that when the aircraft count is lower than 15, these data suggest that it is safe to operate even with a 60 second communication latency.This requirement on communication latency is generous.Whereas, at an aircraft count of 20, the communication delay has to be less than 1 second for safe operation.Unlike the gradual transition in with position error, the probability of loss of separation increases suddenly once the aircraft count is higher than 15. Figure 4(a) shows that the safety measurement (based on loss of separation) is more sensitive to the traffic density, or aircraft count, than the communication latency.Comparing these results to position accuracy, when there are 20 vehicles, the safety measurement is less tolerant to communication latency than to position error (where the operations were still safe within a 40-foot position error).Figure.4(b) presents the percentage of extra flight distance and shows that within the safe operation zone (i.e., the dark blue region in Fig. 4(a)), the inefficiency caused by communication latency is close to 10%, which is slightly higher than what was seen within the safe region for the position uncertainty case. +C. WindWind disturbance becomes a major concern for UAS operation, especially when these small and lightweight UASs operate in low-altitude airspace where the wind gust is much less predictable within the Atmospheric Boundary Layer (ABL).Experiments with different crosswind conditions are explored in this section.The wind condition at any location is defined as a statistical distribution with a mean and standard deviation (intensity).The mean of the crosswind is limited to 0 to 10 mps because the quadrotor vehicle model used in this work cannot conform to its trajectory once the crosswind is higher than 12 mps.The wind intensity is simply defined as one tenth of the mean [12,24].Figure . 5(a) shows the probability of loss of separation at various crosswind speeds.It is noticed that crosswind is not an issue when the aircraft count is lower than 20 (assuming the crosswind is under the 12 mps maximum operational limit).The hypothesis is that the vehicle control system mitigated the trajectory disturbance by applying tight controls to make sure each vehicle conforms to its planned trajectory as much as possible.Therefore, the disturbance caused by crosswind didn't reduce the trajectory projection accuracy enough to affect the performance of the conflict resolution algorithm.Comparing against the previous section, there are not many differences in terms of safe operation area and efficiency, especially at low traffic density levels.However, the safety and efficiency measurements become a bit more sensitive to communication latency when the aircraft count is higher than 15.Similar trends hold for the wind study.As shown in Fig. 8(a) and 8(b), reducing the separation buffer didn't affect safety and efficiency very much.According to this study, it seems that increasing the separation buffer makes the traffic system more robust to position errors, while improving the robustness to communication latency and crosswind less noticeably. +E. DiscussionGiven a specific conflict management model, the above experiments showed that sensitivity analyses of key factors could be examined through the Monte Carlo simulation capability in Fe 3 .Experiments on position accuracy showed that intruder's position accuracy not only affected the safety of a sUAS traffic system but also its efficiency.Data suggest that a sUAS traffic system can safely operate up to 20 vehicles with zero losses of separation if position errors can meet corresponding requirements.The extra energy reserve, derived from extra flight distance, is less than 5%.Studies in communication latency showed two extreme conclusions that high density operations are vulnerable to communication latency, whereas, low density operations are robust to it.Extra energy consumed due to communication latency can reach 10 % within the safe operation zone.Although there may exist a maximum crosswind that a sUAS can tolerate, with an appropriate control system, the crosswind should not be a major concern for maintaining safety and extra flight distance, especially at a relatively low traffic density.However, extra energy may be consumed due to extra drag force caused by wind, even though the extra flight distance is very low.Studies in separation buffer showed that increasing the separation buffer improved the robustness to position errors, whereas it did not drastically improve the robustness to communication latency and wind.Findings in this work may change if the conflict management model (including structure, parameters, and logics) changes.However, given a safety and efficiency threshold, these types of studies (using Fe 3 simulations) can be used to derive and evaluate requirements for a high-density traffic system including equipages and conflict management models. +IV. SummaryThis paper presented a systematic sensitivity analysis of key contributors effects on safety and efficiency in high density UAS operations.The structure and model of the conflict resolution algorithm used in the simulator are presented first to setup the context for experiments in this work.The test scenarios with increasing complexity are then introduced.The probability of loss of separation and the percentage of extra flight distance are used in this work to represent safety and efficiency measurements, respectively.Five key factors including position accuracy, communication latency, wind, separation buffer, and traffic density, were then investigated.Experiments showed that requirements on these factors can be easily derived.For instance, 20 sUASs can safely operate in the given area under the default setup with zero losses of separation as long as the position error is less than 40 feet.For this configuration, the energy reserve must be higher than 5%.Studies also reveal relationships among those factors.For example, increasing the separation buffer improved the system robustness to position errors but doesn't improve its tolerance to communication latency and wind.This type of Fe 3 -based study can be used to define requirements for a high-density traffic system.Future work will extend to sensitivity analyses with conflict management models that are more comprehensive and realistic.Parameters and logics will be studied for evaluating and developing a safe and efficient conflict management model.Meanwhile, different concept of operations, like route structure based operations, will also be investigated to find out if and when route structure based operations are beneficial.Mixing vehicle types and their performance characteristics will also be considered.In addition, using these sensitivity analyses will allow a real-time model to be developed to quickly assess the airborne risk in terms of metrics like the probability of loss of separation and minimum energy reserve., a distance of 30 feet is arbitrarily chosen for d W C this study.The default settings of these parameters are: t L AT = 50 seconds, t CWT = 50 seconds, t C RT = 30 seconds, d W C = 30 feet, and d W C B = 50 feet. +Fig. 2 Test Scenarios +Fig. 33Fig. 3 Impact of Position Errors on Safety and Efficiency +Fig. 44Fig. 4 Impact of Communication Latency on Safety and Efficiency +Figure 5 (5b) confirms that the extra distance is relatively low due to the tight control law utilized by the vehicle.The percentage of extra distance is typically around 2% when the aircraft count is lower than 20.It should be noted that the extra flight distance may not represent the extra energy in this case.Due to the increase of the drag force caused by wind, extra energy will be consumed to produce extra forces, which is not reflected by the extra flight distance in Fig. 5(b). +Fig. 5 Fig. 6 Fig. 7567Fig. 5 Impact of Cross Wind on Safety and Efficiency +Fig. 88Fig. 8 Impact of Separation Buffer on Safety and Efficiency: Wind +Table 1 Test Scenarios1ScenarioIIIIII IVVVIAircraft count510 15 20 25 30Traffic density over the entire simulation period (aircraft/nmi 2 ) 1.6 3.2 4.8 6.5 8.1 9.6Average of instantaneous traffic density (aircraft/nmi 2 )59 69 63 68 72 85Number of potential conflicts1225710 + + + + +Mail Stop 210-15.AIAA senior member. + + + + + + + + + Unmanned Aircraft System Traffic Management (UTM) Concept of Operations + + PKopardekar + + + JRios + + + TPrevot + + + MJohnson + + + JJung + + + JERobinson + + + + 16th AIAA Aviation Technology, Integration, and Operations Conference + Washington, D.C. + + 2016 + + + Kopardekar, P., Rios, J., Prevot, T., Johnson, M., Jung, J., and Robinson, J. E., "Unmanned Aircraft System Traffic Management (UTM) Concept of Operations," 16th AIAA Aviation Technology, Integration, and Operations Conference, Washington, D.C., 2016. + + + + + Enabling Airspace Integration for High-Density On-Demand Mobility Operations + + EricRMueller + + + ParmialHKopardekar + + + KennethHGoodrich + + 10.2514/6.2017-3086 + + + 17th AIAA Aviation Technology, Integration, and Operations Conference + Denver, Colorado + + American Institute of Aeronautics and Astronautics + 2017. 2017 + + + Mueller, P., E. Kopardekar, and Goodrich, K., "Enabling Airspace Integration for High-Density On-Demand Mobility Operations," AIAA Aviation 2017, Denver, Colorado, 2017. + + + + + Communicating with unmanned aerial swarm automatic dependent surveillance transponders + + SreejaNag + + + JaewooJung + + + KarishmaInamdar + + 10.1109/icsens.2017.8234227 + + + 2017 IEEE SENSORS + Glasgow, UK + + IEEE + 2017 + + + Nag, S., Jung, J., and Inamdar, K., "Communicating with unmanned aerial swarm automatic dependent surveillance transponders," IEEE on SENSORS, Glasgow, UK, 2017. + + + + + Communication Density: A Channel Load Metric for Vehicular Communications Research + + DanielJiang + + + QiChen + + + LucaDelgrossi + + 10.1109/mobhoc.2007.4428734 + + + 2007 IEEE Internatonal Conference on Mobile Adhoc and Sensor Systems + Pisa, Italy + + IEEE + 2007 + + + Jiang, D., Chen, Q., and Delgrossi, L., "Communication Density: A Channel Load Metric for Vehicular Communications Research," IEEE International Conference on Mobile Adhoc and Sensor Systems (MASS), Pisa, Italy, 2007. + + + + + Dynamic density: measuring and predicting sector complexity [ATC] + + PKopardekar + + + SMagyarits + + 10.1109/dasc.2002.1067920 + + + Proceedings. The 21st Digital Avionics Systems Conference + The 21st Digital Avionics Systems ConferenceIrvine, CA + + IEEE + 2002 + + + Kopardekar, P., and Magyarits, S., "Measurement and Prediction of Dynamic Density," The 21st Digital Avionics Systems Conference (DASC), Irvine, CA., 2002. + + + + + Describing Air Traffic Complexity Using Mathematical Programming + + MariyaIshutkina + + + EricFeron + + + KarlBilimoria + + 10.2514/6.2005-7453 + + + AIAA 5th ATIO and16th Lighter-Than-Air Sys Tech. and Balloon Systems Conferences + + American Institute of Aeronautics and Astronautics + 2005 + + + Ishutkina, M. A., and Feron, E., "Describing Air Traffic Complexity Using Mathematical Programming," AIAA 5th Aviation, Technology, Integration, and Operations Conference (ATIO), 2005. + + + + + Dynamic Density and Complexity Metrics for Realtime Traffic Flow Management + + AMasalonis + + + MCallaham + + + CWanke + + + 2003 + + + Masalonis, A., Callaham, M., and Wanke, C., "Dynamic Density and Complexity Metrics for Realtime Traffic Flow Management," 2003. + + + + + Air Traffic Complexity Map based on Non Linear Dynamical Systems + + DanielDelahaye + + + StephanePuechmorel + + + JohnHansman + + + JonathanHiston + + 10.2514/atcq.12.4.367 + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 12 + 4 + + 2003 + American Institute of Aeronautics and Astronautics (AIAA) + + + Delahaye, D., Puechmorel, S., Hansman, J., and Histon, J., "Air Traffic Complexity based on Non Linear Dynamical Systems," 2003. + + + + + An Efficient Method for Airspace Analysis and Partitioning Based on Equalized Traffic Mass + + AKlein + + + + Europe Seminar on Air Traffic Management Research and Development + + 2005 + Baltimore, MD + + + 6th USA/ + Klein, A., "An Efficient Method for Airspace Analysis and Partitioning Based on Equalized Traffic Mass," 6th USA/Europe Seminar on Air Traffic Management Research and Development, Baltimore, MD, 2005. + + + + + A Throughput Based Capacity Metric for Low-Altitude Airspace + + VishwanathBulusu + + + RajaSengupta + + + EricRMueller + + + MinXue + + 10.2514/6.2018-3032 + + + 2018 Aviation Technology, Integration, and Operations Conference + Atlanta, GA + + American Institute of Aeronautics and Astronautics + 2018 + + + Bulusu, V., Sengupta, R., Mueller, E. R., and Xue, M., "A Throughput-Based Capacity Metric for Low-altitude Airspace," AIAA Aviation Forum, Atlanta, GA., 2018. + + + + + A Well Clear Recommendation for Small UAS in High-Density, ADS-B-Enabled Airspace + + MatthewDuffield + + + TimWMclain + + 10.2514/6.2017-0908 + + + AIAA Information Systems-AIAA Infotech @ Aerospace + Grapevine, Texas + + American Institute of Aeronautics and Astronautics + 2017 + + + Duffield, M. O., and McLain, T. W., "A Well Clear Recommendation for Small UAS in High-Density ADS-B-Enabled Airspace," AIAA SciTech Forum, Grapevine, Texas, 2017. + + + + + Fe3: An Evaluation Tool for Low-Altitude Air Traffic Operations + + MinXue + + + JosephRios + + + JosephSilva + + + ZhifanZhu + + + AbrahamKIshihara + + 10.2514/6.2018-3848 + + + 2018 Aviation Technology, Integration, and Operations Conference + Atlanta, GA + + American Institute of Aeronautics and Astronautics + 2018 + + + Xue, M., Rios, J., Silva, J., Ishihara, A., and Zhu, Z., "Fe 3 : An Evaluation Tool for Low-Altitude Air Traffic Operations," AIAA Aviation Forum, Atlanta, GA., 2018. + + + + + Flight Demonstration of Unmanned Aircraft System (UAS) Traffic Management (UTM) at Technical Capa... + + MJohnson + + + JLarrow + + 10.2514/6.2020-2851.vid + + 2018 + American Institute of Aeronautics and Astronautics (AIAA) + + + NASA + + + Tech. Rep. Technical Report + Johnson, M., and Larrow, J., "UTM Research Transition Team Sense and Avoid Working Group Technical Work Package #2: UTM Conflict Management Model," Tech. Rep. Technical Report, NASA, 2018. + + + + + Well-Clear Recommendation for Small Unmanned Aircraft Systems Based on Unmitigated Collision Risk + + AWeinert + + + SCampbell + + + AVela + + + DSchuldt + + + JKurucar + + + + AIAA Journal of Air Transportation + + 26 + 3 + 2018 + + + Weinert, A., Campbell, S., Vela, A., Schuldt, D., and Kurucar, J., "Well-Clear Recommendation for Small Unmanned Aircraft Systems Based on Unmitigated Collision Risk," AIAA Journal of Air Transportation, Vol. 26, No. 3, 2018. + + + + + A Fuzzy Logic Approach for Low Altitude UAS Traffic Management (UTM) + + BrandonCook + + + KellyCohen + + + EladHKivelevitch + + 10.2514/6.2016-1905 + + + AIAA Infotech @ Aerospace + San Diego, California + + American Institute of Aeronautics and Astronautics + 2016. 2016 + + + AIAA Science and Technology Forum and Exposition + Cook, B., Cohen, K., and Kivelevitch, E., "A Fuzzy Logic Approach For Low Altitude UAS Traffic Management (UTM)," AIAA Science and Technology Forum and Exposition 2016, San Diego, California, 2016. + + + + + A New Approach for Designing Safer Collision Avoidance Systems + + MykelJKochenderfer + + + JamesPChryssanthacopoulos + + + RolandEWeibel + + 10.2514/atcq.20.1.27 + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 20 + 1 + + 2012 + American Institute of Aeronautics and Astronautics (AIAA) + + + Kochenderfer, M., Chryssanthacopoulos, J. P., and Weibel, R. E., "A new approach for designing safer collision avoidance systems," Air Traffic Control Quarterly, Vol. 20, No. 1, 2012. + + + + + An introduction to ACAS Xu and the challenges ahead + + GuidoManfredi + + + YannickJestin + + 10.1109/dasc.2016.7778055 + + + 2016 IEEE/AIAA 35th Digital Avionics Systems Conference (DASC) + Sacramento, CA + + IEEE + 2016 + + + Manfredi, G., and Jestin, Y., "An Introduction to ACAS Xu and the Challenges Ahead," IEEE/AIAA 35th Digital Avionics Systems Conference, Sacramento, CA., 2016. + + + + + Next-Generation Airborne Collision Avoidance System + + MJKochenderfer + + + JEHolland + + + JPChryssanthacopoulos + + + + Lincoln Laboratory Journal + + 19 + 1 + 2012 + + + Kochenderfer, M. J., Holland, J. E., and Chryssanthacopoulos, J. P., "Next-Generation Airborne Collision Avoidance System," Lincoln Laboratory Journal, Vol. 19, No. 1, 2012. + + + + + Java Architecture for Detect and Avoid Extensibility and Modeling + + CSantiago + + + EMueller + + + MJohnson + + + MAbramson + + + JWSnow + + ARC-E-DAA-TN25940 + + Sep 29 2015 + + + Tech. Rep. + Santiago, C., Mueller, E., Johnson, M., Abramson, M., and Snow, J. W., "Java Architecture for Detect and Avoid Extensibility and Modeling," Tech. Rep. ARC-E-DAA-TN25940, Sep 29 2015. + + + + + ADS-B Surveillance System Performance With Small UAS at Low Altitudes + + MichaelGuterres + + + StanleyJones + + + GregOrrell + + + RobertStrain + + 10.2514/6.2017-1154 + + + AIAA Information Systems-AIAA Infotech @ Aerospace + Grapevine, Texas + + American Institute of Aeronautics and Astronautics + 2017 + + + Guterres, M., Jones, S., Orrell, G., and Strai, R., "ADS-B Surveillance System Performance With Small UAS at Low Altitudes," AIAA SciTech Forum, Grapevine, Texas, 2017. + + + + + Small Unmanned Aerial System (UAS) Flight Testing of Enabling Vehicle Technologies for the UAS Traffic Management Project + + LJGlaab + + + CVDolph + + + SDYoung + + + NCCoffey + + + RGMcswain + + + MJLogan + + + DEHarper + + 10.3030/763719 + NASA TM-2018-219816 + + April 2018 + Publications Office of the European Union + + + Tech. Rep. + Glaab, L. J., Dolph, C. V., Young, S. D., Coffey, N. C., McSwain, R. G., Logan, M. J., and Harper, D. E., "Small Unmanned Aerial System (UAS) Flight Testing of Enabling Vehicle Technologies for the UAS Traffic Management Project," Tech. Rep. NASA TM-2018-219816, April 2018. + + + + + L-band digital aeronautical communications system (LDACS) activities in SESAR2020 + + AGFrequentis + + + ChristophRihacek + + 10.1109/icnsurv.2018.8384970 + + + + 2018 Integrated Communications, Navigation, Surveillance Conference (ICNS) + + IEEE + 2018. April 2018 + + + cited + "L-band Digital Aeronautical Communications System(LDACS)," http://www.ldacs.com/, 2018. [cited April 2018]. + + + + + + GYang + + + XLin + + + YLi + + + HCui + + + MXu + + + DWu + + + HRyden + + + SRedhwan + + arXiv:1803.11048[cs.NI + A Telecom Perspective on the Internet of Drones: From LTE-Advanced to 5G + + 2018 + + + Yang, G., Lin, X., Li, Y., Cui, H., Xu, M., Wu, D., Ryden, H., and Redhwan, S., "A Telecom Perspective on the Internet of Drones: From LTE-Advanced to 5G," arXiv:1803.11048 [cs.NI], 2018. + + + + + Lift Control on Pitching Wings Experiencing Gusts + 10.2514/6.2022-0332.vid + + + 2018. April 2018 + American Institute of Aeronautics and Astronautics (AIAA) + + + cited + "Continuous gusts," https://en.wikipedia.org/wiki/Continuous_gusts, 2018. [cited April 2018]. + + + + + + diff --git a/file808.txt b/file808.txt new file mode 100644 index 0000000000000000000000000000000000000000..b866030881ef0ff8a5ce59e399a93b92a86dae93 --- /dev/null +++ b/file808.txt @@ -0,0 +1,778 @@ + + + + +I. IntroductionA large amount of small Unmanned Aerial Vehicles (sUAVs) are expected to operate in the uncontrolled Class G airspace, which is at or below 500 feet Above Ground Level (AGL), where many static and dynamic constraints exist, such as ground properties and terrain, restricted areas, various wind conditions, manned aircraft operations, and collision avoidance among sUAVs.How to enable safe, efficient, and massive sUAV operations at the low altitude airspace remains a great challenge.NASA's Unmanned aircraft system Traffic Management (UTM) research initiative 1,2 works on establishing infrastructure and developing policies, requirement, and rules to enable safe and efficient sUAVs' operations.While there are lots of research problems need to be addressed, trajectory modeling should serve as the foundation for understanding sUAV operations and developing requirements and rules for UTM system.For manned and unmanned large fixed-wing aircraft, where precision requirements are typically at a nautical-mile level, trajectory errors are dominated by navigation and sensor errors.Despite different controller designs, aircraft control systems are capable of maintaining trajectory deviation at a level that is much lower than a nautical-mile.Especially for en-route traffic, cross-track trajectory errors caused by winds are usually negligible and modeling aircraft controllers becomes unnecessary.This approach greatly simplifies trajectory models for large size fixed-wing aircraft, 3,4 especially when calculating horizontal trajectories.[10][11] While, for sUAV, navigation errors are mitigated through technologies like Differential Global Positioning System (DGPS), 12,13 trajectory errors caused by wind rise because sUAVs are sensitive to wind due to their small size and low operational altitude.Different sUAV controllers usually produce different trajectory responses under the same wind condition and errors and differences are not negligible at the meter-level precision.Therefore, accurate models of sUAV control systems become as critical as dynamic systems when assessing and managing safe and efficient UTM operations.In the past, plenty of research [14][15][16][17] has been done from a vehicle designer's perspective -modeling sUAV's plant/dynamics through system identification and designing control system based on that.However, because manufacturers are usually not willing to share their control designs as a part of intellectual properties, how to model sUAV trajectory with unknown control systems becomes a challenging research question.On the other hand, considering the rapidly growing small UAV market, the low manufacturing cost, and the short length of manufacturing cycle for small UAVs, it might not be feasible to model a sUAV trajectory following the traditional way in manned aviation, where manufacturers provide aerodynamics based on their numerous wind tunnel and flight tests in design phases and trajectory models are usually accurate enough without knowing aircraft's control systems a .Because of the associated costs, efforts, and time, sUAV manufacturers may not conduct enough wind tunnel and flight tests as their conventional aircraft counterparts.This work proposed to use a neural network approach for modelling small unmanned vehicle's trajectory without knowing its control system and bypassing exhaustive efforts for aerodynamic parameter identification.A neural network based approach is proposed in this work to model sUAV's trajectory, where the entire vehicle model including dynamics and control systems is treated as a black box.As a proof of concept, instead of collecting data from flight tests, this work used the trajectory data generated by a mathematical vehicle model for training and testing.The neural network is first trained to learn the vehicle's responses at multiple conditions with identified features.Once being fully trained, given current vehicle states, winds, and desired future trajectory, the neural network will be used to predict the vehicle's future states at next time step.By repeating this step, a complete 4-D trajectory is then generated as time progresses.In this work, Section II describes the method proposed for modeling vehicle dynamics and controls.Section III presents the experiments conducted for initial testing of the proposed neural network method.Section IV draws the conclusions. +II. MethodThis section presents a generalized small UAV vehicle model and proposes a neural network based approach to model sUAV's trajectory. +A. Small UAV vehicle modelA nonlinear dynamic model for sUAVs, both multi-copters and fixed-wing aircraft, can be derived using Newton-Euler equations if the vehicle is assumed to be rigid and symmetric. 17,18 n the vehicle body axis, the model can be written as force, kinematic, and moment equations shown in Eqn. 1, 2, and 3, respectively.(u, v, w) are body axis velocities; (p, q, r) are body axis angular velocity rates; And (φ, θ, ψ) are roll, pitch, yaw angles in body axis, respectively.(F x , F y , F z ) are body axis aerodynamic forces and (M φ , M θ , M ψ ) are body axis moments.J x , J y , and J z are the moments of inertia about the principle axes in the body frame.Then the vehicle location in the Earth frame can be expressed as Eqn. 4, where p n , p e , and h are the north position, east position, and altitude, respectively, and s and c denote sin and cos for simplicity.Although the overall model can be generalized, models of aerodynamic forces (F x , F y , and F z ) and moments (M φ , M θ , and M ψ ) can be quite different because of differences in vehicle types.   u v ẇ   =    rv -qw pw -ru qu -pv    +    -gsinθ gcosθsinφ gcosθcosφ    + 1 m    F x F y F z   (1)   φ θ ψ   =    1 sinφtanθ cosφtanθ 0 cosφ -sinφ 0 sinφ cosθ cosφ cosθ       p q r    (2)    ṗ q ṙ   =    Jy-Jz Jx qr Jz-Jx Jy pr Jx-Jy Jz pq    +    1 Jx M φ 1 Jy M θ 1 Jz M ψ   (3)a Difficulty arises when modeling departure and arrival trajectories because of unknown FMS settings.   ṗn ṗe ḣ    =    cθcψ sφsθcψ -cφsψ cφsθcψ + sφsψ cθsψ sφsθsψ + cφcψ cφsθsψ -sφcψ sθ -sφcθ -cφcθ       u v w   (4) +B. Neural network approach3][24] Considering the challenge associated with collecting vehicle states, this work proposed to apply neural networks to model an entire vehicle including the plant and control system (shown in Fig. 1) to reduce the number of vehicle states required to be collected.In this approach, at each time step, the sUAV's current states( x 0 , ˙ x 0 ), next flight states( x, ˙ x), and wind conditions ( w x , w y ) are used as training input and output pairs for the neural network.For a given vehicle, states are collected at various flight conditions.Once the training is finished, this neural network will be applied to predict future flight states at next time step based on current flight states.Therefore, one neural network should be capable of representing one vehicle model for predicting trajectories.Because the model is used for trajectory prediction, inputs and outputs are limited to positions and speeds.Nine inputs are selected for trajectory calculation: position deviations, speed deviations, and speeds.Six defined outputs are position change rates and speed change rates.This setting leads to a multi-layered neural network composed of nine input neurons, 20 hidden neurons, and six output neurons as shown in Fig. 2.The neural network is trained using the back-propagation approach with learning rate adaptation.The learning rate update rule followed Jacobs' hybrid algorithm, 25 in which a momentum 26 was added into the Delta-Bar-Delta rule.The weight w ij between the ith and jth neurons is updated by Eqn. 5, where E(k) is the mean square sum of the differences between the outputs and the desired target values and α = 0.75.∆w ij (k + 1) = α∆w ij (k) -η k+1 ∂E(k) ∂w ij (k)(5)And the learning rate η k+1 is updated by:η k+1 = η k + ∆η k(6)Where the ∆η k is given by: The δ and δ are govened by:∆η k =      κ, if δk-1 •δ k > 0 -φ•η k if δk-1 •δ k < 0 0 otherwise(7)δ k = ∂E(k) ∂w ij (k) (8)and δk = (1 -θ)δ k + θ δk-1(9)where κ = 0.01, η 0 = 0.75, φ = 0.2, and θ = 0.7. +III. ExperimentsBecause the scope of this paper is limited to concept proof, trajectory data are generated by a mathematical vehicle model instead of being collected from actual flight tests.The generated trajectories are used as both training and testing data for the neural network.Experiments are set up to examine if a neural network can be trained to fit the training trajectory data and to check if the trained neural network can calculated testing trajectories in a sequential manner without diverging away from the truth. +A. Trajectory data generationIn order to generate trajectory data for training and testing, a small UAV dynamic model and a controller model are needed.Without loss of generality, a quadrotor model and a PD controller are selected for generating trajectory data sets.By neglecting Coriolis terms and small angle approximations, a quadrotor dynamics 18,27,28 can be expressed as Eqn. 10, where w n and w e are north and east components of the wind vector and vetical wind component is negelected here.The v n and v e are vehicle velocities in the Earth frame.The force and moments generated by motors can be simplified to Eqn. 11, where k f and k m are the aerodynamic force and moment coefficients for motors.L is the arm length and Ω i is angular velocity of rotor i.              ṗn ṗe vn ve ḧ φ θ ψ               =                v n + w n v e + w e -(cφsθcψ + sφsψ)•F z /m (-cφsθsψ + sφcψ)•F z /m g -cφcθ • F z /m 1 Jx M φ 1 Jy M θ 1 Jz M ψ                (10)      F z M φ M θ M ψ      =      k f (Ω 2 1 + Ω 2 2 + Ω 2 3 + Ω 2 4 ) (-k f Ω 2 2 + k f Ω 2 4 )•L (k f Ω 2 1 -k f Ω 2 3 ))•L (k m Ω 2 1 -k m Ω 2 2 + k m Ω 2 3 -k m Ω 2 4 ))•L     (11)The PD position controller is shown in Eqn. 12 and Eqn.13, where (x d , y d ) and ( ẋd , ẏd ) are desired positions and velocities and k p and k d are associated gains.Eqn. 12 computes desired accelarations, and the desired roll and pitch angles (φ d and θ d ) are then calculated using Eqn.13.Eqn. 14 shows the control law for both roll and pitch angles, where k p,φ , k d,φ , k p,θ , and k d,θ are gains.ẍd ÿd = k p (x d -x) + k d ( ẋd -ẋ) k p (y d -y) + k d ( ẏd -ẏ) (12)φ d θ d = m F z -sinψ -cosψ cosψ -sinψ -1 ẍd ÿd (13)M φ M θ = k p,φ (φ d -φ) + k d,φ ( φd -φ) k p,θ (θ d -θ) + k d,θ ( θd -θ) L (14)This mathematical model was applied to generate trajectory data for training.As an initial effort, the vehicle's horizontal trajectory responses to side wind gust were generated: The vehicle is required to follow a straight line at a constant altitude with various ground speeds and it will enter a constant north wind field, where the wind direction is perpendicular to the flying direction.A total of 35 trajectories were generated by changing vehicle's ground speed and wind speed as shown in Table 1 and flight states were recorded every 0.01s.The current flight states and next flight states at each time step are treated as an input and output pair, respectively.There are a total of 62,850 pairs of training data from these trajectories.In the training process, the training and testing data were applied without specfic ordering and the objective was to reduce overall MSE even though errors at certain data points may not be small.The neural network was trained in 20,000 epochs and Fig. 3 shows the learning evolution for the first 5,000 epochs, where the MSE of the normalized outputs was reduced below 0.005.The MSE was ended at 0.0032 after 20,000 epochs.Fig. 4 presents the histogram of x position training errors.These errors are between unnormalized target outputs and the values generated by the trained neural network.The figure shows that most of the x position training errors are within ±0.2 meters.Fig. 5 presents error histograms of four outputs: x position, y position, forward speed (at x axis), and lateral speed (at y axis).It is noted that training errors of velocities are much higher compared with position errors; sometime they even exceed the ±2 meter per second range. +C. Trajectory predictionIn the process of predicting trajectories, the calculations are performed sequentially starting from the initial flight state.In the prediction experiment presented in this section, the vehicle's ground speed was set to 5m/s and the wind field was assumed to affect the area where x >= 25m.The wind came from the North and the vehicle flied from the West to the East.The wind speed was set to 5m/s.The experiment set-up is similar to the training cases although the values are not the same.Figure 6(a) depicts the comparison between the true and predicted trajectories in the X-Y space.It is noted that predictions are pretty accurate in the spatial dimension.As shown in Fig. 6(b), the lateral errors are less than 0.25 meters over the entire trajectory.It is possible that the trajectory prediction may not converge to the truth as time proceeds because the error may accumulate and cause the prediction to diverge away eventually.The results from Fig. 6(a) also demonstrated that trajectory prediction errors didn't diverge although the training errors (especially for velocities) presented in previous section were not small.However, the condition of stability remains unclear, and needs to be studied in future work.Both spatial and temporal precisions are required for an accurate 4D trajectory prediction, Fig. 7(a) shows the comparison of x position's time histories between the predicted and true trajectories.The x position errors were shown in Fig. 7(b)) and were bounded between 0 and 1.4 meters.This reveals that the predicted trajectory from the neural network is a bit fast and ahead of the true trajectory.Fig. 8(a) shows the comparison of forward speed's time histories between the predicted and true trajectories.The errors were shown in Fig. 8(b)) and were bounded between -0.7 and 0.6 meter per second.The comparison shows that, despite of the similar trend, the temporal difference between forward speeds is relatively high.This neural network was then applied to predict trajectory in various scenarios, where the cross wind is in the range of [0.7, 9.2] and the vehicle desired ground speed is within [2.0 -11.0].Table 2 lists the parameters for 12 cases used for verification and Fig. 9 plots the mean square errors and the maximimum absolute errors (MAE) between the predicted and actual trajectories.As discussed in previous sections, errors associated with the'X-Y' are spatial errors and errors associated with the 'T-X' and 'T-Y' are measured in temporal dimension.Spatial errors (both MSEs and MAEs) are usually smaller than the other errors and most of them were bounded by ±0.5 meters, which means the predicted trajectories are aligned with actual trajectories well if the time was not taken into account.Temporal errors are usually a bit high, especially when the cross wind is strong.Cross track errors are typically less than along-track errors and, not surprisingly, all MAEs are higher than corresponding MSEs.From the figure, it is noted that all errors are well kept below 2 meters, which shows that the neural network approach has great promise in modeling sUAV's trajectories. +IV. ConclusionThis work proposed a neural network approach for modeling sUAV's dynamics and control system for trajectory prediction.The motivation was to model sUAV's trajectory in a practical way such that the needs of intensive wind tunnel tests and flight tests are minimized while the fidelity of trajectory is not reduced much.If successful, this approach will also greatly reduce computational time when calculating trajectories.A neural network composed of nine inputs, 20 hidden neurons, and six outputs was proposed and trained using the back-propagation algorithm with the hybrid delta-bar-delta learning rule.As an initial work, a quadrotor vehicle model associated with a PD controller was used to generate trajectory data for training and testing purposes.Various trajectories were used to train the neural network with different vehicle ground speeds and side wind conditions.After the neural net is fully trained, it is used to predict future flight states at next time step based on current flight state information.This proposed approach was then verified with different but similar set of scenarios.The results show that the neural network can be trained to fit the train data.It can be used in a sequential manner to predict the trajectory step by step without diverging away from the truth.In addition, the 4-D trajectory prediction errors are less than 2.0 meters while the spatial errors are even less.Overall, the proposed approach shows great promise in modeling sUAV trajectories.In the future work, further research/experiments are needed to improve this method.First, training data need to be much selective and representative to increase the efficiency and effectiveness of training.Second, thorough experiments are needed to prove its application with wide-range flight conditions and various vehicle types.Third, the conditions of stable trajectory prediction need to be studied and understood.Finally, different machine learning methods can be explored to identify good machine learning and training algorithms.Figure 1 .1Figure 1.Trajectory modeling using Neural Networks +Figure 2 .2Figure 2. Neural network structure: (9 input units, 20 hidden units, and 6 output units) +Table 1 .Figure 3 .13Figure 3. Mean square error in the training phase +Figure 4 .4Figure 4. Histogram of errors in x position +Figure 5 .5Figure 5. Histogram of errors in x position, y position, forward speed, and lateral speed +Figure 8 .8Figure 8. Temporal comparison of forward speeds (a) predicted and true trajectories (b) prediction errors +Figure 9 .9Figure 9. Errors between the predicted trajectory and the truth in various cases + of 10 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on June 9, 2017 | http://arc.aiaa.org| DOI: 10.2514/6.2017-3072 + of 10 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on June 9, 2017 | http://arc.aiaa.org| DOI: 10.2514/6.2017-3072 + American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on June 9, 2017 | http://arc.aiaa.org| DOI: 10.2514/6.2017-3072 + Downloaded by NASA AMES RESEARCH CENTER on June 9, 2017 | http://arc.aiaa.org| DOI: 10.2514/6.2017-3072 + + + + + + + + + + + Unmanned Aircraft System Traffic Management (UTM) Concept of Operations + + PKopardekar + + + JRios + + + TPrevot + + + MJohnson + + + JJung + + + JERobinson + + + + 16th AIAA Aviation Technology, Integration, and Operations Conference + Washington, D.C. + + June 2016 + + + + Kopardekar, P., Rios, J., Prevot, T., Johnson, M., Jung, J., and Robinson, J. E., "Unmanned Aircraft System Traf- fic Management (UTM) Concept of Operations," 16th AIAA Aviation Technology, Integration, and Operations Conference, Washington, D.C., 13-17 June 2016. + + + + + Unmanned Aerial System Traffic Management (UTM): Enableing Low-altitude Airspace and UAS Operations + + PKopardekar + + NASA TM-2014-218299 + + 2014 + + + Tech. Rep. + Kopardekar, P., "Unmanned Aerial System Traffic Management (UTM): Enableing Low-altitude Airspace and UAS Operations," Tech. Rep. NASA TM-2014-218299, 2014. + + + + + Time Controlled Descent Guidance Algorithm for Simulation of Advanced ATC Systems + + HLee + + + HErzberger + + + + NASA Ames Research Center + + August 1983 + Moffett Field, CA + + + Tech. Rep. Tech. Rep. NASA Technical Memorandum 84373 + Lee, H. and Erzberger, H., "Time Controlled Descent Guidance Algorithm for Simulation of Advanced ATC Systems," Tech. Rep. Tech. Rep. NASA Technical Memorandum 84373, NASA Ames Research Center, Moffett Field, CA, August 1983. + + + + + Ground based concept for time control of aircraft entering the terminal area + + HErzberger + + + JChapel + + 10.2514/6.1985-1888 + + + Guidance, Navigation and Control Conference + Snowmass, CO + + American Institute of Aeronautics and Astronautics + August 1985 + + + + Erzberger, H. and Chapel, J., "Ground Based Concept for Time Control of Aircraft Entering the Terminal Area," AIAA Guidance, Navigance, and Control Conference, Snowmass, CO, 19-21 August 1985. + + + + + Trajectory Synthesis for Air Traffic Automation + + RhondaSlattery + + + YiyuanZhao + + 10.2514/2.4056 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 20 + 2 + + 1997 + American Institute of Aeronautics and Astronautics (AIAA) + + + Slattery, R. and Zhao, Y., "Trajectory Synthesis for Air Traffic Automation," Journal of Guidance, Control, and Dy- namics, Vol. 20, No. 2, 1997. + + + + + Design of Center-TRACON Automation System + + HErzberger + + + TJDavis + + + SMGreen + + + + AGARD Meeting on Machine Intelligence in Air Traffic Management + Berlin, Germany + + May 1993 + + + + Erzberger, H., Davis, T. J., and Green, S. M., "Design of Center-TRACON Automation System," AGARD Meeting on Machine Intelligence in Air Traffic Management, Berlin, Germany, 11-14 May 1993. + + + + + Conflict Detection and Resolution In the Presence of Prediction Error + + HErzberger + + + RPaielli + + + DIsaacson + + + MMEshow + + + + Europe Air Traffic Management Research and Development Seminar + + + June 1997 + Saclay, France + + + 1st USA/ + Erzberger, H., Paielli, R., Isaacson, D., and Eshow, M. M., "Conflict Detection and Resolution In the Presence of Prediction Error," 1st USA/Europe Air Traffic Management Research and Development Seminar , Saclay, France, 17-20 June 1997. + + + + + FACET: Future ATM Concepts Evaluation Tool + + KarlDBilimoria + + + BanavarSridhar + + + ShonRGrabbe + + + GanoBChatterji + + + KapilSSheth + + 10.2514/atcq.9.1.1 + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 9 + 1 + + + American Institute of Aeronautics and Astronautics (AIAA) + + + Bilimoria, K., Sridhar, B., Chatterji, G., Sheth, K., and Grabbe, S., "FACET: Future ATM Concepts Evaluation Tool," + + + + + + + Usa/ + + + + Europe Air Traffic Management R&D Seminar + + + June 2000 + Napoli, Italy + + + USA/Europe Air Traffic Management R&D Seminar , Napoli, Italy, 13-16 June 2000. + + + + + Fast-Time NAS Simulation System for Analysis of Advanced ATM Concepts + + DouglasSweet + + + VikramManikonda + + + JesseAronson + + + KarlinRoth + + + MatthewBlake + + 10.2514/6.2002-4593 + + + AIAA Modeling and Simulation Technologies Conference and Exhibit + Monterey, California + + American Institute of Aeronautics and Astronautics + August 2002 + + + + Sweet, D. N., Manikonda, V., Aronson, J. S., Roth, K., and Blake, M., "Fast-time Simulation System For Analysis of Advanced Air Transportation Concepts," AIAA Modeling and Simulation Technologies Conference and Exhibit, Monterey, California, 5-8 August 2002. + + + + + A review of conflict detection and resolution modeling methods + + JKKuchar + + + LCYang + + 10.1109/6979.898217 + + + IEEE Transactions on Intelligent Transportation Systems + IEEE Trans. Intell. Transport. Syst. + 1524-9050 + + 1 + 4 + + 2000 + Institute of Electrical and Electronics Engineers (IEEE) + + + Kuchar, J. K. and Yang, L. C., "A Review of Conflict Detection and Resolution Modeling Methods," IEEE Transactions On Intelligent Transportation Systems, Vol. 1, No. 4, 2000. + + + + + Protocol-based Conflict Resolution for Finite Information Horizon + + IHwang + + + CTomlin + + + + Proceedings of the American Control Conference + the American Control ConferenceAnchorage, AK + + May 2002 + + + + Hwang, I. and Tomlin, C., "Protocol-based Conflict Resolution for Finite Information Horizon," Proceedings of the American Control Conference, Anchorage, AK, 8-10 May 2002. + + + + + Real-time differential carrier phase GPS-aided INS + + JAFarrell + + + TDGivargis + + + MJBarth + + 10.1109/87.852915 + + + IEEE Transactions on Control Systems Technology + IEEE Trans. Contr. Syst. Technol. + 1063-6536 + + 8 + 4 + + 2000 + Institute of Electrical and Electronics Engineers (IEEE) + + + Farrell, J. A., Givargis, T. D., and Barth, M. J., "Real-Time Differential Carrier Phase GPS-Aided INS," IEEE Trans- actions On Control Systems Technology, Vol. 8, No. 4, 2000. + + + + + Development of a real-time DGPS system in the centimeter range + + HBlomenhofer + + + GWHein + + + ETaveira Blomenhofer + + + WWerner + + 10.1109/plans.1994.303359 + + + Proceedings of 1994 IEEE Position, Location and Navigation Symposium - PLANS'94 + 1994 IEEE Position, Location and Navigation Symposium - PLANS'94Las Vegas, NV + + IEEE + April 1994 + + + + Blomenhofer, H., Hein, G. W., Blomenhofer, E. T., and Werner, W., "Development of a Read-Time DGPS System in the Centimeter Range," IEEE Position Location and Navigation Symposium, Las Vegas, NV, 11-15 April 1994, pp. 532-539. + + + + + Control System Design for Rotorcraft-based Unmanned Aerial Vehicles using Time-domain System Identification + + DHShim + + + HJKim + + + SSastry + + + + Proceedings of the 2000 IEEE International Conference on Control Applications + the 2000 IEEE International Conference on Control ApplicationsAnchorage, Alaska + + September 2000 + + + + Shim, D. H., Kim, H. J., and Sastry, S., "Control System Design for Rotorcraft-based Unmanned Aerial Vehicles us- ing Time-domain System Identification," Proceedings of the 2000 IEEE International Conference on Control Applications, Anchorage, Alaska, 25-27 September 2000. + + + + + System Identification of a Rotary-Wing Micro Air Vehicle + + JosephKConroy + + + JSeanHumbert + + + DarryllJPines + + 10.4050/jahs.56.025001 + + + Journal of the American Helicopter Society + j am helicopter soc + 2161-6027 + + 56 + 2 + + 2011 + American Helicopter Society + + + Conroy, J. K., Humbert, J. S., and Pines, D. J., "System Identification of a Rotory-Wing Micro Air Vehicle," Journal of the American Helicopter Society, Vol. 56, No. 2, 2011. + + + + + Fuzzy adaptive control for a class of nonlinear systems with unknown control gain + + SalimLabiod + + + ThierryMarieGuerra + + 10.1007/s12530-011-9037-y + + + Evolving Systems + Evolving Systems + 1868-6478 + 1868-6486 + + 3 + 1 + + 2012 + Springer Science and Business Media LLC + + + Labiod, S. and Guerra, T. M., "Fuzzy Adaptive Control For A Class Of Nonlinear Systems with Unknown Control Gain," Evolving Systems, Vol. 3, 2012. + + + + + System Identification for Small, Low-Cost, Fixed-Wing Unmanned Aircraft + + AndreiDorobantu + + + AustinMurch + + + BéréniceMettler + + + GaryBalas + + 10.2514/1.c032065 + + + Journal of Aircraft + Journal of Aircraft + 0021-8669 + 1533-3868 + + 50 + 4 + + 2013 + American Institute of Aeronautics and Astronautics (AIAA) + + + Dorobantu, A., Murch, A., Mettler, B., and Balas, G., "System Identification for Small, Low-Cost, Fixed-Wing Unmanned Aircraft," AIAA Journal of Aircraft, Vol. 50, No. 4, 2013. + + + + + Quadrotor Dynamics and Control + + RWBeard + + + October 2008 + + + Tech. rep. + Beard, R. W., "Quadrotor Dynamics and Control," Tech. rep., October 2008. + + + + + Multilayer feedforward networks are universal approximators + + KurtHornik + + + MaxwellStinchcombe + + + HalbertWhite + + 10.1016/0893-6080(89)90020-8 + + + Neural Networks + Neural Networks + 0893-6080 + + 2 + 5 + + 1989 + Elsevier BV + + + Hornik, K., Stinchcombe, M., and White, H., "Multilayer Feedforward Networks are Universal Approximators," Neural Networks, Vol. 2, 1989. + + + + + Approximation capabilities of multilayer feedforward networks + + KurtHornik + + 10.1016/0893-6080(91)90009-t + + + Neural Networks + Neural Networks + 0893-6080 + + 4 + 2 + + 1991 + Elsevier BV + + + Hornik, K., "Approximation Capabilities of Multilayer Feedforward Networks," Neural Networks, Vol. 4, No. 2, 1991. + + + + + Approximation by superpositions of a sigmoidal function + + GCybenko + + 10.1007/bf02551274 + + + Mathematics of Control, Signals, and Systems + Math. Control Signal Systems + 0932-4194 + 1435-568X + + 2 + 4 + + 1989 + Springer Science and Business Media LLC + + + Cybenko, G., "Approximation by Superpositions of a Sigmoidal Function," Mathematics of Control, Signals, and Systems, Vol. 2, No. 4, 1989. + + + + + Nonlinear system identification using additive dynamic neural networks-two on-line approaches + + RGrino + + + GCembrano + + + CTorras + + 10.1109/81.828569 + + + IEEE Transactions on Circuits and Systems I: Fundamental Theory and Applications + IEEE Trans. Circuits Syst. I + 1057-7122 + + 47 + 2 + + 2000 + Institute of Electrical and Electronics Engineers (IEEE) + + + Grino, R., Cembrano, G., and Torras, C., "Nonlinear System Identification Using Additive Dynamic Neural Networks -Two Online Approaches," IEEE Transactions On Circuits And Systems -Part I: Fundamental Theory And Applications, Vol. 47, No. 2, 2000. + + + + + Non-linear system identification using neural networks + + SChen + + + SABillings + + + PMGrant + + 10.1080/00207179008934126 + + + International Journal of Control + International Journal of Control + 0020-7179 + 1366-5820 + + 51 + 6 + + 1990 + Informa UK Limited + + + Chen, S., Billings, S. A., and Grant, P. M., "Non-linear System Identification Using Neural Networks," International Journal of Control, Vol. 51, No. 6, 1990. + + + + + Parameter Estimation in Nonlinear Systems Using Hopfield Neural Networks + + ZhenningHu + + + SivasubramanyaNBalakrishnan + + 10.2514/1.3210 + + + Journal of Aircraft + Journal of Aircraft + 0021-8669 + 1533-3868 + + 42 + 1 + + 2005 + American Institute of Aeronautics and Astronautics (AIAA) + + + Hu, Z. and Balakrishnan, S. N., "Parameter Estimation in Nonlinear Systems Using Hofield Neural Networks," AIAA Journal of Aircraft, Vol. 42, No. 1, 2005. + + + + + Increased rates of convergence through learning rate adaptation + + RobertAJacobs + + 10.1016/0893-6080(88)90003-2 + + + Neural Networks + Neural Networks + 0893-6080 + + 1 + 4 + + 1988 + Elsevier BV + + + Jacobs, R. A., "Increased Rates of Convergence Through Learning Rate Adaptation," Neural Networks, Vol. 1, 1988. + + + + + (1986) D. E. Rumelhart, G. E. Hinton, and R. J. Williams, "Learning internal representations by error propogation," [i]Parallel Distributed Processing: Explorations in the Microstructures of Cognition[/i], Vol. I, D. E. Rumelhart and J. L. McClelland (Eds.) Cambridge, MA: MIT Press, pp. 318-362.(1986) David E. Rumelhart, Geoffrey E. Hinton, and Ronald J. Williams, "Learning representations by back-propogating errors," [i]Nature[/i] 323:533-536. + + DERumelhart + + + JLMcclelland + + 10.7551/mitpress/4943.003.0042 + + + Neurocomputing, Volume 1 + Cambridge MA + + The MIT Press + 1986 + + + + Rumelhart, D. E. and McClelland, J. L., "Learning Internal Representations by Error Propagation," Parallel Distributed Processing: Explorations in the Microstructure of Cognition, MIT Press, Cambridge MA, 1986. + + + + + Full control of a quadrotor + + SamirBouabdallah + + + RolandSiegwart + + 10.1109/iros.2007.4399042 + + + 2007 IEEE/RSJ International Conference on Intelligent Robots and Systems + San Diego, CA + + IEEE + Oct 29 -Nov 2 2007 + + + Bouabdallah, S. and Siegwart, R., "Full Control of a Quadrotor," Proceedings of the 2007 IEEE/RSJ International Conference on Intelligent Robots and Systems, San Diego, CA, Oct 29 -Nov 2 2007. + + + + + Control of a quadrotor helicopter using visual feedback + + EAltug + + + JPOstrowski + + + RMahony + + 10.1109/robot.2002.1013341 + + + Proceedings 2002 IEEE International Conference on Robotics and Automation (Cat. No.02CH37292) + 2002 IEEE International Conference on Robotics and Automation (Cat. No.02CH37292)Washington, D.C. + + IEEE + May 2002 + + + Altug, E., Ostrowski, J. P., and Mahony, R., "Control of a Quadrotor Helicopter Using Visual Feedback," Proceedings of the 2002 IEEE International Conference on Robotics & Automation, Washington, D.C., May 2002. + + + + + + diff --git a/file809.txt b/file809.txt new file mode 100644 index 0000000000000000000000000000000000000000..d7f3511c456f384288755054cc6e378d15b9ee95 --- /dev/null +++ b/file809.txt @@ -0,0 +1,906 @@ + + + + +I. IntroductionAs the ground transportation becomes increasingly congested, the Urban Air Mobility (UAM) concept [1,2] has attracted a great deal of attention for its potential to significantly improve the movement of people and goods.Besides the challenges of building and certifying safe and quiet electric powered vertical takeoff and landing aircraft (eVTOL), developing a safe, efficient, and scalable traffic management system faces great challenges as well.The envisioned high density of traffic operations of diverse aircraft types in complex urban environments mandates a paradigm shift from the conventional human-centric air traffic management.Similarities between UAM and small unmanned aerial vehicle (UAV) operations have led to efforts to adapt the federated autonomous Unmanned aircraft system Traffic Management (UTM) system [3][4][5][6] developed by NASA, FAA, and industry partners to become a foundation for the UAM traffic system.However, there are still many crucial questions needed to be addressed to enable UAM operations, such as how to interact with conventional manned aviation, if/when/where route structures should be placed, and should core functions like conflict resolution be centralized or decentralized?As the core function in airspace management system, the architecture of conflict resolution is more critical for the early stage developement of other systems: for instance, if a decentralized architecture is favored, then the conflict resolution function needs to reside in the UAM vehicle system and a close-range vehicle-to-vehicle(V2V) communication capability becomes paramount.Whereas if a centralized architecture is decided, a reliable and high-bandwidth air-ground communcation becomes more important than the V2V communcation with easier requirements on the vehicle system.To support the development of the UAM concept, this work analyzes and evaluates various architecture options for conflict resolution strategy and service, which is one of the core capabilities to enable UAM operations.To help understand the analysis in this study, the conflict resolution architectures are categorized into four types.The first one is the Centralized architecture, which refers to a conflict resolution architecture where a single centralized and ground-based conflict resolution service sends trajectory amendments to aerial vehicles to resolve projected conflicts with other aircraft.This is analogous to the role of human air traffic controllers in current air traffic management.The second is called Uniform rule-based architecture, which refers to a decentralized conflict resolution architecture where each individual aerial vehicle follows a uniform set of predefined rules or protocols when resolving conflicts."Autonomous Flight Rules" [7], protocol-based conflict resolution [8], distributed conflict resolution with coordinated strategy [9], and agent-based cooperation [10] are all examples of this category.The third group is the Coordination-based architecture, which is similar to the Uniform rule-based strategy except that it allows aircraft to coordinate with each other via a predefined, real-time negotiation mechanism.The difference from the Uniform rule based system is that, for a given conflict situation, resolutions from the Coordination-based system may differ due to the dynamic nature of real-time coordination, whereas the resolutions from the Uniform rule-based system are always the same.The collaborative trajectory options program (CTOP) [11] may be the closest example in the current manned aviation system.The fourth group is called the Mixed rule-based architecture, representing a completely decentralized mechanism with separated objectives or rules for different sets of aircraft.Aircraft-based conflict resolutions in which each aircraft has its own resolution objective [12] is one extreme example.The last three categories, Uniform rule-based, Coordination-based, and Mixed rule-based, are all decentralized; they differ in their levels of decentralization.There have been many studies in comparing centralized and decentralized systems, although most are from outside the aviation domain.For instance, based on mathematical models of incentives, bias, coordination, and communication, Alonso et.al [13] showed that in multidivisional organizations decentralization can outperform centralization even when coordination is extremely important.Using simulations based on a simple sorting problem, Veetil [14] showed that a decentralized system was more robust than a centralized system in the presence of communication errors, whereas the centralized system required less coordination.Although there are many studies proposing different architectures for conflict resolution [7][8][9][10]12], only a few studies quantitatively compared centralized and decentralized systems in the air traffic management domain.Bicchi et.al. [15] showed that a decentralized conflict resolution is more robust to conflict resolution failures from the perspective of trajectory optimization.Conversely, Krozel et.al. [16] showed that centralized separation outperforms decentralized separation in terms of system stability and efficiency metrics proposed by the authors based on the assumption of perfect information exchange and the presence of no errors or delays.This study makes no such assumption, and further extends the study of conflict resolution architectures to small UAM vehicle types.The contribution of this paper is to quantify safety and efficiency performance differences between centralized and decentralized architectures for conflict resolution, and to assess their robustness to input and output errors.To gain better understanding of the differences among these architectures through quantitative analysis, this work investigates three of the above four architectures: Centralized, Uniform rule-based, and Mixed rule-based.The high fidelity Fe 3 [17] simulator, which has the capability of modeling various sources of uncertainty is used to run the experiments.Section II introduces the models and parameters for these three architectures.Section III presents the experiments setup.Section IV analyzes and compares three conflict resolution architectures in terms of system safety, efficiency, robustness, and scalability.Section V presents the findings of this study. +II. Conflict Resolution Architecture ModelingIn general, conflict resolution includes four phases: Strategic deconfliction assigns delays to resolve potential conflicts typically before departure; Separation assurance modifies trajectories after departure to several minutes before potential losses of separation could happen; Detect and avoid initiates tactical maneuvers several minutes before the loss of separation or loss of well-clear; and Collision avoidance activates final maneuvers to avoid near-mid-air-collision(NMAC) once the loss of well-clear/loss of separation happens.Conflict resolution in this study refers to all three airborne phases including Separation assurance, Detect and avoid, and Collision avoidance.This section will describe the models for the three architectures introduced previously: Centralized, Uniform rule-based, and Mixed rule-based. +A. Centralized ArchitectureCentralized architecture is widely used in the national airspace system (NAS) operations, especially at the sector level, where typically one (a radar-side controller) or two (a radar-side controller and a data-side controllers) ground-based human controllers provide conflict resolution services [18].Fig. 1(a) shows a notional picture of a centralized UAM conflict resolution architecture, where the ground-based conflict resolution service collects vehicle information, resolves potential conflicts, and sends trajectory amendments to each individual UAM vehicle.Due to the scalability limitations associated with human-centric conflict resolution, it is envisioned that human controllers may not be responsible for conflict resolution in UAM operations, as in the UTM concept [3,4].Fig. 1(b) shows the information flows in such a centralized architecture.In this figure, the aircraft state information ì x is provided by either surveillance systems or operator (through air-to-ground or ground-to-ground communications) and fed into the centralized conflict resolution service.There will be some errors for the information fed into the conflict resolution service because of imperfect detection, navigation, and communication.Although many tracking and filtering algorithms [19][20][21] have been developed to minimize these errors, they cannot eliminate errors.If the final aircraft state information (such as aircraft position, velocity, intent, etc.) received by the conflict resolution is referred to as ì x representing the approximated aircraft states, then the difference between ìx and ì x is represented as ì x (shown in Eqn. 1 for aircraft i at time t).ì x i (t) = ì x i (t) + ì x i (t)(1)After computing, the centralized service needs to send resolutions, such as waypoints, updated trajectory, heading changes, and speed changes, to remote individual aircraft or operators, typically from a ground-based station.Again due to communication errors, there will be a difference between the true resolutions ì y transmitted by the centralized service and those received by aircraft or operators (represented by ì y).The communication may also be degraded by packet loss.Let the reception probability P{ ì y(t)| ì y(t)} represents the probability of receiving ì y when a resolution ì y is transmitted by the centralized service at time t.Eqn. 2 shows the final conflict resolution instruction ì y i at time t.ì y i (t) = [ ì y i (t) + ì y i (t)] • P{ ì y i (t)| ì y i (t)}(2)To capture the difference in the objectives between centralization and decentralization, it is assumed in this study that the objectives are the same for both centralized and decentralized systems, except that in the centralized service it first prioritizes conflicts based on their estimated time until conflicts and then resolves these conflicts in sequence according to their priorities.For instance, if there are two conflicts, the conflict is projected to occur first will be resolved first before the conflict that is projected to occur second.The coverage for both centralized and decentralized architectures is assumed to be the same long range, which is not the focus of this work. +B. Decentralized with uniform rulesA decentralized system is usually envisioned to provide flexibility, data privacy, and robustness, where each division has the flexibility to optimize its objective function based on its own needs while only sharing necessary data across different divisions.Fig. 2(a) presents a notional picture of a decentralized system with uniform rules and R is the detection/communication range of a given aircraft.Each individual aircraft is required to follow the uniform rules/protocols and determine conflict resolution maneuvers by itself based on predefined rules/protocols when a loss of separation is about to happen.With decentralization, the input flow is the only data flow, either coming from vehicle-to-vehicle communications or onboard sensors as shown in Fig. 2(b).Similar to the input flow in centralization, because of errors in communication, sensor, and navigation, there will exist the error ìx between the approximated states ìx and the true states ì x.As the aircraft resolve conflicts onboard and all the aircraft already have the pre-defined resolution strategies, there is no need to transmit resolutions when using decentralized strategies. +C. Decentralized with mixed rulesAnother decentralized architecture allows multiple different sets of conflict resolution rules operate in one area.Fig. 3(a) shows a notional picture, where the blue aircraft follow conflict resolution rule set A and the green aircraft follow conflict resolution rule set B. An analogy in the UTM concept would be multiple service providers (A, B, ...) providing conflict resolution services to different groups of operators.Although the aircraft abide by one of the two different sets of rules, they are still assumed to share their information with all proximate aircraft regardless of the group to which they belong (as shown in Fig. 3(b)).The data flow and objectives of conflict resolutions are the same as in the decentralized system with uniform rules. +D. Conflict resolution algorithmsThere is a large body of research on en-route conflict resolution algorithms [22].Representative algorithms include the horizontal vector turn method [23,24], the potential field method [25], the rule-based methods [26,27], and the partially observable Markov decision process method [28], and these algorithms cover various time horizons.The scope of this work is to study conflict resolution architectures instead of designing or evaluating a specific conflict resolution algorithm.Therefore, without loss of generality, rule-based algorithms are applied here.In addition to basic conflict resolution parameters, like look-ahead time, conflict detection threshold, etc., a rule-based algorithm also provides a set of flight rules to define the maneuver options for the responsible party under given conditions.If there exist multiple options, a simple objective (e.g., minimum deviation) may be used to decide the final resolution maneuvers.Table 1 shows an example of conflict resolution rules defined in a rule-based algorithm.These rules define the responsibility and specify a range of resolution maneuvers under various conditions.With these rules, a final resolution maneuver is then identified with minimum deviation.In this table, θ RP A represents the relative position orientation of the intruder with respect to the ownship, θ RH A refers to the relative heading angles between two aircraft with the ownship heading as the reference, and φ denotes the action options, which are the allowed heading angle changes of the aircraft.In this paper, two different conflict resolution algorithms -each with their own sets of rules -were implemented as decentralized conflict resolution algorithms.The centralized algorithm was similar to the decentralized algorithm, except that the centralized algorithm prioritized the conflicts first and resolved conflicts sequentially. +III. Experiment Set-upTo conduct generalized comparisons, random and generic scenarios with increasing number of operations were generated.Since UAM vehicles are mostly still under development, vehicle models and separation minimums are not well-defined.In this study, the target ground speeds of flights ranged from 15 to 30 knots and minimum separation was assumed to be 50 feet.The airspace in test scenarios was a simple two-dimensional region (1.5 nmi by 1.5 nmi) designed to be traversed in approximately three minutes of flight time.Although the flight speeds in the experiments were lower than expected for UAM vehicles (e.g., 100 knots) and the minimum separation might be small, the general trends and principles found through the experiments should be able to hold.In the scenarios, flights were required to go through the predefined region with origin and destination outside of the region.To increase the operational density and complexity, all flights in each scenario were set to depart within a three-minute window.Fig. 4 shows a sample scenario with 50 vehicles, where circle and cross markers represent origins and destinations, respectively.An early study [29] showed that the intrinsic complexity of operations for a given scenario doesn't only depend on the number of aircraft.For instance, a scenario with 50 aircraft may not pose any losses of separation without a conflict resolution service, whereas a scenario with 5 aircraft may pose multiple losses of separation if there is no conflict resolution service.However, statistically, a large number of randomly generated scenarios with 50 aircraft should pose more potential conflicts-and hence be more complex for the conflict resolution function-than randomly generated scenarios with 5 aircraft.Therefore, in this study, more than 200 scenarios were randomly generated for each "number of aircraft" test point to ensure that, statistically, scenarios with more aircraft were more complex than scenarios with fewer aircraft.In the experiments, over 3,800 scenarios were created for 19 different levels of density with increased number of aircraft from 3 to 70 aircraft.In the experiments, besides the aforementioned three architectures, the input error and the reception probability of the output(i.e., conflict resolution maneuvers) are also used as independent variables.Table 2 shows the test metrics used in the experiments.Two metrics, loss of separation and extra flight distance caused by resolution maneuvers, are used to quantify safety and efficiency, respectively. +IV. ResultsThis section will analyze performance differences among three architectures based on simulation results.The tolerance of the centralized architecture to input error and output reception probability will also be examined. +A. Baseline case: Centralized vs. decentralized in the absence of input or output errorsIn the absence of communication and information errors, the difference between centralization and decentralization is in their objectives.Centralized conflict resolution cross-checks and tries to deconflict resolution maneuvers with all other proximate aircraft before transmitting the maneuvers to individual aircraft.Conversely, with the decentralized architecture (when there is no real-time negotiation), the conflict resolution function onboard the aircraft relies on the predefined rules and computes and executes its own conflict resolution maneuvers without knowing others' conflict resolution maneuvers. +Fig. 5 Safety comparison between centralized and decentralizedFig. 5 presents the safety metric comparison between centralized and decentralized conflict resolution in terms of the percentage of scenarios with one or more losses of separation.The black curve indicates the percentage of scenarios with losses of separation when conflict resolution services were disabled.As the number of aircraft was increased, the likelihood that a scenario would have a loss of separation also increased.The blue curve represents the percentage of scenarios with one or more losses of separation when conflict resolution services were decentralized with predefined uniform rules, whereas the red curve corresponds to the percentage of such scenarios when conflict resolution services were centralized.It is noted that both architectures were able to resolve 100% of the potential conflicts for the large majority of scenarios.At the test points with the higher aircraft counts, the centralized architecture performed slightly better than the decentralized architecture, albeit only marginal.The centralized conflict resolution produced a total of 3.2% of scenarios with losses of separation (out of 3,800 scenarios), whereas the decentralized architecture produced a total of 3.4% of scenarios with losses of separation. +Fig. 6 Efficiency comparison between centralized and decentralizedFig. 6 presents the efficiency comparison between centralized and decentralized conflict resolution in terms of the percentage of extra flight distance.The blue and red curves represent the decentralized and centralized architectures, respectively.It is not surprising that the centralized outperforms the decentralized with 1.4% improvement at the level of 70 aircraft.In both the safety and efficiency comparisons, the performance advantage of the centralized architecture mostly happened when the number of aircraft exceeded 40 or 50, when the chance of multiple simultaneous conflicts was high.When the number of aircraft was low, the likelihood of multiple conflicts in close proximity was also low and there was no difference between solving conflicts sequentially versus in parallel. +B. Uniform rules vs. mixed rules in decentralized architecturesAs introduced before, another decentralized architecture referred to as the decentralized with mixed rules (shown in Fig. 3(a)) allows mixed conflict resolution rules in the same airspace.To simulate the operations with mixed rules in the same airspace, two different algorithms (named A and B for convenience) with different sets of rules and parameters were applied.In every scenario, 50% of the aircraft were randomly chosen to use algorithm A, and the other 50% of aircraft used algorithm B. Fig. 7 compares safety metrics among three different cases.The blue curve shows the percentage of scenarios with one or more losses of separation when only the algorithm A (or rule set A) was used for conflict resolution.Algorithm A was used to represent the decentralized architecture in the previous results section.The magenta curve shows the percentage of scenarios with one or more losses of separation when only algorithm B was used for confliction resolution.It is noted that algorithm A clearly outperformed algorithm B. However, when both services were deployed in the same area, with half of the aircraft using A and the other half using B, the resulting percentage of scenarios with one or more losses of separation (shown as the black curve) was almost the same as when only algorithm B was deployed.The safety performance was undermined and dominated by the poorer performer, which was algorithm B in this experiment.This result shows that a uniform set of rules is necessary for operations in the same area, otherwise, the performance might be negatively impacted and dominated by the poorer performer.x) Input errors (e.g.aircraft state error) exist in both centralized and decentralized architectures.Though the ground-based centralized service has the advantage of better ground facilities (e.g., advanced radar and powerful computing facility), the information error is still inevitable.Fig. 8 shows the change in the performance advantage of centralized over decentralized when input errors were modeled in both architectures.The blue bars represent the improvement of centralized over decentralized in terms of the percentage of scenarios with losses of separation when there was no input error.When input errors were modeled with a standard deviation of 0.5 feet, the performance advantage quickly reduced from 2% to less than 0.3 % as shown in the brown bars.The possible explanation would be: The centralized service depends on the information collected for all aircraft and makes decisions in sequence to ensure final maneuvers are coordinated, hence, if the aircraft that were involved in the conflicts in the front of the conflict list have imperfect state information, ripple effects will likely pass on to the rest of the conflict list.Since the centralized algorithm will solve the conflicts in the front of the conflict list first, the incorrect resolution maneuvers might affect the following conflicts should there be correlation between these conflicts.Once the centralized service finishes computing resolution maneuvers, it needs to transmit those resolutions to the aircraft through communication links that may experience packet loss or imperfect reception, where the reception probability of these resolutions can be represented as P{ ì y(t)| ì y(t)} as modeled in the previous section.This section investigates the impact of various reception probabilities on the system performance, which is measured by the percentage of scenarios with losses of separation. +Fig. 9 Centralized architectures with different resolution reception probabilityFig. 9 presents the safety performance metric with P{ ì y(t)| ì y(t)} = {99.9%,99.5%, 99%, 98%, 95%}.It shows that when the reception probability was decreased, the percentage of scenarios with losses of separations increases.The increase varied from 0% to 2% when the number of aircraft was lower than 40.Above 50 aircraft, a 2% to 5% degradation in reception probability resulted in 2% to 11% more scenarios having losses of separation.At 70 aircraft, when the reception probability drops from 100% to 95%, the percentage of scenarios with losses of separation went from 14% to 25%, a 78% increase.Converted directly from Fig. 9, Fig. 10 depicts the cumulative percentage of scenarios with losses of separations starting from the lowest number of aircraft to the highest number of aircraft.It can be seen that when errors were not modeled, the centralized architecture (red curve) performed better than the decentralized (shown as a black dashed line).The overall percentages for them are 3.2% and 3.4%, respectively, as discussed in the previous section.As the Fig. 10 Cumulative percentage of scenarios with losses of separation reception probability was reduced, the curve starts to bend upwards and the first crossover point at which the cumulative percentage of scenarios with losses of separation for centralized was worse (higher) than the decentralized occurred with lower and lower aircraft count (i.e., less and less complex scenarios).For instance, with 99.5% of reception probability, centralized became worse than decentralized when the number of aircraft was 22.That point occurred at 20 aircraft when the reception probability was down to 99%.Collectively, these points form a performance frontier between centralized and decentralized architectures as shown in the Fig. 11.The bottom right side of the frontier represents the area where the decentralized conflict resolution outperformed centralized.Since in the gray area when the number of aircraft is less than 18, the centralized and the decentralized performed the same, that leaves the top and right side of the boundary the only region where the centralized performed better.It is noted that when the reception probability of transmitted resolutions was less than 99.5%, the safety performance of the centralized architecture deteriorated quickly.This suggests that the centralized system may not be suitable for high volume traffic operations when the reception probability is not near perfect. +Fig. 11 Performance boundary between centralized and decentralized architecturesAlthough the experiments in this section focused on the normal situations, the results could apply to the abnormal conditions such as the presence of cyber-attacks and the degradation would depend on the severity of the attack.These results suggest that the performance of the centralized architecture would be more vulnerable than decentralized architectures by virtue of the fact that decentralized architectures do not need to communicate resolution maneuvers, because they are computed and executed onboard.vulnerability must be considered when designing the centralized architecture and architecture for conflict resolution. +V. ConclusionsWhile the UAM concept has caught great attention due to its potential to significantly improve the way of moving goods and people, there are still many critical questions needed to be addressed to enable UAM concept development and operations.This work studies one of them: Should the conflict resolution in a UAM traffic system be centralized or decentralized?There are only a few works in comparing centralization and decentralization for air traffic management system, an in-depth study is needed to address not only the impact from the objective perspective but also the sensitivity to information and communication errors.This work modeled and analyzed three conflict resolution architectures: centralized, decentralized with uniform rules, and decentralized with mixed rules.The communication and information errors were also modeled.A high-fidelity Monte Carlo traffic simulator was used to simulate sUAS aircraft in a generic, two-dimensional airspace with increasing traffic complexity.The results showed that when information and communication were perfect, the centralized conflict resolution slightly outperformed decentralized.However, the centralized architecture was found to be sensitive to the modeled input errors, and the results of this study suggest that the marginal performance advantage relative to the decentralized architecture can quickly diminish.More important, the safety performance of the centralized conflict resolution architecture compromised when the output reception error was increased.Additionly, a comparison between decentralized conflict resolution architectures found that, with mixed rules, the system performance was undermined and dominated by the poorer performer.To examine if the findings in this study will still hold in general, future work will explore vehicles with higher speeds with different separation minimums.And advanced centralized and decentralized aglorithms will also be incorporated. +VI. AcknowledgementThe author would like to thank Dr. Husni Indris (NASA) for the discussion of airspace operations under various architectures and Todd Farley (NASA) for valuable feedback during the review process.Fig. 1 Notional1Fig. 1 Notional Graph and Data Flow for the Centralized System +(a) Decentralized with Uniform Rules (b) Data Flow in the Decentralized System +Fig. 22Fig. 2 Notional Graph and Data Flow for the Decentralized System with Uniform Rules +(a) Decentralized with Mixed Rules (b) Data Flow in the Decentralized System +Fig. 33Fig. 3 Notional Graph and Data Flow for the Decentralized System with Mixed Rules +Fig. 4 A4Fig. 4 A Sample Scenario with 50 vehicles +Fig. 77Fig. 7 Comparison between uniform rules and mixed rules in decentralized architectures +Fig. 88Fig. 8 Sensitivity of performance gain to the aircraft state errors + + + + +Table 2 Test metrics2Independent variablesValuesNumber of aircraft3, 5, 7, 9, 11, 13, 16, 19, 22, 25, 30, 35, 40, 45, 50, 55, 60, 65, 70Number of scenarios200 scenarios randomly generated at each density levelArchitecturescentralized, decentralized (uniform rule based), decentralized (mixed rule based)Output reception probability100%, 99.9%, 99.5%, 98%, 95%Input state error (ft)0, 0.5 + + + + + + + + + Urban Air Mobility Airspace Integration Concepts and Considerations + + DavidPThipphavong + + + RafaelApaza + + + BryanBarmore + + + VernolBattiste + + + BarbaraBurian + + + QuangDao + + + MichaelFeary + + + SusieGo + + + KennethHGoodrich + + + JeffreyHomola + + + HusniRIdris + + + ParimalHKopardekar + + + JoelBLachter + + + NatashaANeogi + + + HokKwanNg + + + RosaMOseguera-Lohr + + + MichaelDPatterson + + + SavitaAVerma + + 10.2514/6.2018-3676 + + + 2018 Aviation Technology, Integration, and Operations Conference + Atlanta, Georgia + + American Institute of Aeronautics and Astronautics + 2018. 2018 + + + Thipphavong, D., Apaza, R., Barmore, B., Battiste, V., Burian, B., Dao, Q., Feary, M., Go, S., Goodrich, K., Homola, J., Idris, H., Kopardekar, P., Lachter, J., Neogi, N., Ng, H., Oseguera-Lohr, R., Patterson, M., and Verma, S., "Urban Air Mobility Airspace Integration Concepts and Considerations," AIAA Aviation 2018, Atlanta, Georgia, 2018. + + + + + Enabling Airspace Integration for High-Density On-Demand Mobility Operations + + EricRMueller + + + ParmialHKopardekar + + + KennethHGoodrich + + 10.2514/6.2017-3086 + + + 17th AIAA Aviation Technology, Integration, and Operations Conference + Denver, Colorado + + American Institute of Aeronautics and Astronautics + 2017. 2017 + + + Mueller, P., E. Kopardekar, and Goodrich, K., "Enabling Airspace Integration for High-Density On-Demand Mobility Operations," AIAA Aviation 2017, Denver, Colorado, 2017. + + + + + Unmanned Aircraft System Traffic Management (UTM) Concept of Operations + + PKopardekar + + + JRios + + + TPrevot + + + MJohnson + + + JJung + + + JERobinson + + + + 16th AIAA Aviation Technology, Integration, and Operations Conference + Washington, D.C. + + 2016 + + + Kopardekar, P., Rios, J., Prevot, T., Johnson, M., Jung, J., and Robinson, J. E., "Unmanned Aircraft System Traffic Management (UTM) Concept of Operations," 16th AIAA Aviation Technology, Integration, and Operations Conference, Washington, D.C., 2016. + + + + + Flight Demonstration of Unmanned Aircraft System (UAS) Traffic Management (UTM) at Technical Capa... + 10.2514/6.2020-2851.vid + + + 2018 + American Institute of Aeronautics and Astronautics (AIAA) + + + Unmanned Aircraft System (UAS) Traffic Management (UTM), FAA, 2018. URL https://utm.arc.nasa.gov/docs/2018- UTM-ConOps-v1.0.pdf. + + + + + NASA UAS traffic management national campaign: Operations across Six UAS Test Sites + + JosephRios + + + DanielMulfinger + + + JeffHomola + + + PriyaVenkatesan + + 10.1109/dasc.2016.7778080 + Rep. NASA/TM-2018-220024 + + + 2016 IEEE/AIAA 35th Digital Avionics Systems Conference (DASC) + + IEEE + 2018 + + + NASA + + + Tech. + Rios, J., Smith, I. S., Venkatesan, P., Smith, D. R., Baskaran, V., Jurcak, S. M., Strauss, R., Iyer, S. K., and Verma, P., "UTM UAS Service Supplier Development: Sprint 1 Toward Technical Capability Level 4," Tech. Rep. NASA/TM-2018-220024, NASA, 2018. + + + + + NASA UAS traffic management national campaign: Operations across Six UAS Test Sites + + JosephRios + + + DanielMulfinger + + + JeffHomola + + + PriyaVenkatesan + + 10.1109/dasc.2016.7778080 + Rep. NASA/TM-2018-220050 + + + 2016 IEEE/AIAA 35th Digital Avionics Systems Conference (DASC) + + IEEE + 2018 + + + NASA + + + Tech. + Rios, J., Smith, I. S., Venkatesan, P., Smith, D. R., Baskaran, V., Jurcak, S. M., Iyer, S. K., and Verma, P., "UTM UAS Service Supplier Development: Sprint 2 Toward Technical Capability Level 4," Tech. Rep. NASA/TM-2018-220050, NASA, 2018. + + + + + For Spacious Skies: Self-Separation with "Autonomous Flight Rules" in US Domestic Airspace + + DavidWing + + + WilliamCotton + + 10.2514/6.2011-6865 + + + 11th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Virginia Beach, VA + + American Institute of Aeronautics and Astronautics + 2011 + + + Wing, D. J., and Cotton, W. B., "For Spacious Skies: Self-separation with Autonomous Flight Rules in US Domestic Airspace," 11th AIAA Aviation Technology, Integration, and Operations Conference, Virginia Beach, VA, 2011. + + + + + Protocol-based Conflict Resolution for Finite Information Horizon + + IHwang + + + CTomlin + + + + Proceedings of the American Control Conference + the American Control ConferenceAnchorage, AK + + 2002 + + + Hwang, I., and Tomlin, C., "Protocol-based Conflict Resolution for Finite Information Horizon," Proceedings of the American Control Conference, Anchorage, AK, 2002. + + + + + Provably Safe Coordinated Strategy for Distributed Conflict Resolution + + GillesDowek + + + CesarMunoz + + + VictorCarreno + + 10.2514/6.2005-6047 + + + AIAA Guidance, Navigation, and Control Conference and Exhibit + San Francisco, CA + + American Institute of Aeronautics and Astronautics + 2005 + + + Dowek, G., Munoz, C. A., and Carreno, V. A., "Provably Safe Coordinated Strategy for Distributed Conflict Resolution," AIAA Guidance, Navigation, and Control Conference and Exhibit, San Francisco, CA, 2005. + + + + + Agent-Based Cooperative Decentralized Airplane-Collision Avoidance + + DavidSislak + + + PřemyslVolf + + + MichalPechoucek + + 10.1109/tits.2010.2057246 + + + IEEE Transactions on Intelligent Transportation Systems + IEEE Trans. Intell. Transport. Syst. + 1524-9050 + 1558-0016 + + 12 + 1 + + 2011 + Institute of Electrical and Electronics Engineers (IEEE) + + + Sislak, D., Volf, P., and Pechoucek, "Agent-based Cooperative Decentralized Airplane Collision Avoidance," IEEE Transactions on Intelligent Transportation Systems, Vol. 12, No. 1, 2011, pp. 36-46. + + + + + Impact of Different Trajectory Option Set Participation Levels within an Air Traffic Management Collaborative Trajectory Option Program + + Hyo-SangYoo + + + ConnieBrasil + + + NancyMSmith + + + NathanBuckley + + + GitaHodell + + + ScottKalush + + + PaulULee + + 10.2514/6.2018-3040 + + + 2018 Aviation Technology, Integration, and Operations Conference + Atlanta, GA + + American Institute of Aeronautics and Astronautics + 2018 + + + Yoo, H., Brasil, C. L., Buckley, N., Hodell, G. S., Kalush, S. N., Lee, P. U., and Smith, N. M., "Impact of Different Trajectory Option Set Participation Levels within an Air Traffic Management Collaborative Trajectory Option Program," AIAA Aviation Technology, Integration, and Operations Conference, Atlanta, GA, 2018. + + + + + Optimal Strategies for Free-Flight Air Traffic Conflict Resolution + + PKMenon + + + GDSweriduk + + + BSridhar + + 10.2514/2.4384 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 22 + 2 + + 1999 + American Institute of Aeronautics and Astronautics (AIAA) + + + Menon, P. K., Sweriduk, G. D., and Sridhar, B., "Optimal Strategies for Free-Flight Air Traffic Conflict Resolution," Journal of Guidance, Control, and Dynamics, Vol. 22, No. 2, 1999, pp. 202-211. + + + + + When Does Coordination Require Centralization? + + RicardoAlonso + + + WouterDessein + + + NikoMatouschek + + 10.1257/aer.98.1.145 + + + American Economic Review + American Economic Review + 0002-8282 + + 98 + 1 + + 2008 + American Economic Association + + + Alonso, R., Dessein, W., and Matouschek, N., "When Does Coordination Require Centralization," American Economic Review, Vol. 98, No. 1, 2008, pp. 145-179. + + + + + Coordination in Centralized and Decentralized Systems + + VipinPVeetil + + 10.34196/ijm.00157 + + + International Journal of Microsimulation + IJM + + 10 + 2 + + 2017 + International Microsimulation Association + + + Veetil, V. P., "Coordination in Centralized and Decentralized Systems," International Journal of Microsimulation, Vol. 10, No. 2, 2017, pp. 86-102. + + + + + On optimal cooperative conflict resolution for air traffic management systems + + ABicchi + + + LPallottino + + 10.1109/6979.898228 + + + IEEE Transactions on Intelligent Transportation Systems + IEEE Trans. Intell. Transport. Syst. + 1524-9050 + + 1 + 4 + + 2000 + Institute of Electrical and Electronics Engineers (IEEE) + + + Bicchi, A., and Pallottino, "On Optimal Cooperative Conflict Resolution for Air Traffic Management Systems," IEEE Transactions on Intelligent Transportation Systems, Vol. 1, No. 4, 2000, pp. 221-232. + + + + + System Performance Characteristics of Centralized and Decentralized Air Traffic Separation Strategies + + JimmyKrozel + + + MarkPeters + + + KarlDBilimoria + + + ChangkilLee + + + JosephS BMitchell + + 10.2514/atcq.9.4.311 + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 9 + 4 + + 2001 + American Institute of Aeronautics and Astronautics (AIAA) + Santa Fe, NM + + + Krozel, J., Peters, M., Bilimoria, K., Lee, C., and Mitchell, S. B., "System Performance Characteristics of Centralized and Decentralized Air Traffic Separation Strategies," 4th USS/Europe Air Traffic Management R&D Seminar, Santa Fe, NM, 2001. + + + + + Fe3: An Evaluation Tool for Low-Altitude Air Traffic Operations + + MinXue + + + JosephRios + + + JosephSilva + + + ZhifanZhu + + + AbrahamKIshihara + + 10.2514/6.2018-3848 + + + 2018 Aviation Technology, Integration, and Operations Conference + Atlanta, GA + + American Institute of Aeronautics and Astronautics + 2018 + + + Xue, M., Rios, J., Silva, J., Ishihara, A., and Zhu, Z., "Fe 3 : An Evaluation Tool for Low-Altitude Air Traffic Operations," AIAA Aviation Forum, Atlanta, GA, 2018. + + + + + TRB Special Report 301: Traffic Controller Staffing in the En Route Domain + + RJHansman + + + MSAlcabin + + + MOBall + + + L + + + CMDunlay + + + WJElias + + + ALFearnsides + + + JJLebacqz + + + JVPowderly + + + MJSmith + + + PJTrani + + + AWall + + + RZacharias + + + GLMenzies + + + TRHemel + + + SV + + 10.17226/13022 + + + Tech. Rep. Special Report + + 301 + 2010 + Transportation Research Board + + + Transportation Research Board, National Research Concil of the National Academies + Hansman, R. J., Alcabin, M. S., Ball, M. O., L., C. M., Dunlay, W. J., Elias, A. L., Fearnsides, J. J., Lebacqz, J. V., Powderly, M. J., Smith, P. J., Trani, A., Wall, R., Zacharias, G. L., Menzies, T. R., and Hemel, S. V., "Air Traffic Controller Staffing in the En Route Domain: A Review of the Federal Aviation Administration's Task Load Model," Tech. Rep. Special Report 301, Transportation Research Board, National Research Concil of the National Academies, 2010. + + + + + Real-Time Radar-Based Tracking and State Estimation of Multiple Non-Conformant Aircraft + + BrandonCook + + + TimothyJArnett + + + OwenMacmann + + + ManishKumar + + 10.2514/6.2017-1133 + + + AIAA Information Systems-AIAA Infotech @ Aerospace + Grapevine, Texas + + American Institute of Aeronautics and Astronautics + 2017 + + + Cook, B., Arnett, T., Macmann, O., and Kumar, M., "Real-Time Radar-Based Tracking and State Estimation of Multiple Non-Conformant Aircraft," AIAA SciTech Forum, Grapevine, Texas, 2017. + + + + + Radar Data Tracking Using Minimum Spanning Tree-Based Clustering Algorithm + + ChunkiPark + + + Hak-TaeLee + + + BassamMusaffar + + 10.2514/6.2011-6825 + + + 11th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Virginia Beach, VA + + American Institute of Aeronautics and Astronautics + 2011 + + + Park, C., Lee, H., and Musaffar, B., "Radar Data Tracking Using Minimum Spanning Tree-Based Clustering Algorithm," 11th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, Virginia Beach, VA, 2011. + + + + + Multiple-Target Tracking and Identity Management with Application to Aircraft Tracking + + InseokHwang + + + HamsaBalakrishnan + + + KaushikRoy + + + ClaireTomlin + + 10.2514/1.27366 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 30 + 3 + + 2007 + American Institute of Aeronautics and Astronautics (AIAA) + + + Hwang, I., Balakrishnan, H., Roy, K., and Tomlin, C., "Multiple-Target Tracking and Identity Management with Application to Aircraft Tracking," Journal of Guidance, Control, and Dynamics, Vol. 30, No. 3, 2007, pp. 641-652. + + + + + A review of conflict detection and resolution modeling methods + + JKKuchar + + + LCYang + + 10.1109/6979.898217 + + + IEEE Transactions on Intelligent Transportation Systems + IEEE Trans. Intell. Transport. Syst. + 1524-9050 + + 1 + 4 + + 2000 + Institute of Electrical and Electronics Engineers (IEEE) + + + Kuchar, J. K., and Yang, L. C., "A Review of Conflict Detection and Resolution Modeling Methods," IEEE Transactions On Intelligent Transportation Systems, Vol. 1, No. 4, 2000. + + + + + NASA contract for charged particle spectrometer for manned space vehicle + + CBach + + + RFarrell + + + HErzberger + + 10.1108/aeat.2009.12781dab.019 + AR-2009-214573 + + + Aircraft Engineering and Aerospace Technology + Aircraft Eng & Aerospace Tech + 0002-2667 + + 81 + 4 + 2009 + Emerald + + + Tech. Rep. Technical Report + Bach, C., R. Farrell, and Erzberger, H., "An Algorithm for Level-Aircraft Conflict Resolution," Tech. Rep. Technical Report, AR-2009-214573, NASA, 2009. + + + + + Conflict Detection and Resolution for Free Flight + + JimmyKrozel + + + MarkPeters + + 10.2514/atcq.5.3.181 + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 5 + 3 + + 1997 + American Institute of Aeronautics and Astronautics (AIAA) + + + Krozel, J., and Peters, M., "Conflict Detection and Resolution for Free Flight," Air Traffic Control Quarterly, Vol. 5, No. 3, 1997. + + + + + Chain-based Collision Avoidance for UAS Sense-and-Avoid Systems + + LaithSahawneh + + + RandalWBeard + + + SharathAvadhanam + + + HeBai + + 10.2514/6.2013-4995 + + + AIAA Guidance, Navigation, and Control (GNC) Conference + Boston, MA + + American Institute of Aeronautics and Astronautics + 2013 + + + Sahawneh, L. R., and Beard, R. W., "Chain-based Collision Avoidance for UAS Sense-and-Avoid Systems," AIAA Guidance, Navigation, and Control Conference, Boston, MA, 2013. + + + + + A Fuzzy Logic Approach for Low Altitude UAS Traffic Management (UTM) + + BrandonCook + + + KellyCohen + + + EladHKivelevitch + + 10.2514/6.2016-1905 + + + AIAA Infotech @ Aerospace + San Diego, California + + American Institute of Aeronautics and Astronautics + 2016. 2016 + + + Cook, B., Cohen, K., and Kivelevitch, E., "A Fuzzy Logic Approach For Low Altitude UAS Traffic Management," AIAA Science and Technology Forum and Exposition 2016, San Diego, California, 2016. + + + + + Protocol-based Conflict Resolution for Finite Information Horizon + + IHwang + + + CTomlin + + + + Proceedings of the American Control Conference + the American Control ConferenceAnchorage, AK + + 2002 + + + Hwang, I., and Tomlin, C., "Protocol-based Conflict Resolution for Finite Information Horizon," Proceedings of the American Control Conference, Anchorage, AK, 2002. + + + + + Multi-Rotor Aircraft Collision Avoidance using Partially Observable Markov Decision Processes + + EricRMueller + + + MykelKochenderfer + + 10.2514/6.2016-3673 + + + AIAA Modeling and Simulation Technologies Conference + Washington, D.C. + + American Institute of Aeronautics and Astronautics + 2016 + + + Mueller, E., and Kochenderfer, M. J., "Multi-Rotor Aircraft Collision Avoidance using Partially Observable Markov Decision Processes," AIAA Modeling and Simulation Technologies Conference, Washington, D.C., 2016. + + + + + Scenario Complexity for Unmanned Aircraft System Traffic + + MinXue + + + MinhDo + + 10.2514/6.2019-3513 + + + AIAA Aviation 2019 Forum + Dallas, TX + + American Institute of Aeronautics and Astronautics + 2019 + + + Xue, M., and Do, M., "Scenario Complexity for Unmanned Aircraft System Traffic," AIAA Aviation Forum, Dallas, TX, 2019. + + + + + + diff --git a/file810.txt b/file810.txt new file mode 100644 index 0000000000000000000000000000000000000000..d7162a634ca1545de54688b701c72eaa2685d509 --- /dev/null +++ b/file810.txt @@ -0,0 +1,719 @@ + + + + +I. IntroductionIn conventional aviation, where traffic is managed by air traffic controllers, measuring and predicting airspace complexity can assist planning traffic flow changes, sector level staffing needs, and other operational decisions.For UAS traffic, a good and quick approximation of traffic complexity can help assess traffic scenarios, re-plan flight routes and schedules to alleviate traffic bottleneck, and mitigate operation risk.It can also help categorize traffic senarios for traffic management studies.Real-time or near real-time complexity assessment is needed for both research and operations in UAS traffic management.Simple complexity metrics like aircraft count or traffic density can be misleading, especially for cases when traffic is highly organized, e.g. a set of aircraft fly parallelly where the aircraft count is high but the effort level of managing traffic is low.Over the past years, methods have been developed based on the cognitive complexity, such as Dynamic Density (DD) metric [1][2][3], where key features are weighted according to controller's workload ratings and a weighted sum of these key features is then used as the final measurement.Meanwhile, complexity metrics based on the intrinsic traffic disorder were developed as well to reveal the complexity in a sector regardless of controller's workload.Delahaye et.al. [4] proposed a geometrical approach based on the properties of relative positions and relative speeds of aircraft in a sector to obtain time histories of traffic divergence, convergence, and sensitivity.By modeling aircraft trajectories as a linear dynamical system, Delahaye et.al. [4,5] also developed an aggregate metric using the entropy of the dynamic system.Based on that, the velocity vector field methods [5,6] were developed to compute complexity maps for given traffic scenario snapshots.However, measuring UAS traffic complexity is different from traditional traffic.First, it is likely there will be no human controller to maintain separations, therefore, controller workload rating can not be used to help construct cognitive complexity metrics.Second, while air traffic controllers resolve conflicts based on existing procedures and rules in conventional aviation, autonomous UASs follow their multi-point trajectory plans and perform avoidance maneuvers guided by a predefined conflict management model.Finally, since a large number of UASs are envisioned to operate in a given traffic scenario, the complexity metric needs to be computed near real-time for hundreds of vehicles.In recent UTM related studies, Bulusu et.al. [7] studied the relationship between capacity and flow rate using three conflict management models to obtain an initial understanding of the operational capacity of UAS traffic, however, the scope of the study did not include intrinsic complexity of a given traffic scenario caused by the schedule and layout of flight plans.Meanwhile, the Fe 3 [8] simulator was developed to compute traffic statistics through high-fidelity Monte Carlo simulations.Although the simulator has the capability of providing accurate traffic complexity measurements, its performance may not be suitable for a real-time application.In this paper, a complexity metric is developed based on the number of potential conflicts weighted by the resolution cost associated with each conflict.Using the simulation mesurements from the Fe 3 simulator as the ground truth for compleixty, the proposed metric was compared against several popular metrics based on the hundreds of random-generated scenarios.The results showed that the correlation between the proposed metric is better than the other metrics.In this work, Section II introduces the new complexity metric and the methods used to generate the metric are described as well including the mathematical programming formulation for solving pairwise conflicts.Section III presents the experiments, evaluations, and discussion.Metrics were compared against the measurements from the high-fidelity simulations.Section IV concludes the study. +II. MethodologyThe number of potential conflicts and the number of aircraft within a distance threshold are popular metrics in the past studies [1][2][3]9].In order to differentiate difficulties involved in various encounters, indirect features have been proposed in dynamic density methods [1][2][3], such as the angle of convergence, the time to conflict, the horizontal proximity, and conflict resolution difficulty based crossing angle.In the linear dynamic system approach, Delahaye et.al. [5] constructed matrices for different types of encounters and used eigenvalues of these matrices as an indicator of the degree of organization of each encounter.To safely and efficiently handle a large volume of UAS traffic, the future UAS traffic management system is envisioned to be managed by predefined autonomous conflict managemnet services (centralized or decentralized) instead of air traffic controllers.This opens up the possibility of taking advantage of the autonomous conflict management model when computing the difficulty associated with each conflict.While different encounters or conflicts (e.g.small angle crossing vs. head-on conflict) introduce different complexities, different conflict management models (e.g.different parameter setting, heading change vs. speed change) may also result in difference in complexity.Incorporating conflict management models can help accurately compute the cost involved in an encounter.In this section, the general conflict management structures together with parameters are first introduced.A mathematic programming formulation incorporating vehicle model and conflict management parameters is then presented to compute pairwise conflict resolution maneuvers.Finally, the complexity metric built upon the conflict resolution maneuvers are proposed. +A. Conflict Management ModelA typical conflict management model applies three phases when resolving conflicts [10]: strategic conflict management, separation provision, and collision avoidance.The strategic conflict management changes flight plans to resolve conflicts and normally occurs prior to departure.The separation provision is a tactical process for conflicts predicted while airborne to prevent well-clear violations.The collision avoidance is triggered when the well-clear definition is violated and serves as the last layer of conflict resolution.Figure 1 shows the typical conflict management model structure (not to scale), which includes separation provision and collision avoidance.The Near Mid-Air Collision (NMAC) represents the last layer of separation required to avoid physical contacts.The well-clear threshold d WC provides minimal space for collision avoidance for which both ownship and intruder can take actions as the last effort before colliding with each other and serves as the boundary between the collision avoidance and separation provision phases.The look-ahead time (LAT) defines the time horizon for trajectory prediction.The conflict warning threshold (CWT) defines the time t CWT when a conflict warning becomes effective.The conflict resolution threshold (CRT) defines the time t C RT when the ownship can start to maneuver to resolve the conflict.Although a heading-change resolution was shown in the figure, this structure can be utilized for other conflict resolution options such as speed change and altitude change.Apparently, different settings in conflict management models will yield different complexity.For instance, a larger value of conflict resolution threshold will require more space to resolve a conflict thus yield higher complexity for a given traffic scenario.Also a conflict management model with heading change only will demand a larger airspace than a model with additional speed changes.Since developing the well-clear definition is still an open research question [11,12], a distance of 50 feet is arbitrarily chosen for d WC in this study.The conflict resolution threshold t C RT is set to 30 seconds. +Fig. 1 Structure used in the conflict resolution algorithms B. Method to Compute Pairwise Conflict ResolutionsA mixed integer linear programming was formulated to compute minimum deviations needed for mitigating pairwise conflicts in a traffic scenario when a large amount of vehicles are involved.Over the past years, many Mixed Integer Linear Programming (MILP) based methods have been proposed to solve dynamic obstacle avoidance path planning and conflict-free trajectory planning for UASs [13][14][15][16][17][18][19][20][21].These studies showed that solving conflict-free trajectories in long horizon for multiple aircraft demands expensive computational time, which makes any application with a large number of aircraft prohibitive.However, because only a quick approximation is required for measuring a traffic scenario complexity, computing the minimum deviation for each pairwise conflict should be sufficient.Therefore, a MILP problem is formulated to solve a large amount of pair-wise conflicts to achieve the real-time computation performance. +Vehicle Trajectory ModelThe aircraft trajectory can be characterized by a discrete-time linear model [17,18,20] (Eqn. 1) and associated initial conditions (Eqn.2), where [p x,t , p y,t ] respesent the position vector at time t for an aircraft.[v x,t , v y,t ] and [a x,t , a y,t ] are velocity and acceleration vectors at time t, respectively, and ∆t is the time step used in the formulation.[x 0 y 0 x 0 y 0 ] are initial position and velocity of the aircraft.∀ t ≥ 0,          p x,t+1 p y,t+1 v x,t+1 v y,t+1          =          1 0 ∆t 0 0 1 0 ∆t 0 0 1 0 0 0 0 1                   p x,t p y,t v x,t v y,t          +           1 2 ∆t 2 0 0 1 2 ∆t 2 ∆t 0 0 ∆t           a x,t a y,t(1)p x,0 p y,0 v x,0 v y,0 T = x 0 y 0 x 0 y 0 T(2)In order to capture the aircraft dynamics, dynamics constraints including speed, acceleration, and turn rate are applied.A maximum speed v max is imposed and the quadratic constraints v 2 x + v 2 y ≤ v max are linearized using a K-sided polygon technique proposed in Richards's work [13] to make the computation efficient (shown in Eqn.3), where K can be adjusted to meet a given precision requirement.Ten constraints with K = 10 should provide sufficient approximation in application.∀i∈[1...N], ∀ t∈[t si , t ei ], ∀ k∈[1...K] : v x,i,t • sin( 2πk K ) + v y,i,t • cos( 2πk K ) ≤ v max(3)To introduce a minimum speed v min , the speed vector has to lie outside of a K-sided polygon with at least one of the binary variables c t,k being nonzero as shown in Eqn. 4, where M is a sufficient large positive number to disable a constraint when the binary variable is one.∀ i∈[1...N], ∀ t∈[t si , t ei ], ∀ k∈[1...K] : v x,i,t • sin( 2πk K ) + v y,i,t • cos( 2πk K ) ≥ v min -M•c t,k K k=1 c t,k ≤ K -1 c t,k ∈ [ 0, 1](4)The maximum turn rate ω max is introduced and approximated through the maximum acceleration using the product of the maximum turn rate and the maximum speed [13,15]:a max = v max • ω max(5)And to satisfy the maximum acceleration constraint a max , the acceleration vector needs to stay inside of a K-sided polygon:∀ i∈[1...N], ∀ t∈[t si , t ei ], ∀ k∈[1...K] : a x,i,t • sin( 2πk K ) + a y,i,t • cos( 2πk K ) ≤ a max(6) +Separation ConstraintBesides vehicle dynamics constraints, aircraft have to satisfy a minimum separation distance from each other to avoid collisions.A well-clear distance d wc is required for the minimum separation, therefore, quadratic separation distance constraints are introduced to maintain well-clear among aircraft.The same K-sided polygon technique is applied to linearize the quadratic separation constraints.The relative position vector has to stay outside of the polygon with at least one zero binary varilable c t,k .∀ i, j∈[1...N], ∀ i j, ∀ t∈[t si , t ei ], ∀ k∈[1...K] : (p x,i,t -p x, j,t ) • sin( 2πk K ) + (p y,i,t -p y, j,t ) • cos( 2πk K ) ≥ d wc -M•c t,k K k=1 c t,k ≤ K -1 c t,k ∈ [ 0, 1](7) +Finite HorizonTo reduce the computational time, the finite horizon technique is applied to limit the time horizon of the MILP formulation to a time window during which a conflict resolution maneuver happens.According to the general conflict resolution structure described in Section II.A, the time horizon of each pair-wise conflict can be reduced to a finite range: [t collisiont crt , t collision + t crt ] (shown in Fig. 2), where the t collision is calculated based on the original flight plan. +ObjectiveFinally, within the finite horizon, the objective function is to minimize the deviation from aircraft's original path as shown in Eqn. 8, where t si and t ei denotes the t collisiont crt and t collision + t crt for aircraft i, respectively.And a trajectory point at time t in the original path is represented by q i,t .min p,aJ = N i=1 t ei t=t si (p x,i,t -q x,i,t ) 2 + (p y,i,t -q y,i,t ) 2(8)To linearize the final objective and avoid the quadratic equations for computing distances, the distances are expressed as an additional set of linear constraints for any aircraft i as in Eqn.9:∀ i∈[1, N], ∀ t∈[t si , t ei ], ∀ k∈[1...K] : (p x,i,t -q x,i,t ) • sin( 2πk K ) + (p y,i,t -q y,i,t ) • cos( 2πk K ) ≤ d i,t(9)With the formulation (Eqn.1-7, 9, 10), the conflict resolution maneuvers for a large number of pairwise conflicts can then obtained.Besides t crt and d W C , this formulation can be adjusted easily to a different conflict management model.For instance, if the conflict management model only allows heading changes, it can be achieved by setting v min and v max closely.This MILP model is implemented in C/C++ and solved by Solving Constraint Integer Programs (SCIP) [22,23], an open-source optimization suites supporting MILP, quadratic, and several other types of constraint.The experiments show that for a large random set of scenarios with the number of flights increased from 5 to 100 flights, the median and average running time are 15 and 20 seconds on a Core I7 Mac laptop, respectively.Considering there are many approaches, such as using commerical MILP solvers, parallel computing, etc, reducing the computational time to near real-time should be feasible. +C. Scenario Complexity MetricThe cost of a pairwise conflict is defined as the sum of deviations from the original path in addition to the conflict count (shown in Eqn.11), where C f i and C f j are the deviation costs associated with two involved flights, respectively.The deviation cost for a flight involved in a conflict is defined as the temporal deviation, which is calculated by the percentage of the distance deviation multiplied by the conflict resolution duration.p t s and p t e are the beginning and ending points of the resolution maneuver for a flight, respectively.p t p t+1 denotes the distance between p t and p t+1 , while p t s p t e is the distance between the beginning point p t s and the ending point p t e .C = 1.0 + C f i + C f j C f m =|Figure 3(a), 3(b), and 3(c) present the conflict resolution maneuvers computed by the MILP formulation in three different encounter situations: crossing with a small angle, crossing with a large angle, and a head-on conflict.According to the calculation using Eqn.11, the costs for these three encounters are: 1.26, 1.32, and 1.29, respectively.It shows that with the setup of MILP formulation, the large-angle crossing has a higher cost than the head-on encounter, while the small-angle crossing has a lower cost than the other two.The final measurement of scenario complexity metric SC can then be expressed as Eqn.12, where nc is the number of potential conflicts associated with the original flight schedule in a given scenario without any conflict resolutions.SC = k=1 nc C k(12) +III. ExperimentsIn this section, random scenarios were first generated for experiments.Then the metric called the number of resolution maneuvers was produced by the Fe 3 [8] simulator.Using the simulation generated measurements as the ground truth for the scenario complexity, the proposed complexity metric SC was then analyzed and compared using statistical methods. +A. Test ScenariosTo evaluate complexity metrics, random scenarios with a large variety of complexities need to be generated.Several criteria were used to ensure high traffic intensity and comparability in scenarios: First, a 1.3x1.3nautical mile region is defined (shown as the red box in Fig. 4) and all flights are required to go through the predefined region with origin and destination outside of the region; Second, at most one turning point is allowed other than the origin and destination in a flight plan; Third, all flights are set to depart within a five-minute window; Lastly, the target ground speeds of all flights are in the range of 5 meter per second and 20 meter per second.Fig. 4 shows a sample scenario with 30 vehicles, where circle, cross, diamond markers represent origins, destinations, and mid-points respectively.The number of aircraft in these scenarios varies from 5 to 50 (or in density from 3 to 30 vehicle/nmi 2 ).Fig. 5 shows the percentage of scenarios with and without conflicts during the process of generating scenarios.When the traffic density increases, the likelihood of having conflicts increases and reaches 100 % around 15 vehicle/nmi 2 .As scenarios without conflicts will have zero scenario complexity based on the proposed metric, therefore, only the scenarios have have potential conflicts are used in experiments.As a result, a total of 920 scenarios were created and used in this work with 20 scenarios at each level of density. +B. EvaluationThe Fe 3 [8] simulator has the capability of simulating traffic systems by incorporating critical models in an air traffic system, which includes vehicle models, conflict management models, wind model, communication, navigation, and surveillance models.Detailed description of those models and capabilities can be found in previous work [8].The simulator can provide statistical measurements for metrics such as losses of seperation, number of conflict maneuvers, extra flight distance, number of conflict warning, and energy consumed.Since this study focuses on scenario intrinsic complexity, uncertainties on wind, communication, navigation, and surveilance were not included in the simulations.The deterministic measurements of the number of conflict resolution maneuvers was used as the ground truth for complexity.The number of conflict resolution maneuvers measures the resolution moves issued during the simulation.As the time step size in Fe 3 is 0.5 seconds, the number of conflict resolution maneuvers also reflects the resolution duration.The Pearson method in R [24] was used to compute the correlations.Figure 6(a), 6(b), and 6(c) show the correlation plots for three matrices: the number of flights, the number of potential conflicts, and the proposed complexity metric SC.The correlation coefficients with the number of conflict resolution maneuvers (measurements from the Fe 3 simulator) Fig. 4 A Sample Scenario with 30 vehicles Fig. 5 Likelihood of Conflicts at Different Density Levels are 0.78, 0.9, and 0.9, respectively.p values for all three correlation tests are less than 2.2x10 -16 , which suggests with high significance that all three metrics have positive correlation with the complexity produced by simulations.The number of conflicts and the complexity metric SC have much higher correlation than the number of aircraft.As the Pearson method is designed for checking linear correlations, therefore, a maximal correlation method [25] was used to capture non-linear association and the Alternative Conditional Expections (ACE) package [26] in R was used to compute such maximal correlations.Figure 7(a), 7(b), and 7(c) present the plots after applying transformation in the ACE package.The correlation coefficients for number of flights, number of conflicts, and scenario complexity are then 0.85, 0.908, and 0.913, respectively.The correlation coefficients suggest that the complexity metric SC is more correlated to the number of resolution maneuvers that were generated by the Fe 3 simulator than the number of conflicts. +C. DiscussionThe analysis shows that the number of flights (or density) is much less correlated with the complexity compared to the other two metrics.For instance, as shown in Fig. 6(a), it is common that 40 flights (23 flights/nmi 2 ) may share the same complexity as 5 flights (3 flights/nmi 2 ), which makes the number of flights (or density) a bad indicator of complexity.On the other hand, by taking into account the complexity associated with each conflict, the proposed complexity metric SC showed better correlation than the number of potential conflicts.However, the proposed metric did not improve the correlation coefficient that much as it is essentially derived from the number of potential conflicts and shares a similar pitfall: the consequences of these conflicts are simply missed.For instance, there might exist secondary conflicts as the conflict resolution maneuvers changed the original trajectories.On the other hand, even though the conflict management structure and parameters were considered, it is still not exactly the same as the one applied in the simulations.Many other factors that affect the complexity are also missed or not as accurate as in the simulator, such as vehicle trajectory model, wind, communication, uncertainty and so on.The Fe 3 simulator was developed to address all these factors, whereas, the goal of this study is to identify quickly-computed metrics that can approximate the complexity as close as possible.From this perspective, the proposed complexity metric SC is the best compared to the other two.In addition, using the measurements from high-fidelity simulations as the ground truth, this work developed a mechanism of testing and evaluating any proposed complexity metric for future work. +IV. ConclusionIn UAS traffic management, a quick assessment of complexity for a given traffic scenario can help re-plan flights to alleviate traffic bottleneck and mitigate operation risks, and the traffic complexity measurement can also help categorize traffic scenarios for traffic management studies.In this work a complexity metric was constructed based on the number of potential conflicts weighted by the associated resolution costs.The cost associated with a conflict is calculated based on the corresponding conflict resolution maneuvers.A MILP-based optimization was formulated to obtain the conflict resolution maneuvers.To assess and compare complexity metrices, around 1,000 scenarios at different density level were generated.The Fe 3 simulator was used to run these scenarios and generate a metric measurement called the number of conflict resolution maneuvers for each scenario.The measurements from the simulations were then treated as the ground truth for the scenario complexity.Statistical tools were applied to examine the correlations for three metrics: the number of flights, the number of potential conflicts, and the proposed complexity metric.The analysis showed that the number of flights has much lower correlation with the scenario complexity than the other two according to the correlation coefficients calculated by both Pearson and ACE statistics methods.The ACE maximal correlation method shows that the proposed complexity metric has better correlation with the ground truth than the number of potential conflicts.With the simulation based measurements, future work will focus on investigating if combined simple features can better represent scenario complexity through regressions.Fig. 22Fig. 2 Using finite horizon based on conflict management structure +Fig. 33Fig. 3 Conflict resolution maneuvers in three different encounters +Fig. 6 Pearson 2 3 3 Fig. 7 Maximal62337Fig. 6 Pearson Correlation with the Number of Conflict Resolution Maneuvers + + +t=t s p t p t+1p t s p t e p t s p tt ee | • (t et s ), where m = i or j + + + + + + + + + Airspace Complexity and its Application in Air Traffic Management + + BSridhar + + + KSheth + + + SGrabbe + + + + Europe Air Traffic Management R&D Seminar + + 1998 + Orlando, FL + + + 2nd USA/ + Sridhar, B., Sheth, K., and Grabbe, S., "Airspace Complexity and its Application in Air Traffic Management," 2nd USA/Europe Air Traffic Management R&D Seminar, Orlando, FL, 1998. + + + + + Dynamic density: measuring and predicting sector complexity [ATC] + + PKopardekar + + + SMagyarits + + 10.1109/dasc.2002.1067920 + + + Proceedings. The 21st Digital Avionics Systems Conference + The 21st Digital Avionics Systems ConferenceIrvine, CA + + IEEE + 2002 + + + Kopardekar, P., and Magyarits, S., "Measurement and Prediction of Dynamic Density," The 21st Digital Avionics Systems Conference (DASC), Irvine, CA., 2002. + + + + + Dynamic Density and Complexity Metrics for Realtime Traffic Flow Management + + AMasalonis + + + MCallaham + + + CWanke + + + + Europe Air Traffic Management R&D Seminar + + 2003 + Budapest, Hungary + + + 5th USA/ + Masalonis, A., Callaham, M., and Wanke, C., "Dynamic Density and Complexity Metrics for Realtime Traffic Flow Management," 5th USA/Europe Air Traffic Management R&D Seminar, Budapest, Hungary, 2003. + + + + + Airspace Congestion Metrics + + DanielDelahaye + + + StéphanePuechmorel + + 10.1002/9781118743805.ch7 + + + Modeling and Optimization of Air Traffic + Napoli, Italy + + John Wiley & Sons, Inc. + 2000 + + + + 3rd USA/ + Delahaye, D., and Puechmorel, S., "Airspace Complexity: Towards Intrinsic Metrics," 3rd USA/Europe Air Traffic Management R&D Seminar, Napoli, Italy, 2000. + + + + + Air Traffic Complexity Map based on Non Linear Dynamical Systems + + DanielDelahaye + + + StephanePuechmorel + + + JohnHansman + + + JonathanHiston + + 10.2514/atcq.12.4.367 + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 12 + 4 + + 2004 + American Institute of Aeronautics and Astronautics (AIAA) + + + Delahaye, D., Puechmorel, S., Hansman, J., and Histon, J., "Air Traffic Complexity based on Non Linear Dynamical Systems," Air Traffic Control Quarterly, Vol. 12, No. 4, 2004, pp. 367-388. + + + + + Describing Air Traffic Complexity Using Mathematical Programming + + MariyaIshutkina + + + EricFeron + + + KarlBilimoria + + 10.2514/6.2005-7453 + + + AIAA 5th ATIO and16th Lighter-Than-Air Sys Tech. and Balloon Systems Conferences + + American Institute of Aeronautics and Astronautics + 2005 + + + Ishutkina, M. A., and Feron, E., "Describing Air Traffic Complexity Using Mathematical Programming," AIAA 5th Aviation, Technology, Integration, and Operations Conference (ATIO), 2005. + + + + + A Throughput Based Capacity Metric for Low-Altitude Airspace + + VishwanathBulusu + + + RajaSengupta + + + EricRMueller + + + MinXue + + 10.2514/6.2018-3032 + + + 2018 Aviation Technology, Integration, and Operations Conference + Atlanta, GA + + American Institute of Aeronautics and Astronautics + 2018 + + + Bulusu, V., Sengupta, R., Mueller, E. R., and Xue, M., "A Throughput-Based Capacity Metric for Low-altitude Airspace," AIAA Aviation Forum, Atlanta, GA., 2018. + + + + + Fe3: An Evaluation Tool for Low-Altitude Air Traffic Operations + + MinXue + + + JosephRios + + + JosephSilva + + + ZhifanZhu + + + AbrahamKIshihara + + 10.2514/6.2018-3848 + + + 2018 Aviation Technology, Integration, and Operations Conference + Atlanta, GA + + American Institute of Aeronautics and Astronautics + 2018 + + + Xue, M., Rios, J., Silva, J., Ishihara, A., and Zhu, Z., "Fe3: An Evaluation Tool for Low-Altitude Air Traffic Operations," AIAA Aviation Forum, Atlanta, GA., 2018. + + + + + An Efficient Method for Airspace Analysis and Partitioning Based on Equalized Traffic Mass + + AKlein + + + + Europe Seminar on Air Traffic Management Research and Development + + 2005 + Baltimore, MD + + + 6th USA/ + Klein, A., "An Efficient Method for Airspace Analysis and Partitioning Based on Equalized Traffic Mass," 6th USA/Europe Seminar on Air Traffic Management Research and Development, Baltimore, MD, 2005. + + + + + Flight Demonstration of Unmanned Aircraft System (UAS) Traffic Management (UTM) at Technical Capa... + + MJohnson + + + JLarrow + + 10.2514/6.2020-2851.vid + + 2018 + American Institute of Aeronautics and Astronautics (AIAA) + + + NASA + + + Tech. Rep. Technical Report + Johnson, M., and Larrow, J., "UTM Research Transition Team Sense and Avoid Working Group Technical Work Package #2: UTM Conflict Management Model," Tech. Rep. Technical Report, NASA, 2018. + + + + + A Well Clear Recommendation for Small UAS in High-Density, ADS-B-Enabled Airspace + + MatthewDuffield + + + TimWMclain + + 10.2514/6.2017-0908 + + + AIAA Information Systems-AIAA Infotech @ Aerospace + Grapevine, Texas + + American Institute of Aeronautics and Astronautics + 2017 + + + Duffield, M. O., and McLain, T. W., "A Well Clear Recommendation for Small UAS in High-Density ADS-B-Enabled Airspace," AIAA SciTech Forum, Grapevine, Texas, 2017. + + + + + Well-Clear Recommendation for Small Unmanned Aircraft Systems Based on Unmitigated Collision Risk + + AndrewWeinert + + + ScotCampbell + + + AdanVela + + + DieterSchuldt + + + JoelKurucar + + 10.2514/1.d0091 + + + Journal of Air Transportation + Journal of Air Transportation + 2380-9450 + + 26 + 3 + + 2018 + American Institute of Aeronautics and Astronautics (AIAA) + + + Weinert, A., Campbell, S., Vela, A., Schuldt, D., and Kurucar, J., "Well-Clear Recommendation for Small Unmanned Aircraft Systems Based on Unmitigated Collision Risk," AIAA Journal of Air Transportation, Vol. 26, No. 3, 2018. + + + + + Aircraft trajectory planning with collision avoidance using mixed integer linear programming + + ARichards + + + JPHow + + 10.1109/acc.2002.1023918 + + + Proceedings of the 2002 American Control Conference (IEEE Cat. No.CH37301) + the 2002 American Control Conference (IEEE Cat. No.CH37301)Anchorage, AK + + IEEE + 2002 + + + Richards, A., and How, J. P., "Airspace Trajectory Planning with Collision Avoidacen Using Mixed Integer Linear Programming," Proceedings of the 2002 American Control Conference, Anchorage, AK, 2002. + + + + + Cooperative path planning for multiple UAVs in dynamic and uncertain environments + + JSBellingham + + + MTillerson + + + MAlighanbari + + + JPHow + + 10.1109/cdc.2002.1184270 + + + Proceedings of the 41st IEEE Conference on Decision and Control, 2002. + the 41st IEEE Conference on Decision and Control, 2002Las Vegas, NV + + IEEE + 2002 + + + Bellingham, J. S., Tillerson, M., Alighanbari, M., and How, J. P., "Cooperative Path Planning for Multiple UAVs in Dynamic and Uncertain Environments," Proceedings of the 41st IEEE Conference on Decision and Control, Las Vegas, NV, 2002. + + + + + Receding horizon path planning with implicit safety guarantees + + TSchouwenaars + + + JHow + + + EFeron + + 10.23919/acc.2004.1384742 + + + Proceedings of the 2004 American Control Conference + the 2004 American Control ConferenceBoston, MA + + IEEE + 2004 + + + Schouwenaars, T., How, J., and Feron, E., "Receding Horizon Path Planning with Implicit Safety Guarantees," Proceedings of the 2004 American Control Conference, Boston, MA, 2004. + + + + + MILP optimal path planning for real-time applications + + CSMa + + + RHMiller + + 10.1109/acc.2006.1657504 + + + 2006 American Control Conference + Minneapolis, MN + + IEEE + 2006 + + + Ma, C. S., and Miller, R. H., "MILP Optimal Path Planning for Real-Time Applications," Proceedings of the 2006 American Control Conference, Minneapolis, MN, 2006. + + + + + Optimal search mission with Unmanned Aerial Vehicles using Mixed Integer Linear Programming + + ErikJForsmo + + + EstenIGrotli + + + ThorIFossen + + + TorAJohansen + + 10.1109/icuas.2013.6564697 + + + 2013 International Conference on Unmanned Aircraft Systems (ICUAS) + Atlanta, GA + + IEEE + 2013 + + + Forsmo, E. J., Grotli, E. I., Fossen, T. I., and Johansen, T. A., "Optimal Search Mission with Unmanned Aerial Vehicles using Mixed Integer Linear Programming," 2013 International Conference on Unmanned Aircraft Systems (ICUAS), Atlanta, GA, 2013. + + + + + Dynamic optimal UAV trajectory planning in the National Airspace System via mixed integer linear programming + + MohammadrezaRadmanesh + + + ManishKumar + + + AlirezaNemati + + + MohammadSarim + + 10.1177/0954410015609361 + + + Proceedings of the Institution of Mechanical Engineers, Part G: Journal of Aerospace Engineering + Proceedings of the Institution of Mechanical Engineers, Part G: Journal of Aerospace Engineering + 0954-4100 + 2041-3025 + + 230 + 9 + + 2016 + SAGE Publications + + + Radmanesh, M., Kumar, M., Nemati, A., and Sarim, M., "Dynamic Optimal UAV Trajectory Planning in the National Airspace System via Mixed Integer Linear Programming," Journal Aerospace Engineering, Vol. 230, No. 9, 2016, pp. 1668-1681. + + + + + Exact and Approximate Solving of the Aircraft Collision Resolution Problem via Turn Changes + + AntonioAlonso-Ayuso + + + LaureanoFEscudero + + + FJavierMartín-Campo + + 10.1287/trsc.2014.0557 + + + Transportation Science + Transportation Science + 0041-1655 + 1526-5447 + + 50 + 1 + + 2014 + Institute for Operations Research and the Management Sciences (INFORMS) + + + Alonso-Ayuso, A., Escudero, L. F., and Martin-Campo, F. J., "Exact and Approximate Solving of the Aircraft Collision Resolution Problem via Turn Changes," Transportation Science, Vol. 50, No. 1, 2014, pp. 1-12. + + + + + Flight Formation of Quad-Copters in Presence of Dynamic Obstacles Using Mixed Integer Linear Programming + + MohammadrezaRadmanesh + + + AlirezaNemati + + + MohammadSarim + + + ManishKumar + + 10.1115/dscc2015-9931 + + + Volume 1: Adaptive and Intelligent Systems Control; Advances in Control Design Methods; Advances in Non-Linear and Optimal Control; Advances in Robotics; Advances in Wind Energy Systems; Aerospace Applications; Aerospace Power Optimization; Assistive Robotics; Automotive 2: Hybrid Electric Vehicles; Automotive 3: Internal Combustion Engines; Automotive Engine Control; Battery Management; Bio Engineering Applications; Biomed and Neural Systems; Connected Vehicles; Control of Robotic Systems + Columbus, OH + + American Society of Mechanical Engineers + 2015 + + + Radmanesh, M., Nemati, A., Sarim, M., and Kumar, M., "Flight Formation of Quad-Copters In Presence of Dynamic Obstacles Using Mixed Integer Linear Programming," Proceedings of the ASME 2015 Dynamic Systems and Control Conference, Columbus, OH, 2015. + + + + + Mixed-integer nonlinear programming for aircraft conflict avoidance by sequentially applying velocity and heading angle changes + + SoniaCafieri + + + RiadhOmheni + + 10.1016/j.ejor.2016.12.010 + + + European Journal of Operational Research + European Journal of Operational Research + 0377-2217 + + 260 + 1 + + 2017 + Elsevier BV + + + Cafieri, S., and Omheni, R., "Mixed-integer Nonlinear Programming For Aircraft Conflict Avoidance by sequentially Applying Velocity and Heading Angle Changes," European Journal of Operational Research, Vol. 260, No. 1, 2017, pp. 283-290. + + + + + Constraint Integer Programming: A New Approach to Integrate CP and MIP + + TobiasAchterberg + + + TimoBerthold + + + ThorstenKoch + + + KatiWolter + + 10.1007/978-3-540-68155-7_4 + + + Integration of AI and OR Techniques in Constraint Programming for Combinatorial Optimization Problems + + LPerron + + + MATrick + + Berlin Heidelberg; Berlin, Heidelberg + + Springer Berlin Heidelberg + 2008 + + + + Constraint Integer Programming: A New Approach to Integrate CP and MIP + Achterberg, T., Berthold, T., Koch, T., and Wolter, K., "Constraint Integer Programming: A New Approach to Integrate CP and MIP," Integration of AI and OR Techniques in Constraint Programming for Combinatorial Optimization Problems, edited by L. Perron and M. A. Trick, Springer Berlin Heidelberg, Berlin, Heidelberg, 2008, pp. 6-20. + + + + + SCIP: solving constraint integer programs + + TobiasAchterberg + + 10.1007/s12532-008-0001-1 + + + + Mathematical Programming Computation + Math. Prog. Comp. + 1867-2949 + 1867-2957 + + 1 + 1 + + 2009 + Springer Science and Business Media LLC + + + Achterberg, T., "SCIP: solving constraint integer programs," Mathematical Programming Computation, Vol. 1, No. 1, 2009, pp. 1-41. doi:10.1007/s12532-008-0001-1, URL https://doi.org/10.1007/s12532-008-0001-1. + + + + + Statistical computing + + TeamCore + + 10.1201/b17972-15 + + + + Statistical Computing in Nuclear Imaging + Vienna, Austria + + CRC Press + 2016 + + + + R Core Team, R: A Language and Environment for Statistical Computing, R Foundation for Statistical Computing, Vienna, Austria, 2016. URL http://www.R-project.org/, ISBN 3-900051-07-0. + + + + + Estimating Optimal Transformations for Multiple Regression and Correlation + + LeoBreiman + + + JeromeHFriedman + + 10.1080/01621459.1985.10478157 + + + Journal of the American Statistical Association + Journal of the American Statistical Association + 0162-1459 + 1537-274X + + 80 + 391 + + 1985 + Informa UK Limited + + + Beriman, L., and Friedman, J. H., "Estimating Optimal Transformations for Multiple Regression and Correlation," Journal of the American Statistical Association, Vol. 80, No. 391, 1985, pp. 580-598. doi:10.1080/01621459.1985.10478157. + + + + + Supplementary file 1. Multiple regression of heritability on various gene features. + + PSpector + + + JFriedman + + + RTibshirani + + + TLumley + + + SGBaron + + 10.7554/elife.35471.035 + + null + eLife Sciences Publications, Ltd + + + see the 'Changelog' file (in the package source + Spector, P., Friedman, J., Tibshirani, R., Lumley, T., and andJonathan Baron, S. G., ACE and AVAS for Selecting Multiple Regression Transformations, 2016. R package version 1.4.1 -For new features, see the 'Changelog' file (in the package source). + + + + + + diff --git a/file811.txt b/file811.txt new file mode 100644 index 0000000000000000000000000000000000000000..9ef07ae1111d8d3e560839f6b82807b99f2329c4 --- /dev/null +++ b/file811.txt @@ -0,0 +1,517 @@ + + + + +I. IntroductionContinuous Descent Arrival (CDA) [1][2][3] refers to a procedure that allows aircraft to approach an airport from cruise altitude at near idle engine power.Compared with typical air traffic arrival procedures, which can include multiple level flight segments prior to reaching the crossing altitude at the arrival fix, CDA reduces fuel consumption, emissions, and noise, thus providing both economic and environmental benefits.Continuous descent trajectories are typically computed by the Flight Management System (FMS) and are executed automatically by an autopilot or flown manually by the pilot following flight director guidance.At low traffic levels controllers can safely accommodate pilot requests for CDA's.However, as the traffic density increases controllers find it increasingly difficult to handle CDA requests without the help of decision support tools.The Efficient Descent Advisor (EDA) 3,4 is a tool that has been designed to help controllers manage CDA's safely even during busy traffic periods.EDA accomplishes this by computing CDA solutions that conform with time-based metering schedules computed by Traffic Management Advisor (TMA) 5,6 for balancing traffic demand and capacity.By following advisories generated by EDA controllers are able to issue CDA clearances while maintaining high traffic flow and avoiding frequent conflicts.A key element in EDA is a function for accurately modeling FMS-generated descent profiles.This function is performed by the Trajectory Synthesizer (TS), which is a component shared between TMA and EDA.This paper focuses on methods to improve the prediction accuracy of TS.A sensitivity analysis was first conducted to gain an understanding of error sources and their impact on prediction accuracy.The significance of error sources was prioritized based on error magnitudes observed in actual operations.Then methods of introducing updated descent calibrated air speed (CAS) for non-metered flights and a thrust correction factor for all flights were proposed for improving descent trajectory predictions.It should be noted that the application of adjusting CAS is limited to the uncontrolled arrivals or any prediction computed prior to any EDA control.Trajectories from flights that executed the continuous descents into the Dallas Fort-Worth airport were collected for examining the improvements.It is shown that the proposed methods have the potential to reduce the descent trajectory prediction errors to acceptable ranges, thereby improving the prediction accuracy of TS and the performance of EDA. +II. Model of Continuous Descent TrajectoryTypical vertical profiles of CDA's are shown in Fig. 1.Fig. 1(a) and 1(b) present the spatial and temporal altitude profiles.And Fig. 1(c) and 1(d) show the corresponding speed profiles.During continuous descent, an aircraft starts from its cruise altitude and descends with a constant Mach (if the desired descent CAS is greater than the final cruise CAS) until it reaches the pre-defined CAS (point B in Fig. 1(a)).After that, the aircraft continues its descent with the CAS (from point B to C in Fig. 1).Before it reaches the meter fix, the aircraft decelerates to the final CAS.Typically, an aircraft is required to cross the meter fix at an altitude of 10,000 feet with an indicated airspeed of 250 knots, while in Fig. 1 11,000 feet was required.A trajectory model for calculating descent profiles was developed in the 1980's. 7,8 t typically consists of three segments, an acceleration/deceleration segment to a specified Mach number, followed by a constant CAS segment, and then a deceleration segment.This model not only provides convenience for pilots and controllers but also simplifies the equations of motion to first order differential equations:8 VT = T -D m -g sin γ a -uw cos γ a (1) ẋ = V T cos γ a + u w (2) ż = V T sin γ a(3)where T and D are thrust and drag, respectively.V T is the true airspeed, γ a is the aerodynamic flight path angle, and u w is the horizontal component of wind and g is the acceleration of gravity.It is assumed that there is no vertical wind.x and z are the horizontal and vertical axis in an earth fixed coordinate system, respectively.From Eqn. 1, γ a for the constant Mach and constant CAS segments can be expressed as follows, respectively:γ a(constantM ) = T -D m [g + V T (M da dz + du w dz )] -1 (4) γ a(constantCAS) = T -D m [g + V T ( dV T (V CAS , z) dz + du w dz )] -1 (5)where a is the speed of sound, M is the Mach number, and V CAS is the indicated airspeed.For idle-thrust descent, γ a can be obtained by setting T to idle value.Forward and backward integrations are then used for the first and third segments until the conditions of constant indicated airspeed segment are met.Then the complete descent trajectory can be obtained by integrating the segment of constant indicated airspeed.More details of descent calculation can be found in previous works. 7,8 +III. Sensitivity Study of Trajectory SynthesizerSeveral trajectory sensitivity studies [9][10][11][12][13][14] have been conducted in past years.Most of those examine sensitivities for error in arrival times and along-track position.The purpose of this section is to identify the major factors that affect the prediction of descent trajectories in TS.To facilitate comparisons between trajectories, the descent trajectories are usually treated as two dimensional with horizontal distance to the meter fix and altitude as the two axes.In this section, the distance from the top of descent (TOD) to the meter fix is called TOD distance, and the time that the aircraft flies from the reference point to the meter fix is recorded as meter-fix crossing time.In this section, three typical error sources -descent weight, wind error, and descent speed -are examined.Without loss of generality B737-800 is used as an example. +A. Descent WeightAircraft weight is one of the major concerns in trajectory prediction, especially in the case that datalink is not available or airlines choose not to share the weight.Fortunately, the range of descent weights is much narrower than the range of take-off weights.In nominal situation, assuming descent fuel consumption is negligible, the minimum descent weight can be estimated using Eqn.6, where W OW E is the Operating Weight Empty (OWE), and F res represents minimum reserve fuel required by Federal Aviation Administration (FAA).The reserve fuel is for cruising to the alternate airport and for 45 minutes of airborne holding.Generally the reserve fuel is set to eight percent of the takeoff weight. 15,16 ssuming no extra fuel would be carried by the aircraft, the maximum descent weight is estimated by Eqn. 7, where W M P LD is the maximum payload allowed for the aircraft, and W M LW is the maximum design landing weight which should not be exceeded.Using B737-800 as an example, the minimum descent weight of B737-800 is about 106,457 lbs, and the maximum descent weight is 146,300 lbs, which corresponds to a load factor of 100%.If an aircraft is executing a long range flight, the maximum possible descent weight should be further reduced, which means the load factor will be less than 100%.Currently, in TS, the default descent weight of B737-800 is 126,720 lbs, corresponding to a load factor of 50%.Meter Fix Figure 2 shows the different descent trajectories associated with different descent weights for CAS 280 and standard day without wind.The middle trajectory corresponds the typical descent weight used by TS, the steep one is based on minimum descent weight, and the shallow profile results from maximum descent weight.It is noted that heavily loaded aircraft fly shallower descents with idle thrust than do lightly loaded aircraft.Figure 3(a) shows the difference in TOD range can be as large as 16 nmi from empty payload (load factor 0) to full payload (load factor 100%), which means TOD will be 4.1 nmi further from meter fix for each 10,000 lbs increase in weight.The default weight in the TS is shown as a red dot in the figure, which denotes load factor 50%.The difference in meter-fix crossing time over the weight range is small as shown in Fig. 3(b).To serve the purpose of this paper, it is assumed that the requirement of meter-fix crossing time accuracy is 30 seconds.For the entire range of descent weights, the difference is only about 20 seconds.For each 10,000 lbs increase in weight, the increase in meter-fix crossing time is less than 5 seconds.According to the analysis conducted by the International Air Transport Association (IATA), the average load factor for commercial airlines is about 70%.This further narrows the range of weight and reduces the difference in TOD distance to around 3 nmi.Thus, the prediction errors might be tolerable given the fact that the aircraft descent weight can be well estimated.Since even in actual operations, it is unlikely to obtain aircraft weight information, it is useful to compute accurate descent prediction without weights reported from pilots or airlines.Wind speeds can be as high as 150 knots at high altitudes and are therefore critical to the trajectory calculation.Wind speed error exists due to the inaccurate wind forecast.Although the magnitude of wind speed is high, the error is usually less than 10 knots. 14,17 igure 4 shows the trajectories with different wind speeds.The steep trajectory results from a strong head wind, while the shallow trajectory corresponds to a strong tail wind.Strong head wind shortens the TOD distance to the meter fix. Figure 5(a) presents the different TOD distances due to wind errors.It can be seen that the shift of TOD distance caused by wind speed error is small.The error of 10 knots only corresponds to 1.3 nmi difference in TOD distance.W min = W OW E + F res(6)W max = min{W OW E + F res + W M P LD , W M LW }(7) +Top of Descent +B. Wind Speed +C. Descent SpeedDescent speed is another important factor for descent trajectory calculation.Usually, descent speeds are selected by airlines or pilots for those un-delayed (non-metered) flights.Fig. 6 presents the impacts of descent speed on descent trajectory, and Fig. 7(a) and 7(b) show TOD distances and meter-fix crossing times corresponding to different descent speeds.It was found that descent CAS has the dominant impact due to the length of the segment.A 10 knots difference in CAS causes the differences of 3.2 nmi in TOD distance and 18.1 seconds in meter-fix crossing time.Cruise Mach number has minimal effect of 2.4 seconds and 0.5 nmi per 0.01 Mach.It is noticed that descent CAS in actual operations is different from the default CAS in TS.For instance, the trajectory synthesizer of EDA uses a default descent speed of 280 knots for B737-800, a whereas flights from the actual operational data, which will be discussed in next section, descended at 310 knots.This mismatch alone will cause 9 nmi error in TOD distance and 51 seconds error in meter-fix crossing time, which makes the trajectory prediction unacceptable.Based on this finding, it is strongly recommended that pilot-preferred descent speeds be acquired prior to descent in order to make initial trajectory predictions acceptable.a EDA uses the default speed for the initial, un-delayed trajectory calculation only, once EDA advises a speed profile and controller accepts, descent CAS is no longer an error source +IV. Methods for Improving Prediction AccuracyBased on the impact on the error magnitude and the accessability of the information, the first recommendation for improving un-delayed/initial prediction is to acquire the descent CAS prior to descent.Here it is proposed that the intended descent CAS can be acquired via down linked communication from the aircraft to the air traffic control center.If the intended CAS can NOT be down-linked, the default descent speeds in EDA need to be revised based on the actual operations.Furthermore, a thrust correction factor, which is dependent on aircraft type and airlines, is proposed to mitigate the TOD prediction errors.The thrust correction factor is referred to as τ in Eqns.8 and 9.It is defined as a function of nominal aircraft weight W , and a dimensionless tuning parameter c τ is selected to minimize prediction errors. +VT =T + τ -D m -g sin γ a -uw cos γ a (8)τ = c τ • W(9)Before applying the thrust correction, the impact of thrust changes on the trajectories needs to be understood.Figure 8 shows the impact of a 1.5% thrust correction factor on descent trajectory.A positive thrust correction factor makes the shallow descent trajectory, and the dash-dot line results from a negative thrust correction factor.Figure 9(a) and 9(b) plot the change in TOD location and meter-fix crossing time over a range of ±1.5% thrust correction, which roughly corresponds to ±15 nmi in TOD and ±20 seconds for meter-fix crossing time, respectively. +V. Experiments and ResultsIn order to examine above methods, a limited set of continuous descent trajectories flown by aircraft during regular revenue flights into the Dallas-Fort Worth Airport were recorded and analyzed.Based on the request from NASA, on Feb. 25-26, 2011, controllers acquired FMS-calculated top-of-descent ranges, crossing times and intended descent CAS from pilots who were willing to fly CDA's during light traffic.Prior to such requests, controllers determined that flights had enough space for executing CDA's.There were no pilot briefing or training prior to the short experiment.Meanwhile, the flight track information including aircraft position, altitude, ground speed, and heading was recorded at the NASA's North Texas Experimental Facility, co-located at the Fort Worth Center.The associated wind forecasts were also recorded.Using one flight as an example, here is how the experiment was conducted: On Feb. 26, 2011, at 22:29 GMT (4:29 pm local time), per controller's request, the pilot reported that they were 10 nmi prior to the FMS-calculated TOD, their intended descent CAS was 261 knots, and their arrival time to meter fix YEAGR was estimated to be 22:57 GMT.At 22:46 GMT, the flight was cleared to descend to YEAGER with final speed 250 knots and altitude 9,000 ft, respectively.Overall 14 flights from two major airlines -Airline A and S -were collected and analyzed.They comprised four B737-700 from Airline S, and seven B737-800, two B757-200, and one B737-700 from Airline A. +A. Speed and thrust correctionsIn TS, the default descent CAS for a B737-700 is 280 knots, quite different from 261 knots reported by pilots in Airline S cases.Thus, speed correction was conducted in TS.Fig. 10(a) and 10(b) show spatial and temporal profiles calculated by default TS for flight S101.And Fig. 11(a) and 11(b) show spatial and temporal profiles computed by TS after the speed correction.The blue curves are radar-recorded actual descent trajectories and associated blue dots are the positions where pilots reported their TOD, time, and speed.Blue diamonds and triangles are TOD positions and meter-fix crossing times, respectively, calculated by the onboard FMS and reported by pilots.Yellow diamonds and triangles are actual TOD positions and actual meter-fix crossing time retrieved from radar-recorded track information.Black curves are trajectory predictions as calculated by TS.Temporal profiles show that before speed correction, there is about a 50 second difference between TS prediction (the low end of black curve) and actual crossing time, and actual and FMS crossing times are very close.After speed correction, the difference between TS calculation and actual becomes negligible for this particular set of flights and aircraft type.As shown in Fig. 10(b) and 11(b), after speed correction there still exists a significant gap between the TS-calculated TOD locations and FMS/actual TOD locations.Thus a unified thrust correction factor must be applied on all four B737-700s from Airline S. The thrust correction parameter τ was manually identified as 0.7%.Figure .12(a) and 12(b) present the temporal and spatial profiles after introducing thrust correction for flight S101, which is similar to other Airline S flights as well.It can be seen that the TS prediction of TOD location has been improved significantly after adjustment.Table 1 lists the statistics for all four Airline S flights using the same thrust correction factor τ .The prediction accuracy desired for EDA operations is for TOD and meter-fix crossing time errors to stay within 5 nmi and 30 seconds, respectively.It appears that these error rates are achievable with the proposed methods. +B. Early descent phenomenonAlthough five out of ten Airline A flights behaved similarly to Airline S flights and TS predictions for these flights can be significantly improved by same means, there exists an interesting phenomenon -"early descent" for five Airline A flights. Figure 13 and Fig. 14 present such an example, which shows the comparisons before and after applying thrust correction for a B737-800 of Airline A. The constant thrust correction factor is 1.6%, which was set for all seven Airline A B737-800s.From the figure, it is noticed that although the TS predicted TOD does not match the actual TOD it is a good match for FMS TOD (the blue diamond).Apparently, pilots didn't follow their FMS calculations and chose to descend almost 30 nmi earlier than FMS TOD location.At the end of the early descent segment, the descent trajectory merges with the FMS calculated idle thrust trajectory.Without a debriefing of the pilots after the experiments, the exact reasons for early descents could not be ascertained.However, according to FMS experts, pilots frequently choose the "descend now" option on the FMS before the aircraft reaches the FMS calculated TOD location.The primary reason for pilots selecting "descend now" instead of flying the FMS-calculated idle-thrust descent is to make passengers comfortable.The early descent phenomenon suggest the need to coordinate the descent procedures of airlines with the ground-based EDA tool.Then, the TS in EDA can be adapted to correct for early descent procedure, thereby improving prediction accuracy.Table 2 shows the overall comparisons between TS prediction and actual track.In the table, the errors shown in parentheses are the errors between the TS calculation and the onboard FMS (instead of actual).(Recall that, in early descent cases, TS prediction can still match on board FMS calculations.)The table shows that with speed correction, the meter-fix crossing time errors can be halved to 10 seconds with a standard deviation of 9 seconds.With further thrust correction, the TOD location errors can be reduced from 19 nmi to 1.7 nmi (if taking away the impact of early descents), which is similar to what was found in Airline S cases. +D. DiscussionAlthough it was proposed above that a constant thrust correction factor should be pre-defined based on aircraft type and airlines/operator, the thrust correction can be applied in two different ways: 1) If the FMScalculated TOD location and meter-fix crossing time can be down-linked from an aircraft through Data-link, then in real-time for an individual flight, a thrust correction factor may be calculated based on the downlinked TOD and crossing time.Therefore, an accurate 4D descent trajectory can be calculated by TS, which is required for reliable conflict detection and resolution when flying continuous descent approaches; 2) If the FMS-calculated TOD location and meter-fix crossing time can NOT be down-linked from an aircraft, the thrust correction factor can be used in the way proposed in previous sections: For a given type of aircraft and a given airlines/operator, determine a constant thrust correction factor from analysis of a set of previously recorded descent trajectories.Both methods can improve TS prediction accuracy for continuous descents.However, down-linked information has the advantage in accuracy.It has to be acknowledged that the small sample set is not statistically significant.Additional modification needs to be made in TS, so a large amount of on-line experiments can be conducted in TS for validating the proposed methods.Regarding the early descents, while the TS is architected to model multiple descent segments based on different parameters (e.g.thrust, fixed flight path angle, or rate of descent), efforts are needed to identify appropriate parameters for early descents and to develop methods to obtain such intent. +VI. ConclusionIn order to handle continuous descent procedures during busy traffic conditions, controllers using decision support tools, such as EDA, require accurate prediction of descent trajectories.The prediction function is performed by the trajectory synthesizer (TS).This paper investigated the sensitivity of TS prediction accuracy to various parameters and proposes methods for improving its accuracy based on actual operational flight data.During the sensitivity analysis, the significance of error sources was prioritized based on error magnitudes in actual operations.Assuming that the prediction accuracy requirements of TOD location and meter-fix crossing time are 5 nmi and 30 seconds, respectively, it was found that for the range of expected wind speed and weight errors prediction errors remained within acceptable limits.Since aircraft descent weight lies within a relatively narrow range in practice, it may be acceptable to use an estimated descent weight if the actual weight is not available prior to descent.The FMS-selected descent speed has a significant impact on arrival time prediction for those non-metered flights and should be correctly entered into TS for each aircraft-FMS combination prior to descent.It is suggested to acquire the FMS-selected descent speed for trajectory prediction prior to any EDA speed assignment.At least, the nominal descent speeds in TS need to be calibrated based on actual airline policies.As actual descent trajectories were typically shallower than TS predicted trajectories even after the descent speed correction, a thrust correction factor was introduced to achieve the desired prediction accuracy.Data were collected from a limited set of flights that executed the continuous descents into DFW airport.Analysis shows that introducing a constant thrust correction factor and intended descent CAS correction has the potential to reduce the descent trajectory prediction errors to acceptable/desired ranges.The descent CAS mainly corrected the meter-fix crossing time and the thrust correction factor mainly improved the TOD prediction.These corrections could be achieved by the aircraft down-linking critical FMS parameters prior to the descent.If only descent CAS can be down-linked, a thrust correction factor parameter can be identified based on aircraft type and airlines/operator for future prediction.If TOD locations and meter-fix crossing time can also be down-linked from an aircraft, a thrust correction factor could be computed in real time for accurate 4D trajectory calculations.In both cases, the accuracy of descent trajectories required for reliable conflict detection and resolution will be significantly improved.It is suggested that additional modification can be made in TS such that a large amount of on-line experiments can be carried for further validation.It was also determined, in order to model early descents, appropriate parameters need to be identified for multiple descent segments in the TS model and methods should be developed to obtain the intents of early descents. +VII. AcknowledgementFigure 1 .1Figure 1.Continuous descent arrival: (a) Spatial profile of altitude (b) Temporal profile of altitude (c) Indicated airspeed (d) Mach Number +Figure 2 .Figure 3 .23Figure 2. Impacts of weight on descent trajectory +Figure 4 .Figure 5 .45Figure 4. Impacts of wind on descent trajectory +Figure 5 (5b) shows the effect of wind error on arrival time, typically 11 seconds for every 10 knots. +Figure 6 .Figure 7 .67Figure 6.Impacts of descent speed on descent trajectory +Figure 8 .8Figure 8. Impacts of thrust correction on descent trajectory +Figure 9 .9Figure 9. Impact of thrust correction on (a) TOD distance (b) Meter-fix crossing time. +Figure 10 .Figure 11 .Figure 12 .101112Figure 10.Flight S101 actual trajectory and default TS calculation (a) Spatial profiles (b) Temporal profiles +Figure 13 .Figure 14 .1314Figure 13.Early descent phenomenon: (a) Spatial profiles with NO thrust correction (b) Spatial profiles with thrust correction +Table 1 .1TS predictions errors for Airline S B737-700TOD Error (nmi) Time Error (sec)averagestd.averagestd.Default TS20.316.138.913.0Speed correction5.63.15.42.7Thrust correction1.81.59.47.2 +Table 2 .2TS predictions errors for all flightsTOD Error (nmi) Time Error (sec)averagestd.averagestd.Default TS19.012.222.115.5Speed correction16.111.710.19.3Thrust correction 7.3(1.7) 8.1(1.4)7.87.1 + + + + +The authors gratefully acknowledge the contribution of Mr. Paul Borchers of NASA and Mr. Keenan Roach of University of California Santa Cruz at NASA/FAA Northern Taxes Research Station.They coordinated with air traffic controllers at DFW airport and collected actual operations data for the flights that executed continuous descent arrivals at the airport. + + + + + + + + + Development of an advanced continuous descent concept based on a 737 simulator + + LRAnderson + + + AWWarren + + 10.1109/dasc.2002.1067907 + + + Proceedings. The 21st Digital Avionics Systems Conference + The 21st Digital Avionics Systems Conference + + IEEE + null + 737 + + + Anderson, L. R. and Warren, A. W., "Development of an Advanced Continuous Descent Concept Based on a 737 + + + + + The 21st Digital Avionics Systems Conference [front matter] + + Simulator + + 10.1109/dasc.2002.1052911 + + + Proceedings. The 21st Digital Avionics Systems Conference + The 21st Digital Avionics Systems Conference + + IEEE + 2002 + 1 + + + + Simulator," Proceedings of the 21st Digital Avionics System Conference, Vol. 1, 2002, pp. 1E5-1-1e5-4. + + + + + Community Noise Reduction Using Continuous Descent Approach: A Demonstration Flight Test at Louisville + + KevinElmer + + + JoesephWat + + + BelurShivashankara + + + AnthonyWarren + + + Kwok-OnTong + + + John-PaulClarke + + 10.2514/6.2003-3277 + + + 9th AIAA/CEAS Aeroacoustics Conference and Exhibit + Reston, CA + + American Institute of Aeronautics and Astronautics + 2003 + + + Elmer, K., Wat, J., Shicashankara, B., Clarke, J.-P., Tong, K., Brown, J., and Warren, A., "Community Noise Reduc- tion Using Continuous Descent Approach: A Demonstration Flight Test at Louisville," Proceedings of the 9th AIAA/CEAS Aeroacoustics Conference and Exhibit, Reston, CA, 2003. + + + + + Design and Development of the En Route Descent Advisor (EDA) for Conflict-Free Arrival Metering + + RichardCoppenbarger + + + RichardLanier + + + DougSweet + + + SusanDorsky + + 10.2514/6.2004-4875 + + + AIAA Guidance, Navigation, and Control Conference and Exhibit + Providence, RI + + American Institute of Aeronautics and Astronautics + August 2004 + + + Coppenbarger, R. A., Lanier, R., and Sweet, D., "Design and Development of the En-Route Descent Advisor for Conflict Free Arrival Metering," AIAA Guidance, Navigation, and Control Conference, Providence, RI, August 2004. + + + + + Development and Testing of Automation for Efficient Arrivals in Constrained Airspace + + RACoppenbarger + + + GDyer + + + MHayashi + + + RLanier + + + LStell + + + DSweet + + + + 27th International Congress of the Aeronautical Sciences + Nice, France + + September 2010 + + + Coppenbarger, R. A., Dyer, G., Hayashi, M., Lanier, R., Stell, L., and Sweet, D., "Development and Testing of Automa- tion for Efficient Arrivals in Constrained Airspace," 27th International Congress of the Aeronautical Sciences, Nice, France, September 2010. + + + + + The Traffic Management Advisor + + WilliamNedell + + + HeinzErzberger + + + FrankNeuman + + 10.23919/acc.1990.4790788 + + + 1990 American Control Conference + San Diego, CA + + IEEE + May 1990 + + + Nedell, W. and Erzberger, H., "The Traffic Management Advisor," American Control Conference, San Diego, CA, May 1990. + + + + + Design and Operational Evaluation of the Traffic Management Advisor at the Fort Worth Air Route Traffic Control Center + + HNSwenson + + + THoang + + + SEngelland + + + DVincent + + + TSanders + + + BSanford + + + KHeere + + + + 1st USA/Europe Air Traffic Managemet R&D Seminar + + June 1997 + Saclay, France + + + Swenson, H. N., Hoang, T., Engelland, S., Vincent, D., Sanders, T., Sanford, B., and Heere, K., "Design and Operational Evaluation of the Traffic Management Advisor at the Fort Worth Air Route Traffic Control Center," 1st USA/Europe Air Traffic Managemet R&D Seminar , Saclay, France, June 1997. + + + + + Time Controllerd Descent Guidance Algorithm for Simulation of Advanced ATC systems + + HQLee + + + HErzberger + + + + NASA Ames Research Center + + August 1983 + Moffett Field, CA + + + Tech. Rep. NASA Technical Memorandum 84373 + Lee, H. Q. and Erzberger, H., "Time Controllerd Descent Guidance Algorithm for Simulation of Advanced ATC systems," Tech. Rep. NASA Technical Memorandum 84373, NASA Ames Research Center, Moffett Field, CA, August 1983. + + + + + Ground based concept for time control of aircraft entering the terminal area + + HErzberger + + + JChapel + + 10.2514/6.1985-1888 + + + Guidance, Navigation and Control Conference + Snowmass, CO + + American Institute of Aeronautics and Astronautics + August 1985 + + + + Erzberger, H. and Chapel, J., "Ground Based Concept for Time Control of Aircraft Entering the Terminal Area," AIAA Guidance, Navigation and Control Conference, Snowmass, CO, 19-21 August 1985. + + + + + Sensitivity of Trajectory Prediction in Air Traffic Management + + MichaelRJackson + + + YiyuanJZhao + + + RhondaASlattery + + 10.2514/2.4398 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 22 + 2 + + 1999 + American Institute of Aeronautics and Astronautics (AIAA) + + + Jackson, M. R., Zhao, Y. J., and Slattery, R. A., "Sensitivity of Trajectory Prediction in Air Traffic Management," Journal of Guidance, Control, and Dynamics, AIAA, Vol. 22, No. 2, 1999, pp. 219-228. + + + + + Flight Test Results: CTAS and FMS Cruise/Descent Trajectory Prediction Accuracy + + SMGreen + + + MPGrace + + + DHWilliams + + + + Europe Air Traffic Management R&D Seminar + + + June 2000 + Napoli, Italy + + + 3rd USA/ + Green, S. M., Grace, M. P., and Williams, D. H., "Flight Test Results: CTAS and FMS Cruise/Descent Trajectory Prediction Accuracy," 3rd USA/Europe Air Traffic Management R&D Seminar , Napoli, Italy, 13-16 June 2000. + + + + + Field evaluation of Descent Advisor trajectory prediction accuracy + + StevenGreen + + + RobertVivona + + 10.2514/6.1996-3764 + + + Guidance, Navigation, and Control Conference + San Diego, CA + + American Institute of Aeronautics and Astronautics + July 1996 + + + + Green, S. M. and Vivona, R. A., "Field evaluation of Descent Advisor trajectory prediction accuracy," AIAA Guidance, Navigation and Control Conference, San Diego, CA, 29-31, July 1996. + + + + + Trajectory Modeling Accuracy for Air Traffic Management Decision Support Tools + + SMondoloni + + + MPaglione + + + SGreen + + + + 23rd International Congress of the Aeronautical Sciences + Totonto + + September 2002 + p. ON + + + Mondoloni, S., Paglione, M., and Green, S., "Trajectory Modeling Accuracy for Air Traffic Management Decision Support Tools," 23rd International Congress of the Aeronautical Sciences, Totonto, September 2002, p. ON. + + + + + Call for Papers + + SMondoloni + + + IBayraktutar + + 10.1027/2192-0923/a000067 + + + Aviation Psychology and Applied Human Factors + Aviation Psychology and Applied Human Factors + 2192-0923 + 2192-0931 + + 4 + 2 + + June 2005 + Hogrefe Publishing Group + Baltimore + + + 6th USA/ + Mondoloni, S. and Bayraktutar, I., "Impact of Factors, Conditions and Metrics on Trajectory Prediction Accuracy," 6th USA/Europe Air Traffic Management Research and Development Seminar , Baltimore, June 2005, p. MD. + + + + + Aircraft Trajectory Prediction Errors: Including a Summary of Error Sources and Data + + SMondoloni + + + + Mondoloni, S., "Aircraft Trajectory Prediction Errors: Including a Summary of Error Sources and Data (Version 0.2)," + + + + + + + Faa/ + + + + + Eurocontrol Action Plan + + 16 + July 2006. July 2011 + + + FAA/Eurocontrol Action Plan 16, July 2006, URL: http://acy.ts.faa.gov/cpat/tjm/index.html[cited July 2011]. + + + + + Closed Form Takeoff Weight Estimation Model for Air Transportation Simulation + + H + + + LChatterji + + + GB + + + September 2010 + 16 + Fort Worth, TX + + + 10th AIAA Aviation Techology, Integration, and Operations (ATIO) conference + H., L. and Chatterji, G. B., "Closed Form Takeoff Weight Estimation Model for Air Transportation Simulation," 10th AIAA Aviation Techology, Integration, and Operations (ATIO) conference, Fort Worth, TX, 13-15 September 2010. 16 + + + + + Illustrations of multi-disciplinary high-fidelity analysis capabilities for aero-structural and aero-acoustic aircraft design + + IMKroo + + 10.2514/6.2021-1316.vid + + 2006 + American Institute of Aeronautics and Astronautics (AIAA) + Desktop Aero + + + Kroo, I. M., Aircraft Design: Sysnthesis and Analysis, Desktop Aero, 2006. + + + + + Accuracy of RUC-1 and RUC-2 Wind and Aircraft Trajectory Forecasts by Comparison with ACARS Observations + + BarryESchwartz + + + StanleyGBenjamin + + + StevenMGreen + + + MatthewRJardin + + 10.1175/1520-0434(2000)015<0313:aorarw>2.0.co;2 + + + Weather and Forecasting + Wea. Forecasting + 0882-8156 + 1520-0434 + + 15 + 3 + + 2000 + American Meteorological Society + + + Schwartz, B. E., Benjamin, S. G., Green, S. M., and Jardin, M. R., "Accuracy of RUC-1 and RUC-2 Wind and Aircraft Trajectory Forecasts by Comparison with ACARS Observations," Weather and Forecasting, Vol. 15, No. 3, 2000, pp. 313-326. + + + + + + diff --git a/file812.txt b/file812.txt new file mode 100644 index 0000000000000000000000000000000000000000..a9b64abb3bf6a98c97b97ae4b83306d6d7250d04 --- /dev/null +++ b/file812.txt @@ -0,0 +1,327 @@ + + + + +I. IntroductionThe predicted growth in traffic demand over the next 20 years could result in increased congestion and delays in the National Airspace System (NAS).Under the Dynamic Airspace Configuration (DAC) research sponsored by NASA, 1 two major concepts, dynamic resectorization and corridors-in-the-sky, or tube networks, have been proposed to ease congestion and delay.Developing tube networks is a promising concept due to its function of enabling high-density operations with less Air Traffic Control (ATC) workload. 2Other potential benefits may include lower Traffic Flow Management (TFM) restrictions, better predictability of flight operations, and fuel savings due to shorter routes.Impenetrable tubes may result in additional costs for flights that need to be rerouted to avoid conflict with the traffic within the tubes.Thus, the number of tubes should not be too large.On the other hand, to be beneficial and efficient, the limited number of tubes should be able to accommodate large traffic levels and should not require tube users to fly too much extra distance to utilize the tube.To meet these three requirements: low numbers of tubes, large number of flights in the tubes, and low extra flight distances, it is important to observe and to identify the commonalities of the nationwide flight trajectories.If the number of flights that take common routes is large, constructing tubes based on these common routes will be valuable.Otherwise, tubes might not provide much advantage due to the confliction with the remaining traffic.Several research efforts have focused on tube networks.Alipio et al. 2 and Yousefi et al. 3 suggested tubes based on city pairs by investigating the potential placement of tubes between two major cities such as Chicago and New York.If tubes are constructed based on city pairs, only a small portion of traffic will benefit with a limited number of tubes.Table 1 shows the flights among the top 25 US airports versus the flights over the whole NAS on April 20, 2007.This network requires more than 200 links/tubes and can only serve at most 20% of the flights.Sridhar et al. 4 proposed a tube network interconnecting airports in clusters seeded by major airports to impact a significant amount of traffic.Although approximately 90% of flights are included, fully utilizing the tubes may require too much extra flight distance for tube users, which could make this method prohibitive.A new methodology that proposes a limited number of tubes base on observed patterns of flight trajectories is desired.In this work, consider trajectories that follow the great circle flight routes between city This paper presents a new methodology based on the Hough Transform and Genetic Algorithm to observe and cluster common great circle flight routes.This work provides the basis for future construction by proposing the potential tubes.The results show that a limited number of tubes in the national airspace can accommodate significant flights with low extra flight distances for tube users.Furthermore, dynamic tube recognition can be fulfilled by executing such a procedure periodically. +II. Modelling and Architecture of Designing TubesIn this section, the data used were discussed.The procedures for identifying and proposing the potential tubes that eventually compose the networks are briefly introduced as well. +A. Data AcquisitionTraffic data for this work were obtained from the FAA's Aircraft Situation Display to Industry (ASDI) for the entire day of April 20, 2007.The ASDI file for this day contains 62, 143 flights.Because only the continental US is of interest, a rectangular bounding box surrounding the continental US is formed: the latitude range is [17.2 • N, 50.0 • N ] and longitude range is [225.8• E, 296.7 • E].For each flight, only the origin and destination are used for defining great circle flight routes.Without loss of generality, only flights with tracks above FL290 are taken into account.During preprocessing, the great circle flight trajectories are determined, and those trajectories that do not overfly the continental US region are removed.These preprocessing steps result in a final total of 29, 629 flights to be considered. +B. ProcedureThe procedure to identify and generate tubes is shown in Figure 1.First, the preprocessed great circle trajectories are transformed into points in Hough space.Each point corresponds to one great circle flight route.Next, initial clustering is performed by the voting scheme on the basis of those points, which is essentially a grid based clustering.Because the criterion for clustering the trajectories is minimal excess flight distance instead of the Euclidian distance, the flights are removed from their clusters if there is too much extra flight distance.Based on these initial clusters, the Genetic Algorithm is applied to refine the +III. Methodology and ResultsIn this section the Hough Transform and its parameterization are described first.After that, the initial clustering based on the voting scheme for the points in the Hough space is presented.On the basis of the initial clusters, the Genetic Algorithm is applied to improve the design. +A. Hough Transform for Great Circle RoutesPolar representation of a line is referred to Hough Transform in image processing literature. 9The equation to a line with the shortest distance ρ to the origin and an angle θ between its normal and the x axis is given by: ρ = x cos θ + y sin θThe (x, y) are in Cartesian coordinates, and the (ρ, θ) are in the Hough space.This property is also called point-line duality.Although a great circle line lies in 3-dimensional space, it actually has only 2 degrees of freedom.Thus, it is suitable to apply the Hough transform.Given a flight with the origin at (δ o , λ o ) and the destination at (δ d , λ d ), assuming the Earth is an ideal sphere, the Cartesian coordinates will be:r i =    x i y i z i    = R •    cos λ i • cos δ i sin λ i • cos δ i sin δ i   (2)where i = o, d representing the origin and destination, respectively.R is the radius of the Earth, δ and λ are latitude and longitude, respectively, and x, y, z are the Cartesian coordinates.Any point r = [x, y, z] T on this great circle line should satisfy:r o × r d = r × r d r = R(3)Just as with straight-lines, to transform the great circle lines a reference point has to be defined along with a reference axis.The reference point can be put anywhere inside of the bounding box, and the reference axis can point to any direction.Though different references generate different Hough transform results, they do not affect the nature of clustering as long as the references are fixed during the whole process.In this work, the reference point is set at [33.6 • N, 261.3 • E] -the center of the defined rectangular region, while the reference axis is defined to point the East.Figure 2 shows how ρ is defined as the shortest distance between the reference point and the great circle route and how θ is defined as the cross angle between the reference axis and the tangent vector at the reference point, which is aligned with ρ.Here ρ ∈ [0, ∞] and θ ∈ [0, 2π].Detailed derivation of Hough and DeHough transforms for great circle routes are presented in the Appendix.Figure 3 shows the results after the Hough transform.The 29, 629 great circle flight routes are transformed to points in Hough space.Some of them overlap each other.Apparently, the relationship among the trajectories can be easily told by the points in the Hough space.Notice how the points form many parabolalike curves in the Hough space.This phenomenon can be explained using the point-line duality.Each curve essentially implies an airport, and the points forming the curve correspond to all the trajectories going through that airport.This supposition can be borne out by the following equation:ρ = R • arctan[tan( d R ) • cos(γ -θ)](4)where d is the shortest distance/great circle distance between an airport and the reference point.The variable γ is the angle between the line, which links the airport and the reference point, and the reference axis.The derivation can be found in the Appendix.Given an airport, the values of d and γ, are fixed.Because the radius of Earth R is a constant, parameters ρ and θ have a fixed relationship.For instance, ORD has d = 826 miles and γ = 47.7 • , while JFK has d = 1451.8miles and γ = 26.8• .As shown in Figure 3, all the trajectories going through these two airports fall on the yellow and green curves, respectively. +B. Initial ResultsIn image processing, the continuous (ρ, θ) Hough space is then quantized into suitable size (∆ρ × ∆θ) grids and each grid is associated with an element of an array called the accumulator array.After this, the lines are extracted by the voting scheme.In this work, similar grids are constructed in the Hough space to catch the potential high-density tubes.Without loss of generality, the number of tubes is set to 60 in this work.The resolutions are defined as ∆ρ = 50 miles and ∆θ = 10 • .The 60 grids with the highest point density are shown as red rectangles in Figure 4.They accommodate 59% of the 29, 629 flights.After the Dehough transform, these 60 grids become 60 tubes.A sample tube from one of these grids is presented in Figure 5(a).The red curve/tube is transformed from a midpoint which is the average over all points inside of the grid in the Hough space.The blue curves are the actual great circle flight routes that can utilize this tube.If the Hough coordinates of a flight trajectory are not exactly the same as the weighted center of its corresponding grid, there will be extra flight distance for the flight to use the designated tube.On the other hand, due to the finite length, the great circle routes have largely varied extra distances even if the coordinated in (ρ -θ) space are the same.Therefore, the extra distance, which is assumed to be the main concern of the tube users, should serve as the criteria for clustering instead of the Euclidean distance.The extra distance must be calculated in the original space, because the origin and destination are lost in the the transformation.The extra flight distance, shown in Figure 5(b), is defined as a rate:d extra = d 1 + d 2 D • 100%(5)where d i are the shortest distances to join or leave the tube, and D is the original great circle flight distance.Most of the clustering algorithms suitable for Euclidean distance will not be efficient for the extra flight distance due to the complexity of the relative positions with the extra flight distance.Even a stochasticprocess-based optimization algorithm, like the Genetic Algorithm, cannot determine the optimal solutions in a feasible computational time without a good initial guess.The search will be exhaustive and the required computational time will make it prohibitive.Fortunately, clustering based on Euclidean distance in the Hough space should be rough estimate of the clustering based on the extra distance in the original space.For instance, if two trajectories are far away from each other in the Hough space, they most likely have large extra distance.On the other hand, if they are far from each other in the ρ -θ space, the probability of having low extra distance between them is low.Thus, the clustering results based on Euclidean distance in the Hough space is utilized and the flights that have higher extra distance in original space than a defined threshold are removed.If the tolerable extra flight distance is defined as 5% of the shortest flight distance, the percentage of flights that top 60 tubes can hold drops from 59% to 31% .The large decrease indicates the difference between two clustering criteria.Figure 6 displays the top 60 tubes.In this figure, only the portions that have more than 60 flights are shown.A warmer color denotes higher density of the traffic, whereas a colder color represents lower density.The density of the traffic can be judged by the color bar. +C. Refinement using Genetic AlgorithmTo maximize the number of benefited flights, an optimization algorithms is applied to these initial tubes.In this work, the Genetic Algorithm (GA) is used to optimize the clustering, because it has been widely used as a powerful optimization method.The GA is a stochastic process which models two natural phenomena: genetic inheritance and Darwinian evolution. 11It first creates a population of potential solutions.Each solution is called a "chromosome', represented by a binary string of length m = k i=1 m i , where k is the number of design parameters.The first m i bits of the string correspond to the first design parameter, or"gene".The next group with m i bits will map to the second design parameter, and so on.In each generation, the population of chromosomes will be evaluated by using a cost function.The new population is selected with respect to the probability distribution based on fitness values.Finally, the chromosomes are altered in the new population by mutation and crossover operators.In this work, the weighted centers of the grids in the Hough space are optimized.Because each center has two coordinates, ρ and θ, the number of optimization parameters will be 120.The clustering results from the Hough transform discussed above will serve as initial values for the GA.For the selection process inside of the GA, a roulette wheel with slots sized according to fitness is used.The crossover probability is 0.8 and mutation probability is 0.2.The population size is set to 200 and it will stop after 200 generations.The threshold for the extra flight distance is 5%.After 200 generations, the percentage of flights that the top 60 tubes can accommodate increases from 31% to 44%.The grid centers did not change dramatically, which means the original clustering provided a good start.Figure 7 shows the final top 60 tubes.Only the portions which have more than 60 flights are displayed.On the platform of MacOS with Intel Core 2 Duo Processor 3.0 GHz, 200 generations takes around 10 hours with 4 threads working in parallel.This computation time can be shorten by lowering the number of generations or running the GA in more powerful parallel computing clusters. +D. AnalysisTo get insight of operations in the tubes, the flights within the designed tubes are investigated.Figure 8 shows the flights between any two of the top 25 airports versus the rest in each tube.Similar to the observation in Sridhar's work, 4 it is noticed that flights among these airports do not account for a major portion of the total operations within the tubes.These results further testify the importance of building tubes upon the nature of flight trajectories versus major-city pairs.This study focused on flights over FL290.Several other partitions are explored in a altitude sensitivity analysis.Based on a 5% deviation, Table 2 presents the results for all flights that are flying above a given altitude.The results show that flights at high altitudes have stronger tube conformance than the ones at low altitudes, which implies that it may be more valuable to construct tubes/corridors at high altitudes than at low altitudes.The sensitivity analysis can be conducted if the deviation constraints are varied.Table 3 presents the relationship between the deviation constraints and the number of flights included in the tubes based on flights over FL290.The results demonstrate that the more deviation allowed from the shortest distance, the more flights can be included into the tubes.This provides us a profile of the benefits pool of constructing tube network.Although the largest deviation can yield about 70% flights operated in the tubes, it should be carefully decided in practice and weighed against the cost of deviating. +IV. ConclusionsIdentifying tubes that catch the commonalities of the shortest distance trajectories is necessary for constructing tubes/corridors-in-the-sky structure in future airspace.If the clustering is based on the city-pairs to cluster, only a small amount of traffic per tube might benefit.Building tube network clusters by seeding it with major-cities increases the benefit pool but requires too much extra flight distance for the tube users.Clustering the trajectories directly can identify the potential tubes that provide a compromise between the benefit pool and the extra flight distance required.In this work, a new methodology is developed based on the Hough transform, which enables the clustering of great circle flight routes.Extra flight distance beyond the shortest flight distance serves as the criteria for clustering.There is no single existing clustering algorithm that can solve this problem efficiently, thus, the Hough Transform and Genetic Algorithm are combined.The Hough Transform generates good initial guesses, and thereafter, GA refines the tubes if the restriction of computational time allows.Results show that significant traffic would benefit from the tubes by only using a small number of tubes.Utilizing these tubes only requires five percent extra flight distances.The entire process could be achieved within a feasible computational time.This procedure could be directly utilized for dynamic tube design if the data are fed in during during incremental time periods.The list of tubes proposed by our method can be used as a basis.With further constructional and operational analysis, the final tubes or network can be eventually constructed.This method can be applied to the wind optimal trajectory if the wind velocity is constant for the entire trajectory.In future work, efforts will be put on general wind optimal trajectories. +AppendixAs described in Figure 9, given a great circle flight route composed of two points/airports ( u, v), to get the unique hough transform, a reference point p is first configured, which has geographic coordinates, latitude (δ p ) and longitude (λ p ), and Cartesian coordinates [p 1 , p 2 , p 3 ] T .In this work, the ρ, one parameter in Hough transform, is defined as the shortest distance between point p and the given great circle on the sphere/earth surface.Assume the great circle plane Λ containing u and v has norm n Λ [n Λ1 , n Λ2 , n Λ3 ] T .If we create another great circle plane Θ, which includes reference point p and is perpendicular to existing plane Λ, then ρ will be the great circle distance between p and the closer intersection point f ([δ f , λ f ] and [f 1 , f 2 , f 3 ] T ).Furthermore, we define the tangent vector E p at point p, which points East, to be the base direction.Thus, another parameter in Fough transform θ is defined as the cross angle between vector E p and the tangent vector F p at point p, which lies in the great circle plane Θ. +A. Calculation of ρ in Hough TransformTo compute ρ, we first need to find the intersection points mentioned above.Then, the closer one is chosen.Assume the points are f A (δ f A , λ f A ) and f B (δ f B , λ f B ). δ and λ should satisfy the following equations:cos λ • n Λ1 + sin λ • n Λ2 + tan δ • n Λ3 = 0 (6a)    p 2 • tan δ -p 3 • sin λ p 3 • cos λ -p 1 • tan δ p 1 • sin λ -p 2 • cos λ    •    n Λ1 n Λ2 n Λ3    = 0 (6b) If p 2 • n Λ1 -p 1 • n Λ2 = 0, from Eqn. (6b), it can be derived:tan δ = -sin λ(p 1 n Λ3 -p 3 n Λ1 ) -cos λ(p 3 n Λ2 -p 2 nΛ3) (p 2 n Λ1 -p 1 n Λ2 )(7)by substituting Eqn.(7) in Eqn.(6a) we have:[(p 2 n Λ1 -p 1 n Λ2 )n Λ1 -(p 3 n Λ2 -p 2 n Λ3 )n Λ3 ] cos λ + [(p 2 n Λ1 -p 1 n Λ2 )n Λ2 -(p 1 n Λ3 -p 3 n Λ1 )n Λ3 ] sin λ = 0 (8)By defining:A = (p 2 n Λ1 -p 1 n Λ2 )n Λ2 -(p 1 n Λ3 -p 3 n Λ1 )n Λ3 (9a) B = (p 2 n Λ1 -p 1 n Λ2 )n Λ1 -(p 3 n Λ2 -p 2 n Λ3 )n Λ3(9b)Given the reference point p and US region interested, we have:A 2 + B 2 > 0(10)Eqn 8 then becomes: sin(λ + β) = 0 (11) where:cos β = A √ A 2 + B 2 (12a) sin β = B √ A 2 + B 2(12b)Eventually, we get:λ = kπ -β, k = [• • • -1 0 1 • • •],where λ ∈ [0, 2π].Since there only exists two values between 0 and 2π, define them to be λ 1 and λ 2 , respectively.If p 2 • n Λ1 -p 1 • n Λ2 = 0, from Eqn. (6b), we get:-sin λ(p 1 n Λ3 -p 3 n Λ1 ) -cos λ(p 3 n Λ2 -p 2 n Λ3 ) = 0(13)By defining:C = -(p 1 n Λ3 -p 3 n Λ1 ) (14a) D = -(p 3 n Λ2 -p 2 n Λ3 ) (14b)Again, given the reference point p and the US region of interest, we have:C 2 + D 2 > 0 (15)Then, rewrite Eqn 13 in simpler form: sin(λ + γ) = 0 (16)where:cos γ = C √ C 2 + D 2 (17a) sin γ = D √ C 2 + D 2(17b)Similarly, we have:λ = kπ -γ, k = [• • • -1 0 1 • • •],where λ ∈ [0, 2π].Again, define the two values between 0 and 2π to be λ 3 and λ 4 , respectively.Now, δ can be computed explicitly.If n Λ3 = 0, from (6a), we can derive:δ i = arctan[- cos λ i n Λ1 + sin λ i n Λ2 n Λ3 ](18)where i = 1, 2, 3, 4.If n Λ3 = 0, then the great circle route should have constant longitude, which means the latitudes of intersection point should be equal to δ p .Finally, the latitude and longitude of the intersection point for shortest distance ρ can be concluded:(δ f , λ f ) =                              argmin δ∈[δ1,δ2] λ∈[λ1,λ2] [great circle dist((δ, λ), p)] if p 2 • n Λ1 -p 1 • n Λ2 = 0 and n Λ3 = 0, argmin λ∈[λ1,λ2] [great circle dist((δ p , λ), p)] if p 2 • n Λ1 -p 1 • n Λ2 = 0 and n Λ3 = 0, argmin δ∈[δ3,δ4] λ∈[λ3,λ4] [great circle dist((δ, λ), p)] if p 2 • n Λ1 -p 1 • n Λ2 = 0 and n Λ3 = 0, argmin λ∈[λ3,λ4][great circle dist((δ p , λ), p)] if p 2 • n Λ1 -p 1 • n Λ2 = 0 and n Λ3 = 0,Where the operator great circle dist computes the great circle distance between the pair of geographic coordinates. +B. Calculation of θ in Hough TransformNext, the cross angle θ between vector E p and F p is computed.Assuming the normal of the great circle plane Θ is n Θ , where:n Θ = n Λ × p(20)the vector F p can be computed by:F p = n Θ × p(21)Assuming the great circle plane that E p lies in is plane Γ and its normal is n Γ (actually normalized form is [0, 0, 1] T ), by defining the projection of p in the horizontal plane, which contains the equator, as p [p 1 , p 2 , 0] T , we get:E p = n Γ × p(22)Then, the θ within [0, π 2 ] is computed:θ = arccos( E p • F p E p • F p ) (23)By checking the relative position between point p and f , we can finally get θ explicitly. +C. Dehough TransformGiven a Hough transform point with θ and ρ, by simply reversing the procedures of the above Hough Transform with the same reference point, reference direction and interested region, we can compute the great circle flight route with two end points on the boundary of the defined region.Defining a reference point p (δ p , λ p ) and axis pointing to East, and specifying an airport A δ A , λ A , we could find the relationship between ρ and θ for any great circle routes going through the airport.For any great circle route via airport A, suppose f is the perpendicular foot on the great circle line, which makes pf ⊥ Af .Define the plane that is tangent to the earth at point p as plane Π.Assuming the origin of earth is O, on plane Π, define that A and f are the points collinear with OA and Of , respectively.For the right triangle A f P , we have:ρ = pf = R • tan γ = R • tan( ρ R )(24)andd = pA = R • tan( d R )(25)Assuming the great circle distance/shortest distance between p and A is d, and the angle between the norm of route and the reference axis is γ, the relationship between ρ and θ can be expressed as:ρ = d • cos(γ -θ)](26)⇒ Eqn. 4Figure 1 .1Figure 1.Procedure to identify and generate tubes +Figure 2 .Figure 3 .23Figure 2. Hough Transform for Great Circle Routes +Figure 4 .4Figure 4. ORD and JFK in Hough Space and the Initial Top 60 grids based on Hough Transform +Figure 5 (Figure 6 .56Figure 5. Tube and Extra distance +Figure 7 .7Figure 7. Top 60 tubes after refining using GA +Figure 8 .8Figure 8. Percentage of flights among Top 25 US airports +Figure 9 .9Figure 9. Calculation of Hough Transform +Figure 10 .10Figure 10.Calculation of Relationship for Trajectories via An airport +Table 1 .17lights Among Top 25 US Airports Several image/signal processing techniques for grouping flight patterns were investigated.Robelin et al.5developed an aggregation technique based on Generalized Principal Component Analysis (GPCA)6for constructing a model of air traffic flow directly from Aircraft Situation Display to Industry (ASDI) data.Martinez et al.7also used this aggregation technique to build a network flow graph.In these works, the authors used GPCA to get dominant directions at designated fixes from historical trajectories composedAltitude (ft) Flights among Top 25 airports Flights over US Percentage>07,11760,36211.8%>12,0006,87047,31314.5%>18,0006,67741,22916.2%>24,0006,31335,59017.7%>29,0005,91429,88719.8%>35,0003,68817,91620.6%pairs.of series of waypoints/fixes.In this work, only origin and destination are considered for the great circle flight trajectory.Therefore, dimension reduction techniques like GPCA may not help.In computer vision and pattern recognition, the Hough Transform 8-10 has been widely used for detecting lines, curves, or even geometric shapes that can be defined by parametric equations.Based on the point-line duality, the Hough Transform can be used to transform great circle trajectories to points in another space.Points that are close to each other will likely correspond to great circle trajectories that are close enough to form a tube.Thus, the Hough Transform is a good candidate for clustering great circle flight trajectories. +Table 2 .2Flights Included in Tubes Given 5% DeviationAltitude (ft) Flights involved Percentage>017,27429%>24,00014,48541%>29,00013,01544%>35,0008,65849% +Table 3 .3Flights Included in Tubes Given varied DeviationDeviation Flights involved Percentage1%3,69213%2%6,73623%3%8,99830%4%11,26538%5%13,01544%7%16,08554%9%18,66663%11%20,40769% + + + + + + + + + Initial Concepts for Dynamic Airspace Configuration + + ParimalKopardekar + + + KarlBilimoria + + + BanavarSridhar + + 10.2514/6.2007-7763 + + + 7th AIAA ATIO Conf, 2nd CEIAT Int'l Conf on Innov and Integr in Aero Sciences,17th LTA Systems Tech Conf; followed by 2nd TEOS Forum + Belfast, Northern Ireland + + American Institute of Aeronautics and Astronautics + 18-20 September 2007 + + + P. Kopardekar, K. Bilimoria, and B. Sridhar. Initial concepts for dynamic airspace configuration. In 7th AIAA Aviation Technology, Integration and Operations Conference (ATIO), Belfast, Northern Ireland, 18-20 September 2007. + + + + + Dynamic airspace super sectors (DASS) as high-density highways in the sky for a new US air traffic management system + + JAlipio + + + PCastro + + + HKaing + + + NShahid + + + OSherzai + + + GLDonohue + + + KGrundmann + + 10.1109/sieds.2003.158005 + + + IEEE Systems and Information Engineering Design Symposium, 2003 + + IEEE + October 2003 + + + + J. Alipio, P. Castro, H. Kaing, N. Shahd, O. Sheizai, G.L. Donohue, and K. Grundmann. Dynamic airspace super sectors (DASS) as high-density highways in the sky for a new US air traffic management system. In AIAA/IEEE Digital Avionics Systems Conference, 12-16 October 2003. + + + + + High-volume tube-shape sectors (HTS): a network of high capacity ribbons connecting congested city pairs + + AYousefi + + + GLDonohue + + + LSherry + + 10.1109/dasc.2004.1391296 + + + The 23rd Digital Avionics Systems Conference (IEEE Cat. No.04CH37576) + Salt Lake City, CT + + IEEE + 2004 + + + A Yousefi, G.L. Donohue, and L. Sherry. High-volume tube-shape sectors(HTS): A network of high capacity ribbons connecting congested city pairs. In Proceedings of the 23rd Digital Avionics Systems Conference, Salt Lake City, CT, 2004. + + + + + Initial Study of Tube Networks for Flexible Airspace Utilization + + BanavarSridhar + + + ShonGrabbe + + + KapilSheth + + + KarlBilimoria + + 10.2514/6.2006-6768 + + + AIAA Guidance, Navigation, and Control Conference and Exhibit + Keystone, Colorado + + American Institute of Aeronautics and Astronautics + August 2006 + + + + B. Sridhar, S. Grabbe, K. Sheth, and K.D. Bilimoria. Initial study of tube networks for flexible airspace utilization. In AIAA Guidance, Navigation, and Control Conference and Exhibit, Keystone, Colorado, 21-24 August 2006. + + + + + MILP control of aggregate Eulerian network airspace models + + C-ARobelin + + + Dengfeng Sun + + + Guoyuan Wux + + + AMBayen + + 10.1109/acc.2006.1657558 + + + 2006 American Control Conference + Minneapolis + + IEEE + June 2006 + + + C. Robelin, D. Sun, G. Wu, and A.M. Bayen. MILP control of aggregate eulerian network airspace models. In Proceedings of the American Control Conference, Minneapolis, June 2006. + + + + + Generalized principal component analysis (GPCA) + + RVidal + + + Yi Ma + + + SSastry + + 10.1109/cvpr.2003.1211411 + + + 2003 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2003. Proceedings. + Madison, WI + + IEEE Comput. Soc + June 2003 + + + + R. Vidal, Y. Ma, and S. Sastry. Generalized principal component analysis (GPCA). In IEEE Conference on Computer Vision and Pattern Recognition, pages 621-628, Madison, WI, June 2003. + + + + + A Weighted-Graph Approach for Dynamic Airspace Configuration + + StephaneMartinez + + + GanoChatterji + + + DengfengSun + + + AlexandreBayen + + 10.2514/6.2007-6448 + + + AIAA Guidance, Navigation and Control Conference and Exhibit + + American Institute of Aeronautics and Astronautics + + + + S.A. Martinez, G.B. Chatterji, D. Sun, and A.M. Bayen. A weighted-graph approach for dynamic airspace configuration. + + + + + AIAA Guidance, Navigation and Control Conference and Exhibit + 10.2514/mgnc07 + + + AIAA Guidance, Navigation and Control Conference and Exhibit + Hilton Head, South Carolina + + American Institute of Aeronautics and Astronautics + August 2007 + + + + In AIAA Guidance, Navigation and Control Conference and Exhibit, Hilton Head, South Carolina, 20-23 August 2007. + + + + + Method and means for recognizing complex patterns + + PV CHough + + + 1962 + 3069654 + + + P.V.C. Hough. Method and means for recognizing complex patterns. U.S. Patent No.3069654, 1962. + + + + + Use of the Hough transformation to detect lines and curves in pictures + + RichardODuda + + + PeterEHart + + 10.1145/361237.361242 + + + Communications of the ACM + Commun. ACM + 0001-0782 + 1557-7317 + + 15 + 1 + + 1972 + Association for Computing Machinery (ACM) + + + P.E. Duda, R.O.and Hart. Use of the hough transformation to detect lines and curves in pictures. Communicatons of the ACM, 15:11-15, 1972. + + + + + Introductory Techniques for 3-D Computer Vision + + ETrucco + + + AVerri + + + 1998 + Prentice Hall + + + E. Trucco and A. Verri. Introductory Techniques for 3-D Computer Vision. Prentice Hall, 1998. + + + + + Genetic Algorithms + Data Structures = Evolution Programs + + ZMichaelwicz + + 10.1007/978-3-662-03315-9 + + 1996 + Springer Berlin Heidelberg + Berlin Heidelberg + + + Third Edition + Z. Michaelwicz. Genetic Algorithms + Data Structures = Evolution Programs, Third Edition. Springer-Verlag Berlin Heidelberg, 1996. + + + + + + diff --git a/file813.txt b/file813.txt new file mode 100644 index 0000000000000000000000000000000000000000..4d0ab3379a3d059c79f958d2705d7d8438654b40 --- /dev/null +++ b/file813.txt @@ -0,0 +1,822 @@ + + + + +I. IntroductionThe volume of small Unmanned Aircraft System(sUAS) operations is expected to increase dramatically in the near future. 1 Potential sUAS applications include, but not limited to, search and rescue, inspection and surveillance, aerial photography and video, precision agriculture, and parcel delivery.According to the marketing analysis 2 , the global small UAS market is anticipated to hit 10 billion by 2020.FAA also forecasted that over 7 million sUASs will be sold annually by 2020. 1 The sUAS's low operational altitude, small size, and envisioned scale of operations make Unmanned aircraft system Traffic Management (UTM) quite different from conventional aviation traffic management.In the low altitude airspace, besides fast-changing wind conditions, restricted areas, manned aircraft/helicopters, and tall buildings/terrain impose many constraints in sUAS operations.The sensitive trajectory response and high maneuverability make sUAS different from manned or unmanned large-size fixed-wing aircraft and dramatically change the way traffic system operates.These characteristics and the predicted large scale operations 1 present great challenges in managing safe and efficient traffic operations in low altitude airspace.NASA's UTM research initiative 3,4 is researching and defining requirements and policies for the UTM system to ensure fair, safe, and efficient UAS operations in the future.In order to investigate the impacts of various device parameters, traffic system rules and policies, operational schedules, and wind conditions, especially in dense operations, it is necessary to build an effective fast-time simulation platform that can incorporate different parameters, rules, and models, and evaluate them statistically in realistic environments.This work reviewed past literature, studied key factors/requirements, and presented examples of potential applications of such a UTM fast-time simulation platform.This paper is organized as follows: Section II presents literature review of simulations and trajectory models.Section III analyzes the impacts of several key factors in sUAS trajectory prediction.Section IV discusses the necessity and effectiveness of the Monte Carlo method.Section V presents preliminary simulation experiments and results and Section VI draws conclusions. +II. Literature ReviewSeveral categories of simulations have been widely used in research related to air traffic systems.The first category of simulations is mainly built to study traffic systems that include multiple aircraft operations and rules.For example, NASA developed several high-fidelity fast-time simulation platforms for studying manned aviation, such as the Center TRACON Automation System (CTAS), 5 the Future ATM Concepts Evaluation Tool (FACET), 6 and the Airspace Concept Evaluation System (ACES). 7Each of these platforms has different strengths in aviation simulations.Due to the similarity of trajectory models used for large-size fixed-wing UASs and manned fixed-wing aircraft, the latest ACES incorporated models of large fixed-wing UASs 8 in order to study interactions between Unmanned Aircraft Systems(UAS) and manned aircraft. 9, 10In addition to these large aircraft simulation tools, researchers often developed their own simulation tools for specific research topics.For instance, Cook et.al. 11 defined a set of conflict resolution rules and tested them in simulations with multiple sUASs.Jenie et.al. 12 proposed a method for uncoordinated avoidance maneuvers of UASs and conducted Monte Carlo simulation to verify the proposed method.The second type of simulation deals with encounter models and is usually designated as Conflict Detection and Resolution (CD&R) research.This category of simulations has a much smaller scope than the first one: It typically involves two aircraft and the flight durations are usually short since its purpose is to study encounters between two aircraft. 13,14 n recent literature, Mueller et.al. 15,16 developed a collision avoidance method and included small UASs, especially multi-rotor air vehicles.The third type of simulation is vehicle centric and is mainly developed for studying and simulating the vehicle's model and control.This type of simulator usually includes high-fidelity dynamic system and control model for a specific vehicle.For instance, the Reflection 17 included the Hanger 9 Quarter scale Cessna 172 and was used for autopilot design.In studies of manned aircraft or large fixed-wing UASs, vehicle models usually only refer to vehicles' dynamic systems because their control systems have the capability of tightly following desired trajectories at the nautical-mile level, even in the presence of wind gust.At the cruise phase, the aircraft trajectory lateral (or cross-track) deviation caused by gust is usually ignored because the overshoots caused by wind change are negligible compared to both navigation errors and the nautical-mile separation standards.A typical approach of modeling large fixed-wing aircraft trajectories used point mass equations of motion. 18,19 1][22] Vertical profiles were calculated using flight path angles as control parameters.Whereas, horizontal paths were constructed from straight lines and turn arcs, where turn radii were decided based on bank angles and ground speeds.This simplified horizontal path calculation approach was widely used for enroute trajectory computation and prediction in aviation research.Another approach was specifically developed for CD&R studies to build aircraft encounter models. 13,14 his approach modeled trajectories by constructing a dynamic Bayesian network structure based on historical trajectory data from FAA's radar records.The typical inputs of the Bayesian network are vertical rate, airspace class, turn rate, altitude, and acceleration at the current time step and outputs are vertical rate, turn rate, and acceleration at next time step.Once this Bayesian network is constructed given the above inputs, the Bayesian network should be able to project the next state of the aircraft.Unlike large-size fixed-wing aircraft, whose trajectory errors are dominated by navigation systems, sUAS's trajectories are more sensitive to wind, vehicle speed, and vehicle control system because of their low operational altitude, small size, and limited power, especially when separation distance requirements are at a meter level instead of a nautical-mile level.On the other hand, sUASs are highly maneuverable, which changes the conventional way of conflict resolution because their capability of hovering and flying at low speeds.These unique characteristics demand both sUASs' dynamic sytem models and controller models for an effective UTM fast-time simulation platform, such that evaluations on this platform can provide sufficient accuracy, especially for dense operations.Although recent research 11,12,15,16 started to expand vehicles' speed ranges in an attempt to represent small UASs or multi-rotor vehicles, negligence of modeling sUAS controllers will yield inaccurate trajectory predictions, especially at low altitude airspace where wind changes very often, eventually lead to invalid simulations.Therefore, besides the vehicle dynamic sysmte, vehicle's control system needs to be modeled as well.Table 1 briefly compared the attributes of the aforementioned simulation platforms.This table is not intended to be a complete comparison as it is out of the scope of this paper.The last column of "UTM required attributes" listed the attributes that might be required by an effective sUAS traffic simulation platform.Since the wind effect is closely connected with controllers for sUAS, the UTM required attributes that are missed in many existing simulations can essentially be simplified to trajectory models with controllers and the capability of Monte Carlo simulations. +III. Trajectory SensitivityThis section shows the importance of modeling controllers in trajectory models by discussing sensitive factors in trajectory prediction for small UASs.Although navigation system error still plays an important role in trajectory errors, it will not be considered in this work.A quadrotor is used as a representative sUAS with a predefined controller, which is described in detail in this section.In the following sections, the vehicle is required to follow a straight line with a constant speed and altitude and to fly through a cross wind field.To examine the sensitivity of the trajectory responses, the impact from various wind speeds, vehicle speeds, and controllers are explored. +A. Trajectory ModelMany types of sUASs have been designed and manufactured in past years, such as quadrotors, multirotors, fixed-wing UASs, hybrid UASs.Without loss of generality, a quadrotor dynamics model is used in this work for examination.Quadrotors may have various sizes, weights, shapes, equipment, and control mechanisms; their fundamental dynamics/plant models are the same except for different parameter values.4][25] After neglecting Coriolis terms and applying small angle approximations, the dynamics model can be expressed as Eqn. 1, where φ, θ, and ψ are roll, pitch, and yaw angles in the body frame, and p n , p e , and h are north position, east position, and altitude in the Earth frame.k f and k m are the aerodynamic force and moment coefficients for motors.J x , J y , and J z are vehicle inertia and the vehicle is assumed to be symmetric.Ω i is the angular velocity of rotor i and L is the length of the arms.w n and w e are the north and east components of the wind vector, where the wind effect was simplified to only affect vehicle velocities.It is also noted that drag terms are neglected in this simplified model.The parameters in the model are set as in Table 2.              ṗn pn ṗe pe ḧ φ θ ψ               =                pn + w n -(cos φ sin θ cos ψ + sin φ sin ψ)•F z /m pe + w e (-cos φ sin θ sin ψ + sin φ cos ψ)•F z /m g -cos φ cos θ • F z /m 1 Jx M φ 1 Jy M θ 1 Jz M ψ               (1)Where     F z M φ M θ M ψ      =      k f (Ω 2 1 + Ω 2 2 + Ω 2 3 + Ω 2 4 ) (-k f Ω 2 2 + k f Ω 2 4 )•L (k f Ω 2 1 -k f Ω 2 3 ))•L (k m Ω 2 1 -k m Ω 2 2 + k m Ω 2 3 -k m Ω 2 4 ))•L     (2)Table 2. Dynamics parameters As an initial study, a proportional-derivative (PD) position controller is applied in this work.In order to reach a desired horizontal location (x d , y d ), a quadrotor needs to roll and/or pitch to eliminate the deviation.Usually, a PD position controller calculates the desired accelerations ẍd and ÿd first (shown in Eqn.3), and then desired roll and pitch angles φ d and θ d are derived using Eqn. 4 for the attitude controller to track.This process will be continued with updated states until the desired position is reached.Eqn. 5 shows the controllers for roll and pitch angles.Controller's gains are shown in Table 3, where k p and k d are proportional and derivative gains, respectively.J x J y J z m (kg) k f k m L(ẍd ÿd = k p (x d -x) + k d ( ẋd -ẋ) k p (y d -y) + k d ( ẏd -ẏ)(3)φ d θ d = m U 1 -sinψ -cosψ cosψ -sinψ -1 ẍd ÿd(4)M φ M θ = k p,φ (φ d -φ) + k d,φ ( φd -φ) k p,θ (θ d -θ) + k d,θ ( θd -θ) L(5) +B. Wind SpeedGiven the quadrotor dynamics and control model in the previous section, Fig. 1 presents the trajectory responses when different cross winds are imposed while the quadrotor was trying to follow a straight line trajectory with the speed of 5 meter per second (mps).The north wind caused trajectory deviations and the higher the wind magnitude is, the higher overshoot the vehicle has.As shown in the figure, the overshoot produced by a 5 mps wind reached 5 meters.In addition, the settling time that a vehicle needs to converge to its steady state increases when the cross wind increases.It took the vehicle over 50 meters to recover from the overshoot when it was experiencing 8.7 mps cross wind.Considering the fact that separation standards for multiple sUAS operations might be close to a meter-level precision, these deviations should not be neglected when predicting and calculating trajectories for sUASs in UTM simulations, neither should they be simplified and represented by some statistical distributions. +C. Desired vehicle ground speedIn actual operations, even for the same type of sUASs, different desired vehicle ground speeds may be set up by different operators intentionally or unintentionally as long as speeds are under the maximum value.However, sUAS's trajectories are also sensitive to desired vehicle ground speeds.Fig. 2 presents trajectory responses with the same cross wind but different vehicle desired ground speed.Although lateral overshoots and deviations are similar, the resulted trajectories are quite different.The horizontal distance in the xdirection for the vehicle to recover from overshoots vary from 10 meters to over 50 meters at different vehicle speeds, which will greatly affect the 4D trajectory prediction accuracy and outcomes of collision avoidance algorithms. +D. Control mechanismThe controller might be the most sensitive source for trajectory calculation errors.The difference in the controllers can be a different control gain, a different limit on forces or rotation angles, or a different control function, such as a PD controller vs. a backstepping controller.Even a different range for capturing a waypoint causes discrepancy in trajectories as well.As a simple example, Fig. 3 shows the comparison when different proportional gain k p is applied.It shows that when the proportional gain increases the response time that a vehicle takes to reach the peak deviation decreases while the overshoots stay similar.However, if k p increases too much, the vehicle oscillates around the reference trajectory and needs a long time to settle down to the desired state. +IV. Monte Carlo SimulationStatistical study of parameters and uncertainties is necessary to understand and evaluate the safety and efficiency of the future UTM system.The parameters and uncertainties involve many sources, such as onboard sensors, navigation and communication devices, right of way rules, collision avoidance algorithm and rules, various weather conditions, and vehicle systems.When dealing with a high dimensional problem, or high number of random variables, the Monte Carlo method/simulation 26 can be a very useful tool as it is known to be fairly independent of the problem dimension. 27,28 onte Carlo simulation is characterized by a rate of convergence of order O(1/ √ n), where n is the number of simulations.The relationship between the number of simulations and the percentage error of the mean at a given confidence interval 29 can be explicitly expressed as in Eqn.6, where z c is the confidence coefficient.S x and x are the sample variance and mean, respectively.E = 100z c S x x√ n(6)This property makes Monte Carlo simulation widely used in financial and engineering situations.Application exists in manned aviation as well.For instance, Gravio et.al. 30 applied Monte Carlo method to study safety performance of air traffic management system with about 1, 000 simulations.In order to statistically measure the impacts of parameters and uncertainties in various models in UTM system, it is necessary for an effective fast-time simulation platform to have the capability of Monte Carlo simulations. +V. Preliminary experimentsIn order to conceptually demonstrate how the fast-time simulation can be used for parameter and uncertainty studies, a prototype of a fast-time simulation platform for multiple sUAS operation was implemented for this work.Two sample experiments were conducted to demonstrate the use cases for the fast-time simulation.The experiment set-up is described in the first section. +A. Experiment set-upIn following experiments, a total of six sUASs were planned to fly cross the region.The flight plans were composed of a set of waypoints from origins and destinations with associated time a .For simplicity, the sUASs were assumed to be the same type.The sUASs were assumed to follow the flight plan with a desired vehicle ground speeds at 5 mps.In addition, a narrow rectangular north wind field is added to introduce errors and uncertainty into the simulations.It is assumed that the wind magnitudes in the rectangle follow a normal distribution with a mean value and standard deviation.The wind is the only uncertainty source in this experiment.As the number of Monte Carlo simulations is set to 1, 000, the wind magnitude will vary across different Monte Carlo simulations.Fig. 4 shows the flight plans and wind field.Besides vehicle trajectory models, traffic rules or collision avoidance rules need to be defined before any preliminary experiments can be conducted.In this prototype, conflict detection is assumed to be performed by vehicle-to-vehicle communication.Under this scheme, every sUAS is assumed to broadcast its current position and planned trajectory in future 5 second and the trajectory projection is assumed to be nominal.The detection range was arbitrarily set to 100 meters.The minimum separation requirement is also arbitrarily a A flight plan with waypoints and desired ground speeds is another option.assumed to be 10 meters for this initial study, which means if two vehicles are closer than 10 meters a loss of separation will be recorded.The minimum distance that triggers an avoidance maneuver is set to 20 meters and the minimum distance to the conflict point for an avoidance maneuver is defined to be 30 meters.A de-centralized collision avoidance algorithm is utilized here.The right of way was defined similar to ground transportation, which is that the vehicle coming from the right-hand side has the right of the way.The sUAS who doesn't have the right of way has to yield if there is any incoming conflict.Apparently, this simple rule negelects the head-on encounters as it is just used as an example for preliminary experiments.Three avoidance maneuvers were assumed: left or right turns with constant bank angles and hovering.Obviously, there are numerous parameters and options in setting up traffic rules.For instance, conflict detection can be done by onboard sensors or ground-to-vehicle communications, other than vehicle-to-vehicle communications defined in the prototype.Collision avoidance maneuvers can involve altitude changes and there also exist various methods including centralized algorithms for collision algorithms.Exploration of those parameters, options, and algorithms is out of the scope of this paper although it will be supported by the fast-time simulation architecture in future. +B. Impact of windIn this section, the default avoidance maneuver is defined to be a right turn.Table 4 shows the statistical measurements when different wind speeds were set.Three cases are presented.There is no wind in the first case.The average wind speeds in the second and third case were 3 mps and 5 mps, respectively, and the standard deviations were 2 and 3 mps in the second and third cases, respectively.Two types of metrics are presented.Loss of separation can be a safety metric.And extra flight distance and extra flight time are metrics related to enery consumption or efficiency.Percentage errors are calculated at 99% confidence level according to Eqn. 6.For instance, an error percentage of 3.5% means that it is 99% confident that the calculated mean will not differ by more that 3.5% from the truth.As shown in the table, Case 1 is an ideal case, where vehicles fly at their moderate speeds and there is no wind.Because there is no wind, no errors were introduced in Case 1.Therefore, the 1, 000 simulations are deterministic and identical simulations and standard deviations and errors are zeros.The number of loss of separation reflects how well the collision avoidance scheme works.Case 2 introduced moderate wind with moderate variation; there is still no loss of separation thanks to the conservative set up of collision avoidance scheme.However, the extra flight distance and associated errors increased, so did the variation.The percentage error of 0.17% for the extra flight distance means that there is 99% confidence that the true mean value is within 0.17% of 168.8.In Case 3, the wind's mean and variation were increased.Because the unexpected trajectory deviation increased due to the strong wind, loss of separation happened.A single loss of separation happened in 56 simulations out of a total of 1,000 simulations, which is reflected by the large error percentage of 23.8% at the 99% confidence level.This experiment showed that even with the same vehicles, equipages, and schedules, wind plays an important role on safety and energy consumption.When wind uncertainty is high, the likelihood of loss of separation increases and the required power consumption increases as well. +C. Impact of avoidance maneuverThe experiment in this section compared different avoidance maneuvers as an example of parameter studies that can be performed with this fast-time simulation capability.The purpose of this experiment is not to investigate or validate different maneuvers.The experiment is utilized to present a sample potential application that can be performed on such a fast-time simulation platform and to show the importance and effectiveness of this kind of platform for researching future UTM system.In the first case, a right turn was set as the default collision avoidance maneuver if there is any conflict.A left turn and hover were set as default maneuvers in the second and third cases, respectively.The mean and standard deviation of the wind field were defined to be 3 mps and 1 mps, respectively.Table .5 shows the statistical measurements when different avoidance maneuvers were used.Case 1 has been shown in previous section.Case 2 shows that combining left maneuver with the ground traffic right-ofway is really not a good option.It resulted in a loss of separation in 844 simulations.The percentage error shows that it is almost certain that the loss of separation will happen in any simulation.In Case 3, vehicles used hovers to avoid any detected conflicts.Loss of separation happened in 42 simulations.The extra flight distance is low c , and the extra flight time remains at a level similar to the other two cases. +VI. SummaryThis work presented key factors and requirements of an effective fast-time simulation platform for researching sUAS operations.It first briefly reviewed different capability requirements between UTM simulations where sUASs are dominant and ATM simulations where large size manned and unmanned fixed-wing aircraft are prevailing.Then a trajectory sensitivity study was conducted to demonstrate why the requirements for trajectory models are different in sUAS operations.The study showed that sUAS's trajectory was sensitive to many factors including wind gusts, vehicle speeds, and control systems.The resulted deviations are usually over several meters and should not be ignored when calculating and predicting trajectories for sUASs.The importance and effectiveness of the Monte Carlo method was discussed, which showed that Monte Carlo simulations are suited for UTM traffic problems that involve high-dimensional uncertainty/error sources.Finally, experiments were conducted to demonstrate the impact of wind on the evaluation of sUAS operations.The second experiment presented a sample application of parameter studies with different avoidance maneuvers.The proposed fast-time simulation capability can provide a comprehensive and statistical assessment for the sample parameter study.As a follow-up step, a cloud-based fast-time simulation platform is under development by NASA UTM research teams.To evaluate safety and efficiency metrics for multiple sUAS operations at low altitude airspace, this fast-time simulation capability will address aforementioned requirements by including various sUAS trajectory models and Monte Carlo simulation capability and support studies of paramters, models, rules, and policies.Figure 1 .1Figure 1.Trajectories at various cross wind speeds (vehicle speed = 5 m/s) +Figure 2 .Figure 3 .23Figure 2. Trajectories at various vehicle desired ground speeds ( cross wind speed = 5 m/s) +Figure 4 .4Figure 4. Flight plans and wind field setup +Table 1 .1Brief comparison of functionalities in simulationsSimulationsCTAS/ACESFACETMueller's Jenie'sReflectionUTM required attributesMaxium number of vehicles per scenario> 100> 1002> 1001> 100Fidelity of vehicle modelsmediummediumlowlowhighmedium+Vehicle's controller modeled?××××along-trackalong-trackWind effectalong-trackalong-track××+cross-track+cross-track+vertical+verticalLimited flight duration?×××××Capability of Monte Carlo simulations?×××Small UAS model included?××Collision avoidance algorithm included?×× +Table 3 .3PD controller gains k p k d k roll,p k roll,d k pitch,p k pitch,d7.5 4.24.50.54.50.5 +Table 4 .4Statistical measurements under various wind conditionswind speedloss of separationextra flight distance (m)extra flight time (s)mean std. mean std. error(%) mean std. error(%) mean std. error(%)Case 100000 b165.50.00.031.00.00.0Case 231000168.83.60.1731.00.030.01Case 3520.110.3123.8183.7 27.11.231.33.20.82 +Table 5 .5statistical measurements with various avoidance maneuversavoidanceloss of separationextra flight distance (m)extra flight time (s)maneuver mean std. error(%) mean std. error(%) mean std. error(%)Case 1 right turn000168.83.60.1731.00.030.01Case 2left turn0.847 0.363.4671.023.32.79.53.43.0Case 3hover0.040.2038.95.954.15.620.94.41.72 + of 10 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on June 9, 2017 | http://arc.aiaa.org| DOI: 10.2514/6.2017-3073 + of 10 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on June 9, 2017 | http://arc.aiaa.org| DOI: 10.2514/6.2017-3073 + of 10 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on June 9, 2017 | http://arc.aiaa.org| DOI: 10.2514/6.2017-3073 + of 10 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on June 9, 2017 | http://arc.aiaa.org| DOI: 10.2514/6.2017-3073 + of 10 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on June 9, 2017 | http://arc.aiaa.org| DOI: 10.2514/6.2017-3073 + American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on June 9, 2017 | http://arc.aiaa.org| DOI: 10.2514/6.2017-3073 + b The error percentage was set to zero based on physical meaning, because division by zero happens if following the formula. + Downloaded by NASA AMES RESEARCH CENTER on June 9, 2017 | http://arc.aiaa.org| DOI: 10.2514/6.2017-3073 + + + + + + + + + FAA/NWS aviation route forecast /ARF/ development + + TMitchell + + 10.2514/6.1982-13 + + + 20th Aerospace Sciences Meeting + + American Institute of Aeronautics and Astronautics + 2016-2036, 2016 + + + Tech. Rep. FY + "FAA Aerospace Forecast," Tech. Rep. FY2016-2036, 2016. + + + + + Size (Micro, Mini, Nano), Payload (Sensor, Telemetry System, Camera, Videoscreen, NBC Detection, SAR, Others), Type, Propulsion Systems, & Region -Global Forecasts to 2020 + 10.3030/720041 + + + + Small Drones Market by Application (Defense, Commercial) + + 2016. 2016 + Publications Office of the European Union + + + "Small Drones Market by Application (Defense, Commercial), Size (Micro, Mini, Nano), Payload (Sensor, Telemetry System, Camera, Videoscreen, NBC Detection, SAR, Others), Type, Propulsion Systems, & Region -Global Forecasts to 2020," http://www.marketsandmarkets.com/Market-Reports/small-uav-market-141134567.html, 2016, [online; accessed Oc- tober 2016]. + + + + + Unmanned Aircraft System Traffic Management (UTM) Concept of Operations + + PKopardekar + + + JRios + + + TPrevot + + + MJohnson + + + JJung + + + JERobinson + + + + 16th AIAA Aviation Technology, Integration, and Operations Conference + Washington, D.C. + + June 2016 + + + + Kopardekar, P., Rios, J., Prevot, T., Johnson, M., Jung, J., and Robinson, J. E., "Unmanned Aircraft System Traf- fic Management (UTM) Concept of Operations," 16th AIAA Aviation Technology, Integration, and Operations Conference, Washington, D.C., 13-17 June 2016. + + + + + Flight Demonstration of Unmanned Aircraft System (UAS) Traffic Management (UTM) at Technical Capa... + + PKopardekar + + 10.2514/6.2020-2851.vid + NASA TM-2014-218299 + + 2014 + American Institute of Aeronautics and Astronautics (AIAA) + + + Tech. Rep. + Kopardekar, P., "Unmanned Aerial System Traffic Management (UTM): Enabling Low-altitude Airspace and UAS Op- erations," Tech. Rep. NASA TM-2014-218299, 2014. + + + + + Design of Center-TRACON Automation System + + HErzberger + + + TJDavis + + + SMGreen + + + + AGARD Meeting on Machine Intelligence in Air Traffic Management + Berlin, Germany + + May 1993 + + + + Erzberger, H., Davis, T. J., and Green, S. M., "Design of Center-TRACON Automation System," AGARD Meeting on Machine Intelligence in Air Traffic Management, Berlin, Germany, 11-14 May 1993. + + + + + FACET: Future ATM Concepts Evaluation Tool + + KarlDBilimoria + + + BanavarSridhar + + + ShonRGrabbe + + + GanoBChatterji + + + KapilSSheth + + 10.2514/atcq.9.1.1 + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 9 + 1 + + + American Institute of Aeronautics and Astronautics (AIAA) + + + Bilimoria, K., Sridhar, B., Chatterji, G., Sheth, K., and Grabbe, S., "FACET: Future ATM Concepts Evaluation Tool," + + + + + + + Usa/ + + + + Europe Air Traffic Management R&D Seminar + + + June 2000 + Napoli, Italy + + + USA/Europe Air Traffic Management R&D Seminar , Napoli, Italy, 13-16 June 2000. + + + + + Fast-Time NAS Simulation System for Analysis of Advanced ATM Concepts + + DouglasSweet + + + VikramManikonda + + + JesseAronson + + + KarlinRoth + + + MatthewBlake + + 10.2514/6.2002-4593 + + + AIAA Modeling and Simulation Technologies Conference and Exhibit + Monterey, California + + American Institute of Aeronautics and Astronautics + August 2002 + + + + Sweet, D. N., Manikonda, V., Aronson, J. S., Roth, K., and Blake, M., "Fast-time Simulation System For Analysis of Advanced Air Transportation Concepts," AIAA Modeling and Simulation Technologies Conference and Exhibit, Monterey, California, 5-8 August 2002. + + + + + Modeling and Simulation for UAS in the NAS + + FWieland + + + SAyyalasomayajula + + + RMooney + + NASA/CR-2012-NN11AQ74C + + September 2012 + + + Tech. Rep. + Wieland, F., Ayyalasomayajula, S., and Mooney, R., "Modeling and Simulation for UAS in the NAS," Tech. Rep. NASA/CR-2012-NN11AQ74C, September 2012. + + + + + Call for Papers + + MJohnson + + + EMueller + + + CSantiago + + 10.1027/2192-0923/a000067 + + + Aviation Psychology and Applied Human Factors + Aviation Psychology and Applied Human Factors + 2192-0923 + 2192-0931 + + 4 + 2 + + 2015 + Hogrefe Publishing Group + Lisbon, Portugal + + + 11th USA/ + 9 Johnson, M., Mueller, E., and Santiago, C., "Characteristics of a Well Clear Definition and Alerting Criteria for En- counters between UAS and Manned Aircraft in Class E Airspace," 11th USA/Europe Air Traffic Management Research and Development Seminar , Lisbon, Portugal, 2015. + + + + + Investigating Detect-and-Avoid Surveillance Performance for Unmanned Aircraft Systems + + ChunkiPark + + + SeungManLee + + + EricRMueller + + 10.2514/6.2014-2413 + + + 14th AIAA Aviation Technology, Integration, and Operations Conference + Atlanta, GA + + American Institute of Aeronautics and Astronautics + June 2014 + + + + Park, C., Lee, S. M., and Mueller, E. R., "Investigating Detect-and-Avoid Surveillance Performance for Unmanned Aircraft Systems," 14th AIAA Aviation Technology, Integration, and Operations Conference, Atlanta, GA, 16-20 June 2014. + + + + + A Fuzzy Logic Approach for Low Altitude UAS Traffic Management (UTM) + + BrandonCook + + + KellyCohen + + + EladHKivelevitch + + 10.2514/6.2016-1905 + + + AIAA Infotech @ Aerospace + San Diego, California + + American Institute of Aeronautics and Astronautics + January 2016 + + + + AIAA Science and Technology Forum and Exposition 2016 + Cook, B., Cohen, K., and Kivelevitch, E., "A Fuzzy Logic Approach For Low Altitude UAS Traffic Management (UTM)," AIAA Science and Technology Forum and Exposition 2016 , San Diego, California, 4-8 January 2016. + + + + + Three-Dimensional Velocity Obstacle Method for Uncoordinated Avoidance Maneuvers of Unmanned Aerial Vehicles + + YazdiIJenie + + + Erik-JanVan Kampen + + + CornelisCDe Visser + + + JoostEllerbroek + + + JaccoMHoekstra + + 10.2514/1.g001715 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 39 + 10 + + 2016 + American Institute of Aeronautics and Astronautics (AIAA) + + + Jenie, Y. I., Kampen, E. V., Visser, C. C., Ellerbroek, J., and Hoekstra, J. M., "Three-Dimensional Velocity Obstacle Method for Uncoordinated Avoidance Maneuvers of Unmanned Aerial Vehicles," Journal of Guidance, Control, and Dynamics, Vol. 39, No. 10, 2016. + + + + + Airspace Encounter Models for Estimating Collision Risk + + MykelJKochenderfer + + + MatthewW MEdwards + + + LeoPEspindle + + + JamesKKuchar + + + JDanielGriffith + + 10.2514/1.44867 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 33 + 2 + + June 29 -July 2 2009 + American Institute of Aeronautics and Astronautics (AIAA) + Napa, CA + + + Eighth USA + Kochenderfer, M., Espindle, L. P., Edwards, M., Kuchar, J., and Griffith, J. D., "Airspace Encounter Models for Con- ventional and Unconventional Aircraft," Eighth USA/Europe Air Traffic Management R&D Seminar , Napa, CA, June 29 - July 2 2009. + + + + + Airspace Encounter Models for Estimating Collision Risk + + MykelJKochenderfer + + + MatthewW MEdwards + + + LeoPEspindle + + + JamesKKuchar + + + JDanielGriffith + + 10.2514/1.44867 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 33 + 2 + + 2010 + American Institute of Aeronautics and Astronautics (AIAA) + + + Kochenderfer, M., Edwards, M., Espindle, L. P., Kuchar, J., and Griffith, J. D., "Airspace Encounter Models for Esti- mating Collision Risk," AIAA Journal of Guidance, Control, and Dynamics, Vol. 33, No. 2, 2010. + + + + + Multi-Rotor Aircraft Collision Avoidance using Partially Observable Markov Decision Processes + + EricRMueller + + + MykelKochenderfer + + 10.2514/6.2016-3673 + + + AIAA Modeling and Simulation Technologies Conference + Washington, D.C. + + American Institute of Aeronautics and Astronautics + June 2016 + + + + Mueller, E. and Kochenderfer, M. J., "Multi-Rotor Aircraft Collision Avoidance using Partially Observable Markov Decision Processes," AIAA Modeling and Simulation Technologies Conference, Washington, D.C., 13-17 June 2016. + + + + + Simulation Comparison of Collision Avoidance Algorithms for Small Multi-Rotor Aircraft + + EricRMueller + + + MykelKochenderfer + + 10.2514/6.2016-3674 + + + AIAA Modeling and Simulation Technologies Conference + Washington, D.C. + + American Institute of Aeronautics and Astronautics + June 2016 + + + + Mueller, E. and Kochenderfer, M. J., "Simulation Comparison of Collision Avoidance Algorithms for Multi-Rotor Air- craft," AIAA Modeling and Simulation Technologies Conference, Washington, D.C., 13-17 June 2016. + + + + + An Autonomous Autopilot Control System Design for Small Scale UAVs + + CIppolito + + + + NASA Ames + + December 2006 + + + Tech. Rep. QSS Group + Ippolito, C., "An Autonomous Autopilot Control System Design for Small Scale UAVs," Tech. Rep. QSS Group, NASA Ames, December 2006. + + + + + Time Controlled Descent Guidance Algorithm for Simulation of Advanced ATC Systems + + HLee + + + HErzberger + + + + NASA Ames Research Center + + August 1983 + Moffett Field, CA + + + Tech. Rep. Tech. Rep. NASA Technical Memorandum 84373 + Lee, H. and Erzberger, H., "Time Controlled Descent Guidance Algorithm for Simulation of Advanced ATC Systems," Tech. Rep. Tech. Rep. NASA Technical Memorandum 84373, NASA Ames Research Center, Moffett Field, CA, August 1983. + + + + + Ground based concept for time control of aircraft entering the terminal area + + HErzberger + + + JChapel + + 10.2514/6.1985-1888 + + + Guidance, Navigation and Control Conference + Snowmass, CO + + American Institute of Aeronautics and Astronautics + August 1985 + + + + Erzberger, H. and Chapel, J., "Ground Based Concept for Time Control of Aircraft Entering the Terminal Area," AIAA Guidance, Navigation, and Control Conference, Snowmass, CO, 19-21 August 1985. + + + + + Trajectory Synthesis for Air Traffic Automation + + RhondaSlattery + + + YiyuanZhao + + 10.2514/2.4056 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 20 + 2 + + 1997 + American Institute of Aeronautics and Astronautics (AIAA) + + + Slattery, R. and Zhao, Y., "Trajectory Synthesis for Air Traffic Automation," Journal of Guidance, Control, and Dy- namics, Vol. 20, No. 2, 1997. + + + + + Conflict Detection and Resolution In the Presence of Prediction Error + + HErzberger + + + RPaielli + + + DIsaacson + + + MMEshow + + + + Europe Air Traffic Management Research and Development Seminar + + + June 1997 + Saclay, France + + + 1st USA/ + Erzberger, H., Paielli, R., Isaacson, D., and Eshow, M. M., "Conflict Detection and Resolution In the Presence of Prediction Error," 1st USA/Europe Air Traffic Management Research and Development Seminar , Saclay, France, 17-20 June 1997. + + + + + Improvement of Trajectory Synthesizer for Efficient Descent Advisor + + MinXue + + + HeinzErzberger + + 10.2514/6.2011-7020 + + + 11th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Virginia Beach, VA + + American Institute of Aeronautics and Astronautics + September 2011 + 23 + + + Xue, M. and Erzberger, H., "Improvement of Trajectory Synthesizer for Efficient Descent Advisor," 11th AIAA Aviation Technology, Integration, and Operations Conference(ATIO), Virginia Beach, VA, 20-22 September 2011. 23 + + + + + Quadrotor Dynamics and Control + + RWBeard + + + October 2008 + + + Tech. rep. + Beard, R. W., "Quadrotor Dynamics and Control," Tech. rep., October 2008. + + + + + Full control of a quadrotor + + SamirBouabdallah + + + RolandSiegwart + + 10.1109/iros.2007.4399042 + + + 2007 IEEE/RSJ International Conference on Intelligent Robots and Systems + San Diego, CA + + IEEE + Oct 29 -Nov 2 2007 + 25 + + + Bouabdallah, S. and Siegwart, R., "Full Control of a Quadrotor," Proceedings of the 2007 IEEE/RSJ International Conference on Intelligent Robots and Systems, San Diego, CA, Oct 29 -Nov 2 2007. 25 + + + + + Control of a quadrotor helicopter using visual feedback + + EAltug + + + JPOstrowski + + + RMahony + + 10.1109/robot.2002.1013341 + + + Proceedings 2002 IEEE International Conference on Robotics and Automation (Cat. No.02CH37292) + 2002 IEEE International Conference on Robotics and Automation (Cat. No.02CH37292)Washington, D.C. + + IEEE + May 2002 + + + Altug, E., Ostrowski, J. P., and Mahony, R., "Control of a Quadrotor Helicopter Using Visual Feedback," Proceedings of the 2002 IEEE International Conference on Robotics & Automation, Washington, D.C., May 2002. + + + + + The Monte Carlo Method + + NicholasMetropolis + + + SUlam + + 10.1080/01621459.1949.10483310 + + + Journal of the American Statistical Association + Journal of the American Statistical Association + 0162-1459 + 1537-274X + + 44 + 247 + + 1949 + Informa UK Limited + + + Metropolis, N. and Ulam, S., "The Monte Carlo Method," Journal of the American Statistical Association, Vol. 44, No. 247, 1949. + + + + + Implementing quasi-Monte Carlo simulations with linear transformations + + PiergiacomoSabino + + 10.1007/s10287-009-0104-9 + + + Computational Management Science + Comput Manag Sci + 1619-697X + 1619-6988 + + 8 + 1-2 + + 2011 + Springer Science and Business Media LLC + + + Sabino, P., "Implementing Quasi-Monte Carlo Simulation with Linear Transformations," Computational Management Science, Vol. 8, No. 1-2, 2011. + + + + + Random Number Generation and Quasi-Monte Carlo Methods + + HaraldNiederreiter + + 10.1137/1.9781611970081 + + + SIAM + + 1992 + Society for Industrial and Applied Mathematics + + + Niederreiter, H., Random number generation and quasi-Monte Carlo methods, SIAM, 1992. + + + + + Determining the Number of Iterations for Monte Carlo Simulations of Weapon Effectiveness + + MorrisRDriels + + + YoungSShin + + 10.21236/ada423541 + + April 2004 + Defense Technical Information Center + Monterey, CA + + + Naval Postgraduate School + + + Tech. Rep. Technical Report + Driels, M. R. and Shin, Y. S., "Determining the Number of Iterations for Monte Carlo Simulations of Weapon Effective- ness," Tech. Rep. Technical Report, Naval Postgraduate School, Monterey, CA, April 2004. + + + + + Overall safety performance of Air Traffic Management system: Forecasting and monitoring + + GiulioDi Gravio + + + MaurizioMancini + + + RiccardoPatriarca + + + FrancescoCostantino + + 10.1016/j.ssci.2014.10.003 + + + Safety Science + Safety Science + 0925-7535 + + 72 + + 2015 + Elsevier BV + + + Gravio, G. D., Mancini, M., Patriarca, R., and Costantino, F., "Overall safety performance of Air Traffic Management System," Safety Science, Vol. 72, 2015. + + + + + + diff --git a/file814.txt b/file814.txt new file mode 100644 index 0000000000000000000000000000000000000000..395fe75e1ee270ae1d845f72f79e23ce50662c93 --- /dev/null +++ b/file814.txt @@ -0,0 +1,1171 @@ + + + + +I. IntroductionIn the concepts of unmanned aircraft system traffic management (UTM) 1 and urban air mobility (UAM), 2 aerial vehicles are envisioned to operate much closer to each other in low-altitude airspace than in conventional high-altitude air traffic system.Because of the complex environment (wind, terrain, etc.), the high-density and low-altitude operations not only impose challenges to the vehicle design, such as quiet electric engines, long endurance and safe battery, and fault-tolerant control system, but also bring difficulties into the development of a safe yet efficient low-altitude air traffic system, such as communication, navigation, separation definitions, collision avoidance algorithms/protocols, flight planning and scheduling, risk/safety assessment, and emergency plans.In United States, it took more than 50 years (from 1960 to 2015) for the number of commercial manned aircraft 3 to increase from 2,135 to 6,876 and for the annual fatal accident rate to decrease from 30 per million departures to less than 0.5 per million departures. 4 However, for small unmanned aerial vehicles (UAVs), it only took one year for the number of registered users to reach 626,000 with estimated 1.1 million drones. 5t's unlikely that stakeholders will have decades to evolve the traffic system and mitigate all challenges without significantly impeding the growth of UTM and UAM operations.Previous work 6 evaluated existing tools [7][8][9][10][11] and suggested that a fast-time simulation evaluation capability is needed to study the high-density and low-altitude air traffic system.The Flexible engine for Fast-time evaluation of Flight environments (Fe 3 or Fe-cubed) is such a simulation tool developed at NASA to provide the capability of statistically analyzing the high-density and low-altitude traffic system.With this capability, stakeholders can study the impacts of critical components (such as wind, surveillance, communication, collision avoidace, traffic rules, energy consumption, etc.) in the lowaltitude high-density traffic system, gain insights and help define requirements, policies, and protocols for a safe and efficient traffic system, and assess operational risks and optimize flight schedules.To achieve these goals, Fe 3 is designed to have: 1) the capability of Monte Carlo simulations to support statistical or uncertainty evaluations; 2) high-performance computation ability to finish such simulations in a limited time frame; 3) generalized and high-fidelity models for trajectory calculation of various aerial vehicles, including multi-copter, fixed-wing, and hybrid aerial vehicles; 4) generalized models to accommodate different collision avoidance methods/traffic rules/policies/protocols for comparison; 5) the capability to incorporate low altitude wind models; and 6) the capability to incorporate accurate communication and sensor models.This paper provides an introduction to the Fe 3 simulator including its architecture, build-in models, and sample applications.It starts from the general architecture, and then various models used in the simulator are described.Performance data for the Fe 3 is also presented.Finally sample studies are demonstrated. +II. Fe 3 ArchitectureA conceptual architecture of Fe 3 's core simulation engine is presented in Fig. 1.It is composed of two main functions: trajectory generation and collision avoidance.The details of these two functions are described in the subsections.Databases that are involved are vehicle model parameters, communication and sensor models, terrain, airspace constraints, and avoidance logic/rules.Inputs to Fe 3 include flight plans, wind/weather, and geo-fence data.The simulation configuration input data, which isn't shown in this figure, defines simulation related parameters, such as the number of Monte Carlo simulations, temperature and humidity, the type of avoidance algorithm, and, if needed, it also provide the opportunity of overwriting some vehicle or device model parameters for research purposes.[14][15] +A. Trajectory ModelsTo keep the fidelity at a high level, 6-DOF vehicle trajectory models are implemented in the simulator.To minimize the number of models and optimize the computational performance, vehicle models are categorized into several types, such as the quad-rotor, fixed-wing, and hybrid.Autonomous controllers are included to form complete trajectory models.As an example, Eqns. 1, 2, 3, and 4 present the generalized multi-copter model for vehicle trajectory (in the inertial frame), forces (in the body frame), kinematics (in the body frame), and the moments (in the body frame), respectively. 16x, t, and h are vehicle North, East, and altitude coordinates in the inertial frame.u, v, and w are vehicle speeds in the body frame.φ, θ, and ψ are the vehicle's roll, pitch, and yaw angles, respectively.And p, q, and r are the angular rates.(F x , F y , F z ) are body axis aerodynamic forces and (M φ , M θ , M ψ ) are body axis moments.J x , J y , and J z are the moments of inertia about the principle axes in the body frame.   ẋ ẏ ḣ   =    cθcψ sφsθcψ -cφsψ cφsθcψ + sφsψ cθsψ sφsθsψ + cφcψ cφsθsψ -sφcψ sθ -sφcθ -cφcθ       u v w   (1)   u v ẇ   =    rv -qw pw -ru qu -pv    +    -gsinθ gcosθsinφ gcosθcosφ    + 1 m    F x F y F z   (2)   φ θ ψ   =    1 sinφtanθ cosφtanθ 0 cosφ -sinφ 0 sinφ cosθ cosφ cosθ       p q r   (3)   ṗ q ṙ   =    Jy-Jz Jx qr Jz-Jx Jy pr Jx-Jy Jz pq    +    1 Jx M φ 1 Jy M θ 1 Jz M ψ   (4)For multi-copters, the forces include: aerodynamic drags caused by vehicle's motion and wind, motorgenerated forces that are always in z-direction of the body frame, and the gravity force, which aligns with the z-direction of the inertial frame.The force produced by a motor can be expressed as simple as a function of the motor rotational speed 16, 17 as in Eqn. 5, where k f is a thrust coefficient for a given motor.A more sophisticated quadratic model for a fixed-pitch motor, according to Meyer's work, 18 can be written as Eqn.6, where v 1 is the free stream velocity and C T,i are thrust coefficients.Wind tunnel tests were typically conducted to identified the thrust coefficients for small UAVs. 19,20 t should be noted that these coefficients are proportional to air density as well.F motor = k f • ω 2(5)F motor = C T,0 ω 2 + C T,1 v 1 ω±C T,2 v 1 2(6)The drag forces are usually written as Eqn. 7, where R b i is the rotation matrix from the inertial frame to the vehicle body frame.The v i w is the wind vector in the Earth frame and the v i is essencially [ ẋ ẏ ḣ].The C d contains drag coefficients that are obtained in wind tunnel tests as well.F drag = -C d • R b i • v i -v i w •( v i -v i w )(7)Eqns. 8 and 9 present a sample outer loop controller associated with the multi-copter using the small angle assumption, which transfers the desired way points (x d , y d , and h d ) to the vehicle attitude commands (roll φ and pitch θ in this case).ẍd ÿd = k p (x d -x) + k d ( ẋd -ẋ) k p (y d -y) + k d ( ẏd -ẏ)(8)φ d θ d = m F z -sinψ -cosψ cosψ -sinψ -1 ẍd ÿd(9) +B. Energy Consumption ModelsFrom mechanical perspective, in the case of rotational motion, the analogous calculation for power can be simplified as the product of torque multiplied by the rotational speed 21 as shown in Eqn. 10, where k m is assumed to be a coefficient between motor rotational speed and motor-generated torque.P motor = τ •ω = k m • ω 3(10)A more sophisticated model for a DC motor power consumption is calculated as a product of the voltage V in and current I as shown in Eqn.11.Eqn. 12 shows the expressions for the current, 22 where µ, α M , β M , R A , and k t are motor specific parameters, and ω is the motor rotational speed.P motor = V in • I(11)I = x x + x 2 x + µα M R A , x x = µV in β M -k t ω 2R A .(12)It is noted that, for a given motor, both models show that the consumed energy is mainly a function of the motor's rotational speed, which is driven by the desired thrust and moments.Apparently, even with the same required thrust and moments, different atmospheric conditions, such as wind and air density, result in different rotational speeds, and therefore yield different energy consumptions. +C. Collision Avoidance ModelsMany collision avoidance algorithms have been developed in past decades, 10,[23][24][25][26] such as the potential field (PF) method, 24 the horizontal vector turn (HVT) method, 25 the model predictive control (MPC) method, the protocol-based methods 10,26,27 and the Partially Observable Markov Decision Process(POMDP) method. 28he inputs are typically ownship's position, speed, and intent, and intruder's position, speed, and intent, although some many require information from multiple intruders.The outputs are typically desired way points and/or control commands at next time step.In order to generalize collision avoidance algorithms, slightly different from Albaker's grouping, 29 the avoidance algorithms are categorized to three types in this simulator: trajectory-projection (TP) based, off-line table (OT) based, and force field (FF) based. +Figure 2. Trajectory Projection Based Collision Avoidance AlgorithmA trajectory projection based method (shown in Fig. 2) is defined as a method that predicts intruder's trajectory based on intruder's current states (such as position and velocity) and ranks/identifies resolution(s) with predefined maneuver options and rules.Sample trajectory projection methods are the Detect and Avoid Alerting Logic for Unmanned Systems (DAIDALUS), 27 the Generic Resolution Advisor and Conflict Evaluator (GRACE), 30 and the Auto-Resolver in the Advanced Airspace Concept (AAC). 31The HVT can be treated as a special case of the trajectory projection based method with a unique closed form resolution.The off-line table based method uses predefined tables that are usually empirically generated and abandons the trajectory prediction function.Once the intruder's states (positions and velocities) are received, this group of methods will search their off-line tables and identify the best resolution maneuver(s) (as shown in Fig. 3).Sample off-line table based methods are ACAS-Xu, 32 POMDP, 28 and the Fuzzy-logic based method. 10orce field based methods usually use attractive forces (to stay on original path or to follow original destination) and repulsive forces (to avoid potential conflicts) to yield the control command for the next move of the ownship.The weights of different forces are tuned offline to balance the trade-offs among different forces.Sample methods are the potential field method 33 and velocity obstacle method.The communication and detection capability faces a great challenge in the low-altitude high-density UTM-like operation.The power requirement for the communication is restricted by the vehicle power and size.High transmission power increases the communication channel load and the signal interference, and essentially reduces the signal reception probability.Since the high-density operation leaves a limited space for vehicles to avoid the potential conflicts and the accuracy and update frequency of the intruder's states greatly affect the conflict avoidance algorithm's performance, the communication and detection capability plays a critical role in the analysis of the UTM-type traffic system.Trying not to congest the existing Automatic Dependent Surveillance-Broadcast (ADS-B) channel used for conventional aviation, 35 researchers explored alternative communication devices for the UTM operations 36 such as Dedicated Short Range Communications(DSRC) 37 and L-band Digital Aeronautical Communications System (LDACS). 38Regardless the differences, the reception probability models of various communication devices are normally functions of transmission power and communication density/channel load in a open space for a given orientation.For instance, for DSRC, both the mathematical model 39 and empirical model 40 show that the reception probability model of the DSRC is a function of transmission power, which follows the general Nakagami model, 41 and communication density, which is a product of transmission frequency, vehicle density, and transmission power, for a given message size.Fig. 5 shows a sample reception probability at various communication densities when the message size is 200 bytes.For on-board sensors, researchers are testing and studying devices like Light Detection and Ranging (LIDAR) 22 and Echodyne radar. 42In the simulator, these sensors are assumed to follow Gaussian or Markov probability models for measurements like azimuth angle, elevation angle, detected distance, positions, and velocities with experimentally validated mean, standard deviations, and tau (if applicable). +E. Wind modelsThe High Resolution Rapid Refresh (HRRR) wind model 43 generated by National Oceanic and Atmospheric Administration (NOAA) provides wind information at the low-altitude airspace with great spatial and temporal resolutions.The HRRR wind data covers two altitudes at 10m and 80m.The spatial resolution is 3 km by 3km and its temporal resolution is 15 minutes.Although there exist many other wind model products like the California State University Mobile Atmospheric Profiling System (CSU-MAPS), 44 the wind model in Fe 3 is defined as a spatially discretized database with turbulence intensity/uncertainty associated with every location.The wind condition at each location is defined as a statistical distribution with an empirical mean and deviation/intensity.At low altitude, typical relationship between the turbulence intensity σ w and the wind speed at 20 feet W 20 follows Eqn.13: 45 σ w = 0.1W 20 (13)Since the expected flight times for most UTM and UAM operations are less than one hour, the temporal variation wasn't considered at current stage. +III. ImplementationTo meet the demanding computational performance requirements of Monte Carlo simulations of the traffic system, Fe 3 is highly-parallelized using the CUDA programming language on graphics processing units (GPUs).It is deployed on the Amazon Web Service (AWS) cloud for scalability needs such that the number of GPU instances can be dynamically deployed based on simulation needs. +A. Cloud ArchitectureFigure 6 provides the cloud architecture for Fe 3 .The EC2 web instance handles the web UI and final notification.The AWS SQS is used to maintain two queues for auto scaling and notification, respectively.Fe 3 engine is deployed on AWS GPU instance(s), where the number of instances is determined by the auto scaling.As described previously, three input files are fed into the simulator via the web-based user input interface: flight plan/schedule, wind data, and simulation configuration.Once these inputs are submitted, a job is formed and pushed into the job request queue and waits to be assigned to the next available GPU instances with deployed Fe 3 engine.When the job is done, a notification response will be pushed into the notification queue and the simulation results are stored.The user will then receive the notification and check the results via Fe 3 web-based interface and visualization tool or directly download the final data for further investigation. +B. Web-based output metricsWeb-based output and visualization are implemented to facilitate post analysis.Figures 7 and8 present sample output tables of the metrics that are computed and shown by the simulator after each run.Fig. 7 shows the summary of the overall statistics using metrics like the probability of loss of separation, extra flight distance, number of alerts, trajectory mean deviation, etc.And Fig. 8 provides detailed statistics at the simulation level.With this web-based out interface, users can select any simulation of interest on this table and visualize the playback of this specific simulation. +C. Computational performanceTo show the computing capability of Fe 3 , Table 1 lists sample running times under different scenarios with various combinations of the number of Monte Carlo simulations and the number of small UAVs involved simultaneously.It shows that with 6-DOF trajectory models and various wind conditions, Fe 3 were able to +IV. ExperimentsThis section presents studies to show how the Fe 3 simulation capability can be used to support the system analysis of the new traffic system.In the following experiments, a 6-DOF quad-rotor model with a proportional derivative controller is used as the the small UAV's model.The DSRC reception probability model (shown in Fig. 5) is assumed as the communication method between vehicles.A well-clear violation (WCV) is arbitrarily defined, where the modified horizontal distance (HMOD) is 10 meters and the modified tau (TMOD) is set to zero second.The trajectory projection (TP) based conflict avoidance algorithm is used in the experiments running at an update rate of 2 Hz.In the avoidance algorithm, extra buffers of 10 meters and 25 seconds are added to HMOD and TMOD, respectively.In each traffic scenario, flights are designed to cross from one side to the opposite side for a given area randomly to create conflict traffic scenarios. +A. Pairwise encountersLike in most conflict avoidance algorithm studies, an experiment of sets of pairwise encounters is presented to examine the performance of the avoidance algorithm.In this experiment, a range of relative headings at the Closest Point of Approach (CPA) and a range of vehicle speeds for both ownship and intruder were tested.There were a total of 1,714 encounter scenarios in these experiments with various combinations of relative headings and vehicle speeds.Nominal trajectory duration was set to 120 seconds.Scenarios did not include wind.Also perfect communication and navigation were assumed.The results show that no WCV happened in this experiment, which means this TP collision avoidance algorithm resolved all the conflicts and the algorithm works well in pairwise encounter scenarios.To further investigate the algorithm, separation-deviation measurement are shown in Fig. 9, where the horizontal axis is the mean deviation from the original desired path and the vertical axis is the product of the WCV and the inverse CPA distance.Figure 9 shows all values on the vertical axis are less than 1 since conflicts were resolved and CPA distances are larger than WCV. +B. Communication capabilityIn this section, three communication transmission ranges, 300, 600 and 1,000 meters, and two models, "deterministic" and "probabilistic" were investigated.The "deterministic" model assumes 100% reception within the given range, whereas the "probabilistic" model assumes that the reception probability within the range follows the general Nakagami model shown in Fig. 5. Fig. 10 presents a sample of system safety analysis using the average count of loss of separation (LOS) at various communication capabilities and traffic densities.In this figure, solid curves represent the cases with deterministic models and dashed curves are the cases using probabilistic models.The vertical bars denote the stand deviations σ of the loss of separation counts at different traffic density levels over thousands of Monte Carlo simulations.First, it is noticed that although the conflict avoidance worked perfectly in the pairwise scenarios where the traffic density is low, it has difficulty when the density gets high due to secondary and consequential conflicts.This shows that verifying conflict avoidance algorithms in low density and high density operational environments may be different.Second, with the experimental setup, the overall safe capacity "bottleneck" is formed at 20 vehicles per nmi 2 .In the deterministic cases, loss of separation does not happen until the traffic density reaches about 30 vehicles per nmi 2 .With the probabilistic model, loss of separation happens at a lower traffic density level and the safe capacity upper bound is reduced to 20 vehicles per nmi 2 , which is mainly caused by the reduced accuracy/update rate due to the probabilistic models.Third, in the deterministic cases, with an increased transmission range of 600 meters (shown as the magenta solid curve), the safety capacity upper bound is increased to 35 vehicles per nmi 2 .Mixed situations observed at high density levels in deterministic cases might result from the local selection mechanism in the It is also noted that the loss of separation count is not highly correlated with the traffic density.For instance, the loss of separation count is lower at 50 vehicles per nmi 2 when compared against lower traffic density levels.This indicates that traffic density might not be a good metric for measuring the complexity/safety of a high-density autonomous traffic system.The large range of standard deviation values associated with the probabilistic models further show low correlation between the traffic density and system safety. +C. Wind uncertaintyFig. 11 shows examples with impact using the "600m" and "deterministic" communication model used in previous section.The black curve shows the "600m" and "deterministic" case without wind as in Fig. 10.The blue curve presents the case with average cross wind speed of 5 mps and turbulence intensity or standard deviation of 0.5 mps, and the magenta curve corresponds to a stronger wind condition.It is shown that introducing wind decreases the upper bound of traffic density for safe operations from ∼26 aircraft per nmi 2 to less than 15 aircraft per nmi 2 , which presents a great change from the system safety perspective.The most likely reason is that the wind reduced the accuracy of the trajectory prediction in the collision avoidance algorithm and essentially affected the avoidance algorithm's performance and the system safety measurement. +D. Interaction with Manned AircraftThis section introduces two manned aircraft into the 20 aircraft scenario discussed earlier (corresponding to the case with ∼22 aircraft per nmi 2 ).There are three main differences for manned aircraft operations in this experiment.First, these manned aircraft have the right of the way, which means that it is the sUAV's responsibility to sense and avoid manned aircraft.Second, the WCV is set to 200 meters for manned aircraft (larger than the WCV defined for sUAV).Third, manned aircraft fly faster than the sUAVs.In this experiment, the manned aircraft are assumed to fly at 30 mps compared to 5-15 mps sUAV operations.Figure 12 shows the impact of the transmission power range.Increasing transmission power range clearly reduced the probability of loss of separation in this case.The probability of loss of separation decreased from 90.9% in the 300-meter case, to 21.9% in the 600-meter case, and to 0.4% in the 1,000-meter case.Recall that, in previous section, the finding was that increasing transmission power didn't improve the system safety metric, which is contrary to the finding in this scenario.The main cause is the large WCV and higher speeds of manned aircraft. +V. SummaryThis work presented the Fe 3 simulator, a tool developed for evaluating low-altitude air traffic operations.The concept and architecture of the simulator were demonstrated.Various models were implemented in Fe 3 to incorporate core components in low-altitude high-density UTM or UAM-like operations.The webbased highly parallelized implementation using AWS was introduced and showed acceptable computational performance.Experiments using the simulator were presented to show studies on collision avoidance, communication, wind, and interaction with manned flights.It was shown that the collision avoidance performance might be different in a low-density environment from a high-density environment.The studies presented the impacts of several key factors, such as communication reception models, wind, vehicle speeds, well-clear violation definition, and the collision avoidance algorithm/rules.It is noted that the system analysis is complex as results may change with these factors.Therefore, substantial experiments with the Fe 3 simulator are necessary to gain conclusive insights.Future work will focus on comprehensive and systematic analysis of the low altiude high density air traffic system using the simultor.Figure 1 .1Figure 1.Architecture of Fe 3 's core simulation engine +Figure 3 .3Figure 3. Offline Table Based Collision Avoidance Algorithm +34 +Figure 4 .4Figure 4. Force Field Based Collision Avoidance Algorithm +Figure 5 .5Figure 5. DSRC reception probability at various communication densities +FigureFigure 6.Cloud Architecture +Figure 7 .7Figure 7. Snapshot of Summary over all Monte Carlo Simulations +Figure 8 .8Figure 8. Snapshot of metrics rom individual simulations +Figure 9 .9Figure 9. Loss of separation at various communication capabilities and traffic densities +Figure 10 .10Figure 10.Loss of separation at various communication capabilities and traffic densities +Figure 11 .11Figure 11.Loss of separation at various wind conditions and traffic densities +Figure 12 .12Figure 12.Loss of separation at various wind conditions and traffic densities + +Table Based Collision Avoidance Algorithm +Table 1 .1Computational performance of Fe 3Number of Number of Traffic density Flight time Running timeMC simssUAVs( per nm 2 )(minute)(second)1,000103.87.710.21,0005019.08.754.71,00010038.19.8117.840500190.513.6366.440800304.913.5903.9401,000381.114.01,379.0 + + + + + + + + + Unmanned Aircraft System Traffic Management (UTM) Concept of Operations + + PKopardekar + + + JRios + + + TPrevot + + + MJohnson + + + JJung + + + JERobinson + + + + 16th AIAA Aviation Technology, Integration, and Operations Conference + Washington, D.C. + + June 2016 + + + + Kopardekar, P., Rios, J., Prevot, T., Johnson, M., Jung, J., and Robinson, J. E., "Unmanned Aircraft System Traf- fic Management (UTM) Concept of Operations," 16th AIAA Aviation Technology, Integration, and Operations Conference, Washington, D.C., 13-17 June 2016. + + + + + Enabling Airspace Integration for High-Density On-Demand Mobility Operations + + EricRMueller + + + ParmialHKopardekar + + + KennethHGoodrich + + 10.2514/6.2017-3086 + + + 17th AIAA Aviation Technology, Integration, and Operations Conference + Denver, Colorado + + American Institute of Aeronautics and Astronautics + 2017. June 2017 + + + + Mueller, E. Kopardekar, P. and Goodrich, K., "Enabling Airspace Integration for High-Density On-Demand Mobility Operations," AIAA Aviation 2017 , Denver, Colorado, 5-9 June 2017. + + + + + Europe’s Other Democratic Deficit: National Authoritarianism in Europe’s Democratic Union + + RDanielKelemen + + 10.1017/gov.2016.41 + + + + Government and Opposition + Gov. & oppos. + 0017-257X + 1477-7053 + + 52 + 2 + + 2017. 2017 + Cambridge University Press (CUP) + + + "Number of U.S. Aircraft, Vehicles, Vessels, and Other Conveyances," https://www.rita.dot.gov/bts/sites/rita.dot. gov.bts/files/publications/national_transportation_statistics/html/table_01_11.html, 2017, [online; accessed Octo- ber 2017]. + + + + + Unmanned Aircraft System Operations + 10.1201/b11202-10 + + + + Introduction to Unmanned Aircraft Systems + Statistical Summary of Commerical Jet Airplane Accidents + + CRC Press + 2016. October 2017. 2016. October 2017 + 5 + + + + Unmanned Aircraft Systems + 4 "Statistical Summary of Commerical Jet Airplane Accidents, Worldwide Operations 1959-2016," http://www.boeing. com/resources/boeingdotcom/company/about_bca/pdf/statsum.pdf, 2016, [online; accessed October 2017]. 5 "Unmanned Aircraft Systems," https://www.faa.gov/data_research/aviation/aerospace_forecasts/media/ Unmanned_Aircraft_Systems.pdf, 2016, [online; accessed October 2017]. + + + + + Initial Study of An Effective Fast-time Simulation Platform for Unmanned Aircraft System Traffic Management + + MinXue + + + JosephRios + + 10.2514/6.2017-3073 + + + 17th AIAA Aviation Technology, Integration, and Operations Conference + Denver, Colorado + + American Institute of Aeronautics and Astronautics + June 2017 + + + + 6 Xue, M. and Rios, J., "Initial Study of An Effective Fast-time Simulation Platform for Unmanned Aircraft System Traffic Management," 17th AIAA Aviation Technology, Integration, and Operations Conference, Denver, Colorado, 5-9 June 2017. + + + + + Design of Center-TRACON Automation System + + HErzberger + + + TJDavis + + + SMGreen + + + + AGARD Meeting on Machine Intelligence in Air Traffic Management + Berlin, Germany + + May 1993 + + + + Erzberger, H., Davis, T. J., and Green, S. M., "Design of Center-TRACON Automation System," AGARD Meeting on Machine Intelligence in Air Traffic Management, Berlin, Germany, 11-14 May 1993. + + + + + FACET: Future ATM Concepts Evaluation Tool + + KarlDBilimoria + + + BanavarSridhar + + + ShonRGrabbe + + + GanoBChatterji + + + KapilSSheth + + 10.2514/atcq.9.1.1 + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 9 + 1 + + + American Institute of Aeronautics and Astronautics (AIAA) + + + Bilimoria, K., Sridhar, B., Chatterji, G., Sheth, K., and Grabbe, S., "FACET: Future ATM Concepts Evaluation Tool," + + + + + + + Usa/ + + + + Europe Air Traffic Management R&D Seminar + + + June 2000 + Napoli, Italy + + + USA/Europe Air Traffic Management R&D Seminar , Napoli, Italy, 13-16 June 2000. + + + + + Fast-Time NAS Simulation System for Analysis of Advanced ATM Concepts + + DouglasSweet + + + VikramManikonda + + + JesseAronson + + + KarlinRoth + + + MatthewBlake + + 10.2514/6.2002-4593 + + + AIAA Modeling and Simulation Technologies Conference and Exhibit + Monterey, California + + American Institute of Aeronautics and Astronautics + August 2002 + + + + Sweet, D. N., Manikonda, V., Aronson, J. S., Roth, K., and Blake, M., "Fast-time Simulation System For Analysis of Advanced Air Transportation Concepts," AIAA Modeling and Simulation Technologies Conference and Exhibit, Monterey, California, 5-8 August 2002. + + + + + A Fuzzy Logic Approach for Low Altitude UAS Traffic Management (UTM) + + BrandonCook + + + KellyCohen + + + EladHKivelevitch + + 10.2514/6.2016-1905 + + + AIAA Infotech @ Aerospace + San Diego, California + + American Institute of Aeronautics and Astronautics + January 2016 + + + + AIAA Science and Technology Forum and Exposition 2016 + Cook, B., Cohen, K., and Kivelevitch, E., "A Fuzzy Logic Approach For Low Altitude UAS Traffic Management (UTM)," AIAA Science and Technology Forum and Exposition 2016 , San Diego, California, 4-8 January 2016. + + + + + Simulation Comparison of Collision Avoidance Algorithms for Small Multi-Rotor Aircraft + + EricRMueller + + + MykelKochenderfer + + 10.2514/6.2016-3674 + + + AIAA Modeling and Simulation Technologies Conference + Washington, D.C. + + American Institute of Aeronautics and Astronautics + June 2016 + + + + Mueller, E. and Kochenderfer, M. J., "Simulation Comparison of Collision Avoidance Algorithms for Multi-Rotor Air- craft," AIAA Modeling and Simulation Technologies Conference, Washington, D.C., 13-17 June 2016. + + + + + Unmanned Aircraft System Navigation in the Urban Environment: A Systems Analysis + + MSanfourche + + + JDelaune + + + GLe Besnerais + + + HDe Plinval + + + JIsrael + + + PCornic + + + ATreil + + + YWatanabe + + + APlyer + + + JRAtkins + + + EM + + + + AIAA Journal of Aerospace Information Systems + + 4 + 4 + 2012. 13 Rufa,. 2016 + + + Aerospace Lab Journal + Sanfourche, M., Delaune, J., Le Besnerais, G., De Plinval, H., Israel, J., Cornic, P., Treil, A., Watanabe, Y., and Plyer, A., "Perception for UAV: Vision-Based Navigation and Environment Modeling," Aerospace Lab Journal, Vol. 4, 2012. 13 Rufa, J. R. and Atkins, E. M., "Unmanned Aircraft System Navigation in the Urban Environment: A Systems Analysis," AIAA Journal of Aerospace Information Systems, Vol. 13, No. 4, 2016. + + + + + Vision-Aided Inertial Navigation for Small Unmanned Aerial Vehicles in GPS-Denied Environments + + TianmiaoWang + + + ChaoleiWang + + + JianhongLiang + + + YangChen + + + YichengZhang + + 10.5772/56660 + + + International Journal of Advanced Robotic Systems + International Journal of Advanced Robotic Systems + 1729-8814 + + 10 + 6 + 276 + 2013 + SAGE Publications + + + Wang, T., Wang, C., Liang, J., Chen, Y., and Zhang, Y., "Vision-Aided Inertial Navigation for Small Unmanned Aerial Vehicles in GPS-Denied Environments," International Journal of Advanced Robotic Systems, Vol. 10, No. 276, 2013. 15 + + + + + Vision-Aided Nonlinear Observer for Fixed-Wing Unmanned Aerial Vehicle Navigation + + JesperHosen + + + HåkonHHelgesen + + + LorenzoFusini + + + ThorIFossen + + + TorAJohansen + + 10.2514/1.g000281 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 39 + 8 + + 2016 + American Institute of Aeronautics and Astronautics (AIAA) + + + Hosen, J., Helgesen, H. H., Fusini, L., Fossen, T. I., and Johansen, T. A., "Vision-Aided Nonlinear Observer for Fixed- Wing Unmanned Aerial Vehicle Navigation," AIAA Journal of Guidance, Control, and Dynamics, Vol. 39, No. 8, 2016. 16 + + + + + Quadrotor Dynamics and Control + + RWBeard + + + October 2008 + + + Tech. rep. + Beard, R. W., "Quadrotor Dynamics and Control," Tech. rep., October 2008. + + + + + Full control of a quadrotor + + SamirBouabdallah + + + RolandSiegwart + + 10.1109/iros.2007.4399042 + + + 2007 IEEE/RSJ International Conference on Intelligent Robots and Systems + San Diego, CA + + IEEE + Oct 29 -Nov 2 2007 + + + Bouabdallah, S. and Siegwart, R., "Full Control of a Quadrotor," Proceedings of the 2007 IEEE/RSJ International Conference on Intelligent Robots and Systems, San Diego, CA, Oct 29 -Nov 2 2007. + + + + + Comprehensive Simulation of Quadrotor UAVs Using ROS and Gazebo + + JohannesMeyer + + + AlexanderSendobry + + + StefanKohlbrecher + + + UweKlingauf + + + OskarVon Stryk + + 10.1007/978-3-642-34327-8_36 + + + Simulation, Modeling, and Programming for Autonomous Robots + + Springer Berlin Heidelberg + 2012 + 7628 + + + + Meyer, J., Sendobry, A., Kohlbrecher, S., Klingauf, U., and O., S., "Comprehensive Simulation of Quadrotor UAVs using ROS and Gazebo," Simulation, Modeling, and Programming for Autonomous Robots, Vol. 7628, 2012. + + + + + Wind Tunnel and Hover Performance Test Results for Multicopter UAS Vehicles + + CRRussell + + + JJung + + + GWillink + + + BGlasner + + + + American Helocipter Society 72nd Annual Forum + West Palm Beach, FL + + May 2016 + + + + Russell, C. R., Jung, J., Willink, G., and Glasner, B., "Wind Tunnel and Hover Performance Test Results for Multicopter UAS Vehicles," American Helocipter Society 72nd Annual Forum, West Palm Beach, FL, 16-19 May 2016. + + + + + High-Fidelity Multi-Rotor Unmanned Aircraft System (UAS) Simulation Development for Trajectory Prediction Under Off-Nominal Flight Dynamics + + JohnVFoster + + + DavidHartman + + 10.2514/6.2017-3271 + + + 17th AIAA Aviation Technology, Integration, and Operations Conference + Denver, Colorado + + American Institute of Aeronautics and Astronautics + June 2017 + + + + Foster, J. V. and Hartman, D. C., "High-Fidelity Multirotor Unmanned Aircraft System Simulation Development for Trajectory Prediction Under Off-Nominal Flight Dynamics," 17th AIAA Aviation Technology, Integration, and Operations Conference, Denver, Colorado, 5-9 June 2017. + + + + + Toward Indoor Flying Robots + + DJiang + + + QChen + + + LDelgrossi + + + + IEEE International Conference on Intelligent Robots and Systems + Lausanne, Switzerland + + 30 Sept. -4 Oct 2002 + + + Jiang, D., Chen, Q., and Delgrossi, L., "Toward Indoor Flying Robots," IEEE International Conference on Intelligent Robots and Systems, Lausanne, Switzerland, 30 Sept. -4 Oct 2002. + + + + + A Modeling, Simulation and Control Framework for Small Unmanned Multicopter Platforms in Urban Environments + + CoreyAIppolito + + + SebastianHening + + + ShankarSankararaman + + + VahramStepanyan + + 10.2514/6.2018-1915 + + + 2018 AIAA Modeling and Simulation Technologies Conference + Kissimmee, Florida + + American Institute of Aeronautics and Astronautics + January 2018 + + + + Ippolito, C. A., Krishnakumar, K., Hening, S., and Sankararaman, S., "A Modeling, Simulation and Control Framework for Small Unmanned Multicopter Platforms in Urban Environments," AIAA Modeling and Simulation Technologies Conference, Kissimmee, Florida, 8-12 January 2018. + + + + + A review of conflict detection and resolution modeling methods + + JKKuchar + + + LCYang + + 10.1109/6979.898217 + + + IEEE Transactions on Intelligent Transportation Systems + IEEE Trans. Intell. Transport. Syst. + 1524-9050 + + 1 + 4 + + 2000 + Institute of Electrical and Electronics Engineers (IEEE) + + + Kuchar, J. K. and Yang, L. C., "A Review of Conflict Detection and Resolution Modeling Methods," IEEE Transactions On Intelligent Transportation Systems, Vol. 1, No. 4, 2000. + + + + + Chain-based Collision Avoidance for UAS Sense-and-Avoid Systems + + LaithSahawneh + + + RandalWBeard + + + SharathAvadhanam + + + HeBai + + 10.2514/6.2013-4995 + + + AIAA Guidance, Navigation, and Control (GNC) Conference + Boston, MA + + American Institute of Aeronautics and Astronautics + August 2013 + + + + Sahawneh, L. R. and Beard, R. W., "Chain-based Collision Avoidance for UAS Sense-and-Avoid Systems," AIAA Guid- ance, Navigation, and Control Conference, Boston, MA, 19-22 August 2013. + + + + + NASA contract for charged particle spectrometer for manned space vehicle + + RBach + + + CFarrell + + + HErzberger + + 10.1108/aeat.2009.12781dab.019 + AR-2009-214573 + + + Aircraft Engineering and Aerospace Technology + Aircraft Eng & Aerospace Tech + 0002-2667 + + 81 + 4 + 2009 + Emerald + + + Tech. Rep. Technical Report + Bach, R. Farrell, C. and Erzberger, H., "An Algorithm for Level-Aircraft Conflict Resolution," Tech. Rep. Technical Report, AR-2009-214573, NASA, 2009. + + + + + Protocol-based Conflict Resolution for Finite Information Horizon + + IHwang + + + CTomlin + + + + Proceedings of the American Control Conference + the American Control ConferenceAnchorage, AK + + May 2002 + + + + Hwang, I. and Tomlin, C., "Protocol-based Conflict Resolution for Finite Information Horizon," Proceedings of the American Control Conference, Anchorage, AK, 8-10 May 2002. + + + + + DAIDALUS: Detect and Avoid Alerting Logic for Unmanned Systems + + CesarMunoz + + + AnthonyNarkawicz + + + GeorgeHagen + + + JasonUpchurch + + + AaronDutle + + + MariaConsiglio + + + JamesChamberlain + + 10.1109/dasc.2015.7311421 + + + 2015 IEEE/AIAA 34th Digital Avionics Systems Conference (DASC) + Prague, Czech + + IEEE + October 2015 + + + + Muoz, C., Narkawicz, A., Hagen, G., Upchurch, J., Dutle, A., and Consiglio, M., "DAIDALUS: Detect and Avoid Alerting Logic for Unmanned Systems," In Proceedings of the 34th Digital Avionics Systems Conference, Prague, Czech, 6-8 October 2015. + + + + + Multi-Rotor Aircraft Collision Avoidance using Partially Observable Markov Decision Processes + + EricRMueller + + + MykelKochenderfer + + 10.2514/6.2016-3673 + + + AIAA Modeling and Simulation Technologies Conference + Washington, D.C. + + American Institute of Aeronautics and Astronautics + June 2016 + + + + Mueller, E. and Kochenderfer, M. J., "Multi-Rotor Aircraft Collision Avoidance using Partially Observable Markov Decision Processes," AIAA Modeling and Simulation Technologies Conference, Washington, D.C., 13-17 June 2016. + + + + + A survey of collision avoidance approaches for unmanned aerial vehicles + + BMAlbaker + + + NARahim + + 10.1109/techpos.2009.5412074 + + + 2009 International Conference for Technical Postgraduates (TECHPOS) + + IEEE + 2009 + + + + Albaker, B. M. and Rahim, N. A., "A survey of collision avoidance approaches for unmanned aerial vehicles," Proc Int Conf Tech Postgraduates (TECHPOS), 2009, pp. 1-7. + + + + + The Generic Resolution Advisor and Conflict Evaluator (GRACE) for Detect-And-Avoid (DAA) Systems + + MichaelAbramson + + + MohamadRefai + + + ConfesorSantiago + + 10.2514/6.2017-4485 + + + 17th AIAA Aviation Technology, Integration, and Operations Conference + Denver, Colorado + + American Institute of Aeronautics and Astronautics + June 2017 + + + + Abramson, M., Refai, M., and Santiago, C., "The Generic Resolution Advisor and Conflict Evaluator (GRACE) for Detect-And-Avoid (DAA)," 17th AIAA Aviation Technology, Integration, and Operations Conference, Denver, Colorado, 5-9 June 2017. + + + + + Algorithm and operational concept for resolving short-range conflicts + + HErzberger + + + KRHeere + + 10.1243/09544100jaero546 + + + Proceedings of the Institution of Mechanical Engineers, Part G: Journal of Aerospace Engineering + Proceedings of the Institution of Mechanical Engineers, Part G: Journal of Aerospace Engineering + 0954-4100 + 2041-3025 + + 224 + 2 + + Sept. 2008 + SAGE Publications + Anchorage AK + + + Erzberger, H. and Heere, K. R., "Algorithm and operational concept for resolving short-range conflicts," International Council of the Aeronautical Sciences, Anchorage AK, 14-19 Sept. 2008. + + + + + A New Approach for Designing Safer Collision Avoidance Systems + + MykelJKochenderfer + + + JamesPChryssanthacopoulos + + + RolandEWeibel + + 10.2514/atcq.20.1.27 + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 20 + 1 + + 2012 + American Institute of Aeronautics and Astronautics (AIAA) + + + Kochenderfer, M., Chryssanthacopoulos, J. P., and Weibel, R. E., "A new approach for designing safer collision avoidance systems," Air Traffic Control Quarterly, Vol. 20, No. 1, 2012. + + + + + Chain-based path planning for multiple UAVs + + MatthewArgyle + + + CalebChamberlain + + + RandyBeard + + 10.1109/cdc.2011.6160936 + + + IEEE Conference on Decision and Control and European Control Conference + Orlando, FL + + IEEE + December 2011 + 34 + + + CDC-ECC + + + Argyle, M., Chamberlain, C., and Beard, R., "Chain-based path planning for multiple UAVs," 50th IEEE Conference on Decision and Control and European Control Conference (CDC-ECC), Orlando, FL, 12-15 December 2011. 34 + + + + + Three-Dimensional Velocity Obstacle Method for Uncoordinated Avoidance Maneuvers of Unmanned Aerial Vehicles + + YazdiIJenie + + + Erik-JanVan Kampen + + + CornelisCDe Visser + + + JoostEllerbroek + + + JaccoMHoekstra + + 10.2514/1.g001715 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 39 + 10 + + 2016 + American Institute of Aeronautics and Astronautics (AIAA) + + + Jenie, Y. I., Kampen, E. V., Visser, C. C., Ellerbroek, J., and Hoekstra, J. M., "Three-Dimensional Velocity Obstacle Method for Uncoordinated Avoidance Maneuvers of Unmanned Aerial Vehicles," Journal of Guidance, Control, and Dynamics, Vol. 39, No. 10, 2016. + + + + + Communicating with unmanned aerial swarm automatic dependent surveillance transponders + + SreejaNag + + + JaewooJung + + + KarishmaInamdar + + 10.1109/icsens.2017.8234227 + + + 2017 IEEE SENSORS + Glasgow, UK + + IEEE + 29 Oct. -1 Nov. 2017 + + + Nag, S., Jung, J., and Inamdar, K., "Communicating with unmanned aerial swarm automatic dependent surveillance transponders," IEEE on SENSORS , Glasgow, UK, 29 Oct. -1 Nov. 2017. + + + + + Small Unmanned Aerial System (UAS) Flight Testing of Enabling Vehicle Technologies for the UAS Traffic Management Project + + LJGlaab + + + CVDolph + + + SDYoung + + + NCCoffey + + + RGMcswain + + + MJLogan + + + DEHarper + + + + + L-band Digital Aeronautical Communications System(LDACS) + + April 2018. April 2006. April 2018 + 37 + 591 + + + Tech. Rep. NASA TM-2018-219816 + Vehicle Safety Communications Project -Final Report + Glaab, L. J., Dolph, C. V., Young, S. D., Coffey, N. C., McSwain, R. G., Logan, M. J., and Harper, D. E., "Small Unmanned Aerial System (UAS) Flight Testing of Enabling Vehicle Technologies for the UAS Traffic Management Project," Tech. Rep. NASA TM-2018-219816, April 2018. 37 "Vehicle Safety Communications Project -Final Report," Tech. Rep. DOT HS 810 591, April 2006. 38 "L-band Digital Aeronautical Communications System(LDACS)," http://www.ldacs.com/, [cited April 2018]. + + + + + Communication Density: A Channel Load Metric for Vehicular Communications Research + + DanielJiang + + + QiChen + + + LucaDelgrossi + + 10.1109/mobhoc.2007.4428734 + + + 2007 IEEE Internatonal Conference on Mobile Adhoc and Sensor Systems + Pisa, Italy + + IEEE + October 2007 + + + + 39 Jiang, D., Chen, Q., and Delgrossi, L., "Communication Density: A Channel Load Metric for Vehicular Communications Research," IEEE International Conference on Mobile Adhoc and Sensor Systems (MASS), Pisa, Italy, 8-11 October 2007. + + + + + An Empirical Model for Probability of Packet Reception in Vehicular Ad Hoc Networks + + MoritzKillat + + + HannesHartenstein + + 10.1155/2009/721301 + + + EURASIP Journal on Wireless Communications and Networking + J Wireless Com Network + 1687-1499 + + 2009 + 1 + 2009 + Springer Science and Business Media LLC + + + Killat, M. and Hartenstein, H., "An Empirical Model for Probability of Packet Reception in Vehicular Ad Hoc Networks," EURASIP Journal on Wireless Communications and Networking Volume, 2009. + + + + + Design of 5.9 ghz dsrc-based vehicular safety communication + + DanielJiang + + + VikasTaliwal + + + AndreasMeier + + + WielandHolfelder + + + RalfHerrtwich + + 10.1109/wc-m.2006.250356 + + + + IEEE Wireless Communications + IEEE Wireless Commun. + 1536-1284 + + 13 + 5 + + 2007. April 2018 + Institute of Electrical and Electronics Engineers (IEEE) + + + Jiang, D., Taliwal, V., Meier, A., Hofelder, W., and Herrtwich, R., "Design of 5.9 Ghz DSRC-based Vehicular Safety Communication," IEEE Wireless Communiations, Vol. 13, No. 5, 2007. 42 "Texas A&M UAS Center Tests New UAS Technology To Improve Sky Safety," https://echodyne.com/texas-am-uas- center-tests-new-uas-technology-to-improve-sky-safety/, [cited April 2018]. + + + + + + SSWeygandt + + + TGSmirnova + + + SGBenjamin + + + KJBrundage + + + SRSahm + + + CRAlexander + + + BESchwartz + + The High Resolution Rapid Refresh (HRRR): An Hourly Updated Convection Resolving Model Utilizing Radar Reflectivity Assimilation from the RUC/RR," 23rd Conf. on Weather Analysis and Forecasting/19th Conference on Numerical Weather Prediction + Omaha, NE + + June 2009 + + + + Weygandt, S. S., Smirnova, T. G., Benjamin, S. G., Brundage, K. J., Sahm, S. R., Alexander, C. R., and Schwartz, B. E., "The High Resolution Rapid Refresh (HRRR): An Hourly Updated Convection Resolving Model Utilizing Radar Reflectivity Assimilation from the RUC/RR," 23rd Conf. on Weather Analysis and Forecasting/19th Conference on Numerical Weather Prediction, Omaha, NE, 1-5 June 2009. + + + + + The California State University Mobile Atmospheric Profiling System: A Facility for Research and Education in Boundary Layer Meteorology + + CraigBClements + + + AndrewJOliphant + + 10.1175/bams-d-13-00179.1 + + + + Bulletin of the American Meteorological Society + 1520-0477 + + 95 + 11 + + 2014. April 2018 + American Meteorological Society + + + Continuous gusts. cited + Clements, C. B. and Oliphant, A. J., "The California State University Mobile Atmospheric Profiling System: A Facility for Research and Education in Boundary Layer Meteorology," Bulletin of the American Meteorological Society, Vol. 95, 2014. 45 "Continuous gusts," https://en.wikipedia.org/wiki/Continuous gusts, [cited April 2018]. + + + + + + diff --git a/file815.txt b/file815.txt new file mode 100644 index 0000000000000000000000000000000000000000..e9ef8b21fd2ac76f6804bab602b054b20e8393fc --- /dev/null +++ b/file815.txt @@ -0,0 +1,604 @@ + + + + +I. IntroductionIn order to accommodate soaring commercial interest in small Unmanned Aerial Vehicle (sUAV) operations, NASA has been leading the development of the Unmanned aerial system Traffic Management (UTM) system together with FAA and industry.Within the UTM architecture, besides basic functions or services like authentication, authorization, and strategic deconfliction, many other functions or services are also needed to enable large scale operations of small UAVs.UAV flight planning is one of them.The UTM concept requires that small UAVs operate in Class G airspace, which is typically below 400 feet Above Ground Level (AGL).This restriction, together with urban obstacles and winds, brings challenges to the UAV flight path planning task.Flight path planning algorithms in aviation are typically used to find paths or trajectories that minimize fuel, flight time, and/or propagated noise [1].Three groups of methods are usually applied to solve this type of problem.The first group typically includes trajectory optimization algorithms including indirect methods and direct methods [2].The indirect methods use the calculus of variations or the Maximum Principle of Pontryagin.The direct methods transform the original optimal control problem into a nonlinear parameter optimization problem by using multiple shooting methods and direct collocation methods.This group of methods is normally used to find two dimensional trajectories that can minimize fuel and flight time [3], such as wind-optimal trajectories [4].The second group of methods treats the trajectory optimization as a highly-constrained linear or nonlinear problem and solves them using generalized optimization techniques, such as genetic algorithms, simulated annealing [5], and mixed-integer programming [6].The third group consists of path planning methods.They are typically used in robot motion planning [7] to avoid obstacles including the roadmap method, potential field method, cell decomposition method, probabilistic roadmap method, and the rapidly-exploring random tree [8,9] method.In an obstacle-rich environment where the problem is highly constrained, the path planning methods are widely used as they can find a near optimal shortest path with much less computational time.There has been a lot of research in UAV planning using path planning methods lately [10].Ramana et.al. [11] proposed an obstacle-avoidance path planning algorithm that also takes into account the turn radius and climb rate constraints of a UAV; Primatesta et.al. [12] developed a risk-aware path planning algorithm to minimize the risk to the population on the ground based on a risk-map constructed using probabilities of crashing, impact, and fatality.Chakrabarty et.al. [13] developed a real-time kinematic tree path planner to avoid obstacles when the small UAVs are aloft.Since small UAVs are required to operate below AGL 400 feet, flight planning with a constant Median Sea Level (MSL) as in traditional aviation is not well suited for sUAV operations.An AGL-based flight plan becomes necessary for small UAV operations.This work proposes a path planning algorithm for generating obstacle-free and wind-efficient sUAV paths at a constant AGL in urban environments.By constructing obstacle maps at given AGL levels, this method converts a three dimensional AGL-based flight planning to a two dimensional problem.A quad-tree decomposition is then used to build the search space in terms of obstacle occupancy and wind difference.The wind cost of traveling through each cell is defined based on steady thrusts calculated for a given UAV model under various wind conditions, and a repulsive potential was adopted to make sure the flight plans are not too close to obstacles.With the Theta* search algorithm and post smoothing techniques, an obstacle-free, wind-efficient, and constant AGL flight plan can be quickly generated for small UAV operations in urban environments.In this paper, Section II introduces the proposed flight planning algorithm including search graph construction, cost definition, search algorithm, and post smoothing.Section III presents results and performance of the flight path planning algorithm.Section IV concludes this work. +II. Flight Path Planning AlgorithmIn order to generate an obstacle-free wind-efficient flight path with reasonable computational effort, an algorithm based on path planning techniques was developed.Using the terrain and wind data for San Francisco city as an example, this section presents the graph construction, cost definition, and search method used in this algorithm. +A. Search Graph ConstructionA high-resolution 3D city model for the San Francisco city [14] is shown in Fig. 1(a).To capture details for visualization, triangles were used as basic units in the data set and each terrain object is composed of thousands of triangles.There are two types of objects for the terrain: buildings and land structures.Using this 3D model, a height map with a resolution of three feet by three feet was generated by extracting terrain height information from the provided object file.Figure 1(b) shows the resulting height map, where warm color and cold color represent high and low altitudes, respectively.The total dimensions for the city height map spanned 54,843 feet by 50,872 feet (or 9.02 nmi by 9.02 nmi) after the conversion.The altitude ranges from 0 ft to 934 ft MSL for land, and the highest elevation for man-made structure or buildings is 1,657 ft MSL. +Obstacle Map at Given Above-ground LevelsThe UTM concept limits small UAV operations to Above Ground Level (AGL) 400 feet and below, which in fact allows sUAVs to fly at a high MSL in certain areas.For instance, in areas where the ground elevation is 1,000 ft, if there is no other restriction, a sUAV can fly up to 1,400 ft MSL.Meanwhile, with this restriction, a flight plan with a constant MSL, as in traditional aviation, may not be acceptable for sUAV operations, especially in cities like San Francisco, where ground elevation has up to 934 feet difference from the highest to the lowest location.Therefore, an AGL-based flight plan inevitably becomes the only option for small UAV operations.To generate AGL-based flight plans, a height map needs to be constructed in terms of AGL levels.Figure 2 shows a notional picture of AGL-based flight operations.However, the slope of the ground terrain may exceed the vehicle's maximum climb and descent rates.To investigate this issue, a gradient map of the San Francisco area was generated using algorithms from Canny edge detection [15].Assuming a UAV can climb at a rate of 16.5 feet per second at a cruise speed of 40 knots, the maximum slope for a UAV to operate at its cruise speed would be around 0.3.The yellow spots shown in Fig. 3 represent areas that exceed the range of [-0.3, 0.3] and therefore may not be suitable for constant AGL flights.Some of these areas may even be impossible for UAVs to operate under the current UTM concept, simply because UAVs cannot stay in class G in those areas with their cruise speeds due to their climb and descent rate limits.This concern must be addressed in the future UTM work.In this work, studies are limited to the SF downtown area in the white box in Fig. 3, where the yellow area is not of great concern.To produce flight plans at a constant AGL altitude, obstacle maps at the same AGL altitude were constructed based on the height difference between overall terrain and the ground.Figure 4(a) presents an obstacle map of SF downtown area at 30 ft AGL.With this approach the AGL-based flight path planning can then be simplified to a 2D problem. +Graph for ObstaclesA quad-tree decomposition was used to build a search space for obstacle-free flight planning.In this decomposition method, the obstacle map (as shown in red color in Fig. 4(a)) is treated as a root cell.If a cell is completely occupied by an obstacle, it is marked as "FULL".On the other hand, if a cell doesn't contain any obstacle, it is defined as "EMPTY".If a cell has some free space while partially occupied by obstacles, it is defined as "MIXED".Starting from the root cell, a "MIXED" cell will be recursively decomposed into four quadrants until it becomes either a "FULL" cell or a "EMPTY" cell or the decomposition reaches a pre-defined depth level or cell size.Fig. 4(b) shows the resulting quad-tree decomposition for obstacles with a minimum cell size of 6 feet. +Graph for Wind and ObstaclesTo generate a flight plan that can also minimize wind impact, wind profiles need to be considered as well when constructing the search graph.In this work, a set of high-resolution (1m X 1m) wind profiles were generated for downtown San Francisco area using computational fluid dynamics (CFD) based simulations with a given dominant wind direction.This set of wind profiles range from MSL 0 ft to MSL 360 ft.Fig. 5(a) shows a sample wind profile at MSL 120 ft with magnitude varying from 0 to 14 meters per second, where the dominant wind was assumed to be coming from the East.Similar to constructing the obstacle map, these MSL-based wind profiles are first converted to AGL based.Then, to incorporate wind profile into the search graph, an extra decomposition criteria in the quad-tree decomposition is added: a cell will also be decomposed into four quadrants if the difference of wind vectors inside this cell exceeds a predefined threshold, even if the cell is not "MIXED".The wind-induced decomposition will start +B. Cost definitionOnce the graph is built, coefficients or weights are assigned to each leaf node or cell to help achieve obstacle-free and wind-efficient trajectories.The cost of traveling through the current cell is then calculated by:c step = U rep • w • d i + p i, j(1)where d i denotes the distance needed to travel through the current cell.U rep and w represent the repulsive potential and wind coefficients, respectively, and p i, j is the penalty for unacceptable path angle changes from node j to node i.The definitions of these parameters will be found in the following subsection.Assuming the cost associated with the parent node is g i-1 , the cost for the current node g i can then be written as:g i = g i-1 + c step (2) +Cost to avoid obstacles and penalize unacceptable path angle changesA large coefficient will be assigned to the 'FULL' and 'MIXED' cells to prevent small UAVs from flying through obstacles.For 'EMPTY' cells, a repulsive potential (shown in Eqn. 3) is applied to avoid the flight path getting too close to the obstacles, where the Q * is a constant and can be seen as the buffer size around an obstacle and D is the distance from the sUAV to the obstacle.Once the distance is greater than the threshold Q * , there is no penalty and the coefficient is one.Similarly, the penalty of unacceptable heading angle change between lateral path angle γ j ( at node j) and γ i ( at node i) is defined as in Eqn. 4, which is activated once the path angle change exceeds a predefined Γ.The Γ is set to 45 • in this work.U r ep = Q * -D + 1, if D ≤ Q * . 1, Others.(3)p i, j = ∞, if | γ i -γ j |≥ Γ. 0, Others.(4) +Wind CoefficientsIn this subsection, two types of wind coefficients are derived for minimum energy and minimum time, respectively.To find energy-efficient paths in a wind field, the steady-state thrust T that is applied to maintain a desired ground speed under various wind conditions is calculated using sUAV dynamics [16].The thrust is approximated [16][17][18] as a linear function of the square of motor rotational speed ω as in Eqn. 5, where k f is a constant coefficient for a given motor.The power consumption is then approximated [16,19] as a linear function of the cube of motor rotational speed (shown in Eqn.6), where k m is a constant coefficient between motor rotational speed and motor-generated torque.Using the steady-state thrust, when there is no wind, as a reference T re f , the ratio between thrusts is computed and used to define a wind coefficient w to approximate the energy cost under various wind conditions (shown in Eqn. 7).T = k f • ω 2(5)P = k m • ω 3(6)w = T T re f 1.5 (7) Figure 6(a) shows time histories of thrust for a multi-copter flying with a ground speed of 15 mps (29 knots) in three different wind conditions: 10 mps tail wind, no wind, and 10 mps head wind.The figure shows that the small UAV entered the wind field at the 5th second and then experienced a transition state and finally settled down at a steady state.The steady-state thrusts are 6.39, 7.34, and 9.15 Newton for tail wind, no wind, and head wind, respectively.The energy coefficients for these three wind conditions were then calculated as 0.81, 1.0, and 1.40, respectively.A set of such wind coefficients was then generated for wind vectors with various direction and magnitude.Figure 6(b) presents such a wind coefficient matrix at a desired ground speed of 15 mps, where the coefficient varies from 0.81 to 1.50.In the Calculation of the wind coefficient for finding time-optimal paths in a wind field is straightforward.Assuming that the vehicle airspeed is higher than the wind speed, the coefficient w is simply defined in terms of the vehicle and wind speeds as shown in Eqn. 8, where ì V ac and ì V w are the velocity vectors for aircraft and wind, respectively.This definition will incentivize flying with a tailwind and penalize traveling with a headwind.w = | ì V ac | | ì V ac | + ì V w • ì V a c | ì V a c |(8) +C. Search algorithmOnce the graph is built, the Theta* [20] search algorithm, a variant of A* algorithm, is applied to find the optimal path.The key difference between Theta* and A* is that Theta* allows the parent of a vertex to be any previous predecessor, whereas in A* the parent must be the adjacent predecessor.This change allows any-angle paths and mitigates the rigid path angle constraint introduced by the grid-based search graph.The pseudo code of the Theta* search algorithm is shown in Algorithm 1.The step cost c(S, S ) represents the c step as shown in Eqn. 1.The g-cost is the cost from the start node to current node as in Eqn. 2, where both obstacles and wind condition are taken into account.And the h-value is proportional to the straight-line distance from current node to the goal, while the proportion is set to underestimate the cost from current node to the goal, so the admissibility of the algorithm is preserved.The part between Line 2.19 and 2.24 in Algorithm 1 is the same as in A*, while the part between Line 2.11 and 2.16 is introduced open.add(s, g(s ) + h(s )) +D. Post smoothingAlthough the Theta* method takes heading angle change constraints into account during the search process, a simple and effective post smoothing technique [21] was applied to further smooth the path angle transition while not losing much optimality [20].The pseudo procedure of the post smoothing method is shown below.The main idea is starting from the end node of the path.For each node, if the upstream node of its parent node can be reached in a straight line and can help reduce the overall cost, then this upstream node will be defined as the new parent for the current node. +III. ResultsUsing the method described in the previous section, given an origin and destination pair, an obstacle-free and wind-efficient path is quickly generated.Figure 7(a) and 7(b) show two final paths (shown as orange curves) that are obstacle-free and wind-efficient for operations in downtown San Francisco at AGL 30 ft and 90 ft, respectively.The black circles denote origin, and black crosses denote destinations.The operational environments are quite different at different AGLs: the obstacle field is more complicated at lower AGL and wind is stronger at higher AGL. +A. Computational performanceTo examine the efficiency of the proposed method, experiments were conducted on a MacBook Pro with 2.5 GHz Intel Core i7 and 18 GB memory.The computational times for these cases are presented in Table 1.Case I, II, and III have the same search area, origin, and destination except for different flight altitudes.Without loss of generality, the comparison among Cases I, II, and III shows that when the flight altitude increases, the time spent for building and searching decreases, mainly because the number of obstacles was reduced.On the other hand, Cases IV and V have longer paths than previous cases and the areas increased to 8 km×8 km and 10 km×10 km, respectively.The times consumed at all phases were slightly increased and the overall times are slightly larger (over 9 seconds), which is pretty fast considering the size and complexity of the search space.The breakdown also showed the computational times were almost evenly split among the reading, building, and searching phases.The computational cost of the post smoothing phase was negligible. +B. Vertical profile and lateral path angle changeFigure 8 shows the vertical profile of the final path in Case I.The solid line represents the flight path, and the dashed line denotes the ground terrain.The slopes of the flight path and terrain were well below 10%, which is acceptable for most small UAVs.However, as discussed in the previous section, the terrain slopes in areas highlighted in Fig. 3 would exceed the maximum climb and descent rates if small UAVs were requested to fly there at constant AGL. +Fig. 8 Vertical profile for the final path in Case IFigure 9 presents the absolute values of lateral path angle changes between two consecutive segments of the path in Case I.As expected, all path angle changes are less than 45 • , which was defined as the limit in the cost function.The Theta* method worked well in meeting the constraints for path angle changes. +C. Comparison with the conventional trajectory optimization methodIt is not suprising that the proposed path-planning algorithm can work well in obstacle-rich environments.To examine the effectiveness of the path-planning method in an obstacle-free environment, a multiple-shooting based trajectory optimization method was compared for finding wind-optimal trajectories.An optimization demo tool [22] that uses the multiple-shooting optimization method was applied to produce sample wind fields and minimum time trajectories under wind conditions.The Theta* based path planning method proposed in this work was then applied in the same wind field to find wind-optimal trajectories for comparison.Figure 10(a) shows the trajectories generated by these two different methods.The black curve represents the minimum-time wind-optimal trajectory generated by the ).The magenta curve reprents the path generated by the path planning algorithm using the wind coefficient minimizing the energy (as in Eqn. 7).The final flight times for these three trajectories (black, blue, and magenta) are 85.7s, 85.4s, and 85.7s respectively.The path planning method with time-minimum wind coefficient actually outperforms the multiple-shooting trajectory optimization method.Similar results hold for longer paths as shown in Fig. 10(b), where the wind field is a bit more complicated: the travel times for the multiple-shooting method (black curve) and path-planning method (blue curve) are 213.4sand 212.1s, respectively.With a proper choice of wind coefficient, the Theta* based path planning method performs surprisingly well. +D. DiscussionExperiments in this study showed the grid-based path planning algorithm can solve the flight planning problem efficiently while not sacrificing much optimality.Although only obstacles, wind, and path angle change were incorporated in the search cost in this work, other location-based factors such as dynamic and static restricted airspace, weather, ground risk, noise exposure, population density, and navigation and communication signal strength could also be incorporated to the cost function without much increase in computational time.Like any other multiple-objective optimization problem, how to construct/weigh these multiple costs still needs to be addressed carefully.Once the multiple-objective cost function is constructed, the grid-based Theta* path planning algorithm should be able to handle the search more efficiently than conventional trajectory optimization approaches. +IV. ConclusionsA grid-based Theta* path planning algorithm was introduced to generate obstacle-free and wind-efficient paths for sUAVs at a constant AGL in urban environments.This method first converts the 3D path planning problem to a 2D problem by constructing an obstacle map at a given AGL.A quad-tree decomposition is then used to build the search space in terms of obstacle occupancy and wind difference.The wind cost of traveling through each cell is defined based on UAV power consumption under various wind conditions.A repulsive potential is also adopted to make sure the flight plan stays away from obstacles.The Theta* search algorithm was applied to find the flight path having the lowest cost with the capability of mitigating the path angle change constraints.With the proposed Theta* algorithm and post smoothing techniques, an obstacle-free, wind-efficient, and constant above-ground-level flight plan was efficiently generated for small UAV operations in urban environments while meeting lateral path angle constraints.The results showed that the computational time of this algorithm is reasonable for real-time applications.The vertical profile and path angle change showed the feasibility of resulting paths, although there is a concern for the small UAVs flying at a constant AGL in an area where the slope exceeds the maximum climb/descent rates of the vehicle.Experiments also showed that, with a proper choice of the wind coefficient, the Theta*-based path planning algorithm outperformed the multiple-shooting trajectory optimization method in an obstacle-free environment.Overall, through experiments the proposed algorithm showed efficiency in flight path planning in complex urban environments and potential to incorporate geo-related costs.(a) 3D City Model for San Francisco (b) San Francisco Height Map +Fig. 1 Fig. 212Fig. 1 San Francisco 3D city model and height map +Fig. 33Fig. 3 The map of steep slope areas (shown in yellow) in San Francisco +Fig. 44Fig. 4 Obstacle map and its quad-tree decomposition for downtown San Francisco at AGL30 +Fig. 5 Incorporating wind into quad-tree decomposition for downtown San Francisco at AGL30 +Thrust history at various wind conditions (b) Coeffcient matrix for various wind vectors +Fig. 6 Algorithm 1 :61Fig. 6 Generation of wind coefficients +Algorithm 2 : 3 if 4 g 6 S2346Post Smoothing3.1 s ← s goal 3.2 while s s st art do 3.parent(parent(s)) is visible f rom s and g(parent(parent(s))) + c(parent(parent(s)), s) < g(s) then 3.(s) = g(parent(parent(s))) + c(parent(parent(s)), s) 3.5 parent(s) ← parent(parent(s)) 3.← Parent(S) +Fig. 77Fig. 7 Obstacle-free and wind-efficient sUAV paths with constant AGL over downtown San Francisco +Fig. 99Fig. 9 Heading angle change (absolute) for the final path in Case I +(a) Comparison of short paths (b) Comparison of long paths +Fig. 1010Fig. 10 Comparison of wind optimal trajectories generated with different methods + +Table 1 Computational time breakdown1Case ICase IICase IIICase IVCase V4km×4km 4km×4km 4km×4km 8km×8km 10km×10kmAGL 30ftAGL 60ft AGL 150ft AGL 30ftAGL 30ftReading inputs (s)3.083.143.013.493.78Building (s)1.631.260.731.802.04Searching (s)2.552.601.882.053.91Post-smoothing (s)0.010.010.010.010.03Total (s)7.277.015.637.319.76 + + + + +The authors gratefully acknowledge the contribution of Dr. John Melton and his team in generating the CFD-based wind profiles for this work.The authors would also like to thank Russell Paielli for the discussion regarding operations under the current UTM concept and Dr. Hok Ng for the discussion of aircraft trajectory optimization in winds. + + + + + + + + + Noise-Minimum Runway-Independent Aircraft Approach Design for Baltimore-Washington International Airport + + MinXue + + + EllaMAtkins + + 10.2514/1.15692 + + + Journal of Aircraft + Journal of Aircraft + 0021-8669 + 1533-3868 + + 43 + 1 + + 2005 + American Institute of Aeronautics and Astronautics (AIAA) + + + Xue, M., and Atkins, E. M., "Noise-minimum Runway-Independent Aircraft Approach Design for Baltimore-Washinton International Airport," Journal of Aircraft, Vol. 43, No. 1, 2005, pp. 39-51. + + + + + Survey of Numerical Methods for Trajectory Optimization + + JohnTBetts + + 10.2514/2.4231 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 21 + 2 + + 1998 + American Institute of Aeronautics and Astronautics (AIAA) + + + Betts, J., "Survey of Numberical Methods for Trajectory Optimization," Journal of Guidance, Control, and Dynamics, Vol. 21, 1998, pp. 193-207. + + + + + Generalized gradient algorithm for trajectory optimization + + YiyuanZhao + + + AEBryson + + + RSlattery + + 10.2514/3.20597 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 13 + 6 + + 1990 + American Institute of Aeronautics and Astronautics (AIAA) + + + Zhao, Y., Bryson, A. E., and Slattery, R., "Generalized Gradient Algorithm For Trajectory Optimization," Journal of Guidance, Control, and Dynamics, Vol. 13, No. 6, 1990, pp. 1166-1169. + + + + + Three-Dimensional Trajectory Design for Reducing Climate Impact of Trans-Atlantic Flights + + HokKwanNg + + + BanavarSridhar + + + NeilYChen + + + JinhuaLi + + 10.2514/6.2014-2289 + + + 14th AIAA Aviation Technology, Integration, and Operations Conference + Atlanta, GA + + American Institute of Aeronautics and Astronautics + 2014 + + + Ng, H. K., Sridhar, B., Chen, N. Y., and Li, J., "Three-Dimensional Trajectory Design For Reducing Climate Impact of Trans-Atlantic Flights," AIAA Aviation Forum, Atlanta, GA, 2014. + + + + + Terminal Area Trajectory Optimization Using Simulated Annealing + + MinXue + + + EllaAtkins + + 10.2514/6.2006-1473 + + + 44th AIAA Aerospace Sciences Meeting and Exhibit + Reno, Nevada + + American Institute of Aeronautics and Astronautics + 2006 + + + Xue, M., and Atkins, E. M., "Terminal Area Trajectory Optimization Using Simulated Annealing," 44th AIAA Aerospace Science Meeting and Exhibit, Reno, Nevada, 2006. + + + + + Spacecraft Trajectory Planning with Avoidance Constraints Using Mixed-Integer Linear Programming + + ArthurRichards + + + TomSchouwenaars + + + JonathanPHow + + + EricFeron + + 10.2514/2.4943 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 25 + 4 + + 2002 + American Institute of Aeronautics and Astronautics (AIAA) + + + Richards, A., Schouwenaars, T., How, J. P., and Feron, E., "Spacecraft Trajectory Planning with Avoidance Constraints Using Mixed-Integer Linear Programming," Journal of Guidance, Control, and Dynamics, Vol. 25, No. 4, 2002, pp. 755-764. + + + + + Robot Motion Planning + + Jean-ClaudeCLatombe + + 10.1007/978-1-4615-4022-9 + + 1991 + Springer US + Norwell, MA + + + Latombe, J. C., Robot Motion Planning, Kluwer Academic Press, Norwell, MA, 1991. + + + + + Football Trajectory Monitoring System based on Improved RRT-FN Algorithm + + PCheng + + + ZShen + + + SMLavalle + + 10.1109/aiars57204.2022.00022 + + + 2022 International Conference on Artificial Intelligence and Autonomous Robot Systems (AIARS) + + IEEE + 2001 + 11 + + + + Cheng, P., Shen, Z., and LaValle, S. M., "RRT-based Trajectory Design for Autonomous Automobiles and Spacecraft," Archives of Control Sciences, Vol. 11, No. 3-4, 2001, pp. 167-194. + + + + + Planning Algorithms + + StevenMLavalle + + 10.1017/cbo9780511546877 + + 2006 + Cambridge University Press + Cambridge + + + LaValle, S. M., Planning Algorithms, Cambridge University Press, Cambridge, 2006. + + + + + A Survey of Motion Planning Algorithms from the Perspective of Autonomous UAV Guidance + + CGoerzen + + + ZKong + + + BMettler + + 10.1007/s10846-009-9383-1 + + + Journal of Intelligent and Robotic Systems + J Intell Robot Syst + 0921-0296 + 1573-0409 + + 57 + 1-4 + + 2010 + Springer Science and Business Media LLC + + + Goerzen, C., Kong, Z., and Mettler, B., "A Survey of Motion Planning Algorithms from the Perspective of Autonomous UAV Guidance," Journal of Intelligent and Robotic Systems, Vol. 57, 2010, pp. 65-100. + + + + + Motion Planning for a Fixed-Wing UAV in Urban Environments + + MVRamana + + + SAdityaVarma + + + MangalKothari + + 10.1016/j.ifacol.2016.03.090 + + + IFAC-PapersOnLine + IFAC-PapersOnLine + 2405-8963 + + 49 + 1 + + 2016 + Elsevier BV + Tiruchirappalli, India + + + Ramana, M., Varma, S. A., and Kothari, M., "Motion Planning for a Fixed-Wing UAV in Urban Environments," 4th IFAC Conference on Advances in Control and Optimization of Dynamical Systems (ACODS), Tiruchirappalli, India, 2016. + + + + + A Risk-Aware Path Planning Strategy for UAVs in Urban Environments + + StefanoPrimatesta + + + GiorgioGuglieri + + + AlessandroRizzo + + 10.1007/s10846-018-0924-3 + + + Journal of Intelligent & Robotic Systems + J Intell Robot Syst + 0921-0296 + 1573-0409 + + 95 + 2 + + 2019 + Springer Science and Business Media LLC + + + Primatesta, S., Guglieri, G., and Rizzo, A., "A Risk-Aware Path Planning Strategy for UAVs in Urban Environments," Journal of Intelligent & Robotic Systems, Vol. 95, No. 2, 2019, pp. 629-643. + + + + + Real-Time Path Planning for Multi-copters flying in UTM -TCL4 + + AnjanChakrabarty + + + VahramStepanyan + + + KalmanjeSKrishnakumar + + + CoreyAIppolito + + 10.2514/6.2019-0958 + + + AIAA Scitech 2019 Forum + San Diego, California + + American Institute of Aeronautics and Astronautics + 2019 + + + Chakrabarty, A., Stepanyan, V., and Krishnakumar, K., "Real-Time Path Planning for Multi-copters Flying in UTM-TCL4," AIAA SciTech Forum, San Diego, California, 2019. + + + + + 14. The Lessons of San Francisco + + Browser + + 10.1525/9780520914902-017 + + + + City for Sale + + University of California Press + 2019. May, 2019 + + + + CAD Browser, "San Francisco 3D City Model," https://www.3dcadbrowser.com/3d-model/san-francisco-city, 2019. Online; accessed May, 2019. + + + + + A Computational Approach to Edge Detection + + JohnCanny + + 10.1109/tpami.1986.4767851 + + + IEEE Transactions on Pattern Analysis and Machine Intelligence + IEEE Trans. Pattern Anal. Mach. Intell. + 0162-8828 + + PAMI-8 + 6 + + 1986 + Institute of Electrical and Electronics Engineers (IEEE) + + + Canny, J., "A Computational Approach To Edge Detection," IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 8, No. 6, 1986, pp. 679-698. + + + + + Fe3: An Evaluation Tool for Low-Altitude Air Traffic Operations + + MinXue + + + JosephRios + + + JosephSilva + + + ZhifanZhu + + + AbrahamKIshihara + + 10.2514/6.2018-3848 + + + 2018 Aviation Technology, Integration, and Operations Conference + Atlanta, GA + + American Institute of Aeronautics and Astronautics + 2018 + + + Xue, M., Rios, J., Silva, J., Ishihara, A., and Zhu, Z., "Fe 3 : An Evaluation Tool for Low-Altitude Air Traffic Operations," AIAA Aviation Forum, Atlanta, GA., 2018. + + + + + Wind Tunnel and Hover Performance Test Results for Multicopter UAS Vehicles + + CRRussell + + + JJung + + + GWillink + + + BGlasner + + + + American Helocipter Society 72nd Annual Forum + West Palm Beach, FL + + 2016 + + + Russell, C. R., Jung, J., Willink, G., and Glasner, B., "Wind Tunnel and Hover Performance Test Results for Multicopter UAS Vehicles," American Helocipter Society 72nd Annual Forum, West Palm Beach, FL, 2016. + + + + + High-Fidelity Multi-Rotor Unmanned Aircraft System (UAS) Simulation Development for Trajectory Prediction Under Off-Nominal Flight Dynamics + + JohnVFoster + + + DavidHartman + + 10.2514/6.2017-3271 + + + 17th AIAA Aviation Technology, Integration, and Operations Conference + Denver, Colorado + + American Institute of Aeronautics and Astronautics + 2017 + + + Foster, J. V., and Hartman, D. C., "High-Fidelity Multirotor Unmanned Aircraft System Simulation Development for Trajectory Prediction Under Off-Nominal Flight Dynamics," 17th AIAA Aviation Technology, Integration, and Operations Conference, Denver, Colorado, 2017. + + + + + Toward indoor flying robots + + J-DNicoud + + + J-CZufferey + + 10.1109/irds.2002.1041486 + + + IEEE/RSJ International Conference on Intelligent Robots and Systems + Lausanne, Switzerland + + IEEE + 2002 + + + Nicoud, J. D., and Zufferry, J. C., "Toward Indoor Flying Robots," IEEE International Conference on Intelligent Robots and Systems, Lausanne, Switzerland, 2002. + + + + + Theta*: Any-Angle Path Planning on Grids + + ANash + + + KDaniel + + + SKoenig + + + AFelner + + + + AAAI'07 Proceedings of the 22nd national conference on Artificial intelligence + Vancouver, British Columbia, Canada + + 2007 + + + Nash, A., Daniel, K., Koenig, S., and Felner, A., "Theta*: Any-Angle Path Planning on Grids," AAAI'07 Proceedings of the 22nd national conference on Artificial intelligence, Vancouver, British Columbia, Canada, 2007. + + + + + Near Optimal Hierarchical Path-finding + + ABotea + + + MMuller + + + JSchaeffer + + + + Journal of Game Development + + 1 + + 2004 + + + Botea, A., Muller, M., and Schaeffer, J., "Near Optimal Hierarchical Path-finding," Journal of Game Development, Vol. 1, 2004, pp. 7-28. + + + + + Chapter 5. Using MATLAB optimization toolbox + + MFenelon + + 10.1515/9783110625622-005 + + + + Project Optimization + + De Gruyter + 2020. March 30, 2020 + + + + Fenelon, M., "Demo: Finding an Optimal Path Using MATLAB and Optimization Toolbox," https: //www.mathworks.com/matlabcentral/fileexchange/36321-demo-finding-an-optimal-path-using- matlab-and-optimization-toolbox, 2020. MATLAB Central File Exchange. Retrieved March 30, 2020. + + + + + + diff --git a/file816.txt b/file816.txt new file mode 100644 index 0000000000000000000000000000000000000000..d42fd73f68d82b460f341791e18cb67891c4457e --- /dev/null +++ b/file816.txt @@ -0,0 +1,809 @@ + + + + +In terminal airspace, integrating arrivals and departures with shared waypoints provides the potential of improving operational efficiency by allowing direct routes when possible.Incorporating stochastic evaluation as a post-analysis process of deterministic optimization, and imposing a safety buffer in deterministic optimization, are two ways to learn and alleviate the impact of uncertainty and to avoid unexpected outcomes.This work presents a third and direct way to take uncertainty into consideration during the optimization.The impact of uncertainty was incorporated into cost evaluations when searching for the optimal solutions.The controller intervention count was computed using a heuristic model and served as another stochastic cost besides total delay.Costs under uncertainty were evaluated using Monte Carlo simulations.The Pareto fronts that contain a set of solutions were identified and the trade-off between delays and controller intervention count was shown.Solutions that shared similar delays but had different intervention counts were investigated.The results showed that optimization under uncertainty could identify compromise solutions on Pareto fonts, which is better than deterministic optimization with extra safety buffers.It helps decision-makers reduce controller intervention while achieving low delays. +I. IntroductionIn the National Airspace System (NAS), terminal airspace is often busy and complicated because hundreds of flights must fly through a limited airspace in a short time period and most flights in terminal areas are climbing and descending with varied speeds.Situations can become severe when there are several neighboring busy airports.Bottlenecks can be formed easily and impair the efficiency of air traffic operations.Therefore, improving the operation efficiency in terminal areas with narrow airspace and shared resources is critical for building an efficient air traffic system.2][3][4][5][6] Still others [7][8][9] addressed airport surface management problems, which are correlated to the efficiency of terminal airspace operations.When different arrival and/or departure flows share the same resources such as runways, waypoints and/or route segments, inefficient operations emerge because of the constraints of shared resources.Such interactions can happen among departures, arrivals, or between departures and arrivals, which is common in terminal areas.These interactions are expected to happen more often when continuous descent approaches are deployed in the future.Recent studies [10][11][12][13] showed that optimized integrated arrivals and/or departures in major airports or metroplex areas have promise for improving operation efficiency.However, the benefits from above optimal schedules are calculated under deterministic scenarios and they are usually sensitive to flight time uncertainties, which can be caused by many sources, such as inaccurate wind prediction, error in aircraft dynamics, or human factors.Therefore, the impact of uncertainty must be taken into account when evaluating a schedule's benefits.Incorporating stochastic evaluation as a post-analysis process of deterministic optimization as in previous work 14 is one passive way to learn the impact of uncertainty and to avoid unexpected results.Imposing an extra safety buffer upon separation requirements is another prevalent yet passive way to consider uncertainty.Optimizing schedules under uncertainty is a proactive method that directly takes uncertainty into account.It provides a set of compromise solutions on the Pareto front and helps decision makers avoid unexpected effects from uncertainty.This work presents an optimization-under-uncertainty method for integrated arrival and departure scheduling problems.The impact of uncertainty was evaluated directly in the optimization.The optimization is multi-objective including total delay and controller intervention count.Both costs were evaluated stochastically using Monte-Carlo simulations.To enable the time-consuming optimization, the number of Monte-Carlo simulations were decreased without sacrificing accuracy.The Pareto fronts that contain sets of solutions were presented to show the trade-off between delays and interventions.Solutions with similar de-lays but different controller intervention count were investigated.Furthermore, the solutions provided via deterministic optimization with extra buffers were presented and compared against the Pareto front provided by stochastic optimization or optimization under uncertainty.In the paper, Section II revisits the problem studied in previous work.Section III presents methodology that includes modeling, optimization objectives, uncertainty evaluation approach and implementation.Section III provides the results and analysis.Section IV provides conclusions for this work. +II. Background and ProblemThe interactions between arrivals and departures in Los Angeles terminal airspace were presented in previous work 12 for studying optimal integrated operations in deterministic circumstances.The uncertainty analysis of the deterministic solutions has been conducted in other work 14 as determined from post-processing.This work is focused on the same problem of integrating arrivals and departures in Los Angeles.According to the Standard Terminal Arrival Routes (STARs) and the Standard Instrument Departures (SIDs) of Los Angeles terminal airspace, the arrivals from the FIM fix would follow procedure SADDE6 ( FIM-SYMON-SADDE-SMO) and the departures to the North need to follow procedure CASTA2 (RWY-NAANC-GHART-SILEX) (see Fig. 1).The arrivals are requested to maintain their flight altitudes above 10,000 feet at fix GHART and the departures have to keep theirs at or below 9,000 feet at the same fix to procedurally avoid potential conflicts between arrivals and departures.If there were no interactions, departures to the North and arrivals from FIM would have flown direct routes.As shown in the Fig. 1, the direct routes would be RWY-WPT2-WPT1 and FIM-WPT1-SMO for departures and arrivals, respectively, where WPT1 and WPT2 are made-up fix names for simplicity.Compared to these direct routes, besides flying non-preferred altitudes, individual arrival and departure flights following current procedures will waste approximately 60 and 120 seconds, respectively.Table 1 shows a representative schedule of 14 flights, which covers half an hour of traffic in actual operations.Two flows are included: 6 departures to the North from Runway 24L (RWY) and 8 arrivals from FIM.This schedule used the traffic between 9:00am and 9:30am (local time) on December 4, 2012 as reference.The initial times shown in the table are relative times to simulation start time.The "Order" of each flight is sorted based on intial times. +III. MethodIn this study, the model is constructed similarly to previous works. 12,14 owever, the optimization is formulated as a multiple objective optimization in order to include more uncertainty-related costs.The first objective is to minimize delay and the second objective is to minimize controller intervention, which is believed to correspond with controller workload increase when uncertainty arises.Both objectives were evaluated using Monte Carlo simulations. +A. ModelAs the route structure adopted in this study is the same as in previous work, the model that is built upon the route structure is also the same.Detailed description of the model should be found in previous work. 12For completeness, this model is briefly revisited in this subsection.As a general introduction, in this model, inputs are flights' estimated/scheduled times at entry waypoints.For a departure, this estimated time denotes the taking-off time at the runway.For an arrival, the time refers to the estimated time when the flight enters the designated terminal airspace at certain fixes.Aircraft types are also required inputs in this work.Outputs of the model are essentially decision variables, such as delay times at different waypoints, speeds at various route segments, and route options.They will be described later in this section.By imposing appropriate constraints and/or dynamics, this model provides feasible arrival times at designated waypoints without violating requirements, such as separations and aircraft dynamics.Three separation methods were compared in previous work: spatial, temporal, and hybrid.Spatial separation uses the same strategy as in SIDs and STARs to spatially separate interacting departure and/or arrival flows.Temporal separation utilizes the direct routes with conflicts resolved merely with temporal control.Hybrid separation applies both temporal and spatial separations.This work was focused on hybrid separation.In the formulation of hybrid separation, four design variables were defined for each FIM arrival i: d1 (F IM,i) is the delay before or at FIM; r (F IM,i) is the route option, where 0 denotes the direct route and 1 denotes indirect route; v (F IM,i) is the aircraft speed between FIM and WPT1 for the direct route or the speed between FIM and WPT2 if the indirect route is chosen; d2 (F IM,i) is the delay between WPT1/WPT2 and SUTIE to ensure separation at SUTIE.For a departure flight j, three decision variables were defined: d (DEP,j) is the delay before departure; r (DEP,j) is the route option, where 0 denotes the direct route and 1 denotes the indirect route.v (DEP,j) is the speed from departure to WPT1.Separation requirements were applied as hard constraints at fixes that could have potential violations, such as FIM, RWY, WPT1, WPT2, and SUTIE.Separation requirements were 3 nmi at the RWY and 4 nmi elsewhere.Additionally, an extra safety buffer δ could be added in the separation constraints if required.For example, Eqns. 1, 2, and 3 show separation constraints for crossing flights between FIM arrivals and departures at WPT1, where t F IM (W P T 1,i) is the arrival time of FIM arrivals at fix WPT1 and t DEP (W P T 1,j) is the arrival time of departures at fix WPT1.Eqn. 1 showed that if both arrival flight i and departure flight j took direct routes the separation must be satisfied.The separation requirement is 4.0 nmi in distance and in the time scale the separation depends on the speeds of both flights.(1 -r (F IM,i) )•(1 -r (DEP,j) )•[t F IM (W P T 1,i) -t DEP (W P T 1,j) ] - 4.0×3600.0 V •sinα -δ > 0 (1) V =      v (F IM,i) , if t F IM (W P T 1,i) < t DEP (W P T 1,j) v (DEP,j) , otherwise(2)α =      atan[ v (DEP,j) v (F IM,i) ], if t F IM (W P T 1,i) < t DEP (W P T 1,j) atan[ v (F IM,i) v (DEP,j) ], otherwise(3) +B. Evaluation of uncertaintyTo explain how to evaluate costs caused by uncertainty, three parts are discussed: perturbation error sources, heuristic controller behavior model, and identification of uncertainty cost evaluation method. +Uncertainty sourcesThere exist many sources of uncertainty: inaccurate wind prediction, unknown aircraft weight, error in aircraft dynamics, or inaccurate prediction of flight take-off or arrival time.Because these errors are mainly reflected in flight arrival or departure times in this study, to simulate the uncertainty, errors were added in flight arrival times at waypoint FIM and SUTIE, and departure times at RWY, respectively.The imposed errors were assumed to follow normal distributions.Although, other distribution may be possible, such as Poisson in Mueller's work, 15 it could be straightforward to change in the future when the distribution is finally identified.In this work, the arrival time error follows a standard deviation of 30 seconds with a zero mean, which was commonly used as a desired and reliable prediction accuracy in arrival trajectory prediction studies. 16,17 he departure time error's standard deviation is 90 seconds and the mean value is 30 second, which is based on the Call For Release (CFR) 3-minute compliance window. 18The CFR window is often structured to allow departure 2 minutes prior to or 1 minute later than the target coordinated departure time. +Heuristic controller behavior modelA heuristic model was built to mimic controller intervention behaviors as in previous works. 12,14 en stochastic errors are added in flights' entry times, the heuristic model resolves potential conflicts by imposing extra delays to corresponding aircraft, while keeping the same route options as in the given solution.The conflict resolution simulates controller's behavior by following the First-Come-First-Served (FCFS) rule.At a waypoint, this model sorts the sequence based on the flights' perturbed entry times.Then extra delays are imposed if the separation between any two adjacent aircraft doesn't meet the defined requirement, which is the sum of the separation requirement and an additional buffer if required.The imposed delays are then propagated to following waypoints.In actual operations controllers might use improved heuristics other than this strategy, it should not be a problem to incorporate it into this formulation in the future. +Evaluation methodA traditional approach to evaluate a function with random inputs is the Monte Carlo method, which generates random realizations for the prescribed random inputs and utilizes repetitive deterministic function evaluations for each realization.In this study, the deterministic function is the heuristic controller behavior model.A brute-force Monte Carlo method with K realizations converges asymptotically at a rate 1/ √ K. 19 Groups of methods in the literature have been proposed to speed up the evaluation process while keeping similar accuracy.Many methods were developed for solving stochastic differential equations with uncertain inputs, such as generalized polynomial chaos (gPC), 20 multi-element generalized polynomial chaos (ME-gPC), 21 and multi-element probabilistic collocation method (ME-PCM). 22,23 he probabilistic collocation method (PCM) has been applied successfully to dynamic simulation of power system 24 and geophysical models. 25It has also been tried in certain air traffic problems. 26,27 he main idea of these methods is to speed up the process by applying a smart sampling method with small samples.This is often achieved by a certain type of decomposition which can approximate the random process with desire accuracy.For instance, PCM uses a polynomial expression to represent the random response as a function of orthogonal polynomials of random variables or polynomial chaos expansion.However, although these methods are good at solving stochastic differential equations or time-step simulations, limitation exists when the dimension of the random inputs is high and/or regularity of the function is poor.Experiments were conducted for the problem in this study using PCM, ME-gpc, and MEPCM-A methods.But they were not successful.Poor response accuracies were obtained for the first-order moments or mean values, and even worse precisions were produced for the second-order moments or standard deviations.The reason may be that the problem in this work involves high dimension (number of aircraft) and the response is discontinuous with low regularity.Given the fact that the Monte Carlo method is independent of the dimensionality of the random space, experiments were also conducted using the Monte Carlo method.Directly incorporating full scale Monte Carlo simulations (say 5,000 or 10,000 simulations) into objective evaluation would be computationally expensive and make the optimization impractical.However, experiments showed that the calculations of mean values and standard deviations converge fast when the number of Monte Carlo simulations goes beyond 1,000.Figure 2 shows the differences in the controller intervention costs from 10 to 8,000 simulations during optimization process for around 10,000 different scenarios/cases.The horizontal axis is the number of simulations and the vertical axis denotes the relative differences of mean values against the "truth", which is calculated using 12,000 simulations.The differences are expressed in percentage.It can be seen that at 800 simulations the differences are reduced within 3%, and the progress stabilizes as the number of simulations increase further.Figure 3 presents the changing rates of the differences, or the derivatives of the curve in Figure 2. Apparently, the reduction of the differences is slowed down dramatically after 500.Experiments with different scenarios/cases also showed similar trends.Although, this don't serve as an official proof, it does show that 1,000 Monte Carlo simulations could provide a good approximate to the random process for this work. +C. ObjectivesThe optimization in this work is formulated as multiple objective: minimizing expectation values of total delays and controller interventions, respectively.The objectives are shown in Eqn.4:     J 1 = E[ i t i ] -T unimpeded J 2 = E[ i N i ] (4)J 1 is the expectation value of the total delay over the sampling space caused by uncertainty.The total delay can be expressed as the difference between the sum of flight transit times t i and the sum of the unimpeded transit times.The unimpeded transit time is the transit time when a flight takes a direct route without any delay.As the sum of unimpeded transit times is a constant, its expected value is represented by T unimpeded in the equation.J 2 is the expectation value of the sum of controller intervention count.When a flight needs to be delayed due to a separation violation at a merging or diverging waypoint, the intervention count will be increased be one.Both costs are evaluated over sampling points using aforementioned Monte Carlo simulations.The optimization is solved using a Non-dominated Sorting Genetic Algorithm (NSGA) because of NSGA's ability of handling multiple objective optimization. 12,28 s the costs are handled independently, NSGA will lead its search to a Pareto front, where no solution on the front is dominated by another.The dominance with two objectives is defined as: Assuming the objective is to minimize costs and the constraint function g has to be nonnegative, solution A is dominated by solution B only if:           J 1 (A) > J 1 (B), if g A ≥0 and g B ≥0, or, g A = g B J 2 (A) > J 2 (B), if g A ≥0 and g B ≥0, or, g A = g B g A < g B ,if g A < 0 and g B < 0, or, g B > 0 and g A < 0(5)where J 1 and J 2 are the objectives and g is the constraint value. +D. ImplementationThe optimization was initially implemented using C and Monte Carlo was multi-threaded with pthread.Using 1,000 simulations, this optimization took about 6 hours to solve the 14 aircraft scenario on a MacOS platform with 2x2.66 GHz 6-Core Intel Xeon and 8 GB RAM.The running time is prohibitively long, which makes applications or experiments impossible.To improve the performance, in recent work, 29 the Monte Carlo part was implemented using CUDA programming with GPUs, and the running-time was reduced to around 2.5 minutes on a Linux platform with one GeForce GTX690 GPU.After further code optimization/modification, solving the same scenario now takes around 30 seconds on a Linux platform with 18x2.5 GHz Xeon and 32 GB memory and two GeForce GTX690.This fast-time implementation makes the stochastic optimization a promising method in application. +IV. ResultsThis section presents the solutions optimized under uncertainty.The trade-offs between delay reduction and controller intervention count increase are shown.The solutions that have similar delays and different intervention counts are compared.The advantages of stochastic optimization over deterministic optimization are shown through results. +A. The Pareto frontGiven two objectives of reducing delays and controller intervention counts, NSGA led the search to a Pareto front, where no solution on the front was dominated by other solutions.GA-based optimizations are sensitive to initial guesses, which are decided by initial randomized seeds.Multiple runs are necessary to increase the chance of getting optimal solutions.Ten runs were performed in this work.Solutions in final generations of the evolution process were recorded as they were close to the Pareto front.Figure 4 presents the solutions in final generations ( about 700 solutions) that are on the Pareto front.Each dot (J 1 , J 2 ) corresponds to a solution and its coordinates J 1 and J 2 are the two costs of the solution.The green curve presents coordinates/costs calculated in the optimization process, using 1,000 Monte Carlo simulations.Whereas, the black curve presents coordinates computed in a postprocess using 5,000 Monte Carlo simulations.The percentage differences are 1.4% and 0.2% for intervention count and delay, respectively.The marginal difference further demonstrates that reducing the number of Monte Carlo to 1,000 works well.The big dot at coordinate (1253, 1.23) sets a reference point for intervention counts and delay costs under spatial separation, which can be treated as current ATC procedure.To retrieve the best solution under spatial separation, an optimization was conducted under deterministic scenarios with delay as the single objective.According to the final solutions, it is noted that a clear tradeoff exists between delays and intervention counts.When delay is reduced the chance of controller intervention increases.Variations of intervention counts are large for solutions with similar delays, especially when delays are low.A multiple-objective optimization that incorporates uncertainty can clearly help to find the target delay with minimum intervention or visa versa.If a decision support tool can be built upon this method, in terms of the solutions that are close to the Pareto front, decision makers can choose a "compromise" solution according to their preferences.For instance, if controller intervention is believed to be more important than delay savings, the solutions that have a delay of about 900 seconds and intervention count similar to the spatial solution should be picked.Or, if two costs are weighted similarly, the compromise solutions around the middle (e.g.delay around 500) can be chosen because of their large delay savings and tolerable controller intervention increase.If controller intervention doesn't cause much workload due to advanced equipage in the future, then a solution with 200 second delay can be chosen. +B. Effect of multiple objectivesTo get insight into the effect of the optimization objectives, solutions that shared similar delays but with different controller intervention counts are retrieved for examination.For instance, at delays around 430 seconds, the expectation of intervention count can vary from 2.4 to 3.5.The solution with 3.5 intervention count won't be at the Pareto front.Figures 5 and6 are the time lines for these two solutions at either end of this intervention count range.The small gray boxes are the minimum separation requirements associated with flights.As mentioned previously, the separation requirement is a function of aircraft speed and the type of potential conflicts (crossing or in-trail).Therefore, the gray boxes have different lengths.Any departure following the direct route should go through WPT2, whereas any arrival from the FIM fix flying the direct route would pass WPT1.In terms of costs, these two solutions shared similar delays at around 423 seconds but resulted in different controller intervention efforts.The solution associated with Fig. 6 has an expected intervention count of 2.45, whereas the Fig. 5 solution corresponds an expected intervention count of 3.56 -a 45 % increase.Based on the comparison, the major differences between them are that: The high-controller-intervention solution allowed FIM003 and FIM008 to fly direct routes whereas the low-controller-intervention solution chose FIM007 and DEP006 to fly direct routes (See the flights pointed by arrows in the figures).Those differences in route options would not affect the delay reduction but would lead to different controller intervention.This subtle difference could not be easily predicted without the +C. Stochastic vs. deterministicWhen dealing with uncertainty, one prevalent method in scheduling problems is to add an extra buffer on top of separation requirement in a deterministic optimization.The choice of the buffer size depends on applications and/or users.One advantage of optimization based on this method is its fast running time, because it is a deterministic optimization.However, this method doesn't include any distribution knowledge of error sources, simply and blindly enforcing a constant buffer doesn't usually help to find "best" solutions.In order to demonstrate the difference, Fig. 7 presents the comparisons between stochastic and deterministic optimizations.Because the intervention count cost is zero for a feasible solution in deterministic cases, optimization in deterministic scenarios can only contain one objective -the delay cost.The curve on the lower left side represents the Pareto front generated using stochastic optimization or optimization under uncertainty.Three solutions on the top left side are generated using deterministic optimization with hybrid separation and varied buffer size.The circular dot is the solution with 0s buffer, and the triangle and square solutions are produced with 30s and 60s buffers, respectively.Apparently, the deterministic solutions are not at the Pareto front, which means, with the same delay, stochastic optimization can find better solution with lower controller intervention cost than deterministic optimization.The solution's distances to the front are random, which implies that the size of the buffer produces a random effect on the solution quality.To be complete, Fig. 7 also provides the solutions generated under deterministic optimization and spatial separation with varied extra buffers.It is noted that they are overlapped with each other with pretty close statistic costs.The reason could be that the natural gaps among flights in this scenario with spatial separation are either too big or too small such that imposing extra buffers doesn't affect the final solution too much.This also shows that the effect from the extra safety buffers in deterministic optimization on final solutions can not be easily predicted and the usage is blind.As a comparison, with information of error sources, stochastic optimization can lead the optimization towards the Pareto front with good-quality solutions. +V. ConclusionsIn order to directly take the impact of uncertainty into account, an optimization method of integrated arrivals and departures under uncertainty was developed in this study.A simple controller behavior model was incorporated to compute controller intervention, such that the controller intervention count could be included as another stochastic cost besides delay cost.The problem was then formulated as a multiple objective optimization with delays and intervention count costs.Monte Carlo simulations were utilized to evaluate stochastic costs, because, unlike other prevalent PCM methods, it is independent of the dimensionality and regularity.To enable the time-consuming optimization, the number of Monte Carlo simulations was reduced without losing much accuracy.The Pareto fronts that contain sets of solutions were identified.The trade-off between delays and controller intervention counts was shown and solutions with similar delays but different intervention effort were investigated.Comparison between stochastic and deterministic optimization was conducted.As the Monte Carlo simulations and the NSGA algorithm are all well-suited for parallelization because of their independent calculations and memories, the current implementation with GPUs can solve a typical half an hour scenario in about 30 seconds.Through this study, the optimization under uncertainty for integrated arrivals and departures was found to be feasible with simplified Monte Carlo simulations.Decreasing the number of simulations from 5,000 to 1,000 affected the controller intervention evaluation about 3% or less with given samples, but helped reduce the computational times to a reasonable level.Using this formulation, the method can provide a sweep of solutions at the Pareto front so the decision makers can choose in terms of their preferences.Solutions may have similar delays but cause quite different controller intervention efforts.The subtle difference in solutions/delays can result in a significant difference in intervention counts (e.g.45%), which would not be easily foreseen without the optimization under uncertainty.By the comparison between stochastic optimization and deterministic optimization with a safety buffer, it is shown that stochastic optimization outperforms the latter by providing better solutions located at the Pareto front.Imposing a safety buffer in a deterministic optimization actually produced a random or unpredictable effect on final solutions.Figure 1 .1Figure 1.Interactions between SADDE arrivals and CASTA departures +Figure 2 .2Figure 2. Percentage difference for mean values with varied number of Monte Carlo simulations +Figure 3 .3Figure 3. Changing rate of percentage difference with varied number of Monte Carlo simulations +Figure 4 .4Figure 4. Solutions from stochastic optimization with multiple runs +Figure 5 .5Figure 5.Time line for the solution with high controller intervention count +Figure 6 .6Figure 6.Time line for the solution with low controller intervention count +Figure 7 .7Figure 7. Stochastic optimization versus deterministic optimization +Table 1 .1Scheduled initial timesOrder FIM (sec) RWY (sec)13968244616537283634110652951332161361475183071613NA81770NA + + + + +Mail Stop 210-8.AIAA senior member † Aerospace engineer, Aerospace High Density Operations Branch.Mail Stop 210-10.AIAA senior member + + + + + + + + + The dynamic scheduling of aircraft in the near terminal area + + RGDear + + 10.1016/0041-1647(78)90073-4 + R76-9 + + + Transportation Research + Transportation Research + 0041-1647 + + 12 + 4 + + September 1976 + Elsevier BV + + + Tech. Rep. Flight Transportation Laboratory Report + Dear, R. G., "The Dynamic Scheduling of Aircraft in the Near Terminal Area," Tech. Rep. Flight Transportation Laboratory Report R76-9, MIT, September 1976. + + + + + An algorithm for computer assisted sequencing and scheduling of terminal area operations + + RogerGDear + + + YosefSSherif + + 10.1016/0191-2607(91)90132-a + + + Transportation Research Part A: General + Transportation Research Part A: General + 0191-2607 + + 25 + 2-3 + + 1991 + Elsevier BV + + + Dear, R. G. and Sherif, Y. S., "An Algorithm For Computer Assisted Sequencing and Scheduling of Terminal Operations," Transportation Research A, Vol. 25A, No. 2/3, 1991, pp. 120-139. + + + + + Analysis of Delay Reducing and Fuel Saving Sequencing and Spacing Algorithms for Arrival Spacing + + FNeuman + + + HErzberger + + 103880 + + October 1991 + + + Tech. Rep. NASA Technical Memorandum + Neuman, F. and Erzberger, H., "Analysis of Delay Reducing and Fuel Saving Sequencing and Spacing Algorithms for Arrival Spacing," Tech. Rep. NASA Technical Memorandum 103880, October 1991. + + + + + Scheduling Aircraft Landings Under Constrained Position Shifting + + HamsaBalakrishnan + + + BalaChandran + + 10.2514/6.2006-6320 + + + AIAA Guidance, Navigation, and Control Conference and Exhibit + Keystone, CO + + American Institute of Aeronautics and Astronautics + August 2006 + + + + Balakrishnan, H. and Chandran, B., "Scheduling Aircraft Landings Under Constrained Position Shift- ing," AIAA Guidance, Navigation, and Control Conference, Keystone, CO, 21-24 August 2006. + + + + + Scheduling Aircraft Landings—The Static Case + + JEBeasley + + + MKrishnamoorthy + + + YMSharaiha + + + DAbramson + + 10.1287/trsc.34.2.180.12302 + + + Transportation Science + Transportation Science + 0041-1655 + 1526-5447 + + 34 + 2 + + 2000 + Institute for Operations Research and the Management Sciences (INFORMS) + + + Beasley, J. E., Krishnamoorthy, M., Sharaiha, Y. M., and Abramson, D., "Scheduling Aircraft Land- ings -The Static Case," Transportation Science, Vol. 34, No. 2, 2000, pp. 180-197. + + + + + Scheduling Aircraft Landings to Closely Spaced Parallel Runways + + MKupfer + + + + The Eighth USA/Europe Air Traffic Management Research and Development Seminar + Napa, CA + + June 2009 + + + Kupfer, M., "Scheduling Aircraft Landings to Closely Spaced Parallel Runways," The Eighth USA/Europe Air Traffic Management Research and Development Seminar , Napa, CA, June 2009. + + + + + A Mixed Integer Linear Program for Airport Departure Scheduling + + GautamGupta + + + WaqarMalik + + + YoonJung + + 10.2514/6.2009-6933 + + + 9th AIAA Aviation Technology, Integration, and Operations Conference (ATIO) + Hilton Head, South Carolina + + American Institute of Aeronautics and Astronautics + September 2009 + + + + Gupta, G., Malik, W., and Jung, Y. C., "A Mixed Integer Linear Program for Airport Departure Scheduling," 9th AIAA Aviation Technology, Integration and Operations Conference (ATIO), Hilton Head, South Carolina, 21-23 September 2009. + + + + + A Mixed Integer Linear Program for Solving a Multiple Route Taxi Scheduling Problem + + JustinMontoya + + + ZacharyWood + + + SivakumarRathinam + + + WaqarMalik + + 10.2514/6.2010-7692 + + + AIAA Guidance, Navigation, and Control Conference + Toronto, Canada + + American Institute of Aeronautics and Astronautics + August 2010 + + + + Montoya, J., Wood, Z., Rathinam, S., and Malik, W., "A Mixed Integer Linear Program for Solving a Multiple Route Taxi Scheduling Problem," AIAA Guidance, Navigation, and Control Conference, Toronto, Canada, 2-5 August 2010. + + + + + A Generalized Dynamic Programming Approach for a Departure Scheduling Problem + + SivakumarRathinam + + + ZacharyWood + + + BanavarSridhar + + + YoonJung + + 10.2514/6.2009-6250 + + + AIAA Guidance, Navigation, and Control Conference + Chicago, IL + + American Institute of Aeronautics and Astronautics + August 2009 + + + + Rathinam, S., Wood, Z., Sridhar, B., and Jung, Y. C., "A Generalized Dynamic Programming Ap- proach for a Departure Scheduling Problem," AIAA Guidance, Navigation, and Control Conference, Chicago, IL, 10-13 August 2009. + + + + + A Hybrid Optimization Approach to Air Traffic Management for Metroplex Operations + + BrianCapozzi + + + StephenAtkins + + 10.2514/6.2010-9062 + + + 10th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Fort Worth, Texas + + American Institute of Aeronautics and Astronautics + September 2010 + + + + Capozzi, B. J. and Atkins, S. C., "A Hybrid Optimization Approach to Air Traffic Management for Metroplex Operations," 10th AIAA Aviation Technology, Integration and Operations Conference (ATIO), Fort Worth, Texas, 13-15 September 2010. + + + + + Towards Optimal Routing and Scheduling of Metroplex Operations + + BrianCapozzi + + + StephenAtkins + + + SeongimChoi + + 10.2514/6.2009-7037 + + + 9th AIAA Aviation Technology, Integration, and Operations Conference (ATIO) + Hilton Head, South Carolina + + American Institute of Aeronautics and Astronautics + September 2009 + + + + Capozzi, B. J. and Atkins, S. C., "Towards Optimal Routing and Scheduling of Metroplex Opera- tions," 9th AIAA Aviation Technology, Integration, and Operations Conference(ATIO), Hilton Head, South Carolina, 21-23 September 2009. + + + + + Optimal Integration of Departures and Arrivals in Terminal Airspace + + MinXue + + + ShannonZelinski + + 10.2514/1.60489 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 37 + 1 + + 2014 + American Institute of Aeronautics and Astronautics (AIAA) + + + Xue, M. and Zelinski, S., "Optimal Integration of Departures and Arrivals in Terminal Airspace," AIAA Journal of Guidance, Control, and Dynamics, Vol. 37, No. 1, 2014, pp. 207-213. + + + + + Sequential Dynamic Strategies for Real-Time Scheduling of Terminal Traffic + + HemingChen + + + YiyuanJZhao + + 10.2514/1.c031503 + + + Journal of Aircraft + Journal of Aircraft + 0021-8669 + 1533-3868 + + 49 + 1 + + 2012 + American Institute of Aeronautics and Astronautics (AIAA) + + + Chen, H. and Zhao, Y. J., "Sequential Dynamic Strtegies for Real-Time Scheduling of Terminal Traf- fic," AIAA Journal of Aircraft, Vol. 49, 2012, pp. 237-249. + + + + + Uncertainty Analysis of Integrated Departures and Arrivals: A Los Angeles Case Study + + MinXue + + + ShannonZelinski + + + DanielGMulfinger + + 10.2514/6.2013-4347 + + + 2013 Aviation Technology, Integration, and Operations Conference + Los Angeles, LA + + American Institute of Aeronautics and Astronautics + August 2013 + + + + Xue, M., Zelinski, S., and Mulfinger, D., "Uncertainty Study of Integrated Departures and Arrivals: A Los Angeles Case Study," 13th AIAA Aviation Technology, Integration, and Operations Conference(ATIO), Los Angeles, LA, 12-14 August 2013. + + + + + Analysis of Aircraft Arrival and Departure Delay Characteristics + + EricMueller + + + GanoChatterji + + 10.2514/6.2002-5866 + + + AIAA's Aircraft Technology, Integration, and Operations (ATIO) 2002 Technical Forum + Los Angeles, CA + + American Institute of Aeronautics and Astronautics + 1-3 October 2002 + + + Mueller, E. R. and Chatterji, G. B., "Analysis of Aircraft Arrival And Departure Delay Character- istics," 2nd AIAA Aviation Technology, Integration and Operations Conference (ATIO), Los Angeles, CA, 1-3 October 2002. + + + + + Predictability of Top of Descent Location for Operational Idle-Thrust Descents + + LaurelStell + + 10.2514/6.2010-9116 + + + 10th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Berlin, Germany + + American Institute of Aeronautics and Astronautics + June 2011 + + + Stell, L., "Prediction of Top of Descent Location for Idle-thrust Descents," The Ninth USA/Europe Air Traffic Management Research and Development Seminar , Berlin, Germany, June 2011. + + + + + Improvement of Trajectory Synthesizer for Efficient Descent Advisor + + MinXue + + + HeinzErzberger + + 10.2514/6.2011-7020 + + + 11th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + + American Institute of Aeronautics and Astronautics + + + + Xue, M. and Erzberger, H., "Improvement of Trajectory Synthesizer for Efficient Descent Advisor," + + + + + 11th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + 10.2514/matio11 + + + AIAA Aviation Technology, Integration, and Operations Conference(ATIO) + Virginia Beach, VA + + American Institute of Aeronautics and Astronautics + September 2011 + + + + AIAA Aviation Technology, Integration, and Operations Conference(ATIO), Virginia Beach, VA, 20-22 September 2011. + + + + + Trajectory-Based Takeoff Time Predictions Applied to Tactical Departure Scheduling: Concept Description, System Design, and Initial Observations + + ShawnEngelland + + + RichardCapps + + 10.2514/6.2011-6875 + + + 11th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Virginia Beach, VA + + American Institute of Aeronautics and Astronautics + September 2011 + + + + Engelland, S. A. and Capps, A., "Trajectory-Based Takeoff Time Predictions Applied to Tactical Departure Scheduling: Concept Description, System Design, and Initial Observations," 11th AIAA Aviation Technology, Integration, and Operations Conference(ATIO), Virginia Beach, VA, 20-22 September 2011. + + + + + High-Order Collocation Methods for Differential Equations with Random Inputs + + DongbinXiu + + + JanSHesthaven + + 10.1137/040615201 + + + SIAM Journal on Scientific Computing + SIAM J. Sci. Comput. + 1064-8275 + 1095-7197 + + 27 + 3 + + 2005 + Society for Industrial & Applied Mathematics (SIAM) + + + Xiu, D. and Hesthaven, J. S., "High-order Collocation Methods for Differential Equations with Random Inputs," Society for Industrial and Applied Mathematics Journal on Scientific Computing, Vol. 27, No. 3, 2005, pp. 1118-1139. + + + + + The Wiener--Askey Polynomial Chaos for Stochastic Differential Equations + + DongbinXiu + + + GeorgeEmKarniadakis + + 10.1137/s1064827501387826 + + + SIAM Journal on Scientific Computing + SIAM J. Sci. Comput. + 1064-8275 + 1095-7197 + + 24 + 2 + + 2002 + Society for Industrial & Applied Mathematics (SIAM) + + + Xiu, D. and Karniadakis, G. E., "The Wiener-Askey Polynomial Chaos for Stochastic Differential Equations," Society for Industrial and Applied Mathematics Journal on Scientific Computing, Vol. 24, No. 2, 2002, pp. 619-644. + + + + + An adaptive multi-element generalized polynomial chaos method for stochastic differential equations + + XiaoliangWan + + + GeorgeEmKarniadakis + + 10.1016/j.jcp.2005.03.023 + + + Journal of Computational Physics + Journal of Computational Physics + 0021-9991 + + 209 + 2 + + 2005 + Elsevier BV + + + Wan, X. and Karniadakis, G. E., "An Adaptive Multi-element Generalized Polynomial Chaos Method for Stochastic Differential Equations," Journal of Computational Physics, Vol. 209, 2005, pp. 617-642. + + + + + The multi-element probabilistic collocation method (ME-PCM): Error analysis and applications + + JasmineFoo + + + XiaoliangWan + + + GeorgeEmKarniadakis + + 10.1016/j.jcp.2008.07.009 + + + Journal of Computational Physics + Journal of Computational Physics + 0021-9991 + + 227 + 22 + + 2008 + Elsevier BV + + + Foo, J., Wan, X., and Karniadakis, G. E., "The Multi-element Probabilistic Collocation Method (ME- PCM): Error-analysis and Applications," Journal of Computational Physics, Vol. 227, 2008, pp. 9572-9595. + + + + + Multi-element probabilistic collocation method in high dimensions + + JasmineFoo + + + GeorgeEmKarniadakis + + 10.1016/j.jcp.2009.10.043 + + + Journal of Computational Physics + Journal of Computational Physics + 0021-9991 + + 229 + 5 + + 2010 + Elsevier BV + + + Foo, J. and Karniadakis, G. E., "Multi-element Probabilistic Collocation Method in High Dimensions," Journal of Computational Physics, Vol. 229, 2010, pp. 1536-1557. + + + + + Evaluation of Uncertainty in Dynamic Simulations of Power System Models: The Probabilistic Collocation Method + + JRHockenberry + + + BCLesieutre + + 10.1109/tpwrs.2004.831689 + + + IEEE Transactions on Power Systems + IEEE Trans. Power Syst. + 0885-8950 + + 19 + 3 + + 2004 + Institute of Electrical and Electronics Engineers (IEEE) + + + Hockenberry, J. R. and Lesieutre, B. C., "Evaluation of Uncertainty in Dynamic Simulation of Power System Models: The Probabilistic Collocation Method," IEEE Transactions on Power Systems, Vol. 19, No. 3, 2004, pp. 1483-1491. + + + + + An efficient method for parametric uncertainty analysis of numerical geophysical models + + MennerATatang + + + WenweiPan + + + RonaldGPrinn + + + GregoryJMcrae + + 10.1029/97jd01654 + + + Journal of Geophysical Research: Atmospheres + J. Geophys. Res. + 0148-0227 + + 102 + D18 + + 1997 + American Geophysical Union (AGU) + + + Tatang, M. A., Pan, W., Prinn, R. G., and McRae, G. J., "An Efficient Method for Parametric Uncertainty Analysis of Numerical Geophysical Models," Journal of Geophysical Research, Vol. 102, No. D18, 1997, pp. 21,925-21,932. + + + + + Uncertainty Evaluation Through Mapping Identification in Intensive Dynamic Simulations + + YanWan + + + SandipRoy + + + BernardLesieutre + + 10.2514/6.2008-6480 + + + AIAA Guidance, Navigation and Control Conference and Exhibit + Honolulu, HI + + American Institute of Aeronautics and Astronautics + August 2008 + + + + Wan, Y., Roy, S., and Lesieutre, B., "Uncertainty Evaluation through Mapping Identification in In- tensive Synamic Simulations," AIAA Guidance, Navigation, and Control Conference, Honolulu, HI, 18-21 August 2008. + + + + + Multivariate Probabilistic Collocation Method for Effective Uncertainty Evaluation with Application to Air Traffic Management + + YZhou + + + DRamamurthy + + + YWan + + + SRoy + + + CTaylor + + + CWanke + + + + American Control Conference + + + June 2013 + Washington, DC + + + Zhou, Y., Ramamurthy, D., Wan, Y., Roy, S., Taylor, C., and Wanke, C., "Multivariate Probabilistic Collocation Method for Effective Uncertainty Evaluation with Application to Air Traffic Management," American Control Conference, Washington, DC, 17-19 June 2013. + + + + + A fast and elitist multiobjective genetic algorithm: NSGA-II + + KDeb + + + APratap + + + SAgarwal + + + TMeyarivan + + 10.1109/4235.996017 + + + IEEE Transactions on Evolutionary Computation + IEEE Trans. Evol. Computat. + 1089-778X + + 6 + 2 + + 2002 + Institute of Electrical and Electronics Engineers (IEEE) + + + Deb, K., Pretap, A., Agarwal, S., and Meyarivan, T., "A Fast and Elitist Multiobjective Genetic Algorithm: NSGA-II," IEEE Transactions On Evolutionary Computation, Vol. 6, No. 2, 2002, pp. 182-197. + + + + + GPU-based Parallelization for Schedule Optimization with Uncertainty + + ChristabelleBosson + + + MinXue + + + ShannonZelinski + + 10.2514/6.2014-2024 + + + 14th AIAA Aviation Technology, Integration, and Operations Conference + Atlanta, GA + + American Institute of Aeronautics and Astronautics + June 2014 + + + + Bosson, C., Xue, M., and Zelinski, S. J., "GPU-based Paralellization for Schedule Optimization with Uncertainty," AIAA Aviation, Atlanta, GA, 16-20 June 2014. + + + + + + diff --git a/file817.txt b/file817.txt new file mode 100644 index 0000000000000000000000000000000000000000..4aa9df166f56aef5f9f31dfba9a8f01c7c0b88e0 --- /dev/null +++ b/file817.txt @@ -0,0 +1,819 @@ + + + + +I. IntroductionIn the National Airspace System (NAS) terminal areas, thousands of flights have to depart, arrive, or taxi within short periods of time in the crowded airspace/surface areas every day.High-density operations impose complexity and inefficiency on terminal airspace operations, and form choke points in the system.The situation gets severe at major airports or metroplexes.Improving the operational efficiency in terminal areas is critical for an efficient air traffic system.In the past decade, research has been conducted from different perspectives to improve the efficiency of operations in terminal areas.2][3][4][5][6][7] This research focused on aircraft sequencing problems and used airborne speed or ground push-back time control.Representative tools that were developed by NASA are: the Traffic Management Advisor (TMA) 8,9 for arrival flight scheduling by adjusting speeds; and the Spot and Runway Departure Advisor (SARDA) 10 for surface operations scheduling by runway sequencing and push-back time control.These scheduling algorithms and tools were developed in a segregated fashion.For instance, TMA treats departure slots on runways as constraints when scheduling arrivals, whereas SARDA takes arrival times as hard constraints when sequencing departures.In these circumstances, runways are the competing resources for both departures and arrivals.In addition to runways, waypoints and route segments may also be shared between arrival and departure flows in terminal areas.Inefficient operations often emerge because of the constraints of shared resources.Recently, integrated schedulers [11][12][13][14] were proposed to apply speed controls and route options to optimize the schedules for both arrivals and departures with competing resources.Studies based on problems in San Jose, 11,12 Los Angles 13 , and New York 14 showed promise for improving operation efficiency in the presence of competing resources.On the other hand, benefits from optimal schedules calculated under deterministic scenarios are usually sensitive to uncertainties/errors of estimated arrival/departure times.A study conducted for a virtual single runway sequence optimization in the presence of uncertainties used a two-stage Mixed-Integer Linear Programming (MILP) formulation and Sample Average Approximation (SAA) by adjusting airspeed and push-back time. 15For scheduling integrated operations using both route and speed controls, a stochastic scheduler based on Non-dominated Sorting Genetic Algorithm (NSGA) and Monte Carlo simulations was proposed to identify optimal and robust schedules. 16,17 he optimization algorithm in the scheduler takes uncertainty into account by calculating costs stochastically over thousands of possible estimated arrival/departure times that follow gaussian distributions.Another stochastic scheduler 18 that combines job-shop scheduling method and SAA was also proposed for solving integrated scheduling problems.The scope of the aforementioned integrated schedulers, however, was small and only limited interaction points between certain departure and arrival flows were studied.A more integrated scheduler that coordinates arrivals, departures, and surface operations is necessary to provide more efficiency and/or even consider user preference by removing barriers between different operations. 19This work develops a centralized stochastic scheduler for operations in a terminal area including airborne and surface operations on the basis of previous works 16,17 using NSGA and Monte Carlo simulations.It advances the sequential and stochastic scheduler developed previously and extends its application to arrivals, departures, and surface operations in the entire terminal area.In addition to the inclusion of a subset of competing waypoints between departures and arrivals, this work includes more competing resources between different flows, such as runway allocations (between departure and arrivals), departure fixes (among departures), and runway crossings (between departure and/or arrivals).The Los Angeles (LAX) terminal area was used as an example and experiments were run with a fourhour traffic scenario in LAX.The scheduler was run sequentially to identify the best robust schedule for updated planning horizons.During the four-hour traffic scenario, schedules were updated periodically for each planning horizon.Final schedule solutions included the routes, speed or delays, and runway assignments.In the experiments, the standard deviation values of the departure time uncertainty were time-varied whereas the uncertainty means (arrival and departure) and arrival standard deviations were constant.In this paper, Section II describes the problem and its model.Section III presents the detailed methods proposed for this study.Results are provided and the analysis is presented in Section IV. Conclusions are provided in Section V. +II. Problem and ModelThe interactions between Fillmore arrivals and Northbound departures in Los Angeles terminal airspace (Figure 1) have been investigated in previous works 13,16,17 on scheduling with competing resources.This work extends the scope to the entire LAX airspace by including all arrivals, departures and surface operations.There are typically more than 1,200 flights taking off and landing in a day at LAX, the research goal of this work is to develop an integrated and robust scheduler to statistically operate airborne and surface traffic efficiently.Figure 1 presents the abstract model of LAX terminal airspace and surface/runway layouts used in this work.This layout models the West-flow configuration in LAX, which is reported to be the most common configuration with about 86% of the usage at LAX. 20 Arrival routes are in blue and departure routes are in green.The main differences from actual routes are the short-cut routes added for Fillmore arrivals and northbound departures to improve efficiency.Other than this, the model presented in Fig. 1 should be close to the actual West-flow configuration in LAX.There are eight entry points for arrival flights represented by incoming arrows to the green points.There are also four exit points for departures represented by outgoing arrows.On the surface, four runways 24L, 24R, 25L, and 25R are open for both arrivals and departures.The inner two runways 24L and 25R allow crossings from and to the outer runways through middle and end points.In this model, airborne flights are required to maintain minimum separations at blue, green, and red waypoints/fixes/nodes.Blue and green points are merging or diverging waypoints for traffic flows with the same direction, and red points are crossing waypoints between arrivals and departures.Grey points are either separated by altitudes or modeled as conflict-free points due to the immediate downstream constraints.On the ground, separation is required at departure and arrival nodes and crossing points on runways. +III. MethodIn this section, a method of developing a centralized scheduler for operations in a terminal area including airborne and surface operations is discussed. +A. ObjectivesThe schedule optimization has multiple objectives with respect to total delay and pseudo controller intervention count.For each flight, the delay is defined as the extra flight transit time between entrance and exit points compared with shortest unimpeded transit time.The shortest unimpeded times assume flights take shortest routes at highest feasible speeds as if there are no other aircraft.Pseudo controller intervention count is the number of actions that need to be taken to avoid loss of separations due to prediction errors of flight arrival or departure times.When a flight needs to be delayed due to a separation violation at any merging and/or diverging waypoint including runways, a "controller action" is required to resolve the loss of separation, thus the pseudo controller intervention count will be increased by one.   J 1 = E[ i t i ] -T 0 J 2 = E[ i N i ] (1)The overall objectives are then to minimize expected values of total delays and controller intervention counts, respectively.As shown in Eqn. 1, J 1 is the expected value of total delay over the random sampling space caused by uncertainties or prediction errors.The total delay can be expressed as the difference between the sum of flight transit times t i for all flights and the sum of these flights' shortest unimpeded transit times T 0 .As the sum of unimpeded transit times is a constant, its expected value is represented by T 0 in the equation.J 2 is the expected value of the sum of controller intervention counts for all flights.Both costs are evaluated over thousands sampling points using Monte Carlo simulations.Statistical measurements other than expected values can also be applied in the future if necessary. +B. Decision variablesThere are eight arrival flows with eight respective entry points as shown in the Fig. 1.There are three departure flows corresponding to three directions: North, West, and South/East.Four types of decision variables are defined for every flight: delay, speed, route and runway.Delay may be applied before entry points.For a departure, the entry point is the gate on the surface, and for an arrival, it refers to the arrival meter fix in the airspace.For each flight segment, the speed is discrete and its feasible range is predefined based on preprocessed feasible speed calculated by the Trajectory Synthesizer (TS) tool used in the Center TRACON Automation System (CTAS). 21Route options are designed for arrivals from Fillmore, and Northbound departures where the departures and arrivals can take short cut if conflicts don't happen at shared waypoints between these two flows.Four runways 24L, 24R, 25L, and 25R are available options to all flights, although factors like taxiing distance between runway and gate would affect such assignment. +C. Freeze horizonThe freeze horizon for arrivals is set at arrival meter fixes, which means that decision variables are frozen after arrival flights proceed beyond their entry points.For departures, the freeze horizon is set at the departing gates on the surface.A departure flight is kept at the gate until its decision variables including runway assignment, takeoff slot, and departing route are all decided.Once a departure flight leaves its gate, these assignments are then frozen.The freeze horizons are set up to avoid over-delay due to multiple rescheduling events.For instance, when a departure flight A leaves its gate, its schedule is frozen.If later on, another flight B is ready for scheduling and conflicts with A, then flight B has to yield to flight A in the updated schedule.Otherwise, if flight B overtakes flight A, then flight A can get further delayed.This situation may happen again if other flights cut in.Prior experiments showed that flight A can eventually get significant delay if left unfrozen. +D. ConstraintsAs the aircraft speed options are preprocessed, infeasible speed options are ruled out.Separation requirements are the only constraints in the work.They were applied at fixes, waypoints, and runways, where flights merged and/or diverged.According to FAA's regulations based on wake vortex, 22 minimum separation requirements for airborne flights were listed in Table 1.The separation for departure and crossing flights at the runway is given in Table 2 In this optimization, the separation constraints are not treated as "hard constraints" or inequalities as in usual linear or nonlinear optimization formulations.Because the same amount of separation violation results in different delays in air traffic system, in this model, delay costs were calculated for resolving separation violations.Given decision variables, a method based on constraint algebra 24 was used to resolve aircraft conflicts or constraint violations.With final resolutions from this method, arrival times at exit points were then used to calculated delay costs.In the constraint algebra method, the flight priority is set up according to runway unimpeded estimated times of arrival (ETAs) at the runway for both departures and arrivals.The basic idea of this method is to insert one flight at a time starting from the first flight on the priority list.The insertion should be guaranteed conflict-free at all points.The speed change is not considered in the constraint algebra method, because it is already included in decision variables.If there is any conflict between the inserting flight and existing flights, the new flight is given extra delay at its starting point.For a departure flight, the delay is imposed at its gate as extra gate waiting time.Whereas, for an arrival flight, the delay is propagated back upstream, beyond the arrival meter fix. Figure 2 presents an example of inserting a flight using constraint algebra.Red bars represent the slots occupied by other aircraft.Given a speed profile, the relative time differences between scheduling points are fixed for this flight.The constraint algebra method calculates open slots at each scheduling point, and then finds the feasible starting time slots with the "rigid" speed profile.In this figure, the feasible ranges for inserting the given speed profile are from t 1 to t 2 and any time after t 3 .Therefore, the extra delay cost of inserting this flight is t 1 -t 0 , which would be included in the final delay cost.As the speed is treated as a decision variable, the scheduler will calculate the delay costs for a range of speed profile options in this way during optimization. +E. Uncertainty modelsIn TMA and SARDA, schedules are updated frequently (every 10 seconds) to mitigate the impact of uncertainties/prediction errors.When routes and/or runway assignments need to be decided in schedules for both arrivals and departures, they have to be decided with a certain look-ahead time of at least several minutes.For example, when a departure flight is about to taxi from its gate, pilots need to know the designated runway.To determine the runway, uncertainties in other flights' arrival or departure times have to be taken into account because it would not be feasible to change its runway assignment in next update cycle even if prediction errors are found later.In this circumstance, increasing update frequency is not effective for mitigating the impact of uncertainties.Taking uncertainties directly into account at planning stages is helpful.Arrivals uncertainties must also be taken into account prior to a freeze horizon as runway/route is decided ahead of time.To simulate the real world, this study assumes that "planned times" to entry points are uncertain and follow certain distributions.For a departure flight, the "planned time" is the planned push-back time from its gate.And for an arrival flight, it refers to the estimated arrival time to the arrival meter fix.The ETA of an arrival to entry points is assumed to follow a normal distribution with zero means and 30 second standard deviation regardless of the planning look-ahead times based on trajectory prediction studies. 25Whereas for a departure, it is defined that the estimated time of departure (ETD) from a gate, or the push-back time, follows a normal distribution with a standard deviation linearly increased with the look-ahead time T L where the mean is still constant (see Eqn. 2).This definition and coefficients are set up based on study results from the Surface Decision Support System (SDSS). 26,27 he definition of the look-ahead time T L is described in Fig. 3, in which the ET A 1 , ET A 2 , and ET D 1 are ETAs and ETDs for sample flights with in a time window span of T w .The T L of each flight is the time differences between its ETA/ETD and the planning time t p .According to the equation, when the planning look-ahead time increases, the uncertainty of departure time increases.Therefore, ET D 2 is more uncertain than ET D 1 . +F. Scheduling SchemeTo generate a schedule for a traffic scenario with a large time window, the most practical and efficient approach is to divide it into smaller windows.The relation between planning horizon and actual time is shown in Fig. 4. A sequential dynamic scheduling scheme is adopted, where the schedule for each window with time span T w is identified before the time proceeds to that window at t 1 .If the computational time of the scheduling algorithm T c is considered, the planning should be started at or before t 1 -T c to make sure an updated schedule is available at actual time t 1 .This process will be repeated periodically to provide schedules for all time windows.An overlapping scheduling scheme is similar except that planning horizons overlap by a certain duration. +G. Algorithms and implementationThe NSGA, a variant of Genetic Algorithms, is applied to solve this multi-objective optimization problem due to its promising capability of handling multiple objectives and nonlinear optimizations.The stochastic scheduler combines NSGA and Monte Carlo simulation. 13,16,17,28 Th decision variables including speeds, routes, delays, and runway assignment are coded as "genes", and each solution with a set of decision variables is marked as an "individual".In NSGA, a population with hundreds of "individuals" evolves at each generation in terms of their costs through operations of "crossover", "mutation", "ranking", and "selection".When evaluating costs for each solution or "individual", Monte Carlo simulations are used to identify the statistical measurements.Given a solution with a set of decision variables, its costs are evaluated thousands of times using the constraint algebra in Monte Carlo simulations.In each simulation, the computed costs correspond to one uncertainty sample, which was imposed on the "planned" push-back or arrival times.On top of the optimization core code, scripts that involve the freeze horizon and planning horizon set-ups are then applied to identify outputs for current planning horizon and inputs for next one.This approach is implemented using CUDA programming with GPUs to reduce the significant computational time to a reasonable level.In a Linux platform with 18x2.5 GHz Xeon, 32 GB memory, and two GeForce GTX690 GPUs, a 15-flight scenario scheduling problem takes around 30 seconds to be solved . 28,29 +IV. ResultsA four-hour traffic scenario was built based on historical traffic on Jul. 1, 2014.A total of 315 flights were included composed of 172 departures and 143 arrivals for LAX.In this scenario, information of each flight includes aircraft type, planned time at entry points, gate assignment, and flight routes including arrival/departure meter fixes.Various experiments were then conducted to examine the impacts of difference factors and to analyze the potential benefits introduced by those factors. +A. Planning horizonIn deterministic cases, or stochastic cases with constant means and constant standard deviations, the delay of the best solution decreases when the window size/planning horizon increases.Longer planning horizon helps find solutions closer to global optimum because more information is involved.This may not always be true when means and standard deviations become time-dependent, or in other words, when uncertainties increase with the look-ahead time.Since uncertainties grow in long planning horizons due to long look-ahead times, the increased uncertainties diminish the benefits introduced by expended knowledge because the predicted information becomes useless or even harmful when the look-ahead time is long.Figure 5 shows the minimum delays that can be achieved by solutions generated under different planning The optimizer tries to minimize two costs in the model.Figure 6 shows cost pareto fronts produced by the optimization associated with different planning windows.Each pareto front is composed of a set of points with each point referring to a schedule solution.The coordinates of a point represent two statistical costs associated with the solution as mentioned in previous sections: the first cost is the total delay including airborne and ground delays; And the second cost is the pseudo controller workload (intervention count).The notion of "W" in the figure denotes planning window.For instance, "W20" denotes solutions generated with a 20-minute planning window.Trends observed in Fig. 6 are similar to Fig. 5.When the planning horizon decreases from 20 minutes to eight minutes, small windows outperform large windows along the entire pareto front.Trade-offs rise when the window size decreases from eight minutes to two minutes.Figure 6 shows that with an eight-minute planning window, the scheduler produces solutions with the minimum delay.The duration of planning horizon that produces minimum delay can be affected by several factors: freeze horizon set-up, taxiing time from gate to runway, and relation between uncertainties and time.Here is an example used to compare solutions from eight-minute planning horizon and two-minute planning horizon: A departure Flight A with type A321 is at Gate 42B and according to its plan, it will leave the gate at 08:11:00.Flight B is B737 located at Gate 7. Its planned push-back time is 08:13:30.The taxiing distances from Gate 42B to runway 24L, 24R, 25R, and 25L are 1.3 nmi, 1.42 nmi, 1.32 nmi, and 1.44 nmi, respectively.The corresponding distances from Gate 7 to runways are 0.1 nmi, 0.22 nmi, 2.6 nmi, and 2.72 nmi, respectively.Current scheduling time is at 08:10:00.For a two minute planning horizon, at 08:10:00, the scheduler did not know any information after 08:12:00 including Flight B. Therefore, the scheduler allowed Flight A to leave the runway when it was ready, and assigned runway 24L to the flight, which takes Flight A 312 seconds to taxi from the gate to runway.Then Flight A entered the frozen horizon right after pushing back from gate 42B.When the time proceeded to next planning horizon at 08:12:00, the scheduler finds Flight B, which would be ready in 90 seconds.Its taxiing time to 24L is 24 seconds.Because Flight A was in frozen horizon already, the only option available to Flight B was to wait and follow Flight A, which cost Flight B 169 seconds in waiting time.Whereas, for an eight minute planning horizon, Flight B was within scope when the scheduler was calculating at 08:10:00.Although there was 210 seconds look-ahead time with 86 second standard deviation for Flight B, the scheduler decided to put Flight B in front of Flight A based on statistical cost measurements, which would reduce delay about 160 seconds.Apparently, under current uncertainty assumption, the eight-minute planning horizon is the best candidate if a solution with minimum delay is desired.When the requirement is relaxed by allowing a bit high delay, the scheduler with a two-minute planning horizon can also be used as it can generate solutions with a bit high delay but much lower intervention count.This implies that, in application, any planning horizon from two minutes to eight minutes could be a good candidate.One extreme example is the one-minute planning window, the delay level of the pareto front becomes much greater than other short planning horizons.The hypothesis is that, although time-dependent uncertainties favor short planning horizon, one minute is too short for the given freeze horizon and taxiing time.The lack of flight information in the next planning horizon leads the optimizer to generate solutions with high delay levels. +B. Stochastic vs. deterministicTo examine the difference between deterministic and stochastic schedulers, experiments using a deterministic scheduler were conducted.To mimic the time-dependent uncertainties in deterministic experiments, timedependent errors/noises were imposed to the "planned time" of each flight that is involved in the planning horizon.The deterministic optimization is then conducted on the basis of these noisy "planned times" to generate schedules.A variety of planning strategies were tested by varying planning horizons and update cycles.For example, the black square in Fig. 7 denotes the solution produced by the deterministic scheduler with a 15-minute planning horizon and one minute update cycle (marked as "W15U1").This means the optimizer calculates or updates schedule every minute for the following 15-minute planning horizon.For the sake of comparison, the pareto front generated under the eight-minute planning window using the stochastic scheduler (shown in Fig. 6) is presented as a black curve in Fig. 7.This figure shows clearly that the stochastic scheduler outperforms the deterministic scheduler based on the statistical measurements of both delays and pseudo intervention counts.This conclusion holds for deterministic schedulers with all strategies: overlapped or non-overlapped planning windows, and long or short planning horizons.With the same level of intervention counts, the stochastic scheduler reduces delays anywhere between 28% and 40% when comparing against the deterministic scheduler.The imperfect knowledge of "planned time" and neglect of uncertainty in optimization appear to be the two main factors that contribute to the differences.In deterministic optimization, every piece of information is assumed to be perfect and the evolution at every step in the optimization is built on that assumption, apparently, inaccurate "planned times" can not lead the optimizer to the optimal solution.Whereas, in stochastic optimization, the scheduler assumes that "planned time" is imperfect.By following a certain distribution, the scheduler finds the optimal solution based on statistical measurements.A typical way to deal with uncertainty in deterministic optimization is to impose extra buffers besides basic separation requirements, the green circle in Fig. 7 shows the final solution from a deterministic optimization using a 15-minute planning horizon and one-minute update cycle with a 15-second buffer.The 15-second buffer does reduce the intervention count a lot, however, it also increases the total delay.This implies that stochastic scheduling is an effective way to reduce delay and intervention count statistically, since it takes uncertainty knowledge into account during the optimization.To examine the impact of uncertainty magnitude on final schedules, the linear coefficient 0.41 in Eqn. 2 is reduced to 0.2. Figure 8 presents the results with this lower uncertainty.The labels containing "low" refer to results with low uncertainties.As expected, with more accurate "planned times" both stochastic and deterministic schedulers generated solutions with lower delays.Results from deterministic schedulers improved more than stochastic schedulers, which indicated that deterministic schedulers were more sensitive to uncertainties.In stochastic cases, the big planning horizon benefit more from the accuracy improvement than the small planning horizon.This experiment denotes that the advantage of stochastic schedulers over deterministic schedulers is closely correlated to the uncertainty magnitude. +C. Uncertainty Magnitude +D. Runway usageRunway makespan and occupancy metrics were used to examine runway usage.Runway makespan refers to the time span between first and last flight for a set of operations.When flights are not operated back-toback all of the time, the correlation between the makespans and the effectiveness of schedules may not be tight.However, it is nevertheless a good metric to show the effectiveness of schedules.Figure 9 presents the statistical runway makespans corresponding to three difference schedulers: a stochastic scheduler with an eight-minute planning horizon, a deterministic scheduler with a 15-minute planning horizon with oneminute update cycle, and a deterministic scheduler with a three-minute planning horizon.It shows that the stochastic scheduler produced minimum makespans for all four runways.Compared to the deterministic scheduler with a 15-minute window, the stochastic scheduler finished all flight operations six minutes earlier on runway 24L and 10 minutes earlier on runway 25R.Runway occupancy can serve as another metric to examine the characteristics of runway usage.It is defined as the total percentage of the usage of a runway.When computing the metric, it is assumed that each departure flight uses 40 seconds to takeoff from a runway and each arrival flight takes 50 seconds to land on a runway.Crossing a runway is assumed to take 10 seconds for a flight.Figure 10 presents the statistical runway occupancies corresponding to the same three difference schedulers presented in Fig. 9. Figure 10 shows that all three schedulers utilized 25L/R runways more than 24L/R runways.This could be caused by the nature of the scenario.Flights were typically assigned to the nearest runway in most situations and there are more flights close to 25L/R runways in the experimental scenario.Because all flights have to utilize runway to arrive or to depart, the overall number of operations should be similar with some difference existing in the number of crossings only.It is noted that outer runways were used a bit more in stochastic cases, which implies that there were more runway crossings in stochastic cases than in deterministic cases.The fact that the occupied time of both inner runways in the stochastic case is less than deterministic cases indicates that more departures were assigned to inner runways in the stochastic case.The overall occupancies of the three schedulers are similar, however, the stochastic scheduler occupancies are slightly lower, which indicates that schedules from stochastic optimization utilize runways slightly more efficiently. +V. ConclusionsAn integrated scheduler that coordinates arrivals, departures, and surface operations provides efficiency in terminal areas by removing barriers between different operations.This work developed a centralized stochastic scheduler for operations in a terminal area including airborne and surface operations based on NSGA and Monte Carlo simulations methods.It extended the formulation of the sequential, stochastic, and integrated scheduler to arrivals, departures, and surface operations in an entire terminal environment.In addition to the competing waypoints, this extension included most competing resources in terminal areas between different flows, such as runway allocations, departure fixes, and runway crossings.The Los Angeles terminal area was used as an example and experiments were run with a four-hour traffic scenario in LAX.The scheduler was run sequentially to identify the best robust schedule for the next planning horizon.In the experiments, the standard deviation values of the departure time uncertainty were time-varied whereas the departure and arrival uncertainty means and arrival standard deviations were constant.Final schedules included decisions on routes, speeds or delays, and runway assignments.The algebra constraint method was used to calculate extra costs to satisfy separation requirements, which were eventually amended to the overall costs.Studies on planning horizons showed that trade-offs exist between planning horizons and achievable minimum delays.A 20-minute planning horizon was not a good choice because uncertainties grew with the look-ahead time.Eight minutes was promising for planning as it achieved the lowest delay compared to others.However, the results demonstrated that any duration from two minutes to eight minutes could be a good candidate as well.Experimental results showed that using stochastic schedulers reduced the flight time delay (airborne and ground) 28% to 40% statistically compared to deterministic schedulers with the same level of intervention counts.Experiments on uncertainty magnitude demonstrated the close correlation between uncertainty and the benefit from stochastic schedulers.It was also shown that deterministic schedulers are more sensitive to uncertainty than stochastic schedulers.The results on runway usage showed that using the stochastic scheduler, runway makespans and occupancy were usually slightly lower when compared with deterministic schedulers.Overall, experiments showed that this sequential stochastic scheduler was capable of scheduling arrivals, departures, and surface operations in an integrated fashion.The stochastic scheduler successfully took uncertainty into account and statistical results showed significant delay savings were achieved when the knowledge of uncertainties was involved.Figure 1 .1Figure 1.Los Angles Layout (not to scale) +Figure 2 .2Figure 2. Conflict-free scheduling using Constraint Algebra +Figure 3 .3Figure 3. Definition of the look-ahead time +Figure 4 .4Figure 4. Scheme with nonoverlapping scheduling windows +Figure 5 .5Figure 5. Minimum delays corresponding to various planning windows +Figure 6 .6Figure 6.Pareto fronts for different planning windows +Figure 7 .7Figure 7.Comparison between stochastic and deterministic schedulers +Figure 8 .8Figure 8. Impact of Uncertainty Magnitude +Figure 9 .9Figure 9. Runway Makespans +Figure 10 .10Figure 10.Runway Occupancy +Table 1 .123 converting original distance separation into time-based separation.23MinimumAirborne Aircraft SeparationsSeparation Distance Leading Aircraft: Wake Category(nmi)Heavy B757 Medium SmallTrailingHeavy4433AircraftB7574433WakeMedium5433CategorySmall6533 +Table 2 .2Minimum Runway Separations for Departure and Crossing FlightsSeparation TimeLeading Aircraft Wake Category(sec)Small Large Heavy B757 CrossingTrailingSmall598810911025AircraftLarge59611099125WakeHeavy5961909125CategoryB75759611099125Crossing4040404040 + of 13 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on July 2, 2015 | http://arc.aiaa.org| DOI: 10.2514/6.2015-2273 + of 13 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on July 2, 2015 | http://arc.aiaa.org| DOI: 10.2514/6.2015-2273 + of 13 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on July 2, 2015 | http://arc.aiaa.org| DOI: 10.2514/6.2015-2273 + of 13 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on July 2, 2015 | http://arc.aiaa.org| DOI: 10.2514/6.2015-2273 + of 13 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on July 2, 2015 | http://arc.aiaa.org| DOI: 10.2514/6.2015-2273 + of 13 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on July 2, 2015 | http://arc.aiaa.org| DOI: 10.2514/6.2015-2273 + Downloaded by NASA AMES RESEARCH CENTER on July 2, 2015 | http://arc.aiaa.org| DOI: 10.2514/6.2015-2273 + of 13 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on July 2, 2015 | http://arc.aiaa.org| DOI: 10.2514/6.2015-2273 + + + + +Mail Stop 210-8.AIAA senior member † Aerospace engineer, Aerospace High Density Operations Branch.Mail Stop 210-10.AIAA senior member + + + + + + + + + The dynamic scheduling of aircraft in the near terminal area + + RGDear + + 10.1016/0041-1647(78)90073-4 + R76-9 + + + Transportation Research + Transportation Research + 0041-1647 + + 12 + 4 + + September 1976 + Elsevier BV + + + Tech. Rep. Flight Transportation Laboratory Report + Dear, R. G., "The Dynamic Scheduling of Aircraft in the Near Terminal Area," Tech. Rep. Flight Transportation Laboratory Report R76-9, MIT, September 1976. + + + + + Analysis of Delay Reducing and Fuel Saving Sequencing and Spacing Algorithms for Arrival Spacing + + FNeuman + + + HErzberger + + 103880 + + October 1991 + + + Tech. Rep. NASA Technical Memorandum + Neuman, F. and Erzberger, H., "Analysis of Delay Reducing and Fuel Saving Sequencing and Spacing Algorithms for Arrival Spacing," Tech. Rep. NASA Technical Memorandum 103880, October 1991. + + + + + An algorithm for computer assisted sequencing and scheduling of terminal area operations + + RogerGDear + + + YosefSSherif + + 10.1016/0191-2607(91)90132-a + + + Transportation Research Part A: General + Transportation Research Part A: General + 0191-2607 + + 25 + 2-3 + + 1991 + Elsevier BV + + + Dear, R. G. and Sherif, Y. S., "An Algorithm For Computer Assisted Sequencing and Scheduling of Terminal Operations," Transportation Research A, Vol. 25A, No. 2/3, 1991. + + + + + A Mixed Integer Linear Program for Airport Departure Scheduling + + GautamGupta + + + WaqarMalik + + + YoonJung + + 10.2514/6.2009-6933 + + + 9th AIAA Aviation Technology, Integration, and Operations Conference (ATIO) + Hilton Head, South Carolina + + American Institute of Aeronautics and Astronautics + September 2009 + + + + Gupta, G., Malik, W., and Jung, Y. C., "A Mixed Integer Linear Program for Airport Departure Scheduling," 9th AIAA Aviation Technology, Integration and Operations Conference (ATIO), Hilton Head, South Carolina, 21-23 September 2009. + + + + + A Mixed Integer Linear Program for Solving a Multiple Route Taxi Scheduling Problem + + JustinMontoya + + + ZacharyWood + + + SivakumarRathinam + + + WaqarMalik + + 10.2514/6.2010-7692 + + + AIAA Guidance, Navigation, and Control Conference + Toronto, Canada + + American Institute of Aeronautics and Astronautics + August 2010 + + + + Montoya, J., Wood, Z., Rathinam, S., and Malik, W., "A Mixed Integer Linear Program for Solving a Multiple Route Taxi Scheduling Problem," AIAA Guidance, Navigation, and Control Conference, Toronto, Canada, 2-5 August 2010. + + + + + A Generalized Dynamic Programming Approach for a Departure Scheduling Problem + + SivakumarRathinam + + + ZacharyWood + + + BanavarSridhar + + + YoonJung + + 10.2514/6.2009-6250 + + + AIAA Guidance, Navigation, and Control Conference + Chicago, IL + + American Institute of Aeronautics and Astronautics + August 2009 + + + + Rathinam, S., Wood, Z., Sridhar, B., and C., J. Y., "A Generalized Dynamic Programming Approach for a Departure Scheduling Problem," AIAA Guidance, Navigation, and Control Conference, Chicago, IL, 10-13 August 2009. + + + + + Scheduling Aircraft Landings Under Constrained Position Shifting + + HamsaBalakrishnan + + + BalaChandran + + 10.2514/6.2006-6320 + + + AIAA Guidance, Navigation, and Control Conference and Exhibit + Keystone, CO + + American Institute of Aeronautics and Astronautics + August 2006 + + + + Balakrishnan, H. and Chandran, B., "Scheduling Aircraft Landings Under Constrained Position Shifting," AIAA Guid- ance, Navigation, and Control Conference, Keystone, CO, 21-24 August 2006. + + + + + Design and Operational Evaluation of the Traffic Management Advisor at the Fort Worth Air Route Traffic Control Center + + HSwenson + + + THoang + + + SEnglland + + + DVincent + + + TSanders + + + BSanford + + + KHeere + + + + Europe Air Traffic Management Research and Development Seminar + + + June 1997 + Saclay, France + + + 1st USA/ + Swenson, H., Hoang, T., Englland, S., Vincent, D., Sanders, T., Sanford, B., and Heere, K., "Design and Operational Evaluation of the Traffic Management Advisor at the Fort Worth Air Route Traffic Control Center," 1st USA/Europe Air Traffic Management Research and Development Seminar , Saclay, France, 17-19 June 1997. + + + + + Design and Evaluation of the Terminal Area Precision Scheduling and Spacing System + + HSwenson + + + JThipphavong + + + ASadovsky + + + LChen + + + 2011 + Berlin, Germany + + + Ninth USA/Europe Air Traffic Management Research and Development Seminar + + + Swenson, H., Thipphavong, J., Sadovsky, A., and Chen, L., "Design and Evaluation of the Terminal Area Precision Scheduling and Spacing System," Ninth USA/Europe Air Traffic Management Research and Development Seminar , Berlin, Germany, 2011. + + + + + A Concept and Implementation of Optimized Operations of Airport Surface Traffic + + YoonJung + + + TyHoang + + + JustinMontoya + + + GautamGupta + + + WaqarMalik + + + LeonardTobias + + 10.2514/6.2010-9213 + + + 10th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Fort Worth, TX + + American Institute of Aeronautics and Astronautics + September 13-15 2010 + + + Jung, Y., Huang, T., Montoya, J., Gupta, G., Malik, W., and Tobias, L., "A Concept and Implementation of Optimized Operations of Airport Surface Traffic," 10th AIAA Aviation Technology, Integration and Operations Conference (ATIO), Fort Worth, TX, September 13-15 2010. + + + + + Towards Optimal Routing and Scheduling of Metroplex Operations + + BrianCapozzi + + + StephenAtkins + + + SeongimChoi + + 10.2514/6.2009-7037 + + + 9th AIAA Aviation Technology, Integration, and Operations Conference (ATIO) + Hilton Head, South Carolina + + American Institute of Aeronautics and Astronautics + September 2009 + + + + Capozzi, B. J. and Atkins, S. C., "Towards Optimal Routing and Scheduling of Metroplex Operations," 9th AIAA Aviation Technology, Integration, and Operations Conference(ATIO), Hilton Head, South Carolina, 21-23 September 2009. + + + + + A Hybrid Optimization Approach to Air Traffic Management for Metroplex Operations + + BrianCapozzi + + + StephenAtkins + + 10.2514/6.2010-9062 + + + 10th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + + American Institute of Aeronautics and Astronautics + + + + Capozzi, B. J. and Atkins, S. C., "A Hybrid Optimization Approach to Air Traffic Management for Metroplex Operations," + + + + + Departure Efficiency Benefits of Terminal RNAV Operations at Dallas-Fort Worth International Airport + + RalfMayer + + 10.2514/6.2006-7774 + + + 6th AIAA Aviation Technology, Integration and Operations Conference (ATIO) + Fort Worth, Texas + + American Institute of Aeronautics and Astronautics + September 2010 + + + + AIAA Aviation Technology, Integration and Operations Conference (ATIO), Fort Worth, Texas, 13-15 September 2010. + + + + + Optimal Integration of Departures and Arrivals in Terminal Airspace + + MinXue + + + ShannonZelinski + + 10.2514/6.2012-4977 + + + AIAA Guidance, Navigation, and Control Conference + + American Institute of Aeronautics and Astronautics + 2014 + 37 + + + Xue, M. and Zelinski, S., "Optimal Integration of Departure and Arrivals in Terminal Airspace," AIAA Journal of Guidance, Control, and Dynamics, Vol. 37, No. 1, 2014. + + + + + Sequential Dynamic Strategies for Real-Time Scheduling of Terminal Traffic + + HemingChen + + + YiyuanJZhao + + 10.2514/1.c031503 + + + Journal of Aircraft + Journal of Aircraft + 0021-8669 + 1533-3868 + + 49 + 1 + + 2012 + American Institute of Aeronautics and Astronautics (AIAA) + + + Chen, H. and Zhao, Y. J., "Sequential Dynamic Strategies for Real-Time Scheduling of Terminal Traffic," AIAA Journal of Aircraft, Vol. 49, No. 1, 2012. + + + + + Runway Operations Optimization in the Presence of Uncertainties + + GustafSolveling + + + SenaySolak + + + John-PaulClarke + + + EllisJohnson + + 10.2514/1.52481 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 34 + 5 + + 2010 + American Institute of Aeronautics and Astronautics (AIAA) + + + Solveling, G., Solak, S., Clarke, J. P., and Johnson, E., "Runway Operations Optimization in the Presence of Uncertain- ties," AIAA Journal of Guidance, Control, and Dynamics, Vol. 34, No. 5, 2010. + + + + + Optimization of Integrated Departures and Arrivals Under Uncertainty + + MinXue + + + ShannonZelinski + + 10.2514/6.2013-4322 + + + 2013 Aviation Technology, Integration, and Operations Conference + Los Angeles, LA + + American Institute of Aeronautics and Astronautics + August 2013 + + + + Xue, M. and Zelinski, S., "Optimization of Integrated Departures and Arrivals Under Uncertainty," 13th AIAA Aviation Technology, Integration, and Operations Conference(ATIO), Los Angeles, LA, 12-14 August 2013. + + + + + Dynamic stochastic scheduler for integrated arrivals and departures + + MinXue + + + ShannonZelinski + + 10.1109/dasc.2014.6979399 + + + 2014 IEEE/AIAA 33rd Digital Avionics Systems Conference (DASC) + Colorado Springs, CO + + IEEE + October 2014 + + + + Xue, M. and Zelinski, S., "Dynamic Stochastic Scheduler For Integrated Arrivals and Departures," 33rd Digital Avionics Systems Conference (DASC), Colorado Springs, CO, 5-9 October 2014. + + + + + Optimizing integrated terminal airspace operations under uncertainty + + ChristabelleBosson + + + MinXue + + + ShannonZelinski + + 10.1109/dasc.2014.6979545 + + + 2014 IEEE/AIAA 33rd Digital Avionics Systems Conference (DASC) + Colorado Springs, CO + + IEEE + October 2014 + + + + Bosson, C., Xue, M., and Zelinski, S., "Optimizing Integrated Terminal Airspace Operations Under Uncertainty," 33rd Digital Avionics Systems Conference (DASC), Colorado Springs, CO, 5-9 October 2014. + + + + + A framework for integrating arrival, departure, and surface operations scheduling + + ShannonZelinski + + 10.1109/dasc.2014.6979543 + + + 2014 IEEE/AIAA 33rd Digital Avionics Systems Conference (DASC) + Colorado Springs, CO + + IEEE + October 2014 + + + + Zelinski, S., "A Framework for Integrating Arrival, Departure, and Surface Operations Scheduling," 33rd Digital Avionics Systems Conference (DASC), Colorado Springs, CO, 5-9 October 2014. + + + + + Super Density Operations Airspace Modeling for the Southern California Metroplex + + SebastianTimar + + + GauravNagle + + + AdityaSaraf + + + PeterYu + + + PeterHunt + + + AndrewTrapani + + + NickJohnson + + 10.2514/6.2011-6535 + + + AIAA Modeling and Simulation Technologies Conference + Portland, Oregon + + American Institute of Aeronautics and Astronautics + August 2011 + + + + Timar, S. D., Nagle, G., Saraf, A., Yu, P., Hunter, P., Trapani, A., and Johnson, N., "Super Density Operations Airspace Modeling for the Southern California Metroplex," AIAA Modeling and Simulation Technology Conference, Portland, Oregon, 08-11 August 2011. + + + + + Design of Center-TRACON Automation System + + HErzberger + + + TJDavis + + + SMGreen + + + + 22 Aeronautical Information Manual (AIM): Official guide to basic flight information and ATC procedure, Federal Aviation Administration (FAA) + Berlin, Germany + + May 11-14 1993. 2014 + + + ARARD Meeting on Machine Intelligence in Air Traffic Management + Erzberger, H., Davis, T. J., and Green, S. M., "Design of Center-TRACON Automation System," ARARD Meeting on Machine Intelligence in Air Traffic Management, Berlin, Germany, May 11-14 1993. 22 Aeronautical Information Manual (AIM): Official guide to basic flight information and ATC procedure, Federal Aviation Administration (FAA), 2014. + + + + + Incorporating Active Runway Crossings in Airport Departure Scheduling + + GautamGupta + + + WaqarMalik + + + YoonJung + + 10.2514/6.2010-7695 + + + AIAA Guidance, Navigation, and Control Conference + Toronto, Canada + + American Institute of Aeronautics and Astronautics + August 2010 + + + + Gupta, G., Malik, W., and Jung, Y. C., "Incorporating Active Runway Crossings in Airport Departure Scheduling," AIAA Guidance, Navigation, and Control Conference, Toronto, Canada, 2-5 August 2010. + + + + + A Closed-Form Solution to Multi-Point Scheduling Problems + + LarryMeyn + + 10.2514/6.2010-7911 + + + AIAA Modeling and Simulation Technologies Conference + Toronto, Ontario, Canada + + American Institute of Aeronautics and Astronautics + 2-5 August 2010 + + + Meyn, L., "A Closed-Form Solution to Multi-Point Scheduling Problems," AIAA Modeling and Simulation Technologies Conference, Toronto, Ontario, Canada, 2-5 August 2010. + + + + + Predictability of Top of Descent Location for Operational Idle-Thrust Descents + + LaurelStell + + 10.2514/6.2010-9116 + + + 10th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Berlin, Germany + + American Institute of Aeronautics and Astronautics + June 2011 + + + Stell, L., "Prediction of Top of Descent Location for Idle-thrust Descents," The Ninth USA/Europe Air Traffic Manage- ment Research and Development Seminar , Berlin, Germany, June 2011. + + + + + Characterization of Tactical Departure Scheduling in the National Airspace System + + RichardCapps + + + ShawnEngelland + + 10.2514/6.2011-6835 + + + 11th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + + American Institute of Aeronautics and Astronautics + + + + 11th + Capps, A. and Engelland, S., "Characterization of Tactical Departure Scheduling in the National Airspace System," 11th + + + + + 11th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + 10.2514/matio11 + + + Integration and Operations Conference (ATIO) + Virginia Beach, VA + + American Institute of Aeronautics and Astronautics + September 20-22 2011 + + + AIAA Aviation Technology, Integration and Operations Conference (ATIO), Virginia Beach, VA, September 20-22 2011. + + + + + Surface Management System Field Trial Results + + StephenAtkins + + + YoonJung + + + ChristopherBrinton + + + LStell + + + TCarniol + + + SRogowski + + 10.2514/6.2004-6241 + + + AIAA 4th Aviation Technology, Integration and Operations (ATIO) Forum + Chicago, IL + + American Institute of Aeronautics and Astronautics + September 20-22 2004 + + + Atkins, S., Jung, Y., Brinton, C., Stell, L., and Rogowski, S., "Surface Management System Field Trial Results," 4th AIAA Aviation Technology, Integration and Operations Conference (ATIO), Chicago, IL, September 20-22 2004. + + + + + Integrated Arrival- and Departure-Schedule Optimization Under Uncertainty + + MinXue + + + ShannonZelinski + + 10.2514/1.c032957 + + + Journal of Aircraft + Journal of Aircraft + 0021-8669 + 1533-3868 + + 52 + 5 + + 2015 + American Institute of Aeronautics and Astronautics (AIAA) + + + accepted + Xue, M. and Zelinski, S., "Integrated Arrival and Departure Schedule Optimization Under Uncertainty," AIAA Journal of Aircraft, 2015 (accepted). + + + + + GPU-based Parallelization for Schedule Optimization with Uncertainty + + ChristabelleBosson + + + MinXue + + + ShannonZelinski + + 10.2514/6.2014-2024 + + + 14th AIAA Aviation Technology, Integration, and Operations Conference + Atlanta, GA + + American Institute of Aeronautics and Astronautics + June 16-20 2014 + + + Bosson, C., Xue, M., and Zelinski, S., "GPU-based Parallelization for Schedule Optimization with Uncertainty," AIAA Aviation and Aeronautics Forum and Exposition, Atlanta, GA, June 16-20 2014. + + + + + + diff --git a/file818.txt b/file818.txt new file mode 100644 index 0000000000000000000000000000000000000000..3f835d313a504b2fb0a78938be8770d12f71fa20 --- /dev/null +++ b/file818.txt @@ -0,0 +1,662 @@ + + + + +IntroductionIn the National Airspace System (NAS) terminal airspace, hundreds of flights must depart or arrive within a short time period every day.Limited airspace, complicated operations, and diverse uncertainties can easily impose inefficiency on terminal airspace operations, and form choke points in the system.Improving the operation efficiency in constrained terminal airspace in the presence of uncertainty becomes critical for achieving an efficient air traffic system.In order to improve the efficiency, researchers have addressed different perspectives of terminal airspace research, including arrival scheduling, departure scheduling, surface scheduling, and corresponding uncertainty analyses [1][2][3][4][5][6][7][8][9].Additionally, when arrivals and departures share the same resources such as runways, waypoints and route segments, recent studies [10][11][12][13] found that optimized integrated arrivals and/or departures in major airports showed promise for improving operation efficiency.Besides the constrained airspace, uncertainty sources, such as flight time and weather, play an important role in scheduling problems in terminal airspace.The benefits from optimal schedules calculated under deterministic scenarios are usually sensitive to these uncertainties.A conventional way to deal with uncertainty is to use extra buffers in addition to the required aircraft separations.Previous work [14] proposed a proactive method -a stochastic scheduler -that directly takes uncertainty into account by optimizing integrated arrivals and departures under uncertainty.Compared with deterministic optimization using extra buffers, the stochastic scheduler can find better solutions under the guidance of uncertainty costs.This work extends the previous stochastic scheduler by integrating dynamic capability such that the scheduler can be applied to real-time traffic sequentially and seamlessly.With this dynamic stochastic scheduler, the investigations of integrated arrivals and departures are enabled for traffic with much longer time windows, while the capability of optimization under uncertainty is still intact.The Los Angeles terminal area was used as an example for demonstrating integrated arrivals and departures, because previous works showed potential benefits if departures and arrivals were scheduled to share waypoints.Experiments were run with given daily traffic in the LAX terminal area.Current terminal arrival and departure procedures along with firstcome-first-serve (FCFS) like scheduling represented a baseline, and the proposed dynamic stochastic scheduler, which performs optimization for integrated operations, was compared against the baseline.The outputs from these two methods were evaluated using a standalone simulator, in which flight departure and arrival times were assumed to be uncertain.In the experiments, different separation buffer sizes were used.Results showed hours of flight time reduction in a typical day when comparing with current arrival and departure procedures.Also time-varied uncertainties were introduced into departure times.These uncertainties were assumed to increase with look-ahead time.Experiments were then conducted with various window sizes to gain insights into the impact of time-dependent uncertainties.In the paper, Section II describes the problem studied in the work.Section III presents the methodology including the sliding window method, trajectory synthesizer, and standalone simulations.Section III provides the results and analysis.Section IV concludes this work. +Problem and BackgroundLos Angeles terminal airspace is a challenging airspace to study because there exists potential inefficiency due to interactions between arrivals and departures.This type of interactions between different flows may happen more often in the future due to the deployment of Optimized Profile Descents (OPD).Recent studies [15][16] are focusing on potential arrival and/or departure interactions in NYC when executing OPD.Although the Los Angeles terminal airspace is studied in this work, the methodology can be applied to other areas.According to the Standard Terminal Arrival Routes (STARs) and the Standard Instrument Departures (SIDs) of Los Angeles terminal airspace, the arrivals from the FIM fix are required to follow procedure SADDE6 (FIM-SYMON-SADDE-SMO) and the departures to the North need to use procedure CASTA2 (RWY-NAANC-GHART-SILEX) (see Figure 1).The arrivals are requested to maintain their flight altitudes above 10,000 feet at fix GHART and the departures have to keep theirs at or below 9,000 feet at the same fix to procedurally avoid potential conflicts between arrivals and departures.If there were no interactions, departures to the North and arrivals from FIM would have flown direct routes.The direct routes would be RWY-WPT2-WPT1 and FIM-WPT1-SMO for departures and arrivals, respectively, as shown in the figure .WPT1 and WPT2 are made-up fix names for simplicity.Compared to these direct routes, besides flying nonpreferred altitudes, individual arrival and departure flights following current procedures will waste approximately 60 and 120 seconds, respectively. +Figure 1. Interactions Between Arrivals and Departures at Los Angeles Terminal AirspaceIn order to study the benefit of sharing waypoints, a hybrid approach that included temporal control and routing was proposed such that both departures and arrivals can have two route options besides metering.Previous work [13] showed great benefits of delay savings when comparing against the current procedure without route options.Subsequently, a stochastic scheduler was developed [14] to cope with schedule optimization under uncertain environment.The results showed promising statistical delay/fuel savings over current procedures even with the same statistical controller intervention level as using current procedures, where the controller intervention was a cost purely caused by uncertainty in flight departure/arrival times.In that study, it was also found that, with the guidance of uncertainty costs, stochastic optimization can achieve better solutions than deterministic optimization in which extra buffers were used to deal with uncertainty. +MethodThe static stochastic scheduler in previous work [14] was developed for a single fixed short time window (e.g. 30 minutes).It was assumed that the flights hadn't entered the route structure when scheduling.Therefore, only three categories of flights were formulated: departure flights to the North, arrival flights via Fillmore/FIM, and arrival flights from the East.For each FIM arrival that has not reached FIM, there are four design variables: delay before FIM, route option, speed between FIM and WPT1 for the direct route or speed between FIM and WPT2 for the indirect route, and delay between WPT1/WPT2 and SUTIE.For a departure flight that hasn't departed yet, three decision variables are defined: delay before departure, route option, and speed.For arrivals from the East, only one decision variable is defined, which is the delay before SUTIE.In this multiple-objective problem, there are two stochastic costs defined.One is the total delay compared to unimpeded operations where no separation is required.Another is the total count of controller interventions, that is the number of times a controller would be required to intervene to maintain separation.To introduce uncertainty, perturbations are imposed on the flight entry times: takeoff times for departures and entry waypoint (e.g.FIM or SUT) arrival times for arrivals.In both current and previous works, the perturbations follow a normal distribution.Unless otherwise defined, departures have a mean of -30 seconds and a standard deviation of 1.5 minutes, whereas arrivals are zero mean with a standard deviation of 30 seconds.It can be changed to other distributions if it is needed in the future.Given a solution, the costs are evaluated statistically over thousands of perturbed inputs/entry times.Both costs are evaluated stochastically with thousands of simulations.In previous work [14], an optimization method based on Non-dominated Sorting Genetic Algorithm (NSGA) and Monte Carlo simulation was developed for the stochastic scheduler and its implementation using Graphics Processing Units (GPUs) [17][18] enabled fast-time applications after dramatic reduction of computational time.The core scheduler in this work will adopt the same implementation. +Sliding WindowIn order to investigate the benefit for daily traffic, and furthermore, to enrich the scheduler with the capability of scheduling continuously in real application for traffic with long periods, a dynamic stochastic scheduler that can sequentially provide updated optimal schedules is desired.To achieve that capability, a sliding window method is utilized.The main idea of the sliding window is to divide a large time frame (e.g. a day) into small time windows (e.g. 30 minutes) and then to apply the scheduler to these small windows sequentially.In this work these small windows are chosen to have no overlaps.When scheduling for a small time window, the flights that come from the previous time window but still remain in the airspace would be included in the updated flight states.These updates will be fed into the optimization for next window.Three categories of flight state in the previous static scheduler will not be enough as the categories only cover flights that haven't entered the entry fixes/waypoints.In the sequential scheduler, a total of nine flight state categories were formulated.For instance, flights that are between FIM and WPT1 would be one category with three decision variables, and flights that are between WPT1 and SMO will be in another category with one less decision variable.With this enriched formulation, the scheduler is ready to handle flights at any stage on the designated route structure. +Trajectory SynthesizerThe Trajectory Synthesizer (TS), which is the trajectory calculation engine of the Center TRACON Automation System (CTAS) [19], is applied before the optimization process to generate various feasible high-fidelity trajectory profiles for arrivals and departures.When producing those profiles, the route structures associated with waypoints, altitude ranges, and speed requirements are imposed on the inputs of TS.TS will only generate trajectory profiles for feasible speed and route options based on the aircraft aerodynamics.These pre-processed feasible trajectory profiles, including speed options and fuel consumptions for flights with different aircraft types, will then be fed into the optimization and postanalysis to increase the fidelity of the results. +Standalone SimulatorTo set-up a complete and seamless experiment environment for the sequential stochastic scheduler, a standalone simulation platform, which includes Monte Carlo simulations, is needed to provide objective updated flight states by mimicking reality through incorporating randomness to flight entry times for next time window.To perform this functionality, the number of simulation is set to one to generate traffic scenario for next window.On the other hand, a standalone platform is also necessary for testing or verifying the benefits claimed in the optimization process.In the optimization process, only 1,000 Monte Carlo simulations are conducted when evaluating an individual solution for a given time window of flight states because of the limits on running time for real-time application feasibility.Because the run-time requirements are not as strict for post-processing, the standalone simulation platform used 5,000 Monte Carlo simulations in postprocessing to verify the solution benefits. +Figure 2. Work Flow for the Sequential Stochastic SchedulerThe overall process is shown in Figure 2. The traffic including flight information is split to small time windows.After incorporating updated flight information (including uncertain departure/arrival times) provided by the standalone simulator, the traffic scenario is passed to the static scheduler, where trajectory/speed options provided by TS for each flight would be used in the optimization.The output schedule is recorded and it is tested using the standalone simulator.After that, the standalone simulator will provide: flight departure/arrival times based on the given distribution, updated flight states for flights that are still remaining on the route, and flights to be scheduled in next window.The process will be repeated till the time reaches the end of the traffic. +ResultsIn these experiments, a daily schedule along with aircraft types was built according to the traffic on December 4, 2012 (Tuesday).There are a total of 378 flights, including 290 arrivals from Fillmore and the East and 88 northbound departures in that day.The separation criteria between different aircraft follow the published FAA ATC regulations [20] as shown in Table 1.Experiments were run on HP Z820 workstation with multicores and 32GB memory.Two GTX690 with four GPUs were installed in the system.And expected values (averages) were used as the measurement of the stochastic results. +Table 1. Separation Based on Wake Category +Pareto FrontPareto fronts representing ranges of solutions minimizing flight time and controller interventions were used to evaluate optimization outputs.In this work, Pareto fronts were obtained for each 30-minute time window.At the end of the process, dozens of Pareto fronts were generated corresponding to all 30minute windows in a day.To obtain a single front, the collection of Pareto fronts were combined and ranked one by one in a temporal order.Figure 3 shows the final augmented Pareto front, which is represented by a curve composed of black dots.The coordinates of these dots are expected values of the costs associated, and they are evaluated in the optimization process.The curve composed of blue dots is based on re-evaluated costs of the same solutions using the standalone simulator with large number of Monte Carlo simulations.It is noted that these two curves almost overlap with each other, which further testifies that reducing the number of Monte Carlo simulations to one thousand in optimization doesn't sacrifice the cost evaluation accuracy too much while saving running time of optimization as pointed out in previous work [14].It is shown that the front from stochastic optimization with hybrid approach covers a large range of delays and there is clearly a trade-off between delay and controller intervention count.When the solution with the least delay is chosen, the average delay is less than 100 minutes in a day relative to unimpeded operations, whereas the expected value of controller intervention count would be over 100.That means, to achieve the least delay, if flights follow the optimized schedule at the beginning of every 30-minute window, controllers need to statistically take 100 actions on average during the day to prevent unexpected loss of separations due to departure and arrival time uncertainty. +Stochastic vs. DeterministicA popular approach of deterministic schedulers to deal with uncertainty is to add a buffer to the separation requirement.Figure 3 also shows solutions from a deterministic method with different buffers.As identified in the figure, the left circular, triangular, and square dots were produced using deterministic optimization with 0s, 30s, and 60s buffer, respectively.The data suggests that, under the guidance of stochastic costs, stochastic optimization performs better than the deterministic method.With the same controller intervention levels, stochastic optimization can provide an extra 50 to 150 minutes of delay savings.The right circular, triangular, and square dots were generated from deterministic optimization using only segregated procedural routes as in the published STARs and SIDs without direct route options.The differences between the stochastic hybrid method and the deterministic spatial method demonstrate the benefits brought by combining the hybrid separation approach with stochastic optimization.With the same level of controller intervention, hybrid separation approach with stochastic optimization can provide an extra 150 to 250 minutes of flight time savings in a day over hybrid separation with deterministic optimization.Select delays from the stochastic hybrid approach and deterministic spatial approach are shown in Figure 4 and5.For the former approach, the solutions with the least delay were chosen.These figures present comparisons using different extra buffers.It is noted that the total daily delay savings corresponding to different buffers vary from 5 hours to 7.5 hours.Whereas despite various buffers, the daily fuel savings stabilize around 32,000 lbs/4712 gallons per day, which can be translated to 10 million dollars saving per year based on the average Jet A fuel price ($6.08/gal) in LAX area. +Window size vs. UncertaintyIn previous experiments, it was assumed that the uncertainties of both departure times and arrival times had constant means and standard deviations in spite of the look-ahead time.This may be a reasonable assumption for arrival times based on arrival time prediction research [21] [22], but not for departures where uncertainty always increases with look-ahead time.In the work done by Capps et al [23], a relationship between wheels off prediction accuracy at DFW and look-ahead time in June 2011 was presented, where the wheels off time was predicted using the Surface Decision Support System (SDSS) [24] [25].Although that relationship may change with traffic scenarios, available information, and airport, it is nevertheless a good reference for setting up experiments in this work.The simple linear relationships shown in Eq.( 1) and (2) were applied in this work, where the mean and standard deviation were assumed to linearly increase with look-ahead time ( T L ).Mean = 0.39 × T L (1) Std.dev.= 0.41 × T L (2)Figure 6 shows the Pareto fronts for the optimizations conducted under time-varied means and standard deviations that follow Eq. ( 1) and ( 2).The blue curve is the Pareto front produced using 20-minute windows.Whereas the green curve and black curve are Pareto fronts obtained using 30 and 60 minute windows, respectively.In deterministic cases or stochastic cases with constant uncertainty, the optimality of the results should increase with the window size because larger window size allows more space for optimization.This may not always be true when uncertainties grow with look-ahead time.The enlarged uncertainties will contradict the benefit brought by a large planning window and reduce the likelihood of planned benefits.As shown in Figure 6, at the high delay cost part of the front, large planning windows still produces better solutions than small ones.Whereas, at the low delay cost part of the front, the small window tends to be a better choice.The hypothesis is that when delay costs are relatively high, gaps between flights are large in associated solutions and uncertainty is less critical than planning duration.However, when delay costs are low, spaces between flights are tight such that uncertainty becomes the key impact on the final costs.Figure 6 also shows that the margins between different window sizes are relatively small when the delay costs are low.The reason could be that the time-varied uncertainties were only applied to 88 departure flights, while the remaining 290 arrival flights still held constant uncertainties.The impact of time-varied uncertainties was only strong enough to eliminate the advantage brought by large window size but not strong enough to reverse it much.The other take-away message from this figure is that time-dependent uncertainties should not be a concern when delay saving are moderate as the spaces between flights can absorb the impact of varied uncertainties.When schedules with aggressive delay saving are desired, a small window may be a better choice than a large window as it finds better solutions with less computational time. +Figure 6. Pareto Fronts Under Different Uncertainties Using Expected Values +ConclusionsThis work extends the static stochastic scheduler with dynamic capability such that the scheduler can be applied to real-time traffic continuously.Using this dynamic stochastic scheduler enabled investigations of optimized integrated arrivals and departures under uncertainty for extended traffic periods.sliding window method was proposed to divide a large time frame into small time windows to which the scheduler can be applied.In order to increase the fidelity of results, TS was used to calculate possible trajectory profiles in preprocess and resulting speed options were then fed into the schedule optimization.Additionally, a standalone simulator was developed to simulate the effects of uncertainty and to provide updated flight states for following time window.The standalone simulator was also used to verify the results from optimization.The Los Angeles terminal area was used as a test location for demonstrating integrated arrivals and departures.Experiments were run using given daily traffic in LAX terminal area, which covered 30% of arrival traffic and 10% of departures.With a 30-minute time window, the stochastic scheduler can find better solutions than the deterministic scheduler using buffers to account for stochastic impacts.The results show that with the same controller intervention level, additional delay savings vary from 50 to 150 minutes.Compared to the baseline, which utilized current terminal arrival and departure procedures with first-come-first-serve (FCFS) like scheduling, the stochastic scheduler can provide an extra 150 to 250 minutes delay savings at the same controller intervention level.If an aggressive solution is allowed, an average of 5.2 hours can be saved in a day in LAX.The expected value of annual fuel saving would be more than 10 million dollars.However, the cost of potential controller intervention, which results from the uncertainty of estimated departure and arrival times, will increase by 50% on average.If a moderate solution is chosen instead, with the same level of controller intervention, there is still more than four hours of delay saving.Time-varied uncertainties were introduced to departure times, where uncertainties increased with the look-ahead time.Using constant uncertainties for arrival times and time-varying uncertainties for departure times, experiments were conducted with 20, 30, and 60-minute windows.It was shown that to search for solutions with moderate delays, window size played a more important role than uncertainty.When the target was to find solutions with aggressive delay savings, the impact of uncertainty dominated, which made small windows a better choice.The sequential stochastic scheduler developed in this work showed its promising capability in continuously optimizing schedules under timevaried uncertainties.In future work, the scheduler will be extended to cover all departures and arrivals in LAX for both airborne and surface operations.Figure 3 .3Figure 3. Solutions Using Different Approaches +Figure 4 .Figure 5 .45Figure 4. Daily Delay Comparisons + + October 5-9, 2014 + + + + +AcknowledgementsThe authors gratefully acknowledge the contribution of Dr. Minghong (Gilbert) Wu of University Affiliated Research Center, University of California at Santa Cruz.Dr. Wu guided the usage of TS and associated tools, and helped generate appropriate trajectory profiles for different aircraft and speeds. + + + + + + + + + The dynamic scheduling of aircraft in the near terminal area + + RGDear + + 10.1016/0041-1647(78)90133-8 + + + Transportation Research + Transportation Research + 0041-1647 + + 12 + 3 + + September 1976 + Elsevier BV + + + Dear, R. G., September 1976, "The Dynamic Scheduling of Aircraft in the Near Terminal Area", Tech. Rep. Flight Transportation Laboratory Report R76-9, MIT. + + + + + An algorithm for computer assisted sequencing and scheduling of terminal area operations + + RogerGDear + + + YosefSSherif + + 10.1016/0191-2607(91)90132-a + + + Transportation Research Part A: General + Transportation Research Part A: General + 0191-2607 + + 25 + 2-3 + + 1991 + Elsevier BV + + + Dear, R. G., Y. S. Sherif, 1991, "An Algorithm For Computer Assisted Sequencing and Scheduling of Terminal Operations", Transportation Research A, Vol. 25A, No. 2/3, pp. 120-139. + + + + + Analysis of Delay Reducing and Fuel Saving Sequencing and Spacing Algorithms for Arrival Spacing + + FNeuman + + + HErzberger + + Memorandum 103880 + + October 1991 + + + Tech. Rep. NASA Technical + Neuman, F., H. Erzberger, October 1991, "Analysis of Delay Reducing and Fuel Saving Sequencing and Spacing Algorithms for Arrival Spacing", Tech. Rep. NASA Technical Memorandum 103880. + + + + + Scheduling Aircraft Landings Under Constrained Position Shifting + + HamsaBalakrishnan + + + BalaChandran + + 10.2514/6.2006-6320 + + + AIAA Guidance, Navigation, and Control Conference and Exhibit + Keystone, CO + + American Institute of Aeronautics and Astronautics + August 2006 + + + + Balakrishnan, H., B. Chandran, 21-24 August 2006, "Scheduling Aircraft Landings Under Constrained Position Shifting", AIAA Guidance, Navigation, and Control Conference, Keystone, CO. + + + + + Scheduling Aircraft Landings—The Static Case + + JEBeasley + + + MKrishnamoorthy + + + YMSharaiha + + + DAbramson + + 10.1287/trsc.34.2.180.12302 + + + Transportation Science + Transportation Science + 0041-1655 + 1526-5447 + + 34 + 2 + + 2000 + Institute for Operations Research and the Management Sciences (INFORMS) + + + Beasley, J. E., M. Krishnamoorthy, Y. M. Sharaiha, D. Abramson, 2000, "Scheduling Aircraft Landings -The Static Case", Transportation Science, Vol. 34, No. 2, pp. 180-197. + + + + + Scheduling Aircraft Landings to Closely Spaced Parallel Runways + + MKupfer + + + + Europe Air Traffic Management Research and Development Seminar + + June 2009 + Napa, CA + + + The Eighth USA + Kupfer, M., June 2009, "Scheduling Aircraft Landings to Closely Spaced Parallel Runways", The Eighth USA/Europe Air Traffic Management Research and Development Seminar, Napa, CA. + + + + + A Mixed Integer Linear Program for Airport Departure Scheduling + + GautamGupta + + + WaqarMalik + + + YoonJung + + 10.2514/6.2009-6933 + + + 9th AIAA Aviation Technology, Integration, and Operations Conference (ATIO) + Hilton Head, South Carolina + + American Institute of Aeronautics and Astronautics + 21-23 September 2009 + + + A Mixed Integer Linear Program for Airport Departure Scheduling + Gupta, G., W. Malik, Y. C. Jung, 21-23 September 2009, "A Mixed Integer Linear Program for Airport Departure Scheduling", 9th AIAA Aviation Technology, Integration and Operations Conference (ATIO), Hilton Head, South Carolina. + + + + + A Mixed Integer Linear Program for Solving a Multiple Route Taxi Scheduling Problem + + JustinMontoya + + + ZacharyWood + + + SivakumarRathinam + + + WaqarMalik + + 10.2514/6.2010-7692 + + + AIAA Guidance, Navigation, and Control Conference + Toronto, Canada + + American Institute of Aeronautics and Astronautics + 2-5 August 2010 + + + Montoya, J., Z. Wood, S. Rathinam, W. Malik, 2-5 August 2010, "A Mixed Integer Linear Program for Solving a Multiple Route Taxi Scheduling Problem", AIAA Guidance, Navigation, and Control Conference, Toronto, Canada. + + + + + A Generalized Dynamic Programming Approach for a Departure Scheduling Problem + + SivakumarRathinam + + + ZacharyWood + + + BanavarSridhar + + + YoonJung + + 10.2514/6.2009-6250 + + + AIAA Guidance, Navigation, and Control Conference + Chicago, IL + + American Institute of Aeronautics and Astronautics + 10-13 August 2009 + + + Rathinam, S., Z. Wood, B. Sridhar, Y. C. Jung, 10-13 August 2009, "A Generalized Dynamic Programming Approach for a Departure Scheduling Problem", AIAA Guidance, Navigation, and Control Conference, Chicago, IL. + + + + + A Hybrid Optimization Approach to Air Traffic Management for Metroplex Operations + + BrianCapozzi + + + StephenAtkins + + 10.2514/6.2010-9062 + + + 10th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Fort Worth, Texas + + American Institute of Aeronautics and Astronautics + September 2010 + + + + Capozzi, B. J., S. C. Atkins, 13-15 September 2010, "Hybrid Optimization Approach to Air Traffic Management for Metroplex Operations", 10th AIAA Aviation Technology, Integration and Operations Conference (ATIO), Fort Worth, Texas. + + + + + Towards Optimal Routing and Scheduling of Metroplex Operations + + BrianCapozzi + + + StephenAtkins + + + SeongimChoi + + 10.2514/6.2009-7037 + + + 9th AIAA Aviation Technology, Integration, and Operations Conference (ATIO) + Hilton Head, South Carolina + + American Institute of Aeronautics and Astronautics + 21-23 September 2009 + + + Capozzi, B. J., S. C. Atkins, 21-23 September 2009, "Towards Optimal Routing and Scheduling of Metroplex Operations", 9th AIAA Aviation Technology, Integration, and Operations Conference (ATIO), Hilton Head, South Carolina. + + + + + Sequential Dynamic Strategies for Real-Time Scheduling of Terminal Traffic + + HemingChen + + + YiyuanJZhao + + 10.2514/1.c031503 + + + Journal of Aircraft + Journal of Aircraft + 0021-8669 + 1533-3868 + + 49 + 1 + + 2012 + American Institute of Aeronautics and Astronautics (AIAA) + + + Chen, H., Y. J. Zhao, 2012, "Sequential Dynamic Strategies for Real-Time Scheduling of Terminal Traffic", American Institute of Aeronautics and Astronautics, Journal of Aircraft, Vol. 49, No. 1. pp.237-249. + + + + + Optimal Integration of Departures and Arrivals in Terminal Airspace + + MinXue + + + ShannonZelinski + + 10.2514/1.60489 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 37 + 1 + + 2014 + American Institute of Aeronautics and Astronautics (AIAA) + + + Xue, M., S. Zelinski, 2014, "Optimal Integration of Departures and Arrivals in Terminal Airspace", Journal of Guidance, Control, and Dynamics, AIAA, Vol.37, No.1, pp.207-213. + + + + + Optimization of Integrated Departures and Arrivals Under Uncertainty + + MinXue + + + ShannonZelinski + + 10.2514/6.2013-4322 + + + 2013 Aviation Technology, Integration, and Operations Conference + Los Angeles, CA + + American Institute of Aeronautics and Astronautics + August 12-14, 2013 + + + Xue, M., S. Zelinski, August 12-14, 2013, "Optimization of Integrated Departures and Arrivals Under Uncertainty", AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, Los Angeles, CA. + + + + + Development of a route crossing tool for shared airspace environments + + DaphneRein-Weston + + + EricChevalley + + 10.1109/dasc.2014.6979655 + + + 2014 IEEE/AIAA 33rd Digital Avionics Systems Conference (DASC) + Colorado Springs, Colorado + + IEEE + October 5-9, 2014 + + + Rein-Weston, D., E. Chevalley, A. Globus, R. Jacoby, E. Palmer, October 5-9, 2014, "Development of Route Crossing Tool for Shared Airspace Environments", 33 rd Digital Avionics Systems Conference, Colorado Springs, Colorado. + + + + + Improving departure throughput by dynamically adjusting inter-arrival spacing + + Hyo-SangYoo + + + PaulLee + + + EverettPalmer + + 10.1109/dasc.2014.6979546 + + + 2014 IEEE/AIAA 33rd Digital Avionics Systems Conference (DASC) + Colorado Springs, Colorado + + IEEE + October 5-9, 2014 + + + Yoo, H. S., E. Palmer, October 5-9, 2014, "Improving Departure Throughput by Dynamically Adjusting Inter-Arrival Spacing", 33 rd Digital Avionics Systems Conference, Colorado Springs, Colorado. + + + + + GPU-based Parallelization for Schedule Optimization with Uncertainty + + ChristabelleBosson + + + MinXue + + + ShannonZelinski + + 10.2514/6.2014-2024 + + + 14th AIAA Aviation Technology, Integration, and Operations Conference + Atlanta, Georgia + + American Institute of Aeronautics and Astronautics + June 16-20, 2014. 2014 + + + Bosson, C., M. Xue, S. Zelinski, June 16-20, 2014, "GPU-based Parallelization for Schedule Optimization with Uncertainty", AIAA Aviation and Aeronautics Forum and Exposition 2014, Atlanta, Georgia. + + + + + Integrated Arrival- and Departure-Schedule Optimization Under Uncertainty + + MinXue + + + ShannonZelinski + + 10.2514/1.c032957 + + + Journal of Aircraft + Journal of Aircraft + 0021-8669 + 1533-3868 + + 52 + 5 + + 2014 + American Institute of Aeronautics and Astronautics (AIAA) + + + Submitted + Xue, M., S. Zelinski, 2014, "Integrated Arrival and Departure Schedule Optimization Under Uncertainty", Journal of Aircraft, AIAA, Submitted. + + + + + Design of Center-TRACON Automation System + + HErzberger + + + TJDavis + + + SMGreen + + + + AGARD Meeting on Machine Intelligence in Air Traffic Management + Berlin, Germany + + May 1993 + + + + Erzberger, H., T. J. Davis, S. M. Green, 11-14 May 1993, "Design of Center-TRACON Automation System", AGARD Meeting on Machine Intelligence in Air Traffic Management, Berlin, Germany. + + + + + Use of NEXRAD to Meet FAA and Aviation Needs + + MichaelDEilts + + + JTJohnson + + 10.2514/atcq.2.1.33 + FAA Order JO 7110.65 + + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 2 + 1 + + + American Institute of Aeronautics and Astronautics (AIAA) + + + FAA Order JO 7110.65s: Air Traffic Control, http://www.faa.gov/documentlibrary/media/order/7 110.65s.pdf + + + + + Improvement of Trajectory Synthesizer for Efficient Descent Advisor + + MinXue + + + HeinzErzberger + + 10.2514/6.2011-7020 + + + 11th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Virginia Beach, VA + + American Institute of Aeronautics and Astronautics + September 20-22, 2011 + + + Xue, M., H. Erzberger, September 20-22, 2011, "Improvement of Trajectory Synthesizer for Efficient Descent Advisor", 11th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, Virginia Beach, VA. + + + + + Predictability of Top of Descent Location for Operational Idle-Thrust Descents + + LaurelStell + + 10.2514/6.2010-9116 + + + 10th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Berlin, Germany + + American Institute of Aeronautics and Astronautics + June 2011 + + + The Ninth USA/Europe Air Traffic Management Research and Development Seminar + + + Stell, L., June 2011, "Prediction of Top of Descent Location for Idle-thrust Descents", The Ninth USA/Europe Air Traffic Management Research and Development Seminar, Berlin, Germany. + + + + + Characterization of Tactical Departure Scheduling in the National Airspace System + + RichardCapps + + + ShawnEngelland + + 10.2514/6.2011-6835 + + + 11th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Virginia Beach, VA + + American Institute of Aeronautics and Astronautics + September 20-22, 2011 + + + Capps, A., S. A. Engelland, September 20-22, 2011, "Characterization of Tactical Departure Scheduling in the National Airspace System", AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, Virginia Beach, VA. + + + + + Surface Management System Field Trial Results + + StephenAtkins + + + YoonJung + + + ChristopherBrinton + + + LStell + + + TCarniol + + + SRogowski + + 10.2514/6.2004-6241 + + + AIAA 4th Aviation Technology, Integration and Operations (ATIO) Forum + Chicago, IL + + American Institute of Aeronautics and Astronautics + September 20-22, 2004 + + + Atkins, S., Y. C. Jung, C. Brinton, L. Stell, S. Rogowski, September 20-22, 2004, "Surface Management System Field Trial Results", AIAA 4 th Aviation Technology, Integration, and Operations (ATIO) Conference, Chicago, IL. + + + + + Field test results of Collaborative Departure Queue Management + + ChrisBrinton + + + SteveLent + + + ChrisProvan + + 10.1109/dasc.2010.5655527 + + + 29th Digital Avionics Systems Conference + Salt Lake City, Utah + + IEEE + October 3-7, 2010 + + + Brinton, C., S. Lent, C. Provan, October 3-7, 2010, "Field Test Results of Collaborative Departure Queue Management," 29 th Digital Avionics Systems Conference, Salt Lake City, Utah. + + + + + + diff --git a/file819.txt b/file819.txt new file mode 100644 index 0000000000000000000000000000000000000000..d20a15f7962ad112340506ca4d7303161c0724ed --- /dev/null +++ b/file819.txt @@ -0,0 +1,365 @@ + + + + +I. IntroductionIn a metroplex or high density terminal operations, typically within 100 nautical miles of an airport or a major airport in a metroplex, resources, such as route segments, fixes, and runway, are normally spatially segregated in order to reduce interactions between different traffic flows and to guarantee separation between aircraft.Such separation may introduce unnecessary inefficiency due to lengthened routes or undesired altitude constraints, introducing integrated arrivals and/or departures with shared resources may help improve the efficiency.In past years, in order to improve efficiency of terminal airspace operations, researchers treated the arrival and departure scheduling problems as separate ones.Many algorithms were developed, such as constrained position shifting (CPS), [1][2][3] CPS with dynamic programming (DP), 4 mixed integer linear programming (MILP), 5 and basic genetic algorithm (GA) 6 for optimizing schedules.Recently, spatial and temporal usage of shared resources started to draw researchers' attention.A couple of approaches were proposed to solve the optimal routing and scheduling problem.In 2009, Capozzi et.al 7 introduced a MILP formulation and applied it to coordinate SFO arrivals and SJC departures.The MILP formulation was found suitable for solving small scale problems but it required significant computational time if the number of flights was greater than 10.Later, the authors 8 further proposed a hybrid algorithm, which combined basic GA and MILP.The GA was used to solve a high level problem (route assignment and sequencing) while MILP was applied to the low level problems.Significant reduction in computational time was achieved when the algorithm was applied to a small problem, but further research is needed for solving realistic and complicated problems.In this paper, a new formulation using a Non-dominated Sorting Genetic Algorithm (NSGA) was introduced because of its ability to handle multi-objective optimization and multiple constraints.The NSGA was demonstrated on a test problem first, then it was applied to an observed problem in LAX terminal airspace.Currently, interactions between LAX arrivals and departures are resolved by spatially segregating arrival and departure routes, which may introduce inefficiency.In this paper, three different separation approaches to the LAX problem were examined including the hybrid separation which combines temporal and spatial separation.The temporal and hybrid separations were formulated and solved using the NSGA algorithm.Furthermore, a First-Come-First-Serve (FCFS) based heuristic method was applied to formulate the hybrid separation to compare with the NSGA algorithm.In this paper, Section II introduces the NSGA algorithm.In section III, a test problem is set up to examine the performance of the new algorithm.In Section IV, the NSGA algorithm is applied to solve the interactions between arrivals and departures in LAX terminal airspace in three different ways.Analysis is then conducted to examine the benefits brought by integrating arrivals and departures spatially and temporally.Comparison between the NSGA and the heuristic method is also conducted.Section V discusses conclusions of the study. +II. MethodologyIn the terminal airspace, methods of integrating flights with shared resources include routing, sequencing, and scheduling.The objective is usually to minimize total delay time for a given set of flights while maintaining separation constraints and achieving proper sequence of landings.The problems are typically highly constrained due to aircraft separation requirements.According to Capozzi's work 8 the MILP formulation using the CPLEX solver requires a large amount of computational time, which might prohibit further application.In this work, a variation of GA is investigated.GAs 9 have been quite successful in a great range of problems.These groups of algorithms are stochastic processes that model two natural phenomena: genetic inheritance and Darwinian evolution.Evolutionary operators include selection, crossover, and mutation.In the basic GA, the selection is based on the fitness functions of the population in the generation.Typical approaches to handle constraints are rejecting or penalizing infeasible individuals.The rejection of infeasible individuals is easy and popular, but it may get stuck when the feasible search space is not convex or the search space is highly constrained.Penalizing infeasible individuals relaxes the constraints, but it is hard to decide the penalty.NSGA II, 10 the NSGA 11 variant used in this research was developed in recent years in order to improve GA's performance on multi-objective optimization and multiple constraint handling.Each individual has two attributes: fitness and errors.Fitness is calculated based on the objective, whereas errors are calculated if constraints are violated.Compared with the basic GA, the only change in NSGA is the selection process.Instead of fitness, the population is evaluated and ranked based on the ordering of their dominance (Pareto dominance) and is sorted into a hierarchy of subgroups.Assuming the objective is to minimize and the constraint function g has to be nonnegative, individual A is dominated by individual B if:   f A > f B , if g A ≥0 and g B ≥0, or, g A = g B g A < g B , if g A < 0 and g B < 0, or, g B > 0 and g A < 0 (1)Where f is the objective and g is the constraint value.In order to estimate the density of solutions surrounding a particular solution in the population, an average distance between two neighboring solutions on either side of the particular solution along each objective is calculated and is termed as the crowding distance.Crowding distance between members of each sub-group introduces diversity among nondominated solutions.Figure 1 demonstrates the selection procedure in NSGA II.At generation t, assuming P is the parent population and Q is the offspring population.Both P and Q have N individuals.They are first combined to a 2N population R t , then individuals in R t are sorted based on their Pareto dominance.The best nondominated solutions are in set F 1 .After calculating crowding distances of set F 1 , if the size of F 1 is smaller than N, F 1 will be added into the new population P t+1 .Then, solutions from set F 2 are added to the new population, and so on.This process will continue unless the size of P t+1 exceeds N .In order to make sure that the size of P t+1 reaches N , the solutions from the next F i will be partially chosen in terms of their crowding distances .The order in the flow is defined as Eqn. 2. After the selection process, the remaining processes are the same as basic GAs.A B if rank A < rank B , or (rank A = rank B and crowding distance A > crowding distance B )(2)Combine parent and offspring populationR t =P t U Q t Non-dominated-sort(R t ) →F (F 1 F 2 F 3 …) P t+1 = Ø & i = 1 size(P t+1 +F i ) ≤N Crowding-distance-assignment(F i ) P t+1 =P t+1 U F i i = i + 1 Sort(F i , p ) P t+1 = P t+1 U F i [1:( N-size(P t+1 )]Selection: generation = tCrowding-distance-assignment(F i ) In this work, bit strings were used to represent solutions.Therefore, the search space is discrete and it helps speed up the process without sacrificing much optimality.Usually, the population size was set to 800 and the maximum number of generations was defined to be 300.Because random initialization was used, any problem set in this study was usually run over three times and the best results were reported.All problems in this work were solved on a MacOS platform with 2x2.66GHz 6-Core Intel Xeon and 8GB RAM. +III. A Test ProblemTo test the formulation using the NSGA algorithm, a test problem was rebuilt based on Capozzi's papers. 7,8 n this problem, two departure flows are assumed to come from two neighboring airports OAK and SJC, respectively.As shown in Fig. 2, the two departure fixes W 0 and W 1 are defined as shared resources.The distance for both W 1 to OAK (route R 4 ) and W 0 to OAK (route R 3 ) is 45 miles; The distance for both W 1 to SJC (route R 2 ) and W 0 to SJC (route R 1 ) is 56 miles.The flight departure schedule is described in Table 1, which is the same as the "phased peak demand" mentioned in Capozzi's paper. 7Aircraft were assumed to be of the same type.The minimum speed of aircraft was defined to be 140 knots and the maximum speed was 180 knots.The minimum separation required at runway D r was set to 3 nautical miles for all aircraft pairs and the minimum separation at departure fixes D f was defined to be 4 nautical miles.The maximum delay was assumed to be 200 seconds.Flight time uncertainty was included using a time error (δ) of 60 seconds.Two scenarios of fix usage were examined: segregated vs. shared.In the "segregated" case, only R 1 and R 4 are available, whereas in the "shared" case, all four routes are available.In this problem formulation, decision variables for each flight i are ground delays (d i ), aircraft speeds (v i ), and route options (r i ).For each flight, its time to exit a departure fix (tp i ) and its time to take off (tr i ) can be expressed as in Eqn. 3, where t i0 is the scheduled departure time as shown in Table 1.There are two routes for either SJC or OAK departures: r i = 0 represents the default route (R 1 for SJC departures and R 4 for OAK departures) and r i = 1 represents the second route (R 2 for SJC departures and R 3 for OAK departures).R d means the default route and R s is the second route.Variable v i denotes the airspeed.Therefore, in NSGA, there are three genes (d i , r i , v i ) for each flight.   tr i = t i0 + d i , tp i = tr i + ((1 -r i ) • R d + r i • R s )/v i (3) f = tp i (4)The objective is minimizing the total time as in Eqn. 4. The constrains are: the separation at departure fixes (Eqn.5) and separation at runways (Eqn.6).The problem with 20 flights was modelled and solved without multiple windows that were used in MILP formulations. 7The formulations in Eqn. 3 and 5 are built for the case of shared fixes.For the case of segregated fixes, r i can be simply fixed at zero.|tp i -tp j |•[r i •r j + (1 -r i )•(1 -r j )] -D f /v k -δ≥0, (i =j, if tp i > tp j , k = j otherwise k = i) (5) |tr i -tr j | -D r /v k -δ≥0, (i =j, if tp i > tp j , k = j otherwise k = i)(6)Table 2 shows the resulting delays with segregated and shared fixes, respectively.A reduction of 65% was achieved by shared departure fixes.The total delay saving was 440 seconds (from 677 seconds to 237 seconds) over segregated fixes.Among them, 516 seconds were saved from ground delay with increased airborne delay of 76 seconds.Makespans were the same because it was constrained purely by flight schedule in this problem.Figure 3(a) and 3(b) presented the results for segregated and shared fixes, respectively.The vertical axes are way points.The times of SJCRWY and OAKRWY represent the scheduled departure times, while the times of SJC and OAK are departure times computed by the algorithm.The differences denotes the suggested delays.Note that many SJC departures used W 1 which wasn't available to them in the "segregated" case and six out of ten OAK departures used W 0 which wasn't available to them in the "segregated" case.Sharing departure fixes provides flexibility in route options, and the departure fixes can therefore be fully used to improve operation efficiency.This test showed that the new formulation with the NSGA performs well.It solved the 20 flight problem in around 30 seconds without any parallelization.Because one of the strengths of GA-like algorithms is parallelization, significant reduction in computational time could be easily realized. +IV. Case Study: LAX Terminal AirspaceThe interactions between arrivals and departures in LAX terminal airspace was identified to be a potential scheduling problem that could be solved more efficiently than current procedures.This section describes the model, method, and analysis for this problem. +A. DescriptionAccording to the Standard Terminal Arrival Routes (STARs) of Los Angeles terminal airspace, arrivals to Los Angeles airport (LAX) from the North are required to take procedure SADDE6, which is to fly from Fillmore(FIM) to Santa Monica(SMO) via SYMON and SADDE fixes.Based on the Standard Instrument Departures (SIDs), LAX departures to the North need to follow procedure CASTA2, which is to take off from Runway 24R to WPT1 a through NAANC and GHART (see Fig. 4 ).In order to spatially segregate these two flows, arrival flights from FIM are required to maintain their flight altitudes above 12,000 feet at Fix GHART, while departures have to keep theirs at or below 9,000 feet at the same fix.As studied by Timar, 12 approximately 28.1% of LAX arrivals use the SADDE6 procedure and 10.4% of LAX departures use the CASTA2 procedure.In a typical day, this can be translated to 220 arrival flights and 80 departure flights.The total cost or delay in a day due to the arrival departure interaction is approximately 380 minutes with spatially segregated routes in the SIDs and STARs.Could it be improved?The following study addresses this using three different methods: spatial, temporal and hybrid separation.Spatial separation uses the same strategy as in SIDs and STARs.Temporal separation utilizes the direct routes with conflicts resolved solely with temporal controls.As in Fig. 4, the direct route for departures is RWY-WPT2-WPT1, and for FIM arrivals it is FIM-WPT1-SMO-SUTIE.Hybrid separation applies both temporal and spatial separation. +B. ModelingThree flows are taken into account in this work: arrivals from FIM (Set A 1 ), departures from Runway 24L(Set D), and arrival flights from the East towards SUTIE(Set A 2 ).Table 3 shows scheduled arrival times (t i0 ) at FIM, RWY, and SUTIE, respectively.Historical traffic schedules between 18:30 pm to 19pm (UTC time) on March 5, 2010 were used as a reference for generating the schedules.There are a total of 15 flights a Points WPT1 and WPT2 are made up for simplicity.including five arrivals from FIM, six arrivals from east of LAX, and four departures from Runway 24L.In this work, flights were assumed to be the same type. +Decision VariablesAssume that route R 1 refers to RWY-WPT2-WPT1 (direct route for departures), R 2 represents RWY-NAANC-GHART-AJAYE-WPT1 (lengthened route for departures), R 3 denotes FIM-WPT1-SMO-SUTIE (direct route for arrivals), and R 4 is the route of FIM-SADDE-GHART-WPT2-SMO-SUTIE (lengthened route for arrivals).In the formulation of hybrid separation, four design variables were defined for each arrival flight in Set A 1 :• d1 i -The delay at or before FIM.• r i -If r i = 0, the direct route R 3 will be chosen, otherwise, R 4 is selected.• v i -The aircraft speed between FIM and WPT1 when flying the direct route or the speed between FIM and WPT2 if flying the indirect route.• d2 i -The delay at or before SUTIE to ensure separation at SUTIE.For a departure flight in set D, three decision variables were defined:• d i -The delay before departure.• r i -If r i = 0, the direct route R 1 will be chosen, otherwise R 2 is selected.• v i -The speed from departure to WPT1.Only one decision variable exists for an arrival flight in Set A 2 :• d i -The delay time at or before SUTIE to ensure separation with A 1 at SUTIE.In the case of temporal separation, route options (r i ) in Set D and Set A 1 are fixed at zero so both departures and arrivals take direct routes.The only way to meet the separation requirements is to use time control.In the study, two scenarios were set up.In scenario one, no aircraft is allowed to arrive/depart early or speed up.In scenario two, aircraft are allowed to arrive/depart early or speed up up to 30 seconds. +ConstraintsEqn. 7 shows the expression for FIM arrivals.Let R 3 denote route FIM-WPT1 and R 4 represents route FIM-SADDE-WPT2, which are partial routes of R 3 and R 4 , respectively.Variable L R i represents the length or distance of route R i .Variable t F IM (F IM,i) is defined to be the time when flight i arrives at FIM. Variable t F IM (W P T,i) denotes the arrival time of flight i at WPT1 if a direct route is chosen, or the arrival time of flight i at WPT2 if the indirect route is selected.Variable t F IM (SU T IE,i) refers to the arrival time of flight i at SUTIE.The minimum travel time between WPT1 to SUTIE is defined to be 290 seconds.When R 4 is activated, the minimum travel time between WPT2 to SUTIE is set to 220 seconds.       t F IM (F IM,i) = t i0 + d1 i F IM (W P T,i) = t F IM (F IM,i) + [(1 -r i ) • L R 3 + r i • L R 4 ]/v i t F IM (SU T IE,i) = t F IM (W P T,i) + (1 -r i )•(d2 i + 290) + r i •(d2 i + 220)(7)Eqn. 8 shows the expression for departures, where t DEP (RW Y,j) represents the time flight j departs from RWY, variable t DEP (W P T 2,j) denotes the time flight j arrives at WPT2, and R 1 refers to the route RWY-WPT2.Variable t DEP (W P T 1,j) represents the time flight j arrives at WPT1.       t DEP (RW Y,j) = t i0 + d j t DEP (W P T 2,j) = t DEP (RW Y,j) + (1 -r j ) • L R 1 /v j t DEP (W P T 1,j) = t DEP (RW Y,j) + [(1 -r j ) • L R1 + r j • L R2 ]/v j(8)Eqn. 9 presents the expression for A 2 arrivals with simply one decision variable.t SU T IE (SU T IE,k) = t i0 + d k(9)Separation constraints were applied at fixes that could have potential violations, such as FIM, RWY, WPT1, WPT2, and SUTIE.Separation requirements were 3 nmi at the runway and 4 nmi elsewhere.As in the previous section, an uncertainty buffer of δ was added in the separation constraints for a sensitivity study. +ObjectiveThe objective is to minimize the sum of exit times, as shown in Eqn.10.For departures it is the time when a flight leaves the waypoint WPT1.For arrivals, it is the time when a flight reaches waypoint SUTIE.J = i,j,k t F IM (SU T IE,i) + t DEP (W P T 1,j) + t SU T IE (SU T IE,k)(10) +C. ResultsIn this section, results using three different separation methods are presented and compared.Different buffers are set up to deal with uncertainty and the impacts are studied.In addition, two scenarios are defined: only delays are considered in the first one, while early arrivals are allowed in the second case.For any flight, its unimpeded flight time (fly via direct route without any consideration of separation from other flights) is treated as a baseline.Beyond that, any extra flight time will be called delay. +Comparison of separation methodsThis section compares separation methods when uncertainty buffers were defined to be zero.In the case of spatial separation, because the indirect route was the only option, there exists associated extra flight time of 771 seconds.Because all flights were assumed to be the same aircraft type, the total delay can be computed manually.Including route-caused delay the total delay is 1,001 seconds as shown in the Table 4.It can be seen that without uncertainty buffers the delay with hybrid separation was 357 seconds -a reduction of 64% or 10.7 minutes compared to the 1,001 second delay with spatial separation.While in this case, the temporal separation also achieved much less delay than the spatial separation.The reduction was 59% or 9.8 minutes.Tables 5, 6 and 7 show individual flight results for sets A 1 , D 1 , and A 2 , respectively, when the hybrid separation is applied.Table 5 provides the results for arrivals from FIM.It indicates that three of five arrival flights can make use of the direct route to reduce overall delay.In Table 6, all four departures flew their direct routes.Among the arrivals from the East, two flights were assigned some delays.The resulting delay under hybrid separation was reduced to 357 seconds as shown in Table 4. +Impact of uncertaintyThe results of schedulers could be sensitive to the uncertainty of flight times.Robustness is required in actual operations.The easy and popular way to increase the robustness is to introduce an uncertainty buffer for flight times.As a trade-off, adding buffers causes additional delays.In this study, the buffers of 30 and 60 seconds were applied.Table 4 shows the results.In the table, the temporal separation introduced much less delay than the spatial separation in the deterministic case, but when the uncertainty buffer increased to 60 seconds, the temporal separation caused more delay than spatial separation.This showed that temporal separation was sensitive to the uncertainty buffer and corresponding schedules might be undesired in actual operations.The hybrid approach generated the least delay compared with the other two approaches, although the reduction decreased to 183 seconds when the buffer was 60 seconds.Results in Table 4 show the trade-off between buffer size and delays.In order to find out the best balance, an uncertainty study is required for future work.and6 show the time lines for all metering points when the buffer is zero and 30 seconds, respectively, where the hybrid approach was applied.SUT stands for SUTIE for simplicity.Each flight has a safe zone shown as a grey box in front of its arrival time.These safe zones can be packed in Fig. 5, but they are well separated by a 30 second buffer in Fig. 6.Note how flight loading of WPT1 and WPT2 changes significantly between 0 and 30 seconds.When the buffer is zero, FIM001, FIM004, and FIM005 are proposed to fly the direct route, but when the buffer increases to 30 seconds, the other FIM arrivals -FIM002 and FIM003 -are proposed to take the short cut. +Impact of early arrivalsAllowing early arrival or speeding up increases flexibility and reduces delay.In the previous cases, no early arrival or speeding up was allowed.In this section, early arrival is allowed up to 30 seconds.The results are presented in Table 8.For spatial separation, the delay was simply calculated by shifting each aircraft 30 seconds earlier than the previous case.The temporal and hybrid separations were solved using the NSGA algorithm as in previous sections.As shown in the table, when the uncertainty buffer was zero, total delays caused by both temporal and hybrid separation were negative, which means on average aircraft arrived early.The delay reduction from spatial separation to hybrid separation was 12.5 minutes when the uncertainty buffer is zero and the reduction decreased to 3 minutes when the buffer increased to 60 seconds.On the other hand, the uncertainty buffer in flight time still plays an important role.When the buffer increased, the delay required increased quickly, especially for the temporal separation.The performance of the NSGA algorithm is compared with that of a FCFS based heuristic method.Because this problem involves route option and multiple scheduling points, the rules of FCFS may not be straightforward and have to be clarified:• The estimated or scheduled entering times are used as references for setting up priority.For FIM arrivals, the entering times are the arrival times to fix FIM.For departures, they are the estimated times of leaving RWY.And the estimated SUTIE arrival times are used as references for arrivals from the East.Each flight decides its route based on the FCFS rule in the order of their entering times.The route that causes the lowest delay at the time will be chosen.• The conflicting flights in metering points WPT1, WPT2, and SUTIE are resolved based on the estimated arrival times, not their entering times.For instance, assume flight A has an earlier entering time than flight B, but the estimated arrival time of flight A at fix F is later than flight B. If flight A would conflict with flight B at fix F, then flight A would be delayed before it reaches fix F. With this rule, the arrival sequences at these metering points are actually allowed to be changed.This may be different from the strict FCFS rule.• No flight should be delayed more than M seconds at any fix.M is set to 200 in this work.• The order of flights in the same flow should be kept, e.g.FIM arrivals, SUTIE arrivals, and departures.In the NSGA algorithm, the resolution of delay was approximately five to ten seconds.Table 9 presents the results using the heuristic method with a delay resolution of one second.When the uncertainty buffer is zero, the results generated by the NSGA algorithm can save about 42% over the heuristics method even though the latter has high resolution.When the buffer size increases, that saving was reduced to 20% or 9%, which was probably due to the decreased solution space.Overall, the proposed NSGA algorithm outperformed the heuristic method with a great difference.Unlike conventional scheduling problems, the optimization method showed greater advantage over heuristics due to the complicated solution space.On the other hand, it is also noticed that when the buffer size is large enough, the benefit of integration of arrivals and departures could disappear and the heuristic method becomes a good choice due to its computational easiness. +V. ConclusionThe integration of departures and arrivals seems promising in improving operational efficiency in terminal airspace.The problem combines routing and scheduling problems, which further challenges the MILP solver.This work introduced a variation of genetic algorithm -NSGA.The NSGA was used because it is better than basic GA in handling constraints.Results with a test problem showed that the new formulation with NSGA can solve the problem in a fast time fashion.A potential application of integrated arrivals and departures was identified in LAX terminal airspace.The arrival and departure route structures were modelled with three different strategies: spatial separation, temporal separation, and hybrid separation.A problem was set up based on a historical traffic schedule with a total of 15 flights and three flows included.The results showed that although the temporal separation introduced much less delay than the spatial separation, it caused more delay than the latter when the uncertainty buffer increased to 60 seconds.The hybrid separation outperformed both the temporal and the spatial separations: It reduced unnecessary delay by 64% or 10.7 minutes if no early arrival/departure or speeding up was allowed; And if early arrival/departure or speeding up was allowed, the saving increased to 12.5 minutes.Compared with a FCFS based heuristic method, the schedules produced by the NSGA saved flight time up to 42%, which showed greater advantage over FCFS than typically seen in conventional scheduling problems.Overall, this study showed that it is promising to improve operation efficiency in LAX terminal airspace by integrating departures and arrivals using hybrid separation with the NSGA algorithm.Apparently, such efficiency may vary with aircraft departure and arrival schedules.The proposed method can be applied in a fast time fashion to decide if benefit exists and how to quantify it.Therefore, it can help decision makers to operate properly.In order to achieve this goal, an analysis needs to be completed for the uncertainty in the schedules.In the future work, such uncertainty analysis will be conducted.The robustness of the benefits and controllers' workload will be examined by imposing flight time perturbations.The balancing point for the trade-off between robustness and delay will be studied.The schedulers using different separation approaches and different algorithms will be investigated and compared.Figure 1 .1Figure 1.NSGA II selection procedure +4 Figure 2 .42Figure 2. Shared departure fixes for SJC and OAK +Figure 3 .3Figure 3. Schedules for (a) segregated departure fixes and (b) shared departure fixes +Figure 4 .4Figure 4. Interactions between SADDE arrivals and CASTA departures +Figure 5 .5Figure 5.Time Line with hybrid separation and buffer = 0 seconds +Figure 6 .6Figure 6.Time Line with hybrid separation and buffer = 30 seconds +Table 1 .1Scheduled departure time +Table 2 .2Comparison of delaysGround delay(sec) Airborne delay (sec) Total delay (sec) Makespan (sec)Segregated645326772900Shared1291082372900OAKRWYOAKW1W0SJC0 SJCRWY50010001500200025003000Time (second) +Table 3 .3Scheduled arrival timesOrder FIM (sec) RWY (sec) SUITE (sec)103043021352986713263540107048601240121051230NA13766NANA1780 +Table 4 .4Total delay with different separation methodsuncertainty buffer Spatial Temporal Hybrid0 s1,001s413 s357 ss1,163 s776 s759 s60 s1,673 s1,808 s1,490 s +Table 5 .5Results for A 1 arrivals with hybrid separation and zero uncertaintyA 1 Arrivals delay at FIM (s)RouteSpeed (kt) delay at SUITE (s)Flight 10direct34939Flight 268indirect34415Flight 30indirect31912Flight 40direct3500Flight 50direct3500 +Table 6 .6Results for departures (D 1 ) with hybrid separation and zero uncertaintyDepartures delay before departure (s) Route Speed (kt)Flight 10direct250Flight 20direct250Flight 38direct250Flight 40direct250 +Table 7 .7Results for A 2 arrivals with hybrid separation and zero uncertaintyA 2 Arrivals delay before SUITE (s)Flight 10Flight 20Flight 38Flight 40Flight 562Flight 64 +Table 8 .8Total delay when early arrival is alloweduncertainty buffer Spatial Temporal Hybrid0 s551 s-170 s-195 s30 s713 s393 s309 s60 s1,223 s1,338 s1,041 s4. NSGA algorithm v.s Heuristic method +Table 9 .9Delay with hybrid separation using different methodsuncertainty buffer NSGA algorithm Heuristic Difference0 s357 s611 s42%30 s758 s950 s20%60 s1,490 s1,638 s9% + + + + + + + + + The dynamic scheduling of aircraft in the near terminal area + + RGDear + + 10.1016/0041-1647(78)90133-8 + R76-9 + + + Transportation Research + Transportation Research + 0041-1647 + + 12 + 3 + + September 1976 + Elsevier BV + + + Tech. Rep. Flight Transportation Laboratory Report + Dear, R. G., "The Dynamic Scheduling of Aircraft in the Near Terminal Area," Tech. Rep. Flight Transportation Laboratory Report R76-9, MIT, September 1976. + + + + + An algorithm for computer assisted sequencing and scheduling of terminal area operations + + RogerGDear + + + YosefSSherif + + 10.1016/0191-2607(91)90132-a + + + Transportation Research Part A: General + Transportation Research Part A: General + 0191-2607 + + 25 + 2-3 + + 1991 + Elsevier BV + + + Dear, R. G. and Sherif, Y. S., "An Algorithm For Computer Assisted Sequencing and Scheduling of Terminal Operations," Transportation Research A, Vol. 25A, No. 2/3, 1991. + + + + + Analysis of Delay Reducing and Fuel Saving Sequencing and Spacing Algorithms for Arrival Spacing + + FNeuman + + + HErzberger + + 103880 + + October 1991 + + + Tech. Rep. NASA Technical Memorandum + Neuman, F. and Erzberger, H., "Analysis of Delay Reducing and Fuel Saving Sequencing and Spacing Algorithms for Arrival Spacing," Tech. Rep. NASA Technical Memorandum 103880, October 1991. + + + + + Scheduling Aircraft Landings Under Constrained Position Shifting + + HamsaBalakrishnan + + + BalaChandran + + 10.2514/6.2006-6320 + + + AIAA Guidance, Navigation, and Control Conference and Exhibit + Keystone, CO + + American Institute of Aeronautics and Astronautics + August 2006 + + + + Balakrishnan, H. and Chandran, B., "Scheduling Aircraft Landings Under Constrained Position Shifting," AIAA Guid- ance, Navigation, and Control Conference and Exhibit, Keystone, CO, 21-24 August 2006. + + + + + Scheduling Aircraft Landings—The Static Case + + JEBeasley + + + MKrishnamoorthy + + + YMSharaiha + + + DAbramson + + 10.1287/trsc.34.2.180.12302 + + + Transportation Science + Transportation Science + 0041-1655 + 1526-5447 + + 34 + 2 + + 2000 + Institute for Operations Research and the Management Sciences (INFORMS) + + + Beasley, J. E., Krishnamoorthy, M., Sharaiha, Y. M., and Abramson, D., "Scheduling Aircraft Landings -The Static Case," Transportation Science, Vol. 34, No. 2, 2000. + + + + + Scheduling Aircraft Landings to Closely Spaced Parallel Runways + + MKupfer + + + + The Eighth USA/Europe Air Traffic Management Research and Development Seminar + Napa, CA + + June 2009 + + + Kupfer, M., "Scheduling Aircraft Landings to Closely Spaced Parallel Runways," The Eighth USA/Europe Air Traffic Management Research and Development Seminar , Napa, CA, June 2009. + + + + + Towards Optimal Routing and Scheduling of Metroplex Operations + + BrianCapozzi + + + StephenAtkins + + + SeongimChoi + + 10.2514/6.2009-7037 + + + 9th AIAA Aviation Technology, Integration, and Operations Conference (ATIO) + Hilton Head, South Carolina + + American Institute of Aeronautics and Astronautics + September 2009 + + + + Capozzi, B. J. and Atkins, S. C., "Towards Optimal Routing and Scheduling of Metroplex Operations," 9th AIAA Aviation Technology, Integration, and Operations Conference(ATIO), Hilton Head, South Carolina, 21-23 September 2009. + + + + + A Hybrid Optimization Approach to Air Traffic Management for Metroplex Operations + + BrianCapozzi + + + StephenAtkins + + 10.2514/6.2010-9062 + + + 10th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + + American Institute of Aeronautics and Astronautics + + + + Capozzi, B. J. and Atkins, S. C., "A Hybrid Optimization Approach to Air Traffic Management for Metroplex Operations," + + + + + Departure Efficiency Benefits of Terminal RNAV Operations at Dallas-Fort Worth International Airport + + RalfMayer + + 10.2514/6.2006-7774 + + + 6th AIAA Aviation Technology, Integration and Operations Conference (ATIO) + Fort Worth, Texas + + American Institute of Aeronautics and Astronautics + September 2010 + + + + AIAA Aviation Technology, Integration and Operations Conference (ATIO), Fort Worth, Texas, 13-15 September 2010. + + + + + Genetic Algorithms + Data Structures = Evolution Programs + + ZMichaelwicz + + 10.1007/978-3-662-03315-9 + + 1996 + Springer Berlin Heidelberg + Berlin Heidelberg + + + Third Edition + Michaelwicz, Z., Genetic Algorithms + Data Structures = Evolution Programs, Third Edition, Springer-Verlag Berlin Heidelberg, 1996. + + + + + A fast and elitist multiobjective genetic algorithm: NSGA-II + + KDeb + + + APratap + + + SAgarwal + + + TMeyarivan + + 10.1109/4235.996017 + + + IEEE Transactions on Evolutionary Computation + IEEE Trans. Evol. Computat. + 1089-778X + + 6 + 2 + + 2002 + Institute of Electrical and Electronics Engineers (IEEE) + + + Deb, K., Pretap, A., Agarwal, S., and Meyarivan, T., "A Fast and Elitist Multiobjective Genetic Algorithm: NSGA-II," IEEE Transactions On Evolutionary Computation, Vol. 6, No. 2, 2002. + + + + + Muiltiobjective Optimization Using Nondominated Sorting in Genetic Algorithms + + NSrinivas + + + KalyanmoyDeb + + 10.1162/evco.1994.2.3.221 + + + Evolutionary Computation + Evolutionary Computation + 1063-6560 + 1530-9304 + + 2 + 3 + + 1995 + MIT Press - Journals + + + Srinivas, N. and Deb, K., "Multi-Objective Function Optimization using Non-dominated Sorting Genetic Algorithms," Evolutionary Computation, Vol. 2, No. 3, 1995. + + + + + Super Density Operations Airspace Modeling for the Southern California Metroplex + + SebastianTimar + + + GauravNagle + + + AdityaSaraf + + + PeterYu + + + PeterHunt + + + AndrewTrapani + + + NickJohnson + + 10.2514/6.2011-6535 + + + AIAA Modeling and Simulation Technologies Conference + Portland, Oregon + + American Institute of Aeronautics and Astronautics + August 2011 + + + + Timar, S. D., Nagle, G., Saraf, A., Yu, P., Hunter, P., Trapani, A., and Johnson, N., "Super Density Operations Airspace Modeling for the Southern California Metroplex," AIAA Modeling and Simulation Technology Conference, Portland, Oregon, 08-11 August 2011. + + + + + + diff --git a/file820.txt b/file820.txt new file mode 100644 index 0000000000000000000000000000000000000000..038f22eb3e98cdece14823471bd8630c1db27098 --- /dev/null +++ b/file820.txt @@ -0,0 +1,434 @@ + + + + +I. IntroductionIn the National Airspace System (NAS) terminal airspace is often busy and complicated because hundreds of flights must fly through a limited airspace in a short time period and most flights in terminal areas are climbing and descending with varied speeds.Situations can become severe when there are several neighboring busy airports.Bottlenecks can be formed easily and impair the efficiency of air traffic operations.Therefore, improving the operation efficiency in terminal airspace is critical for building an efficient air traffic system.2][3][4][5][6] Still others [7][8][9] addressed airport surface management problems, which are correlated to the efficiency of terminal airspace operations.When different arrival and/or departure flows share the same resources such as runways, waypoints and/or route segments, inefficient operations emerge because of the constraints of shared resources.Such interactions can happen among departures, arrivals, or between departures and arrivals.Recent studies [10][11][12] showed that optimized integrated arrivals and/or departures in major airports or metroplex areas have promise for improving operation efficiency.However, the benefits from optimal schedules calculated under deterministic scenarios are usually sensitive to flight time uncertainties, which can be caused by many sources, such as inaccurate wind prediction, error in aircraft dynamics, or human factors.Therefore, the impact of uncertainty must be taken into account when evaluating a schedule's benefits.Incorporating stochastic evaluation as a post-analysis process of deterministic optimization as in previous work 13 is one passive way to learn the impact of uncertainty and to avoid unexpected results.Optimizing integrated arrivals and departures under uncertainty is a proactive method that directly takes uncertainty into account.A set of compromise solutions can be identified and offered directly to decision makers to avoid unexpected effects from uncertainty.This work presents an optimization method of integrating arrivals and departures under uncertainty, in which the impact of uncertainty was measured directly in the optimization.The optimization is multi-objective including total delay and controller intervention count.Both costs were evaluated using Monte-Carlo simulations.To enable the time-consuming optimization, the number of Monte-Carlo simulations were decreased.The impact of simplified Monte-Carlo simulation was examined.The Pareto fronts that contain sets of solutions were presented to show the trade-off between delays and interventions.Solutions with similar delays but different controller intervention count were investigated further.In the paper, Section II revisits the problem modelled in previous work.Section II also presents the multiple objective optimization method that involves Monte Carlo simulations.Section III provides the results and analysis.Section IV provides conclusions for this work. +II. Background and ProblemThe interactions between arrivals and departures in Los Angeles terminal airspace were presented in previous work 12 for studying optimal integrated operations in deterministic circumstances.The uncertainty analysis of the determinstic solutions has been conducted in other work 13 as determined from post-processing.This work is focused on the same problem of integrating arrivals and departures in Los Angeles.According to the Standard Terminal Arrival Routes (STARs) and the Standard Instrument Departures (SIDs) of Los Angeles terminal airspace, the arrivals from the FIM fix would follow procedure SADDE6 ( FIM-SYMON-SADDE-SMO) and the departures to the North need to follow procedure CASTA2 (RWY-NAANC-GHART-SILEX) (see Fig. 1).The arrivals are requested to maintain their flight altitudes above 12,000 feet at fix GHART and the departures have to keep theirs at or below 9,000 feet at the same fix to procedurally avoid potential conflicts between arrivals and departures.If there were no interactions, departures to the north and arrivals from FIM would have flown direct routes.As shown in the Fig. 1, the direct routes would be RWY-WPT2-WPT1 and FIM-WPT1-SMO for departures and arrivals, respectively, where WPT1 and WPT2 are made-up fix names for simplicity.Compared to these direct routes, besides flying non-preferred altitudes, individual arrival and departure flights following current procedures will waste approximately 60 and 120 seconds, respectively.Table 1 shows a representative schedule of 14 flights, which covers half an hour traffic in actual operations.Two flows are included: 6 departures to the North from Runway 24L (RWY) and 8 arrivals from FIM.This schedule used the traffic between 9:00am and 9:30am (local time) on December 4, 2012 as reference.The initial times shown in the table are relative times to simulation start time.The "Order" of each flight is sorted based on intial times. +III. MethodIn this study, the model is constructed similarly to previous works. 12,13 nlike pervious work, the optimization is formulated as a multiple objective optimization.The first objective is to minimize delay and the second objective is to minimize controller intervention, which is believed to associate with controller workload increase.Both objectives were evaluated using Monte Carlo simulations. +A. ModelThree separation methods were compared in previous work: 12 spatial, temporal, and hybrid.Spatial separation uses the same strategy as in SIDs and STARs to spatially separate interacting departure and/or arrival flows.Temporal separation utilizes the direct routes with conflicts resolved merely with temporal control.Hybrid separation applies both temporal and spatial separations.This work was focused on hybrid separation.In the formulation of hybrid separation, four design variables were defined for each FIM arrival i: d1 (F IM,i) is the delay before or at FIM; r (F IM,i) is for route option, where 0 denotes the direct route and 1 denotes indirect route; v (F IM,i) is the aircraft speed between FIM and WPT1 for the direct route or the speed between FIM and WPT2 if the indirect route is chosen; d2 (F IM,i) is the delay between WPT1/WPT2 and SUTIE to ensure separation at SUTIE.For a departure flight j, three decision variables were defined: d (DEP,j) is the delay before departure; r (DEP,j) is the route option, where 0 denotes the direct route and 1 denotes the indirect route.v (DEP,j) is the speed from departure to WPT1.Separation requirements were applied as hard constraints at fixes that could have potential violations, such as FIM, RWY, WPT1, WPT2, and SUTIE.Separation requirements were 3 nmi at the RWY and 4 nmi elsewhere.Additionally, an uncertainty buffer δ was added in the separation constraints.For example, Eqns. 1, 2, and 3 show separation constraints for crossing flights between FIM arrivals and departures at WPT1, where t F IM (W P T 1,i) is the arrival time of FIM arrivals at fix WPT1 and t DEP (W P T 1,j) is the arrival time of DEP departures at fix WPT1.Eqn. 1 showed that if both arrival flight i and departure flight j took direct routes the separation must be satisfied.The separation requirement is 4.0 nmi in distance and in time scale the separation depends on the speeds of both flights.(1 -r (F IM,i) )•(1 -r (DEP,j) )•[t F IM (W P T 1,i) -t DEP (W P T 1,j) ] - 4.0×3600.0 V •sinα -δ > 0 (1) V =    v (F IM,i) , if t F IM (W P T 1,i) < t DEP (W P T 1,j) v (DEP,j) , otherwise(2)α =    atan[ v (DEP,j) v (F IM,i) ], if t F IM (W P T 1,i) < t DEP (W P T 1,j) atan[ v (F IM,i) v (DEP,j) ], otherwise(3) +B. ObjectivesThe optimization in previous work was deterministic with a single objective.The objective was to minimize the total delay of all departures and arrivals.In this work, the problem is formulated as a multiple objective optimization, where controller intervention is considered as another objective.The objectives are shown in Eqn. 4, where N is the controller intervention count.   J 1 = i,j,k t F IM (SU T IE,i) + t DEP (W P T 1,j) J 2 = i,j,k N (4)Delay cost J 1 is equivalent to the sum of flight's exit times.For arrivals and departures, they are the times when flights reach SUTIE and WPT1, respectively.The difference from previous study 12 is that the delay in this work is evaluated stochastically using Monte Carlo simulations instead of evaluating deterministically.In Monte Carlo simulations, errors that follow normal distributions are introduced in flights' entry times.For a FIM arrival, the perturbation is introduced into the estimated arrival time at FIM.For a westbound arrival, the error is added to the estimated arrival time at SUTIE.Whereas for a departure, such error is imposed on the estimated takeoff time from the runway.As in previous works, 12,13 a heuristic model was built to mimic controller intervention behaviors.When stochastic errors are added in flights' entry times, the heuristic model resolves potential conflicts by imposing extra delays to corresponding aircraft, while keeping the same route options as in the given solution.The extra delays are then propagated to flights' next waypoints, and resolutions would happen in chronological order.Meanwhile, the controller intervention count is accumulated.If there was no perturbation in entry times like in deterministic cases, no extra delay and controller intervention should be imposed.In each simulation, the heuristic model is called to check if there is any extra delay and controller intervention.With Monte Carlo simulations, statistics can be calculated so the objectives are evaluated stochastically.The optimization is implemented using a Non-dominated Sorting Genetic Algorithm (NSGA), 14 because of NSGA's ability of handling multiple objective optimization.Because the costs are handled independently, NSGA will lead its search to a Pareto front, where no solution on the front is dominated by another.The dominance with two objectives is defined as: Assuming the objective is to minimize costs and the constraint function g has to be nonnegative, solution A is dominated by solution B only if:       J 1 (A) > J 1 (B), if g A ≥0 and g B ≥0, or, g A = g B J 2 (A) > J 2 (B), if g A ≥0 and g B ≥0, or, g A = g B g A < g B ,if g A < 0 and g B < 0, or, g B > 0 and g A < 0(5)where J 1 and J 2 are the objectives and g is the constraint value.In other words, if all three conditions are true, solution A cannot be at the Pareto front. +C. Simplified Monte Carlo simulationsDirectly incorporating a large scale set of Monte Carlo simulations (like 5,000 simulations) into objective evaluation would be computationally expensive and make the optimization infeasible.Therefore, the number of simulations must be decreased.However, the robustness of stochastic cost evaluation might decrease together with the number of simulations due to the reduced sampling size.For instance, the mean values of extra delay and controller intervention will fluctuate more in the simplified Monte Carlo than in the full-scale Monte Carlo.The inconsistent evaluations make the generated costs inaccurate and may result in incorrect ranking.For instance, solution A is supposed to have a higher rank than solution B because of its low cost, but the opposite situation happened due to the inconsistent evaluations.The more fluctuated the evaluation is, the higher likelihood the ranking is incorrect.Therefore, the number of simulations shouldn't be too low, otherwise, the NSGA's searching process will be misled and good solutions can not be achieved.In this work the number of simulations was reduced to 1,000, which seemed to be a good compromise between quality and computational time.Figure 2(a) shows the differences in controller intervention between 1,000 and 5,000 simulations during optimization process for around 25,000 solutions.The horizontal axis is the order of each simulation and the vertical axis denotes the differences measured based on the number of controller interventions.The average of intervention errors is 0.19 and the standard deviation is 0.065.The controller intervention counts are averaged at 2.0, therefore, the intervention errors between 1,000 and 5,000 simulations are about 9% in terms of average value.Figure 2(b) shows the differences in extra delays between 1,000 and 5,000 simulations.The average of errors is about 2.9 seconds and standard deviation is 1.8 seconds.Given the average delay of 542 seconds, the difference is negligible. +IV. ResultsThis section presents the solutions optimized under uncertainty.The trade-offs between delay reduction and controller intervention count increase are shown.The solutions that have similar delays and different intervention counts are compared.Both costs were evaluated stochastically.Monte Carlo simulations were multi-threaded and the optimization took about 6 hours on a MacOS platform with 2x2.66 GHz 6-Core Intel Xeon and 8 GB RAM. +A. The Pareto frontGiven two objectives of reducing delays and controller intervention counts, NSGA led the search to a Pareto front, where no solution on the front was dominated by other solutions.Solutions in final generations of the evolution process were recorded as they were close to the Pareto front.Figure 3 presents these solutions in final generations ( 1,000+ solutions).Each dot (J 1 , J 2 ) corresponds to a solution and its coordinates J 1 and J 2 are the two costs of the solution.Both costs were evaluated stochastically using simplified Monte Carlo simulation.The red diamond at coordinates (1227.0,1.2) sets a reference point for intervention counts and delay costs under spatial separation for which a low intervention count is expected.The costs associated with the reference point were generated as averages from full-scale Monte Carlo simulations of the optimal spatial separation solution, which was produced under deterministic scenarios with delay as the single objectives.The horizontal red dotted line marks the red diamond's intervention cost which serves as a reference representing low controller intervention.Similarly, the black dotted line at 300 seconds serves as a reference.These dotted lines and diamond solution were used as references for clarity throughout figures in this section.The accuracy of the costs of solutions in Fig. 3 needs to be examined because simplified Monte Carlo simulations were used for evaluation.Therefore, in the post-process, the costs of the same solutions were re-evaluated using full-scale Monte Carlo simulations.The rectified costs are shown in Fig. 4. Using the vertical dotted line as a reference, it can be seen that the solutions in Fig. 3 were shifted up and right in Fig. 4 because of the impact of the robustness imposed on delays and controller intervention.But the marginal difference demonstrated that simplified Monte Carlo simulation works well.GA-based optimizations are sensitive to initial guesses, which are decided by randomized seeds.Multiple runs are necessary to increase the chance of getting optimal solutions.In this work, ten runs were performed.Figure 5 shows the solutions in final generations for each run.It is noted that different initial guesses led to different solutions in spite of NSGA's effort, especially when the average delay is low.Figure 6 presents the rectified costs using full-scale simulations in a post-process.As expected, most solutions were shifted up and right.According to the final solutions, it is noted that a clear tradeoff exists between delays and intervention counts.When delay is reduced the chance of controller intervention increases.Variations of intervention counts are large for solutions with similar delays, especially when delays are low.A multipleobjective optimization that incorporates uncertainty can clearly help to find the reduced delay with minimum intervention.If a decision support tool can be built upon this method, in terms of the solutions that are close to the Pareto front, decision makers can choose a "compromise" solution according to their preferences.For instance, if controller intervention is believed to be more important than delay savings, the solutions that have a delay of about 900 seconds and intervention count similar to the spatial solution (red diamond solution) should be picked.Or, if two costs are weighted similarly, the compromise solutions around the middle (e.g.delay around 500) can be chosen because of their large delay savings and tolerable controller intervention increase. +B. Effect of controller interventionAccording to Fig. 6, solutions that shared similar delays may cause quiet different controller intervention counts.For instance, at delays around 430 seconds, the average intervention can vary from 2.4 to 3.5.Figures 7 and8 are the time lines for solutions at either end of this intervention count range.The small gray boxes are the minimum separation requirements associated with flights.As mentioned previously, the separation requirement is a function of aircraft speed and the type of potential conflicts (crossing or in-trail).Therefore, the gray boxes have different lengths.As defined, the boxes are separated by an extra 30 second buffer.Any departure following the direct route should go through WPT2, whereas any arrival from the FIM fix flying the direct route would pass WPT1.In terms of costs, these two solutions shared similar delays at around 423 seconds but resulted in different controller intervention efforts.The solution associated with Fig. 8 has an average intervention count of 2.45, whereas the Fig. 7 solution corresponds an average intervention count of 3.56 -a 45 % increase.Based on the comparison, the major differences between them are that: The high-controller-intervention solution allowed FIM003 and FIM008 to fly direct routes whereas the low-controller-intervention solution chose FIM007 and DEP006 to fly direct routes (See the flights pointed by arrows in the figures).Those differences in route options would not affect the delay reduction but would lead to different controller intervention.Apparently, this subtle difference could not be easily predicted without the optimization under uncertainty. +C. DiscussionThe proposed method presents a set of Pareto solutions to decision makers/air traffic controllers and can help them quickly identify potential benefits and trade-offs for integrated arrival and departure operations.Although this work is focused on LAX, the methodology should be easily applied to different major airports where similar inefficiency exists.In this work, it took about 6 hours to finish the optimization process, which seemed to be a problem for real-time application.But because both Monte Carlo simulations and the NSGA algorithm are well-suited for parallelization due to their independent calculations and memories, the optimization method proposed in this paper has the potential to be sped up by hundreds of times using GPU implementation 15 and thus may be a candidate for a real-time/fast-time tool.In addition, the searching ranges of decision variables may be smaller in actual operations.For example, the ranges of delays for arrivals are connected to certain traffic control actions like vectoring, holding, and slowing down.Instead of continuous decision variables as in this work, discrete variables that are associated with smaller searching spaces may be used in actual operations.This will help speed up the optimization process as well. +V. ConclusionsIn order to directly take the impact of uncertainty into account, an optimization method of integrated arrivals and departures under uncertainty was developed in this study.The impact to controller intervention was included in the optimization.The problem was formulated as a multiple objective optimization with delays and intervention count as the costs.Monte Carlo simulations were utilized to stochastically evaluate both costs.To enable the time-consuming optimization, the number of Monte Carlo simulations were reduced.The Pareto fronts that contain sets of solutions were identified.The trade-off between delays and controller intervention counts was shown and solutions with similar delays but different intervention effort were investigated.Through this study, the optimization under uncertainty for integrated arrivals and departures were found to be feasible with simplified Monte Carlo simulations.Decreasing the number of simulations from 5,000 to 1,000 affected the controller intervention evaluation 9% but reduced the computational times to a reasonable level.Using this formulation, the method can provide a sweep of solutions that are close to the Pareto front so the decision makers can choose in terms of their preferences.Solutions may have similar delays but cause quite different controller intervention efforts.The subtle difference in solutions/delays can result in a significant difference in intervention counts (e.g.45%), which would not be easily foreseen without the optimization under uncertainty.As the Monte Carlo simulations and the NSGA algorithm are all wellsuited for parallelization because of their independent calculations and memories, the proposed optimization algorithm is expected to be sped up by hundreds of times using GPU implementation and becomes a realtime/fast-time application.Figure 1 .1Figure 1.Interactions between SADDE arrivals and CASTA departures +Figure 2 .2Figure 2. Differences in computing 1,000 simulations vs. 5,000 simulations: (a) Controller intervention count (b) Delay +Figure 3 .Figure 4 .34Figure 3. Costs of optimized solutions from a single run where costs were evaluated using simplified Monte Carlo simulations +Figure 5 .Figure 6 .56Figure 5. Costs of optimized solutions from ten runs where costs were evaluated using simplified Monte Carlo simulations +Figure 7 .Figure 8 .78Figure 7. Time line for the solution with high controller intervention count +Table 1 .1Scheduled initial timesOrder FIM (sec) RWY (sec)13968244616537283634110652951332161361475183071613NA81770NA + of 10 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on August 14, 2013 | http://arc.aiaa.org| DOI: 10.2514/6.2013-4322Thismaterial is declared a work of the U.S. Government and is not subject to copyright protection in the United States. + of 10 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on August 14, 2013 | http://arc.aiaa.org| DOI: 10.2514/6.2013-4322Thismaterial is declared a work of the U.S. Government and is not subject to copyright protection in the United States. + of 10 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on August 14, 2013 | http://arc.aiaa.org| DOI: 10.2514/6.2013-4322Thismaterial is declared a work of the U.S. Government and is not subject to copyright protection in the United States. + of 10 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on August 14, 2013 | http://arc.aiaa.org| DOI: 10.2514/6.2013-4322Thismaterial is declared a work of the U.S. Government and is not subject to copyright protection in the United States. + of 10 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on August 14, 2013 | http://arc.aiaa.org| DOI: 10.2514/6.2013-4322Thismaterial is declared a work of the U.S. Government and is not subject to copyright protection in the United States. + of 10 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on August 14, 2013 | http://arc.aiaa.org| DOI: 10.2514/6.2013-4322Thismaterial is declared a work of the U.S. Government and is not subject to copyright protection in the United States. + + + + + + + + + The dynamic scheduling of aircraft in the near terminal area + + RGDear + + 10.1016/0041-1647(78)90133-8 + R76-9 + + + Transportation Research + Transportation Research + 0041-1647 + + 12 + 3 + + September 1976 + Elsevier BV + + + Tech. Rep. Flight Transportation Laboratory Report + Dear, R. G., "The Dynamic Scheduling of Aircraft in the Near Terminal Area," Tech. Rep. Flight Transportation Laboratory Report R76-9, MIT, September 1976. + + + + + An algorithm for computer assisted sequencing and scheduling of terminal area operations + + RogerGDear + + + YosefSSherif + + 10.1016/0191-2607(91)90132-a + + + Transportation Research Part A: General + Transportation Research Part A: General + 0191-2607 + + 25 + 2-3 + + 1991 + Elsevier BV + + + Dear, R. G. and Sherif, Y. S., "An Algorithm For Computer Assisted Sequencing and Scheduling of Terminal Operations," Transportation Research A, Vol. 25A, No. 2/3, 1991. + + + + + Analysis of Delay Reducing and Fuel Saving Sequencing and Spacing Algorithms for Arrival Spacing + + FNeuman + + + HErzberger + + 103880 + + October 1991 + + + Tech. Rep. NASA Technical Memorandum + Neuman, F. and Erzberger, H., "Analysis of Delay Reducing and Fuel Saving Sequencing and Spacing Algorithms for Arrival Spacing," Tech. Rep. NASA Technical Memorandum 103880, October 1991. + + + + + Scheduling Aircraft Landings Under Constrained Position Shifting + + HamsaBalakrishnan + + + BalaChandran + + 10.2514/6.2006-6320 + + + AIAA Guidance, Navigation, and Control Conference and Exhibit + Keystone, CO + + American Institute of Aeronautics and Astronautics + August 2006 + + + + Balakrishnan, H. and Chandran, B., "Scheduling Aircraft Landings Under Constrained Position Shifting," AIAA Guid- ance, Navigation, and Control Conference, Keystone, CO, 21-24 August 2006. + + + + + Scheduling Aircraft Landings—The Static Case + + JEBeasley + + + MKrishnamoorthy + + + YMSharaiha + + + DAbramson + + 10.1287/trsc.34.2.180.12302 + + + Transportation Science + Transportation Science + 0041-1655 + 1526-5447 + + 34 + 2 + + 2000 + Institute for Operations Research and the Management Sciences (INFORMS) + + + Beasley, J. E., Krishnamoorthy, M., Sharaiha, Y. M., and Abramson, D., "Scheduling Aircraft Landings -The Static Case," Transportation Science, Vol. 34, No. 2, 2000. + + + + + Scheduling Aircraft Landings to Closely Spaced Parallel Runways + + MKupfer + + + + The Eighth USA/Europe Air Traffic Management Research and Development Seminar + Napa, CA + + June 2009 + + + Kupfer, M., "Scheduling Aircraft Landings to Closely Spaced Parallel Runways," The Eighth USA/Europe Air Traffic Management Research and Development Seminar , Napa, CA, June 2009. + + + + + A Mixed Integer Linear Program for Airport Departure Scheduling + + GautamGupta + + + WaqarMalik + + + YoonJung + + 10.2514/6.2009-6933 + + + 9th AIAA Aviation Technology, Integration, and Operations Conference (ATIO) + Hilton Head, South Carolina + + American Institute of Aeronautics and Astronautics + September 2009 + + + + Gupta, G., Malik, W., and Jung, Y. C., "A Mixed Integer Linear Program for Airport Departure Scheduling," 9th AIAA Aviation Technology, Integration and Operations Conference (ATIO), Hilton Head, South Carolina, 21-23 September 2009. + + + + + A Mixed Integer Linear Program for Solving a Multiple Route Taxi Scheduling Problem + + JustinMontoya + + + ZacharyWood + + + SivakumarRathinam + + + WaqarMalik + + 10.2514/6.2010-7692 + + + AIAA Guidance, Navigation, and Control Conference + Toronto, Canada + + American Institute of Aeronautics and Astronautics + August 2010 + + + + Montoya, J., Wood, Z., Rathinam, S., and Malik, W., "A Mixed Integer Linear Program for Solving a Multiple Route Taxi Scheduling Problem," AIAA Guidance, Navigation, and Control Conference, Toronto, Canada, 2-5 August 2010. + + + + + A Generalized Dynamic Programming Approach for a Departure Scheduling Problem + + SivakumarRathinam + + + ZacharyWood + + + BanavarSridhar + + + YoonJung + + 10.2514/6.2009-6250 + + + AIAA Guidance, Navigation, and Control Conference + Chicago, IL + + American Institute of Aeronautics and Astronautics + August 2009 + + + + Rathinam, S., Wood, Z., Sridhar, B., and C., J. Y., "A Generalized Dynamic Programming Approach for a Departure Scheduling Problem," AIAA Guidance, Navigation, and Control Conference, Chicago, IL, 10-13 August 2009. 10 + + + + + A Hybrid Optimization Approach to Air Traffic Management for Metroplex Operations + + BrianCapozzi + + + StephenAtkins + + 10.2514/6.2010-9062 + + + 10th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + + American Institute of Aeronautics and Astronautics + + + + Capozzi, B. J. and Atkins, S. C., "A Hybrid Optimization Approach to Air Traffic Management for Metroplex Operations," + + + + + Departure Efficiency Benefits of Terminal RNAV Operations at Dallas-Fort Worth International Airport + + RalfMayer + + 10.2514/6.2006-7774 + + + 6th AIAA Aviation Technology, Integration and Operations Conference (ATIO) + Fort Worth, Texas + + American Institute of Aeronautics and Astronautics + September 2010 + + + + AIAA Aviation Technology, Integration and Operations Conference (ATIO), Fort Worth, Texas, 13-15 September 2010. + + + + + Towards Optimal Routing and Scheduling of Metroplex Operations + + BrianCapozzi + + + StephenAtkins + + + SeongimChoi + + 10.2514/6.2009-7037 + + + 9th AIAA Aviation Technology, Integration, and Operations Conference (ATIO) + Hilton Head, South Carolina + + American Institute of Aeronautics and Astronautics + September 2009 + + + + Capozzi, B. J. and Atkins, S. C., "Towards Optimal Routing and Scheduling of Metroplex Operations," 9th AIAA Aviation Technology, Integration, and Operations Conference(ATIO), Hilton Head, South Carolina, 21-23 September 2009. + + + + + Optimal Integration of Departures and Arrivals in Terminal Airspace + + MinXue + + + ShannonZelinski + + 10.2514/6.2012-4977 + + + AIAA Guidance, Navigation, and Control Conference + Minneapolis, MN + + American Institute of Aeronautics and Astronautics + August 2012 + + + + Xue, M. and Zelinski, S., "Optimal Integration of Departure and Arrivals in Terminal Airspace," AIAA Guidance, Navigation, and Control Conference, Minneapolis, MN, 13-16 August 2012. + + + + + Uncertainty Analysis of Integrated Departures and Arrivals: A Los Angeles Case Study + + MinXue + + + ShannonZelinski + + + DanielGMulfinger + + 10.2514/6.2013-4347 + + + 2013 Aviation Technology, Integration, and Operations Conference + Los Angeles, LA + + American Institute of Aeronautics and Astronautics + August 2013 + + + + Xue, M., Zelinski, S., and Mulfinger, D., "Uncertainty Study of Integrated Departures and Arrivals: A Los Angeles Case Study," 13th AIAA Aviation Technology, Integration, and Operations Conference(ATIO), Los Angeles, LA, 12-14 August 2013. + + + + + A fast and elitist multiobjective genetic algorithm: NSGA-II + + KDeb + + + APratap + + + SAgarwal + + + TMeyarivan + + 10.1109/4235.996017 + + + IEEE Transactions on Evolutionary Computation + IEEE Trans. Evol. Computat. + 1089-778X + + 6 + 2 + + 2002 + Institute of Electrical and Electronics Engineers (IEEE) + + + Deb, K., Pretap, A., Agarwal, S., and Meyarivan, T., "A Fast and Elitist Multiobjective Genetic Algorithm: NSGA-II," IEEE Transactions On Evolutionary Computation, Vol. 6, No. 2, 2002. + + + + + Parallel hybrid evolutionary algorithms on GPU + + TVLuong + + + NMelab + + + ETalbi + + + + IEEE Congress on Evolutionary Computation (CEC) + + + 2010. July 2010 2010 + Barcelona + + + Luong, T. V., Melab, N., and Talbi, E., "Parallel hybrid evolutionary algorithms on GPU," 2010 IEEE Congress on Evolutionary Computation (CEC), Barcelona, 18-23 July 2010 2010. + + + + + + diff --git a/file821.txt b/file821.txt new file mode 100644 index 0000000000000000000000000000000000000000..605dbe5aa649a0cc09f46a8a864c84866eccf804 --- /dev/null +++ b/file821.txt @@ -0,0 +1,260 @@ + + + + +IntroductionIn the National Airspace System (NAS), arrival meter fixes are fixes over which aircraft are metered prior to entering terminal airspace.Using meter fixes is critical for air traffic controllers to manage arrival traffic safely and efficiently.When aircraft funnel through the same meter fix they are required to maintain minimum separation for safety reasons.However, for an airport, high demand at a few arrival meter fixes can cause unacceptable arrival scheduling delay even though runways and the rest of the meter fixes are still underutilized.Many factors, including but not limited to, uncoordinated flight plans, geometric locations of airports, uncertainty in flight times and departure times, and weather, contribute to the imbalance among meter fixes and runways.The Optimized Route Capability (ORC) concept [1] proposed by the FAA was envisioned to assist Traffic Management Units (TMUs) identify and resolve arrival flight delays caused by imbalance between meter fixes and runways by collecting and interpreting data, predicting and calculating arrival flight times, and evaluating and optimizing route options.NASA teamed with the FAA to implement an initial algorithm and evaluate the ORC concept [2].The ORC concept takes advantage of precise predictions from automation functions.These predictions include the estimated times of arrivals (ETAs) at meter fixes and arrival schedules inside the Terminal Radar Approach Control (TRACON) areas.In the NAS, they can be obtained by using the Traffic Flow Management System (TFMS) and Time Based Flow Management (TBFM) automation tools [3], respectively.The core function of ORC monitors arrival flight delays at meter fixes based on predicted meter fix ETAs and arrival schedules.If the estimated arrival flight delays at meter fixes exceed a predefined delay threshold, a route option search engine of ORC function will be triggered.The search function will evaluate eligible re-routes and identify the best routes to alternative meter fixes.The resulting reroutes will be presented to decision makers.The initial implementation and evaluation were reported in previous work [2], which showed substantial flight time saving.In the initial implementation, the route option search function utilized a brute-force search method based on a heuristically prioritized flight list and the route option search space was limited to be the spatial dimension only.This work extends previous work by introducing an optimization search method and adding a temporal dimension in reroute search space.The optimization search method was used to identify reroutes with more flight time savings than the brute force method.The temporal dimension was added in reroute options to identify the best time to reroute, which was expected to further improve results.This paper is organized as follows: Section II describes the concept and implementation architecture of ORC; Section III presents the proposed approach; Section IV & V provide the experiment setup, results and analysis; Section VI concludes this work. +Optimized Route CapabilityWhen imbalance at the meter fixes occurs, traffic managers may not be able to notice the imbalance in time and it is also hard for traffic managers to identify efficient re-route candidates as they don't have adequate time for conducting many evaluations.The ORC concept [1][2] is proposed to build a tool to assist traffic managers to identify and resolve arrival flight delays caused by such imbalance. +Concept +Figure 1. ORC concept diagramFigure 1 presents the main functions in the ORC concept.The predictions of ETAs at meter fixes and arrival schedules inside the TRACON areas are first obtained to predict congestion at overloaded arrival meter fixes.Once arrival fix overloads are predicted, a set of pre-filtered reroute options will be fed into a search function.The search function will identify the best re-route combinations after evaluating possible delays associated with each given re-route option. +ImplementationIn the implementation, the Future ATM Concepts Evaluation Tool (FACET) [4] tool was used to serve as a trajectory prediction function to estimate unimpeded meter fix ETAs, where the simulation cycle is one minute.A simplified model of the TBFM arrival scheduler, which was developed to reduce computation time for iterative use, was applied to emulate TBFM to predict the schedules inside of the TRACON airspace [2][5].This arrival scheduler takes flights' ETAs as inputs and allocates the scheduled times of arrival (STAs) to flights based on separation requirements, runway assignment logic and adaption data similar to the existing TBFM tool [3].Fig. 2 shows the implementation diagram for ORC evaluation experiments.The diagram shows a self-contained close-loop design.The rerouting solution, if it exists, suggested by the ORC algorithm is automatically accepted by the system and fed into the next time-step (e.g., 1 minute) of simulation. +Figure 2. ORC implementation diagramAs shown in this figure, the traffic data and airport related data, such as the published standard terminal arrival routes (STARs) will be fed into the system.The 'Reroute Generator' function generates cloned flights with flight plans re-directed to different arrival fixes.For eligibility of reroutes to alternative meter fixes, certain rules were applied: If a flight is within 5 minutes travel time to any center boundary, it will not be eligible to reroute; If a flight's ETA to its arrival fix is more than 90 or 120 minutes or if it's 700 miles away from the airport, it will be not eligible.There are also geographical restrictions on reroutes a flight can only reroute to its neighboring arrival fixes.The flight plan list, including those candidate reroute flight plans, will be fed into FACET all together.Corresponding meter fix arrival times will be calculated at once for all flights to save computational time.The information of actual flights will be input into the arrival scheduler to see if there is any non-absorbable delay at arrival fixes.If the arrival fix delay is over a predefined threshold, the search engine is triggered to find a rerouting solution to offload the congested arrival fix.The search engine will try to replace current routes with alternative routes and identify a re-routing option that can reduce the arrival fix delay to less than the predefined threshold.If no alternative route can cut the delay to the threshold, then a partial solution that can reduce the delay the most will be output.Or there will be no solution at all if none of alternative routes could reduce the delay.Either a complete or partial solution will be fed into next step, affecting the next cycle of reroute generation. +Planning HorizonsThere are two planning horizons in the ORC implementation.One decides when a flight is eligible for rerouting, which is called eligibility horizon.The perception horizon, which is usually further away in time than the range of eligibility horizon, dictates when a flight should be taken into account in the arrival scheduler, even if it is out of scope for reroute eligibility.These two horizons can be the same and can be defined using either a distance or a time.In previous work, a 90-minute range was utilized for both, which means flights beyond 90 minutes from the airport are not eligible for rerouting nor can they be observed and considered in the arrival scheduler.Shortcomings may arise when both ranges are set the same.Because the information of the trailing flights is unknown, it affects the optimality of the reroute solution. +MethodIn the ORC concept the best re-routes and the best time for re-routing when resolving overloads at meter fixes will be identified.Re-routing flights right after they are eligible may cause a "first-come-firstreroute" phenomenon, which is due to the absent information about the trailing flights that are out of the detection or perception horizon. +Expansion of reroute optionsIn this work, a temporal dimension was added in the search space.Figure 2 shows an example of adding a temporal dimension in reroute options.The current position of the flight is shown as a green triangle and its current flight plan (a white dash line) shows the flight is going to fly to WHACK meter fix via SWB.Meanwhile, this flight is eligible to reroute to the meter fixes MPORT and LINKK via CVE and LFT, respectively, as shown in yellow dash lines.Adding temporal reroute options will allow this flight to be rerouted in any future time at specific Δt minute intervals.This extension of the reroute solution space provides an extra degree of freedom in rerouting by allowing flights to go to alternative fixes at a future time.Therefore, the optimizer doesn't have to rush into a sub-optimal reroute solution at planning time. +Figure 2. Temporal dimension in reroutes +OptimizationThe search method applied in the initial algorithm [2] was developed based on a brute-force search with a heuristic priority flight list for rerouting.Two shortcomings may arise from the initial method: the brute-force method will encounter difficulties when the re-route options are large due to increased numbers of flights or enlarged planning horizon; Second, in the future, when introducing uncertainties into the model, a powerful optimization method is definitely desired due to an even larger search space.In order to develop an ORC algorithm with a high capability of searching, such that alternative routes with more flight time saving can be identified in a fast-time or real-time fashion, a genetic algorithm (GA) method [6] is developed and implemented based on a Non-dominated Sorting Genetic Algorithm (NSGA) [7] for the ORC concept.A Genetic Algorithm (GA) mimics the process of natural selection by selection, mutation, and crossover.It is widely applied because of its capability of handling nonlinear optimization problems and parallel computing.(1)To achieve the maximum runway throughput, another cost function (Eqn.2), which is to minimize the total of the runway landing times rwySTA i , is also tried in this work.(2) +ExperimentsExperiments were set up to examine the proposed method.As described previously, the experiments are designed to be closed-loop.Any reroute solution recommended by the algorithm will be executed in the next simulation cycle except for the reroutes planned in a future time.A future reroute will be re-evaluated in the next simulation cycle until it becomes a present reroute. +Traffic ScenarioA traffic scenario was created based on 2-hour historical arrival traffic into George Bush Intercontinental Houston Airport on Oct 22, 2014.On this day, Houston airport had high traffic volume, low weather impacts and was dominated by the W3 runway configuration, which is West flow and three arrival runways in operation.Flight plans were slightly changed to make sure flights would go through one of the four major meter fixes, WHACK (Northeast), MPORT (Northwest), GMANN (Southwest) and LINKK (Southeast).The traffic was increased by adding some flights to create a challenging rush to WHACK.A total of 163 flights were included in the traffic scenario and 12 of them were added manually to increase the load at WHACK.According to the original arrival fix ETAs, the traffic reaches its peak around 11:45 am, when the load at WHACK is about 15 flights in a 15minute interval while MPORT, GMANN, and LINKK have 5, 1, and 7 flights, respectively. +Horizon Set-upIn the experiments, the proposed approach used different values for the eligibility horizon and perception horizon.Because of the capability of finding solutions in a future time, the new approach can perceive and utilize the information of trailing flights beyond the eligibility horizon without rushing into a 'first-come-first-reroute' situation. +ResultsExperiments were conducted to evaluate the proposed new approach along with different parameter settings.The study cases including their search methods and associated parameters are shown in Table 1.The 'ORC0' case is the same approach used in previous work [2].The 'ORC1' case utilized a different cost function to increase runway throughput (Eqn.2).The 'ORC2' case increased the perception horizon to 120 minutes to include trailing flight information to improve planning.The 'ORC3' case introduced the temporal dimension. +Delay savingsThe delay refers to the delay at the runway, which is the difference between unimpeded runway landing time and the actual/simulated runway landing time.Figure 5 shows the comparison of delay savings among different cases.The comparison between 'ORC0' and 'ORC1' shows that minimizing the total runway landing times allows 18 minutes more delay savings, and therefore, increases the runway throughput.By increasing the perception horizon in 'ORC2', the delay savings were further improved by 6 minutes.Using the ultimate approach, 'ORC3' can plan reroutes in a future time to increase the delay savings another 6 minutes to 143 minutes. +Figure 5. Comparison of delay savings +Delay distributionsWhen there is no ORC, flights stay with their original arrival fixes.Imposed delays at arrival meter fixes can be called scheduled delay and have to be absorbed through holding or vectoring before entering TRACON.When ORC moves flights, extra flight time may be imposed when flights are directed to alternative meter fixes.Since the time to fly (TTF) would be different between different meter fixes to runways, extra flight time may also be introduced here as well.Figure 6 shows the total delay distributions associated with original routing and the ORC cases.The remainder of the ORC comparisons will be between 'ORC2' and 'ORC3' to focus on the effects of adding the temporal dimension.Results demonstrated that the only delay source for the 'Original' was the scheduled delay.ORC algorithms introduced extra time due to the change of meter fixes even though they managed to reduce the overall delays.The delays caused by extra distances to alternative meter fixes for 'ORC2' and 'ORC3' are 54 and 43 minutes, respectively.The difference between the ORC cases is actually caused by one less reroute in 'ORC3', which will be further investigated in the next section. +Figure 6. Comparison of delay sources +Reroute solutionsBesides the delay savings, the number of flights moved is another important metric.Figure 7 presents the time history of reroutes suggested by 'ORC2' and 'ORC3'.Orange boxes identify reroutes to MPORT, whereas the rest are to LINKK.This figure demonstrates that 'ORC3' suggested moving eight flights to alternative meter fixes, with two to MPORT and six to LINKK.'ORC2' moved nine flights with three to MPORT and six to LINKK.Fewer moves are preferred in actual operations due to the operational complexity associated with each move.Fewer moves also lead to less extra time to alternative fixes as shown in Fig. 6 in previous section.The main difference between these two cases is the timing of those moves.Apparently, most of moves from 'ORC3' happened later than those from 'ORC2'.The average time frame of the route changes was postponed around 30 minutes.Because the temporal dimension was added into the rerouting solutions of 'ORC3', it had the capability to consider trailing flights for future reroutes even though they were still out of the eligibility horizon.'ORC3' can consider solutions in future times to improve the optimality and stability of the route change suggestions.Therefore, 'ORC3' can wait for the best moves, which enhances ORC's capability in actual operations.Controllers prefer to make these decisions as late as possible, because, in the real world, the uncertainty is usually high when the look-ahead time is early and it decreases with time. +Meter fix loadsFigure 9 presents the meter fix loads that resulted from different cases between 11:00 to 13:00 in 15-minute intervals.Both ORC cases managed to offload WHACK traffic to alternative meter fixes around 11:52 and reduced the load at WHACK from 22 in the 'Original' case to less than 15.Most of these reroutes were to LINKK, which matches the solutions described in the previous section.Meanwhile, slight differences exist between 'ORC2' and 'ORC3', including numbers of reroutes and destination meter fixes. +Figure 9. Comparison of meter fix loads +Airport loadsFigure 10 presents the airport loads that resulted from different cases between 11:00 to 14:00 in 15minute intervals.The figure shows that both ORC cases shifted the airport load forward from 'Original' by moving flights to alternative arrival fixes to utilize the airport capacity, which was not fully used in the 'Original' case due to the saturated WHACK fix.This shift happened around 13:00, when both ORC cases pushed the runway loads to about 28 per 15-minute interval.The 'Original' underutilized runways with less then 20 landings in the same period because WHACK fix was congested and formed the bottleneck. +Figure 10. Comparison of airport loadsBoth ORC cases showed great similarity in airport loads.However, 'ORC3' was able to push a couple more flights earlier than 'ORC2' around 13:00 to further utilize the runway capacity. +DiscussionAdding a temporal dimension in reroute options enables the optimizer to identify the best time to move flights and enhance the optimality and stability of the reroutes.Although adding a temporal dimension introduces a large number of reroute options, it can be handled well by the GA optimization algorithm.Including uncertainty in future work will be important and necessary in order to handle departure time and weather impacts precisely.Since GA-type optimization can be parallelized and combined with Graphics Processing Unites (GPU) implementation [8], the proposed approach will enable the introduction of uncertainties in future research, where uncertainties can be injected to route options in terms of their corresponding look-ahead times. +SummaryThe ORC concept was proposed to facilitate traffic managers to fully utilize airport runway capacity when bottlenecks are formed in certain meter fixes.By making full use of the data and prediction capability provided by automation tools and optimization/search capability by advanced algorithms, the ORC concept can ease the burden on traffic mangers while improving the efficiency in TRACON operations.On the basis of initial implementation of the ORC concept, a temporal dimension was added to route options, which provides the opportunity for the ORC tool to find the best time to change routes.Meanwhile, a GA-based optimization algorithm was developed to enable ORC to solve the large solution space due to the expanded route options.Experiments were conducted using the same 2hour traffic scenario used in previous work.Results showed that the new approach further improved total delay savings from 113 to 143 minutes.More importantly, the suggested reroutes from the proposed approach were on average 30 minutes later than previous ORC approaches, which may be preferred in actual operations to reduce uncertainty.Also, the number of route changes recommended was less than previous approaches, which would reduce complexity associated with rerouting flights.This work offers promise in improving the efficiency and stability of reroutes in the NAS.Figure 3 .Figure 4 .34Figure 3. Sample reroute options +Figure 7 .Figure 878Figure 7.Comparison of reroutes (orange boxes represent reroutes to MPORT) Figure 8 provides the runway landing time differences among individual flights when comparing 'ORC2' and 'ORC3'.Positive values denote that the flight landed earlier in 'ORC3' and negative values appear when the flight landed earlier in 'ORC2'.Out of 163 flights, 17 flights are presented with different runway landing times.Although the overall delay reductions are only slightly different as shown in Fig.6, the individual flight landing times for these 17 flights are quite different due to the different route choices. +Figure 8 .8Figure 8.Comparison of flight landing times + +Table 1 . Case study set up1CaseCost functionSearchTemporal dimensionEligibility horizonPerception horizonORC0Eqn. 1Brute-force + priority flight listNo90 min90 minORC1Eqn. 2GANo90 min90 minORC2Eqn. 2GANo90 min120 minORC3Eqn. 2GAYes90 min120 min + + + + +AcknowledgementsThe authors would like to acknowledge the contributions from Philip Basset of FAA, Christina Young and Albert Schwartz from the FAA Technical Center in generating the traffic scenario for the ORC evaluation experiments. + + + + + + + + + Optimized Route Capability (ORC) Intelligent Offloading of Congested Arrival Routes + + ShannonZelinski + + + MinXue + + + PhilipBassett + + 10.2514/6.2016-4357 + + + 16th AIAA Aviation Technology, Integration, and Operations Conference + Washington, D.C + + American Institute of Aeronautics and Astronautics + September 2015 + + + Federal Aviation Administration, September 2015, "Optimized Route Capability -A Revised Concept of Operations", Washington, D.C. + + + + + Optimized Route Capability (ORC) Intelligent Offloading of Congested Arrival Routes + + ShannonZelinski + + + MinXue + + + PhilipBassett + + 10.2514/6.2016-4357 + + + 16th AIAA Aviation Technology, Integration, and Operations Conference + Washington, D.C + + American Institute of Aeronautics and Astronautics + June 2015 + + + Zelinski, S., Xue, M., and Bassett, P., June 2015, "Optimized Route Capability Intelligent Offloading of Congested Arrival Routes", AIAA Aviation 2016, Washington, D.C. + + + + + A framework for decision making in extended enterprises: The FAA NextGen case + + HamidRDarabi + + + MoMansouri + + + NazaninAndalibi + + + EileenPara + + 10.1109/icumt.2010.5676566 + + + + International Congress on Ultra Modern Telecommunications and Control Systems + + IEEE + + + + FAA NextGen Decision Support Systems, https://www.faa.gov/nextgen/update/progress_and_ plans/decision_support_systems/ + + + + + FACET: Future ATM Concepts Evaluation Tool + + KarlDBilimoria + + + BanavarSridhar + + + ShonRGrabbe + + + GanoBChatterji + + + KapilSSheth + + 10.2514/atcq.9.1.1 + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 9 + 1 + + June 2000 + American Institute of Aeronautics and Astronautics (AIAA) + Napoli, Italy + + + 3 rd USA + Bilimoria, K., Sridhar, B., Chatterji, G., Sheth, K., and Grabbe, S., June 2000, "FACET: Future ATM Concepts Evaluation Tool", 3 rd USA/Europe Air Traffic Management R&D Seminar, Napoli, Italy. + + + + + A Closed-Form Solution to Multi-Point Scheduling Problems + + LarryMeyn + + 10.2514/6.2010-7911 + + + AIAA Modeling and Simulation Technologies Conference + Toronto, Canada + + American Institute of Aeronautics and Astronautics + August 2010 + + + Meyn, L., August 2010, "A Closed-Form Solution to Multi-Point Scheduling Problems", AIAA Modeling and Simulation Technologies Conference, Toronto, Canada. + + + + + Genetic Algorithms + Data Structures = Evolution Programs + + ZMichaelwicz + + 10.1007/978-3-662-03315-9 + + 1996 + Springer Berlin Heidelberg + Berlin Heidelberg + + + Michaelwicz, Z., 1996, "Genetic Algorithms + Data Structures = Evolution Programs", Springer- Verlag, Berlin Heidelberg. + + + + + A fast and elitist multiobjective genetic algorithm: NSGA-II + + KDeb + + + APratap + + + SAgarwal + + + TMeyarivan + + 10.1109/4235.996017 + + + IEEE Transactions on Evolutionary Computation + IEEE Trans. Evol. Computat. + 1089-778X + + 6 + 2 + + 2003 + Institute of Electrical and Electronics Engineers (IEEE) + + + Deb, K., Pretap, A., Agarwal, S., and Meyarivan, T., "A Fast and Elitist Multi-objective Genetic Algorithm: NSGA-II", IEEE Transaction On Evolutionary Computation, Vol.6, No.2, 2003, pp.182-197. + + + + + Integrated Arrival- and Departure-Schedule Optimization Under Uncertainty + + MinXue + + + ShannonZelinski + + 10.2514/1.c032957 + + + Journal of Aircraft + Journal of Aircraft + 0021-8669 + 1533-3868 + + 52 + 5 + + 2015. 2016 + American Institute of Aeronautics and Astronautics (AIAA) + + + 35th Digital Avionics Systems Conference September 26-29 + Xue, M., and Zelinski, S., "Integrated Arrival and Departure Schedule Optimization Under Uncertainty", AIAA Journal of Aircraft, Vol.52, No.5, 2015, pp.1437-1443. 35th Digital Avionics Systems Conference September 26-29, 2016 + + + + + + diff --git a/file822.txt b/file822.txt new file mode 100644 index 0000000000000000000000000000000000000000..1271f546219ebe64a112368de3b9005af9db8d99 --- /dev/null +++ b/file822.txt @@ -0,0 +1,386 @@ + + + + +I. IntroductionThe corridors-in-the-sky 1 imitate highways in ground transportation by grouping a large number of flights with similar trajectories.The concept is expected to increase overall airspace capacity by allowing high traffic density in corridors with negligible controller workload while reducing the complexity of underlying sectors.Many studies [2][3][4][5][6][7] have shown that a significant number of flights take common trajectories (candidate corridors), and the number increases if the users are willing to travel a little extra distance to join the corridor.In addition, Wing et.al. 8 conducted an analysis at a conceptual level for a Dynamic Multi-track Airways (DMA) concept, which is similar to the corridor concept.However, the key components on which corridor benefits rely have not been well studied, such as the complexity of handling high density traffic, the extra fuel consumption taken by corridor users, and the complexity of handling non-corridor traffic in underlying sectors.Although Wing et.al 8 mentioned the challenge of interactions between DMA (corridor) and non-DMA (non-corridor) traffic, there is no quantitative analysis of the complexity caused by such interactions.Conflict count with no Traffic Flow Management (TFM) actions 8 and peak aircraft count 9 have been utilized to assess non-DMA or non-corridor traffic complexity.However, the accuracy of using these metrics to represent complexity is questionable.In this work, conflict resolution counts are used to measure complexity.This work first presents a simulation developed for imitating conflict-free operations in corridors.In the simulation, the behaviors of corridor users are approximated by several prescribed conflict resolution maneuvers.Different lane options and policies for resolving conflicts are also explored to find the lowest number of required resolutions.Then corridor complexity is measured and the associated extra fuel consumption is calculated.Additionally, to find the complexity of non-corridor traffic including the complexity caused by the interactions, the Airspace Concept Evaluation System (ACES) 10 simulation tool with Automated Airspace Concept (AAC) 11,12 plug-in are applied.Based on a 1X traffic scenario, the results show that with appropriate operational polices and lane options, the complexity of handling heavy traffic is low.For instance, with 608 flights, only 84 actions need to be taken in a 24-hour period to resolve the conflicts for an 8-lane corridor.The simulation of corridor traffic shows that the total extra fuel for corridor flights is 26,373 gallons, or 2.76%, compared with their great circle trajectories.The simulation of non-corridor traffic shows that the complexity of underlying sectors is increased by 9.14%.The results indicates that the benefit of the corridor is its capability of handling large amount of traffic with relative low complexity and the 2.76% extra fuel may ba acceptable.However this experiment doesn't show benefit in the complexity of underlying sectors.The paper will be arranged as follows: Section II describes the corridor model and traffic data.Section III presents the procedures and policies for the simulations of corridor traffic.Section IV shows the results for corridor traffic.After a brief introduction of the ACES AAC simulation, Section V presents the complexity results for non-corridor traffic.Finally, Section VI concludes the paper. +II. Corridor Model and Traffic DataThe algorithm from previous work 6 identified the top corridor candidate.This was a 2D great circle corridor location such that the number of flights accommodated in the corridor was maximized and the extra flight distance for a flight to join the corridor was less than 5% of its total flight distance.Based on the historical traffic data taken from Apr.20, 2007, a corridor travelling approximately between San Francisco and New York City via Chicago was identified as the top candidate.A total of 621 flights were identified to be potential users.Because the above algorithm provides only 2D location, the corridor occupying flight levels must be defined separately.In the study, the corridor occupied three flight levels, FL340, FL350, and FL360.Among them, FL350 served as a passing level.A corridor flight would join the corridor from the passing level and enter another levels when possible.Corridor lane profiles are shown in Fig. 1.Assuming corridor users can perform Required Navigation Performance(RNP) 1.0, which means their lateral deviations from the nominal flight path must stay in ±RNP or ±2.0 nautical miles during 99% of flight time, each lane had a width of 9 nautical miles.Half of the corridor was for west bound traffic and the other half was for east bound traffic.The number of lanes was a decision variable and was explored in the experiments.The safety zone for individual aircraft was defined as 5 nmi in the horizontal direction and 1,000 ft in the vertical direction. +Passing LanesWest-bound East-bound To be practical, aircraft that could not reach the aforementioned levels were ruled out.Furthermore, every flight's climb and descent segments were removed.Only the portion between top of climb (TOC) and top of descent (TOD) were considered for corridor operation.The TOC and TOD were computed according to the aircraft's nominal climb and descent rate.Fig. 2 illustrates sample TOC and TOD segments for a flight.Furthermore, if the distance between TOC and TOD was less than 50 nmi, the corresponding flight was also ruled out.608 out of 621 flights satisfied these constraints.275 flights were east bound and 333 flights were west bound. +III. Procedure for Simulating Aircraft BehaviorsIn order to get conflict resolution counts, traffic behaviors are approximated by prescribed maneuvers.Six prescribed maneuvers were applied: climb, descent, left turn, right turn, slowing down, and default cruise maneuver.The left and right turns were defined to be 45 o turns.Climb and descent assumed a nominal rate.The "slowing down" forced the aircraft to fly at the same speed as the flight in front of it such that a conflict would be avoided.The default maneuver was the nominal flight status when there was no conflict or abnormal situation.A typical default maneuver is to fly forward at the cruising speed.All aircraft performances were calculated based on the Base of Aircraft Data (BADA) 13 developed by Eurocontrol.To simplify the conflict detection, all maneuvers were descretized into defined time intervals.In this work, the time interval was 0.3 minutes.Each maneuver was divided into several 0.3 minute segments, each lasting 0.3 minutes.Because the duration of a maneuver might not be an exact multiplier of 0.3 minutes, Figure 6 presents the work flow of the simulation.Because the durations of prescribed maneuvers normally covered several 0.3 minute steps, a maneuver would not be "conflict-free" unless no conflict occurred during the time steps covering the entire maneuver.As soon as a "conflict-free" maneuver was decided, the postmaneuver position was pre-allocated to prevent any future conflict.Additionally, when a "slowing down" maneuver was chosen, all flights following the aircraft were re-checked for conflicts.Several viable resolution maneuvers may exist for a conflict depending on operational policies.Two types of operational policies were explored to examine their effects on conflict resolution.1. "Order-based": the maneuver will be chosen based the following order: default maneuver, left turn, right turn, climb, descent, and slowing down.If the aircraft is at a passing flight level, the order will change to climb, descent, default maneuver, left turn, right turn, and slowing down.2. "Load-based": instead of a fixed order, the maneuver that balances the lane loads within a given range will be picked.Any maneuver that causes a secondary conflict before the end of the resolution's tail segment is skipped.Two "load-based" ranges were simulated: "load-based:short" checks lane loads from 10 nmi behind the aircraft to 30 nmi ahead of it, and "load-based:long" checks lane loads form 10 nmi behind to 50 nmi ahead. +IV. Simulations and Results of Corridor TrafficThe simulation was implemented in C language and displayed with Matlab.It was run on a Mac-Pro computer with a dual-core CPU at 2.8MHz.Figure 7 The experiments explored three operational polices and a various number of lanes.Figure 8 presents the comparison among operation policies and the number of lanes.When there are only two lanes for each direction, the number of resolutions for all three policies are high.Because aircraft are only allocated to two lanes, there is little room for improvement using the "load-based" policies.But the load-based policies still have much lower numbers of resolutions than the "order-based" policy.When the number of lanes is doubled to four lanes each way, the resolutions decrease dramatically for all three polices.The "loadbased:long" policy needs 84 resolutions whereas "order-based" requires 654.Obviously, the "load-based" policies outperformed the "order-based" policy.If the number of lanes is 6 in each direction, the resolutions reach minimums at 36 and 43 for "load-based:long" and "load-based:short", respectively.However, the resolution efforts with "order-based" reach a minimum at 654 because the added lanes may not be used at all.These comparisons demonstrate that different rules and different lanes options have a big impact on complexity.On average a sector performs 50 resolutions with peak aircraft less than 20.Therefore, 84 resolutions is attractive for an airspace with peak aircraft count of 100.This indicates that handling a large traffic volume traffic in corridors is less complex than in classic sectors.Figure 9 provides the time history of resolutions with the "load-based:long" policy and the 8 lane option, and the peak number of resolutions is only four.The "load-based:long" policy and the 8 lane option will be kept in the following study.Based on the simulated conflict-free trajectories, the fuel consumption of corridor flights was calculated using BADA3.6. 13As a baseline, fuel consumption based on shortest paths (great circle trajectories) and cruise altitudes were calculated.Table 1 lists the number of corridor users and extra fuel consumption corresponding to different corridor altitudes.For simplicity, the flight levels shown in the table are the corridor passing lanes.When the corridor altitude is above FL350, many flights are not able to reach the corridor.Whereas when the corridor altitude drops to FL300, although there are 608 aircraft in the corridor, the extra fuel is 8.99%.Because many big jets were forced to fly at altitudes that are far off their optimal cruise altitudes, the extra fuel consumption is high.In the table, the differences between flying the corridor and flying filed flight plans based on historical data are also listed.Flying the FL350 corridor requires less fuel burn than flying filed flight plans.When the altitude of corridor drops to FL300, flying the corridor requires 5.78% more than flying filed flight plan, which may not be acceptable in operation.The comparisons indicate that the altitude plays an important role in corridor design and analysis, it has to be carefully chosen to maximize the number of corridor users and minimize extra fuel burn.In this experiment, FL350 seems to be the best choice, burning only 2.76% extra fuel over flying shortest paths. +V. Complexity of Non-corridor TrafficThe complexity of non-corridor traffic in underlying sectors must be examined to find out the impact of the corridor.Because such large scale human-in-the-loop simulations require lot of resources and resolving conflicts is the most important task for a controller, a preliminary analysis was done by using the number of conflict resolutions as a measure of complexity.The ACES tool was used to simulated the traffic without capacity constraints.In order to perform conflict resolution, the AAC tool was applied as a plug-in of ACES.The AAC simulates the behaviors of resolving conflicts on the basis of predefined maneuvers.The AAC tool used in this work is comprehensive, it can solve three types of problems: conflict resolution, arrival management, and weather avoidance.Because the AAC tool performs 2D maneuvers for conflicts between aircraft and weather polygons and 3D maneuvers for other types of conflicts, modifications were made for this study to allow the tool handle 3D maneuvers for conflicts between aircraft and the designated corridor airspace.Because running the ACES AAC simulation for the entire national airspace is time-consuming, only three centers affected by the corridor were chosen and studied.They are Chicago center (ZAU), Cleveland center (ZOB), and New York center (ZNY).Figure 10 shows these three centers and the location and size of the corridor.There were 568 out of 608 corridor flights in these three centers.To get an in-depth analysis, four cases were set up to explore the contribution of complexity.Case I simulated all traffic, including corridor and non-corridor traffic, without any airspace or airport capacity constraints.Case II simulated non-corridor traffic without any capacity constraints and flights flying in the corridor were simply removed.On the basis of Case II, Case III includes the corridor airspace as an obstacle but still no corridor flights.In case III, the corridor was treated as a restricted airspace such that non-corridor traffic need to be capped or tunnelled to avoid the corridor airspace.Finally, Case IV included the climbing and descending portions of corridor traffic.In corridor design the climbing and descending segments of corridor traffic should be taken care of by controllers in surrounding sectors.Among these four cases, Case I represents the current system without corridors.Case IV represents a system surrounding a corridor.In this experiment, the conflict resolutions required in the corridor was excluded.In Table 2, the complexity in Case II dropped 18.0% by removing corridor flights, and then it increased a bit in Case III to 3,996 due to the imposed impenetrable corridor airspace.In the experiment, non-corridor flights that would have crossed the corridor were assumed to change their flight plan before departure, so corridor airspace avoidance was not counted as the part of controllers' complexity.However, when climb and descent portions of corridor traffic were involved, in Case IV, the complexity dramatically increased to 5,300.Among the 5,300, 4,070 resolutions happened among non-corridor flights, 855 resolutions were between climb/descent segments and non-corridor flights, and 375 resolutions happened among climb/descent segments.By comparing Case IV and Case I, the complexity increased by 9.14%.Table 3 provides the complexities of the top 5 busiest sectors in Case I. Sector ZNY75 reduced complexity by 44% by simply removing corridor flights from the sector, but in taking climb and descent segments of corridor flights into account, the complexity of ZNY75 actually increased by 72.0%.Similar situations happened in other sectors too.Based on these comparisons, the interaction between climb and descent segments of corridor flights and non-corridor flights plays a critical role.It eventually eliminated the benefit of corridors from a complexity perspective.Obviously, in this experiment, the corridor did not show the benefit of reducing complexity. +VI. ConclusionThis work studied the complexity of traffic in a selected corridor airspace.The corridor's capability of handling high density traffic with negligible controller workload, the acceptance of extra fuel or distance, and complexity reduction in underlying sectors was investigated.A conflict-free traffic simulation is developed for studying the corridor traffic.Prescribed conflict resolution maneuvers mimic corridor users' behaviors and the count of conflict resolutions measures complexity.Different lane options and operational policies were proposed to examine their impacts on complexity.Fuel consumption was calculated and compared for corridor traffic.To investigate the complexity of non-corridor traffic in underlying sectors, the ACES tool was used together with AAC to examine the number of conflict resolutions.With traffic simulated based on historical flight schedule, the results showed that a high traffic volume can be handled with much lower complexity in the corridor airspace than in classic sectors.For instance, with peak aircraft count of 100, only 84 actions need to be taken in a 24-hour period to resolve the conflicts in an 8-lane corridor.Compared with optimal fuel consumptions with great circle trajectories, the total extra fuel for corridor flights was 26,373 gallons, or 2.76%, which was 0.38% less than flying filed flight plan.Furthermore, according to the simulations of non-corridor traffic, without taking climb and descent portions of corridor traffic into account, the complexity of underlying sectors was reduced by 17.71%.However, when the climb and descent segments were included, the complexity increased by 9.14% compared to the operations without the corridor.Although in this experiment with a 1X traffic scenario, the selected corridor didn't reduce the complexity, further study is needed to determine if the complexity will be reduced when weather is involved or traffic volume reaches 2X or more.Figure 1 .1Figure 1.Lane Profiles (a) 4 lanes (b) 8 lanes (c) 12 lanes +Figure 2 .Figure 3 .23Figure 2. Calculation of TOC and TOD for corridor traffic +Figure 4 .4Figure 4. Flight time in the corridor +Figure 5 .5Figure 5. Prescribed maneuvers (a) 45 degree left turn (b) climb +Figure 6 .6Figure 6.Procedure for Simulating Corridor Traffic +provides a snapshot of an 8-lane corridor simulation based on the "load-based:long" policy.The horizontal axes are the distances to one end of the corridor (the west end in this case) and the units are nautical miles.For instance, Chicago O'Hare International airport locates at around 2,550 nautical miles away from the west end.The vertical axis in the upper figure is in nautical mile too.The upper figure (Fig. 7(a)) is a top view of 8 lanes (4 lateral, by 2 vertical) of traffic (only 4 can be seen).Among these 4 lanes, the upper 2 lanes are for westbound traffic, and the lower 2 lanes are for eastbound traffic.The lower figure (Fig. 7(b)) presents the side view of the corridor traffic.A red circle or square represents an aircraft in the process of a resolution.A blue circle represents an aircraft during default maneuver. +Figure 7 .7Figure 7. Snapshot of a simulation (a) top view (b) side view +Figure 8 .Figure 9 .89Figure 8.Comparison of Polices and Number of Lanes +Figure 10 .10Figure 10.Three Centers and the selected Corridor +Table 1 .1Comparison of Corridors at Different Flight LevelMiddle flight level Number of attendees Extra fuel over great circle Extra fuel over filed flight planFL3803720.91%-4.69%FL3704351.12%-3.56%FL3506082.76%-0.38%FL3306084.75%1.66%FL3006088.99%5.78% +Table 2 .2Comparison of complexitiesCase I Case II Case III Case IVNumber of resolutions4,8563,9783,9965,300Difference from Case I0%-18.08% -17.71%9.14% +Table 3 .3Complexity of top 5 busiest sectorsSectorCase ICase IICase IVZNYJF342246 (-28.0%) 346 (1.2%)ZNY75193108 (-44.0%) 332 (72.0%)ZAU33163140 (-14.1%) 190 (16.6%)ZOB6811698 (-15.5%)110 (-5.2%)ZNY7411385 (-24.8%) 146 (29.2%) + + + +VII. AcknowledgementThe authors gratefully acknowledge the contribution of Mr. Chok Fung Lai of University of California at Santa Cruz.Mr. Lai modified the AAC tool and ran all simulations using ACES with AAC plug-in for non-corridor traffic. + + + + + + + Initial Concepts for Dynamic Airspace Configuration + + ParimalKopardekar + + + KarlBilimoria + + + BanavarSridhar + + 10.2514/6.2007-7763 + + + 7th AIAA ATIO Conf, 2nd CEIAT Int'l Conf on Innov and Integr in Aero Sciences,17th LTA Systems Tech Conf; followed by 2nd TEOS Forum + Belfast, Northern Ireland + + American Institute of Aeronautics and Astronautics + 18-20 September 2007 + + + Kopardekar, P., Bilimoria, K., and Sridhar, B., "Initial Concepts for Dynamic Airspace Configuration," 7th AIAA Aviation Technology, Integration and Operations Conference (ATIO), Belfast, Northern Ireland, 18-20 September 2007. + + + + + Dynamic airspace super sectors (DASS) as high-density highways in the sky for a new US air traffic management system + + JAlipio + + + PCastro + + + HKaing + + + NShahid + + + OSherzai + + + GLDonohue + + + KGrundmann + + 10.1109/sieds.2003.158005 + + + IEEE Systems and Information Engineering Design Symposium, 2003 + + IEEE + October 2003 + + + + Alipio, J., Castro, P., Kaing, H., Shahd, N., Sheizai, O., Donohue, G., and Grundmann, K., "Dynamic Airspace Super Sectors (DASS) As High-Density Highways in The Sky for A New US Air Traffic Management System," AIAA/IEEE Digital Avionics Systems Conference, 12-16 October 2003. + + + + + High-volume tube-shape sectors (HTS): a network of high capacity ribbons connecting congested city pairs + + AYousefi + + + GLDonohue + + + LSherry + + 10.1109/dasc.2004.1391296 + + + The 23rd Digital Avionics Systems Conference (IEEE Cat. No.04CH37576) + Salt Lake City, CT + + IEEE + 2004 + + + Yousefi, A., Donohue, G., and Sherry, L., "High-Volume Tube-Shape Sectors(HTS): A Network of High Capacity Ribbons Connecting Congested City Pairs," Proceedings of the 23rd Digital Avionics Systems Conference, Salt Lake City, CT, 2004. + + + + + Initial Study of Tube Networks for Flexible Airspace Utilization + + BanavarSridhar + + + ShonGrabbe + + + KapilSheth + + + KarlBilimoria + + 10.2514/6.2006-6768 + + + AIAA Guidance, Navigation, and Control Conference and Exhibit + Keystone, Colorado + + American Institute of Aeronautics and Astronautics + August 2006 + + + + Sridhar, B., Grabbe, S., Sheth, K., and Bilimoria, K., "Initial Study of Tube Networks for Flexible Airspace Utilization," AIAA Guidance, Navigation, and Control Conference and Exhibit, Keystone, Colorado, 21-24 August 2006. + + + + + Freeways in the Sky: Exploring Tube Airspace Design Through Mixed Integer Programming + + GautamGupta + + + BanavarSridhar + + + AvijitMukherjee + + 10.2514/6.2008-6824 + + + AIAA Guidance, Navigation and Control Conference and Exhibit + Washington, D.C. + + American Institute of Aeronautics and Astronautics + October 2008 + + + Gupta, G., Sridhar, B., and Mukherjee, A., "Freeways in the Sky: Exploring Tube Airspace design through Mixed Integer Programming," INFORMS Annual Meeting, Washington, D.C., October 2008. + + + + + High-Capacity Tube Network Design Using the Hough Transform + + MinXue + + + ParimalKopardekar + + 10.2514/1.40386 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 32 + 3 + + 2009 + American Institute of Aeronautics and Astronautics (AIAA) + + + Xue, M. and Kopardekar, P., "High-Capacity Tube Network Design using the Hough Transform," Journal of Guidance, Control, and Dynamics, Vol. 32, No. 3, 2009, pp. 788-795. + + + + + Optimization Based Tube Network Design for the Next Generation Air Transportation System (NextGen) + + PankitKotecha + + + InseokHwang + + 10.2514/6.2009-5860 + + + AIAA Guidance, Navigation, and Control Conference + Chicago, IL + + American Institute of Aeronautics and Astronautics + August 2009 + + + + Kotecha, P. and Hwang, I., "Optimization based Tube Network Design for the Next Generation Air Transportation System (NEXTGEN)," AIAA Guidance, Navigation, and Control Conference and Exhibit, Chicago, IL, 10-13 August 2009. + + + + + Analysis of a Dynamic Multi-Track Airway Concept for Air Traffic Management + + DJWing + + + JCSmith + + + MGBallin + + + 2008 + Langley Research Center + Hampton, Virginia + + + Tech. Rep. NASA/TP-2008-215323 + Wing, D. J., Smith, J. C., and Ballin, M. G., "Analysis of a Dynamic Multi-Track Airway Concept for Air Traffic Management," Tech. Rep. NASA/TP-2008-215323, Langley Research Center, Hampton, Virginia, 2008. + + + + + Design Analysis of Corridors-in-the-Sky + + MinXue + + 10.2514/6.2009-5859 + + + AIAA Guidance, Navigation, and Control Conference + Chicago, IL + + American Institute of Aeronautics and Astronautics + August 10-13 2009 + + + Xue, M., "Design Analysis of Corridors-in-the-sky," AIAA Guidance, Navigation, and Control Conference and Exhibit, Chicago, IL, August 10-13 2009. + + + + + Build 4 of the Airspace Concept Evaluation System + + LarryMeyn + + + RobertWindhorst + + + KarlinRoth + + + DonaldVan Drei + + + GregKubat + + + VikramManikonda + + + SharleneRoney + + + GeorgeHunter + + + AlexHuang + + + GeorgeCouluris + + 10.2514/6.2006-6110 + + + AIAA Modeling and Simulation Technologies Conference and Exhibit + Keystone, Colorado + + American Institute of Aeronautics and Astronautics + August 12-24 2006 + + + Meyn, L. and Windhorst, R., "Build 4 of the Airspace Concept Evaluation System," AIAA Modeling and Simulation technologies Conference and Exhibit, Keystone, Colorado, August 12-24 2006. + + + + + Concept for Next Generation Air Traffic Control System + + HeinzErzberger + + + RussellAPaielli + + 10.2514/atcq.10.4.355 + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 10 + 4 + + August 30 2004 + American Institute of Aeronautics and Astronautics (AIAA) + Yokohama, Japan + + + Erzberger, H., "Transforming the NAS: The Next Generation Air Traffic Control System," the 25th International Congress of the Aeronautical Sciences(ICAS), Yokohama, Japan, August 30 2004. + + + + + Automated conflict resolution, arrival management, and weather avoidance for air traffic management + + HErzberger + + + TALauderdale + + + Y-CChu + + 10.1177/0954410011417347 + + + Proceedings of the Institution of Mechanical Engineers, Part G: Journal of Aerospace Engineering + Proceedings of the Institution of Mechanical Engineers, Part G: Journal of Aerospace Engineering + 0954-4100 + 2041-3025 + + 226 + 8 + + September 2010 + SAGE Publications + Nice, France + + + Erzberger, H., Lauderdale, T. A., and Cheng, Y., "Automated Conflict Resolution, Arrival Management and Weather Avoidance for ATM," the 27th International Congress of the Aeronautical Sciences(ICAS), Nice, France, September 2010. + + + + + BADA Performance Modelling Report + + EECentre + + + + November 2009 + + + Website + Centre, E. E., "BADA Performance Modelling Report," Website, November 2009, http://www.eurocontrol.int/eec/ public/standard_page/ETN_2009_1_BADA.html. + + + + + + diff --git a/file823.txt b/file823.txt new file mode 100644 index 0000000000000000000000000000000000000000..9953930a43759b21e8153f12b17c81d57804e214 --- /dev/null +++ b/file823.txt @@ -0,0 +1,531 @@ + + + + +I. IntroductionIn terminal areas, especially in metroplexes where several busy airports are close to each other, hundreds of flights have to fly through a narrow area for departing or arriving in a short time period.8][9] Another class of terminal airspace problems arises when different departure and/or arrival flows in a terminal airspace share the same resources such as waypoints, fixes or/and routes.The interactions can happen among departures, arrivals, or between departures and arrivals.Recent studies have shown that integrated arrivals and/or departures in major airports or metroplex areas [10][11][12] may have the potential of improving operational efficiency.However, in scheduling problems, results are usually sensitive to flight time uncertainties that are caused by many sources, such as inaccurate wind prediction, error in aircraft dynamics, or human factors.Uncertainty analysis is necessary to evaluate the robustness of the benefits.For arrival scheduling problems, Thipphavong et al 13 studied the relationship between uncertainty and system performance using Stochastic Terminal Arrival Scheduling Software (STASS). 14Mulfinger et al 15 also analyzed scheduling benefits of reduced arrival time uncertainty using expanded STASS.Because the interactions between departures and arrivals occur at both merging and diverging points, the impact of the uncertainty may be different from the impact in pure arrival scheduling problems.Therefore, uncertainty analysis is necessary to evaluate the expected benefits for integrated departures and arrivals.Two different scenarios were investigated.The solutions were first generated for deterministic scenarios, then flight entry times were perturbed in Monte Carlo simulations.Given the deterministic solution, a heuristic controller model was used to resolve conflicts caused by perturbations.The impact of the uncertainties on total delays and controller intervention was analyzed based on the outcomes of resolved conflicts.A sensitivity study of the delays and controller interventions with varied precisions of arrivals and departures was then conducted.In the paper, Section II revisits the problem modelled in previous work.The solutions generated under deterministic cases are presented.Section III presents the method for the uncertainty study, including Monte Carlo simulation set up.Section IV provides the analysis and results. +II. Problem and ModelThe interactions between arrivals and departures in Los Angeles terminal airspace were presented in previous work 12 for studying optimal integrated operations in deterministic circumstances.Because the uncertainty investigation in this work is based on the same problem, the model and problem statement from previous work are described in this section. +II.A. ProblemBased on the Standard Terminal Arrival Routes (STARs) and the Standard Instrument Departures (SIDs) of Los Angeles terminal airspace, the arrivals from FIM would follow procedure SADDE6 (FIM-SYMON-SADDE-SMO) and the departures to the North need to follow procedure CASTA2 (Runway-NAANC-GHART-SILEX) (see Fig. 1).The arrivals are requested to maintain their flight altitudes above 12,000 feet at Fix GHART and the departures need to remain at or below 9,000 feet at the same fix in order to procedurally avoid potential conflicts between arrivals and departure.If there was no interaction, departures to the north and arrivals from FIM would have flown direct routes.As shown in the Fig. 1, the direct routes would be RWY-WPT2-WPT1 and FIM-WPT1-SMO for departures and arrivals, respectively, where WPT1 and WPT2 are made-up fix names for simplicity.Compared to these direct routes in ideal situations, individual arrival and departure flights following current procedures will approximately fly an extra 60 and 120 seconds, respectively, and would also be flying non-preferred altitudes. +II.B. ModelingIn previous work, 12 three different separation methods were compared.They are spatial, temporal and hybrid separations.Spatial separation uses the same strategy as in the SIDs and STARs to spatially separate interacting flows, which can be treated as current operations.Temporal separation utilizes the direct routes with conflicts resolved solely with temporal controls.Hybrid separation applies both temporal and spatial separations.Three flows were taken into account in the problem: arrivals from FIM, northbound departures from Runway 24L (shown as "RWY" in Fig. 1 ), and westbound arrival flights towards SUTIE.In the formulation of hybrid separation, four design variables were defined for each FIM arrival: d1 i is the delay before or at FIM; r i is the route option, where 0 denotes the direct route and 1 denotes the indirect route; v i is the aircraft speed between FIM and WPT1 for the direct route or the speed between FIM and WPT2 if the indirect route is chosen; d2 i is the delay before or at SUTIE to ensure separation at SUTIE.For a departure flight, three decision variables were defined: d j is the delay before departure; r j is the route option, where 0 denotes the direct route and 1 denotes the indirect route.v j is the speed from departure to WPT1.There is only one decision variable for each arrival flight from the east, d k , which is the delay time at or before SUTIE.The separation requirements at all fixes are formulated as hard constraints.The objective is to minimize the total delay of all departures and arrivals.More details of the model including route structures and constraints can be found in a previous paper. 12 +III. Deterministic Case StudyTo perform the uncertainty study, deterministic solutions are first generated using a non-dominated sorting genetic algorithm (NSGA) 16 as in previous work. 12To be generalized, two typical cases were chosen from historical traffic data for this study.Case I includes three flows: departures to the North from Runway 24L, arrivals from FIM, and arrivals from the East.Case II contains only two flows without the flow from the East departures.The initial times are relative times to simulation start time.The "Order" of each flight is sorted based on initial times. +III.A. Initial condition +III.B. Deterministic solutionsTable 3 and Table 4 show the deterministic solutions from optimizations.Table 3 presents the total delays in Case I with varied uncertainty buffers and separation methods.Table 4 presents the total delays in Case II.Without the arrival flow from the East, Case II showed higher delay savings than Case I.In Case II, the hybrid separation achieved 94%, 83%, and 65% savings over the spatial separation for uncertainty 0s, 30s, and 60s, respectively.In this work, the route options from the deterministic solutions are used and fixed.And if necessary, additional delays are imposed to maintain required separation. +IV. Monte Carlo SimulationThe stochastic behavior of the flight time errors in arrival and departure flights will result in unexpected separation loss.Therefore, controllers have to spend extra effort to resolve unexpected conflicts, which unavoidably increase total delays and controller interventions.It will essentially reduce benefits claimed under deterministic scenarios.In order to study such behavior, a Monte Carlo simulation was implemented.Heuristic behaviors were modeled to mimic the controller intervention to prevent separation loss and to measure the extra delays. +IV.A. Perturbation in timesAs shown in Fig. 2, error sources that follow normal distributions were added in flight arrival times at waypoints FIM and SUTIE and departure times at departure runway, respectively.They will affect FIM arrivals, arrivals from the east, and departures, respectively.By default, the arrival time error has a standard deviation of 30 seconds and a mean of zero second, which has been commonly used as a desired prediction accuracy in many arrival trajectory prediction studies. 17,18 he departure time error's standard deviation is 90 seconds and the mean value is 30 second, which is based on the Call For Release (CFR) 3-minute compliance window.The window is often structured to allow departure 2 minutes prior to or 1 minute later than the target coordinated departure time. 19 +IV.B. Heuristic controller behavior modelIn order to simulate controllers' behavior of resolving separation loss and to measure the unexpected delay as a result of the intervention, a heuristic model was implemented in the Monte Carlo simulation.The number of times that controllers need to step in to resolve potential conflicts is associated with the increase of controller workload.In this model, the aircraft routes were decided by the solution already generated The first box contains the schedule optimization that developed and described in previous work. 12The second box shows the uncertainty study.The "Heuristic Conflict Solver" is the function that mimic a controller's intervention.The Monte Carlo simulation in this work includes 5,000 simulations and it can be run in less than one second on a MacOS platform with 2x2.66GHz 6-Core Intel Xeon and 8GB RAM.By feeding in live traffic, the first part can find deterministic solutions and the second part can be used as a quick reference to see if uncertainty can reduce or eliminate the benefit.The decision maker can thus decide if suggested operations should be executed. +V. ResultsThis section presents the impact of flight time uncertainty on delay reduction and controller intervention.The robustness of these performance measures to the departure and arrival time precision is also investigated. +V.A. Delay distributionAs in previous work, 12 three separation strategies were investigated.Besides the minimum separation requirement, uncertainty buffers of 0, 30, or 60 seconds were included when generating the schedule solution in deterministic scenarios.Figures 4(a) and 4(b) present the comparison among different strategies with varied buffers for Case I and II, respectively.In the horizontal axis, letters "S", "H", and "T", represent spatial, hybrid, and temporal separations, respectively.The numbers that follow these letters denote the values of extra buffers in seconds.For instance, "H30" means hybrid separation with 30 seconds buffer.The results in the figures are shown in Box-and-Whisker plots.In each case, the top and bottom horizontal lines are the maximum and minimum delays.The top and bottom boundaries of the narrow long box represent 90 th and 10 th percentiles, respectively.And the top and bottom boundaries of the wide short box are 75 th and 25 th percentiles, respectively.The horizontal line in the wide box is the mean value of the Monte Carlo simulation.From the figures, it is noted that the hybrid separation showed greater delay reduction compared with spatial separation.In the cases of 60 seconds buffer, Case II showed 55% saving at 90 th percentiles.There was still 20% delay savings even when comparing the worst case of hybrid solutions with the best case of spatial solutions according to the 5,000 Monte Carlo simulations.Although in Case I, the delays from the hybrid and spatial separations are similar in the worst case, the hybrid separation still has at least 9% delay savings over the spatial separation 90% of the time even with 60 second buffer.It is also noticed that generally as the buffer increases, the uncertainty of the delay reduction decreases, which means the benefit is more robust to flight time uncertainty if a larger extra buffer is applied in the deterministic solutions.As a trade-off, the differences of mean values between "spatial" and "hybrid" decreases, as do the benefits from hybrid separation over spatial separation.According to these uncertainty analysis charts, in both cases the integrated arrivals and departures showed advantages in delay saving, but the levels of savings are quite different in different scenarios. +V.B. Controller intervention distributionBy counting the times that controllers have to intervene to resolve separation loss, a controller intervention distribution can be generated under stochastic scenarios.Figure 5(a) and 5(b) show the comparison among different strategies and uncertainty buffers for Case I and Case II, respectively.By increasing uncertainty buffers in Case I, the average intervention drops.However, the uncertainty characteristics showed great connection to the schedules.For instance, comparing "S0" and "S60" in Fig. 5(a), the controller intervention decreases from 3 to 1. Whereas Case II's trend wasn't as strong, "S0", "S30", and "S60" showed similar mean controller intervention.On the other hand, a trade-off exists between delay reduction and intervention.A significant benefit was gained in delay reduction as discussed in the previous section, whereas controller intervention increases as shown in Fig. 5(a) and 5(b).For example, comparing "S0" and "H0" in Case I, the controller intervention changed from 3 to 5 with about 60% increase.A similar situation happened in Case II.Based on the comparison, a buffer of 30 or 60 seconds might be recommended to balance the trade-offs. +V.C. Sensitivity to departure and arrival time precisionsThe above studies assumed that the standard deviation of departure times is 90 seconds with a mean of 30 seconds and the standard deviation of arrival times is 30 seconds with a zero mean.Questions may arise regarding the impact of departure and arrival time precisions.How would departure and arrival time precisions affect delay reduction and controller intervention?In this section, a constant buffer of 30 seconds was used but arrival and departure time standard deviations were varied.To be conservative, the difference between the 75 th percentile of the "H30" distribution and the 25 th percentile of the "S30" was used to calculate delay robustness.(a) (b) As shown in Figs.6(a) and 6(b), cold color represents high delay savings and warm denotes low delay savings.The saving ranges anywhere from 140 seconds to 340 seconds in Case I and from 740 seconds to 940 seconds in Case II.In Fig. 6(a), the delay saving would be 5 minutes for an arrival deviation of 30 seconds and a departure deviation of 60 seconds.In Fig. 6(b), the same precision yielded 14 minutes savings.It is noted that in Case I, the delay reduction is less sensitive to the departure time precision than arrival time precision when the departure deviation exceeds 80 seconds.For example, with an arrival time precision of 30 seconds, the difference of delay reductions at departure time precisions of 2 minutes and above are similar.In Case II, similar phenomenon showed up when the departure deviation exceeded 120 seconds.One explanation for this phenomenon could be the relatively bigger gaps among departures than gaps in arrivals.For instance, the departure and arrival intervals in Case II are 352 s and 247 s respectively.The big gaps in departure flows provided the flexibility in absorbing delays caused by flight time errors.Another explanation would be the relatively low number of departure flights.Although these hypotheses need to be further verified, this kind of information is helpful to traffic managers when they are making decisions.show the delay reduction and controller intervention increase for 60 seconds buffer in both cases.Although the delay savings decreased and the controller intervention increased, similar trends still persist with 60 seconds buffer.However, the impact of departure precision on delay reduction has slightly increased compared with the results under 30 second buffer.This change may result from the decreased gaps between departure flights due to the increased buffers.Therefore, the ability to absorb delays has been weakened.It can be expected that the delay will be more sensitive to departure precision when the departure flights get closer or the buffer increases. +V.D. DiscussionThis post analysis of the uncertainty presents a clear picture of the stochastic characteristics of the solution generated under deterministic scenarios.This stochastic analysis is supplemental and necessary to the deterministic optimization.For example, in Case I, with a 60 second buffer, a great delay savings with more than 90% chance should be expected.A second example is: given an arrival deviation of 30 seconds, if the departure precision varies from 2 minutes (-1.5 minutes/+2.5 minutes) to 3 minutes (-2.5 minutes/+3.5 minutes), the expected delay savings in Case I should not change too much according to the uncertainty analysis.Compared to Case I, Case II still has significant delay savings even in the worst case, which makes Case II a good candidate for integrated operations.The study of these two scenarios demonstrated that analysis results are highly correlated to the scenario setups.These case dependent results suggest the necessity of an advisory tool which has the capability of real-time uncertainty analysis.If the uncertainty analysis can be combined with the schedule optimization in an advisory tool, according to the results provided by such a tool, decision makers would be able to issue the "H30" solutions in both cases with great confidence in expecting good delay savings and a manageable controller intervention increase. +VI. ConclusionsIntegrated operations between arrivals and/or departures provides a way to improve operation efficiency in terminal airspace.Previous studies 12 showed great benefits in deterministic circumstances.Because benefits from schedulers could be sensitive to flight time uncertainty, the robustness of the gained benefits must be investigated under uncertainty scenarios.This paper presents a method and analysis for the uncertainty study of integrated operations using two representative Los Angeles cases.Perturbations were incorporated into flight entry times in Monte Carlo simulations.Solutions generated in deterministic scenarios were used as references.In each simulation, the routes were fixed but extra delays were imposed if necessary to avoid separation loss caused by the pertubations in flight entry times.Impacts of the uncertainties on total delays and controller interventions were then presented and analyzed.A sensitivity study of the delays and interventions with varied precisions of arrivals and departures was also carried out.The study showed that in the two sample cases, deterministic solutions using hybrid separation still had substantial delay reduction over the spatial separation solutions 90% of the time when a 60 buffer was included.As a trade-off the controller intervention increased.When the buffer decreased, the intervention increase was high.For instance, in Case I, when the buffer was reduced to zero, the intervention increased from 3 in spatial separation to 5 in hybrid separation -a 60% increase.But in terms of absolute values, the intervention should still be manageable in these two sample cases.The sensitivity analysis of departure and arrival time precisions showed that the departure time precision had less impact on delay reduction than the arrival time precision, which might be caused by a sparse departure queue.But the departure time precision presented strong correlation with the controller intervention possibly due to the unavoidable need for conflict resolution.These analysis results are dependent on scenario setups, and they are supplemental and necessary components to the deterministic optimization and could be helpful to decision makers.Figure 1 .1Figure 1.Interactions between SADDE arrivals and CASTA departures +Figure 2 .Figure 3 .23Figure 2. Error sources in Monte Carlo simulation set up +Figure 33Figure3describes the entire work flow.The first box contains the schedule optimization that developed and described in previous work.12The second box shows the uncertainty study.The "Heuristic Conflict Solver" is the function that mimic a controller's intervention.The Monte Carlo simulation in this work includes 5,000 simulations and it can be run in less than one second on a MacOS platform with 2x2.66GHz 6-Core Intel Xeon and 8GB RAM.By feeding in live traffic, the first part can find deterministic solutions and the second part can be used as a quick reference to see if uncertainty can reduce or eliminate the benefit.The decision maker can thus decide if suggested operations should be executed. +Figure 4 .4Figure 4. Delay distribution under uncertainty (σ dep = 90s, µ dep = 30s; σarr = 30s, µarr = 0s) (a) Case I (b) Case II. +Figure 5 .5Figure 5. Controller intervention distribution under uncertainty (σ dep = 90s, µ dep = 30s; σarr = 30s, µarr = 0s) (a) Case I (b) Case II. +Figure 6 .6Figure 6.Delay reduction under varied uncertainties (extra separation buffer = 30 s) (a) Case I (b) Case II. +Figure 7 .7Figure 7. Controller intervention increase under varied uncertainties (extra separation buffer = 30 s) (a) Case I (b) Case II. +Figure 8 .8Figure 8. Delay reduction under varied uncertainties (extra separation buffer = 60 s) (a) Case I (b) Case II. +Figures 7 (7Figures 7(a) and 7(b) present the controller intervention with varied departure and arrival time precisions.Cold colors represent low controller intervention increase when comparing hybrid with spatial, and warm colors denote high intervention increase.The controller intervention showed a different pattern from the delay savings.The patterns in both figures implied that departure time precision dominated impacts on controller interventions, whereas arrival time precision showed much less effect.For instance, in Case I, with the arrival time deviation of 30 seconds, the controller intervention increase over "spatial" is around 1.3 (shown as negative 1.3 in the figure) when the departure time deviation is 60 seconds.The controller intervention increase changes to 1.6 when the departure time deviation is 90 seconds.Whereas, the controller intervention increase stayed the same when increasing arrival time deviation to 60 seconds and keeping departure time deviation at 60 seconds.One explanation could be that, although big gaps in departure flows provided the +Table 1 .1Initial times in Case IOrder FIM (sec) RWY (sec) SUTIE (sec)103043021352986713263540107048601240121051230NA13766NANA1780 +Table 2 .2Initial times in Case IIOrder FIM (sec) RWY (sec)13968244616537283634110652951332161361475183071613NA81770NA +Table 11shows the initial times for Case I based on traffic data between 10:30AM and 11:00AM (local time) on March 5, 2010.A total of 15 flights were involved including 5 FIM arrivals, 4 departures, and 6 westbound arrivals to SUTIE.Table2shows the initial times of 14 flights for Case II based on the traffic data between 9:00AM and 9:30AM (local time) on December 4, 2012.It includes 8 FIM arrivals and 6 +Table 3 .3Total delay with different separation methods in Case IUncertainty buffer Spatial Temporal Hybrid0 s1,001s334 s275 s30 s1,163 s805 s778 s60 s1,673 s1,694 s1,408 s +Table 4 .4Total delay with different separation methods in Case IIUncertainty buffer Spatial Temporal Hybrid0 s1,185s61 s73 s30 s1,185 s336 s207 s60 s1,195 s781 s423 s + + + + + + + + + + + The dynamic scheduling of aircraft in the near terminal area + + RGDear + + 10.1016/0041-1647(78)90133-8 + R76-9 + + + Transportation Research + Transportation Research + 0041-1647 + + 12 + 3 + + September 1976 + Elsevier BV + + + Tech. Rep. Flight Transportation Laboratory Report + Dear, R. G., "The Dynamic Scheduling of Aircraft in the Near Terminal Area," Tech. Rep. Flight Transportation Laboratory Report R76-9, MIT, September 1976. + + + + + An algorithm for computer assisted sequencing and scheduling of terminal area operations + + RogerGDear + + + YosefSSherif + + 10.1016/0191-2607(91)90132-a + + + Transportation Research Part A: General + Transportation Research Part A: General + 0191-2607 + + 25 + 2-3 + + 1991 + Elsevier BV + + + Dear, R. G. and Sherif, Y. S., "An Algorithm For Computer Assisted Sequencing and Scheduling of Terminal Operations," Transportation Research A, Vol. 25A, No. 2/3, 1991. + + + + + Analysis of Delay Reducing and Fuel Saving Sequencing and Spacing Algorithms for Arrival Spacing + + FNeuman + + + HErzberger + + 103880 + + October 1991 + + + Tech. Rep. NASA Technical Memorandum + Neuman, F. and Erzberger, H., "Analysis of Delay Reducing and Fuel Saving Sequencing and Spacing Algorithms for Arrival Spacing," Tech. Rep. NASA Technical Memorandum 103880, October 1991. + + + + + Scheduling Aircraft Landings Under Constrained Position Shifting + + HamsaBalakrishnan + + + BalaChandran + + 10.2514/6.2006-6320 + + + AIAA Guidance, Navigation, and Control Conference and Exhibit + Keystone, CO + + American Institute of Aeronautics and Astronautics + August 2006 + + + + Balakrishnan, H. and Chandran, B., "Scheduling Aircraft Landings Under Constrained Position Shifting," AIAA Guid- ance, Navigation, and Control Conference, Keystone, CO, 21-24 August 2006. + + + + + Scheduling Aircraft Landings—The Static Case + + JEBeasley + + + MKrishnamoorthy + + + YMSharaiha + + + DAbramson + + 10.1287/trsc.34.2.180.12302 + + + Transportation Science + Transportation Science + 0041-1655 + 1526-5447 + + 34 + 2 + + 2000 + Institute for Operations Research and the Management Sciences (INFORMS) + + + Beasley, J. E., Krishnamoorthy, M., Sharaiha, Y. M., and Abramson, D., "Scheduling Aircraft Landings -The Static Case," Transportation Science, Vol. 34, No. 2, 2000. + + + + + Scheduling Aircraft Landings to Closely Spaced Parallel Runways + + MKupfer + + + + The Eighth USA/Europe Air Traffic Management Research and Development Seminar + Napa, CA + + June 2009 + + + Kupfer, M., "Scheduling Aircraft Landings to Closely Spaced Parallel Runways," The Eighth USA/Europe Air Traffic Management Research and Development Seminar , Napa, CA, June 2009. + + + + + A Mixed Integer Linear Program for Airport Departure Scheduling + + GautamGupta + + + WaqarMalik + + + YoonJung + + 10.2514/6.2009-6933 + + + 9th AIAA Aviation Technology, Integration, and Operations Conference (ATIO) + Hilton Head, South Carolina + + American Institute of Aeronautics and Astronautics + September 2009 + + + + Gupta, G., Malik, W., and Jung, Y. C., "A Mixed Integer Linear Program for Airport Departure Scheduling," 9th AIAA Aviation Technology, Integration and Operations Conference (ATIO), Hilton Head, South Carolina, 21-23 September 2009. + + + + + A Mixed Integer Linear Program for Solving a Multiple Route Taxi Scheduling Problem + + JustinMontoya + + + ZacharyWood + + + SivakumarRathinam + + + WaqarMalik + + 10.2514/6.2010-7692 + + + AIAA Guidance, Navigation, and Control Conference + Toronto, Canada + + American Institute of Aeronautics and Astronautics + August 2010 + + + + Montoya, J., Wood, Z., Rathinam, S., and Malik, W., "A Mixed Integer Linear Program for Solving a Multiple Route Taxi Scheduling Problem," AIAA Guidance, Navigation, and Control Conference, Toronto, Canada, 2-5 August 2010. + + + + + A Generalized Dynamic Programming Approach for a Departure Scheduling Problem + + SivakumarRathinam + + + ZacharyWood + + + BanavarSridhar + + + YoonJung + + 10.2514/6.2009-6250 + + + AIAA Guidance, Navigation, and Control Conference + Chicago, IL + + American Institute of Aeronautics and Astronautics + August 2009 + + + + Rathinam, S., Wood, Z., Sridhar, B., and C., J. Y., "A Generalized Dynamic Programming Approach for a Departure Scheduling Problem," AIAA Guidance, Navigation, and Control Conference, Chicago, IL, 10-13 August 2009. 10 + + + + + A Hybrid Optimization Approach to Air Traffic Management for Metroplex Operations + + BrianCapozzi + + + StephenAtkins + + 10.2514/6.2010-9062 + + + 10th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + + American Institute of Aeronautics and Astronautics + + + + Capozzi, B. J. and Atkins, S. C., "A Hybrid Optimization Approach to Air Traffic Management for Metroplex Operations," + + + + + Departure Efficiency Benefits of Terminal RNAV Operations at Dallas-Fort Worth International Airport + + RalfMayer + + 10.2514/6.2006-7774 + + + 6th AIAA Aviation Technology, Integration and Operations Conference (ATIO) + Fort Worth, Texas + + American Institute of Aeronautics and Astronautics + September 2010 + + + + AIAA Aviation Technology, Integration and Operations Conference (ATIO), Fort Worth, Texas, 13-15 September 2010. + + + + + Towards Optimal Routing and Scheduling of Metroplex Operations + + BrianCapozzi + + + StephenAtkins + + + SeongimChoi + + 10.2514/6.2009-7037 + + + 9th AIAA Aviation Technology, Integration, and Operations Conference (ATIO) + Hilton Head, South Carolina + + American Institute of Aeronautics and Astronautics + September 2009 + + + + Capozzi, B. J. and Atkins, S. C., "Towards Optimal Routing and Scheduling of Metroplex Operations," 9th AIAA Aviation Technology, Integration, and Operations Conference(ATIO), Hilton Head, South Carolina, 21-23 September 2009. + + + + + Optimal Integration of Departures and Arrivals in Terminal Airspace + + MinXue + + + ShannonZelinski + + 10.2514/6.2012-4977 + + + AIAA Guidance, Navigation, and Control Conference + Minneapolis, MN + + American Institute of Aeronautics and Astronautics + August 2012 + + + + Xue, M. and Zelinski, S., "Optimal Integration of Departure and Arrivals in Terminal Airspace," AIAA Guidance, Navigation, and Control Conference, Minneapolis, MN, 13-16 August 2012. + + + + + Design Considerations for a New Terminal Area Arrival Scheduler + + JaneThipphavong + + + DanielMulfinger + + + AlexanderSadovsky + + 10.2514/6.2010-9290 + + + 10th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Fort Worth, Texas + + American Institute of Aeronautics and Astronautics + September 2010 + 14 + + + Thipphavong, J. and Mulfinger, D., "Design Consideration for a New Terminal Area Arrival Scheduler," 10th AIAA Aviation Technology, Integration and Operations Conference (ATIO), Fort Worth, Texas, 13-15 September 2010. 14 + + + + + Airport Arrival Capacity Benefits Due to Improved Scheduling Accuracy + + LarryMeyn + + + HeinzErzberger + + 10.2514/6.2005-7376 + + + AIAA 5th ATIO and16th Lighter-Than-Air Sys Tech. and Balloon Systems Conferences + Arlington, Virginia + + American Institute of Aeronautics and Astronautics + September 2005 + 15 + + + Meyn, L. and Erzberger, H., "Airport Arrival Capacity Benefits Due to Improved Scheduling Accuracy," 5th AIAA Aviation Technology, Integration and Operations Conference (ATIO), Arlington, Virginia, 26-28 September 2005. 15 + + + + + Design and Evaluation of a Stochastic Time-Based Arrival Scheduling Simulation System + + DanielMulfinger + + + AlexanderSadovsky + + 10.2514/6.2011-6874 + + + 11th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Virginia Beach, VA + + American Institute of Aeronautics and Astronautics + September 2011 + + + + Mulfinger, D. and Sadovsky, A., "Design and Evaluation of a Stochastic Time-Based Arrival Scheduling Simulation Sys- tem," 11th AIAA Aviation Technology, Integration and Operations Conference (ATIO), Virginia Beach, VA, 20-22 September 2011. + + + + + A fast and elitist multiobjective genetic algorithm: NSGA-II + + KDeb + + + APratap + + + SAgarwal + + + TMeyarivan + + 10.1109/4235.996017 + + + IEEE Transactions on Evolutionary Computation + IEEE Trans. Evol. Computat. + 1089-778X + + 6 + 2 + + 2002 + Institute of Electrical and Electronics Engineers (IEEE) + + + Deb, K., Pretap, A., Agarwal, S., and Meyarivan, T., "A Fast and Elitist Multiobjective Genetic Algorithm: NSGA-II," IEEE Transactions On Evolutionary Computation, Vol. 6, No. 2, 2002. + + + + + Improvement of Trajectory Synthesizer for Efficient Descent Advisor + + MinXue + + + HeinzErzberger + + 10.2514/6.2011-7020 + + + 11th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Virginia Beach, VA + + American Institute of Aeronautics and Astronautics + September 2011 + 18 + + + + Xue, M. and Erzberger, H., "Improvement of Trajectory Synthesizer for Efficient Descent Advisor," 11th AIAA Aviation Technology, Integration, and Operations Conference(ATIO), Virginia Beach, VA, 20-22 September 2011. 18 + + + + + Predictability of Top of Descent Location for Operational Idle-Thrust Descents + + LaurelStell + + 10.2514/6.2010-9116 + + + 10th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Berlin, Germany + + American Institute of Aeronautics and Astronautics + June 2011 + + + Stell, L., "Prediction of Top of Descent Location for Idle-thrust Descents," The Ninth USA/Europe Air Traffic Manage- ment Research and Development Seminar , Berlin, Germany, June 2011. + + + + + Trajectory-Based Takeoff Time Predictions Applied to Tactical Departure Scheduling: Concept Description, System Design, and Initial Observations + + ShawnEngelland + + + RichardCapps + + 10.2514/6.2011-6875 + + + 11th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Virginia Beach, VA + + American Institute of Aeronautics and Astronautics + September 2011 + + + + Engelland, S. A. and Capps, A., "Trajectory-Based Takeoff Time Predictions Applied to Tactical Departure Scheduling: Concept Description, System Design, and Initial Observations," 11th AIAA Aviation Technology, Integration, and Operations Conference(ATIO), Virginia Beach, VA, 20-22 September 2011. + + + + + + diff --git a/file825.txt b/file825.txt new file mode 100644 index 0000000000000000000000000000000000000000..51b6dda6fb7d336488bd6f715600de882bd0d32a --- /dev/null +++ b/file825.txt @@ -0,0 +1,484 @@ + + + + += average ρ(s)∀s ∈ N s (r) σ ρ (r)= standard deviation of ρ(s)∀s ∈ N s (r) (r)= average total delay for all flights in region r R (r) = recovered average total delay for region r R t (r)= recovered average throughput for region r [x 1 ... Several algorithms have been developed for repartitioning airspace into sectors given a set of flight tracks.Each algorithm attempts to laterally partition a layer of airspace to minimize and/or balance controller workload.They approach the problem in different ways and produce sectors that are different.Previous analyses assessed algorithmic success at a regional scale using different historical flight track data.Reference [1] compares a current day sectorization with those generated from two different algorithms.Reference [2] evaluates sectorizations generated from a single algorithm using different optimization functions.Both analyses are constrained to the Fort-Worth Center and compare aircraft-count metrics based on historical flight track data.Because of their use of historical flight track data only, there have been no analyses performed for expected future traffic levels.Thus far, a number of algorithms, including those cited above, have been developed but their relative value is not known at a national scale or for future traffic levels.This paper assesses the relative value of several algorithms for future traffic levels on the national level.That is, a sideby-side comparison using the same flight track data examined the comparative value of each algorithm.Also, in addition to considering current traffic levels, simulations at 1.5 time todays AF2009164 traffic level were conducted.These simulations expanded the previous area of consideration from the regional level to the entire continental United States airspace.This paper is organized as follows.Section II presents an overview of the sectorization algorithms.The experiment design is presented in section III.Section IV describes the metrics used to compare the sectorizations.Section V discusses results.Finally, concluding remarks are presented in section VI. +II SECTORIZATION ALGORITHMSThis section presents an overview of the three algorithms used to produce sectorizations.They all try to produce a sectorization that minimizes or balances a workload metric while conforming to traffic flows.The main objective function, method of conforming to flows, and method for determining the number of sectors are described for each. +A Flight Clustering AlgorithmThe Flight Clustering algorithm [3] groups flight tracks together to sectorize airspace.It is constrained to a maximum specified number of flight tracks per cluster.Sector boundaries are then formed around the clusters.This limits the amount of workload required to control a sector.The clustering algorithm approach allows Dynamic Density [11] metrics to be implicitly manipulated.It attempts to partition airspace in such a manner that acceptable Dynamic Density levels are achieved, and the impact on user-preferred flight routes is minimized.Flight route segments are clustered according to distances of clustering criteria from the cluster center.These criteria are selected and weighted to achieve control over the Dynamic Density of the resulting airspace partition. +1) Objective Function:The flight clustering objective is to minimize the sum of the 'distance' metric between flight tracks and the assigned cluster center based on the selected clustering distance criteria.The clustering criteria include current and future distance, lateral and vertical speed, and heading.Another clustering criteria, referred to as the 'Corridor' criteria, includes the perpendicular distance of the flight position from the major axis of the group of flight positions associated with the cluster and the flight's heading difference with respect to the major axis.All the above clustering criteria were used for this experiment.2) Flow Conformance: All of the clustering criteria attempt to group flight tracks that belong to the same flow, especially the corridor criteria. +3) Number of Sectors:The number of sectors is determined by a user defined target number-of-flight-tracks-percluster parameter.Flight tracks are clustered into groups within a minimum and maximum threshold of the target number. +B Voronoi Genetic AlgorithmThe Voronoi Genetic algorithm [2] uses a Voronoi Diagram to partition the airspace and a genetic algorithm to optimize the partitions.The Voronoi Diagram decomposes a space into subdivisions around given generating points.All coordinates within a region associated with a specific generating point are closer to that generating point than any other generating point.The genetic algorithm is a guided random search based on the principals of genetic inheritance and Darwinian evolution.Here, the genetic algorithm is used to find the set of Voronoi Diagram generating points that optimize given parameters.This algorithm can use any number of objective funtions.1) Objective Function: For this experiment, the genetic algorithm objective function is to maximize a capacity estimate minus peak aircraft count of each sector.The capacity estimate for each sector is directly related to the average flight time through that sector.2) Flow Conformance: Because the capacity estimate for each sector is directly related to the average flight time through that sector, sectors with longer average flight times will have higher capacity and work toward the objective function.Sectors with longer average flight times also tend to align to major flows. +3) Number of Sectors:The number of resulting sectors is determined by the number of Voronoi Diagram generating points chosen to partition a region of airspace.Assuming a goal average capacity of 15 aircraft per sector, the desired number of sectors for a given region is estimated by dividing the peak traffic count for the region by 15. +C Mixed Integer Programming AlgorithmThe Mixed Integer Programming (MIP) algorithm [4] discretizes the airspace into hexagonal cells and clusters the cells according to workload and connectivity.The workload of a cell is the number of flight track counts within that cell.Connectivity from cell i to a neighboring cell j is the total number of fights that travel from cell i to the cell j.Thus, connectivity is an abstract quantity of workload flow.Flow enters each cell from at least one of it's neighbors and exits into exactly one neighboring cell.The workload of each cell is added to the flow, which is finally absorbed by a sink cell.A sector consists of all cells whose flows converge to one sink.Potential sink cells are chosen at random.Improvements to the MIP algorithm were made in [5] by making connectivity between cells symmetric.By redefining the connectivity between two neighboring cells i and j to include both flights traveling from cell i to cell j and from cell j to cell i, flows become bidirectional and give the optimization more options to find a feasible solution with a lower objective function. +1) Objective Function:The MIP objective function is to balance the total number of flights tracks of a cluster of cells while minimizing the connectivity between cells in different clusters.2) Flow Conformance: Becasue the algorithm attempts to minimize the flow between cells in different sectors, the cell clusters will tend to be oriented along the dominant traffic flows 3) Number of Sectors: The number of resulting sectors is determined a priori. +III EXPERIMENT DESIGNThis section discusses how the experiment was designed.The goal was to perform a side-by-side comparison of the current day sectorization and several algorithm generated sectorizations for the continental US airspace.Therefore, each sectorization is generated according to the same guidelines using the same flight track data.Rather than using historical data to evaluate the sectorizations, flight track data is generated through simulation to remove the effects of current sector constraints and traffic management initiatives.Simulating the flight track data also allows the sectorizations to be evaluated for projected future traffic levels that do not exist in historical data.Simulations were completed using the Airspace Concept Evaluation System (ACES) [6].ACES models gate-to-gate flight operations on airport surfaces and in terminal and enroute airspaces.These include gate pushback and arrival, taxi, runway takeoff and landing, local approach and departure, climb, decent, transition, and cruise.Air traffic control and traffic flow management models control flights during these operations to ensure that airport and airspace capacity constraints are not violated.Fig. 1 shows the process used to generate sectors, estimate capacities, and simulate traffic for each algorithm.Each of the stages in this process are discussed in the following subsections. +A Unconstrained SimulationThe first stage of the experiment process is to simulate unconstrained flight tracks.The unconstrained simulation flies each flight without any airport or sector capacity constraints according to it's flight plan contained in the flight schedule.The flight schedule consisted of all the flight plans departing within 24 hours from a single high-traffic, good-weather day starting at 4/21/2005 8AM GMT.The last flight plan submitted before departure for each flight was used.The unconstrained simulation produced flight tracks for every minute of each flight.These flight tracks were used to generate new sectorizations with new sector capacities. +B Generating SectorizationsThe second stage of the experiment process generates new sectorizations using the unconstrained flight tracks produced by the first stage.Due to the infinite number of ways each sectorization algorithm could generate new airspace partitions, the algorithms are made to follow guidelines that make the resulting sectorizations more comparable.These are described below.The algorithms all produce lateral airspace divisions for a given set of flight track data.The current airspace sectorization is far more complicated than a single layer of sectors spanning the nation.First, national airspace is divided into regions called centers, and then, it is subdivided into sectors.Sectors are also stratified into low, high, and super-high altitudes.Sectors within the same stratum may have different altitude ranges.Future airspace operational concepts do not necessarily preclude the airspace from being redesigned irrespective of center boundaries.However, in order to make the sectorization comparable at a more regional level, all sectorization algorithms were constrained to redesigning the airspace within the current center boundaries shown in Fig. 2. A current day sectorization from 2005 was modified to be comparable with the algorithm generated sectorizations.By design, sectors were already organized within centers, but altitude ranges were not consistent between altitude stratum.To ensure that at least the same overall volume of airspace was compared, the 2005 sectorization was truncated below 23,950 feet.All sectors with a maximum altitude greater than 23,950 feet were given a minimum altitude of 23,950 feet.Any sector with a minimum altitude of 23,950 feet was classified as high altitude, and any sector that did not have another sector above it was classified as super-high altitude.Some sectors covered the entire altitude range and were classified as both high and super-high altitude sectors, even though they were not split. +C Estimating Sector CapacitiesAfter new sectorizations were generated, their sector capacities were estimated.Several methods for estimating sector capacity have been proposed in the literature [7,8].However, the most straightforward capacity estimation method, and the one used for this experiment, is the method used by the FAA to determine Monitor Alert Parameter (MAP) values [9].The MAP formula is a function of the average flight time of aircraft in the sector between 7AM and 7PM local time.δ(s) = 1 n f (s) i (t e (s, i) -t i (s, i))∀i ∈ N f (s) (1) c(s) =    5 if 5 3 δ(s) ≤ 5 5 3 δ(s) if 5 < 5 3 δ(s) < 18 18 if 5 3 δ(s) ≥ 18(2)These equations were used to calculate c(s) for all high and super-high altitude sectors in a current day sectorization from 2005 and for each of the new sectorizations.The FAA allows c(s) to be adjusted nominally +/-3 when needed based on the judgment of traffic management representatives.Therefore, the c(s) values calculated for the 2005 sectorization are different from their published MAP values. +D Constrained SimulationThe final stage of the experiment process is to simulate constrained flight tracks for each sectorization.Constraints are applied to airport and sector capacities in an ACES simulation.Air traffic control and traffic flow management models control flights to ensure that capacity constraints are not violated by delaying flights along their filed flight plan.For the purposes of this experiment, only capacity constraints on sectors within redesigned airspace are applied.Airport capacities and capacities for airspace outside the scope of this experiment are left unconstrained.The sector capacity constrained simulations result in a unique set of constrained flight track data for each sectorization simulated.These flight tracks and simulated delays are used to generate metrics with which to compare the sectorizations. +IV METRICSThis section discusses the metrics used to compare the sectorizations.All of the metrics can be applied to a group of sectors.The basic metrics are number of sectors, demand, throughput, capacity, complexity, and delay.The demand/capacity and throughput/capacity ratios are also of interest because ideally, capacity should be placed where demanded. +A Number of SectorsThere were no restrictions made on the number of sectors, n s (r), generated for a given region r in the new sectorizations.Everything else being equal, a lower n s is desirable to make more efficient use of controller resources.There is an inherent tradeoff between n s and capacity.As n s (r) increases, the sum capacity in r would be expected to increase as well.However, increasing n s reduces the average sector size which may make average δ(s) and c(s) for each sector lower.At some point, increasing the number of sectors will reduce average sector capacity so much that the sum capacity of the sectors does not increase. +B Demand and ThroughputDemand and throughput metrics are based on average instantaneous sector flight counts within quarter-hour intervals.Demand and throughput are computed from unconstrained and constrained flight track data, respectively.Demand and throughput metrics are computed for a region of airspace, r, by averaging the average instantaneous flight counts per quarterhour over the mid 8 hours of the day when traffic is highest.A region can comprise any group of sectors, such as a single sector, center, altitude layer, or NAS-wide.Let m(r, k) be the average flight count in r for quarter-hour, k.The average sector flight count in r is given bym(r) = 1 32 k0+32 k=k0 m(r, k).(3)Let the subscripts d and t denote demand and throughput.Let M d (r) and M t (r) be the average m d (s) and m t (s) for all s in r.Let σ m d (r) and σ mt (r) be the standard deviations for m d (s) and m t (s) for all s in r.Throughput is desired to be as close as possible to demand.The ratio of throughput to demand should be as close to 1 as possible.Recovered throughput is a metric designed to evaluate how much a new sectorization increases the throughout/demand ratio from the Current Day sectorization.Let R t (r) be the recovered average throughput in region r given byR t (r) = 1 - m d (r) -m t (r) m d0 (r) -m t0 (r)(4)where the subscript 0 denotes the Current Day sectorization.Assuming that maximum quarter-hourly flight count is a major component of controller workload, it is desirable for σ m d (r) and σ mt (r) to be low in order to balance the workload. +C CapacityThe capacity of each sector is defined according to (1) and (2).The c(s)s are used as constant constraints for the sector capacity constrained ACES simulations.Capacity sum, average, and standard deviation for regions or sectors are computed as well.Let Σ c (r), C(r), and σ c (r) be the sum, average, and standard deviation of c(s) for all s in r.Higher Σ c (r) and C(r) are desirable to enable increased throughput.Intuitively, a lower σ c (r) should be desirable.However, capacities themselves are designed to keep workload within acceptable limits.Therefore, there may be a tradeoff between balancing workload metrics and capacity.It is more desirable to balance workload metrics than capacity. +D Capacity RatiosA set of metrics that is perhaps more relevant than demand and throughput, or capacity, are the ratios of demand and throughput to capacity.Ideally, capacity should be placed where the demand is in order to maximize overall throughput.The demand/capacity ratio is a measure of how well the sectors accommodate the traffic, and the throughput/capacity ratio is a measure of workload levels.Let ρ d (s) and ρ t (s) be the average maximum quarter-hourly demand/capacity and throughput/capacity ratios for s over the mid 8 hours of the day given byρ d (s) = m d (s) c(s)(5)ρ t (s) = m t (s) c(s) .(6)Average and standard deviations of capacity ratios for regions of sectors are also computed.Let P d (r) and P t (r) be the average ρ d (s) and ρ t (s) for all s in r.Let σ ρ d (r) and σ ρ d (r) be the standard deviations of ρ d (s) and ρ t (s) for all s in r. +E ComplexityThere has been a lot of research done to develop complexity metrics that measure controller workload [10,11,12,13,14,15].These efforts concentrate on identifying and validating up to 52 quantifiable complexity variables based on the factors that contribute to workload.References [13], [14], and [15] present the most simplified subset of complexity metrics referred to as Simplified Dynamic Density (SDD) metrics.SDD metrics comprise just seven components that can be derived from historical track data.The combined SDD metric was chosen to represent complexity in this experiment.The seven components (x 1 through x 7 ) of SDD are occupancy count, proximity, altitude transition, sector boundary crossing, aircraft per sector volume, heading variance, and cruise speed variance.These are calculated per quarter-hour and are combined in a weighted sum.Component weights were taken from [15].Each component is described below.2) Proximity (x 2 ): Proximity events of different severity levels are calculated for all aircraft pairs within 10 nmi.Proximity severity levels in [13], [14], and [15] were designed to account for location uncertainty, time-stamps being recorded at different times, and one minute granularity.The proximity severity computations for simulation data have been simplified with respect to time, because the one minute time-stamps for each flight are produced for the exact same time.Table I shows the criteria for calculating proximity severity level between a pair of aircraft within the same time-stamp.The combined proximity component, x 2 (s, k), is defined as follows.x2 (s, k) = 1 4 (4p 1 + 2p 2 + p 3 + p 4 )(7)where p 1 , p 2 , p 3 , and p 4 indicate the number of proximities counted in s during k for each corresponding severity level.3) Altitude Transitions (x 3 ): Altitude transitions are counted for tracks that climb or descend more than 500 feet within a minute.x 3 (s, k) is the sum of all tracks within s during k that are not within 500 feet of their last track.4) Sector Boundary Crossings (x 4 ): Every time a flight crosses a sector boundary, a boundary crossing is counted for both the outbound and inbound sector.x 4 (s, k) is the combined number of flights that enter or exit s within k. +5) Heading variance (x 6) and speed variance (x 7 ): Heading and speed variances, x 6 (s, k) and x 7 (s, k), are calculated for the set of tracks in sector s within k.Variances are calculated for heading in degrees and for groundspeed in knots. +6) Combined SDD Components:The seven SDD components described above were combined in a weighted sum as follows.x = [x 1 , x 2 , x 3 , x 4 , x 5 , x 6 , x 7 ] (8) w = [2.2,.4,.3,.5, 30000, .0005,.0005](9) χ = w • x(10)where w weights were taken from [15].The average complexity for a given sector, χ(s), is calculated similar to m(s) as follows.χ(s) = 1 32 k0+32 k=k0 χ(s, k)(11)where k 0 is the first quarter hour in the 8 hours for which s has the highest average traffic counts.Average and standard deviations of complexity for regions of sectors are also computed.Let X(r) be the average χ(s) for all s in r.Let σ χ (r) be the standard deviation of χ(s) for all s in r. +F DelayACES collects the time for various events of each simulated flight.The difference between event times for unconstrained and constrained simulations is delay.Traffic flow management models may apply delay multiple times to the same flight in order to meet multiple constraints.Sector delays are computed as average total delay for all flights flying through the sector.Delays by region are computed similarly.The total delay for a single flight is the difference between it's constrained and unconstrained gate arrival time.Let (r) be the average total delay for all flights in region r given by(r) = 1 n f (r) i (t c (i) -t u (i))∀i ∈ N f (r) (12)where t c (i) and t u (i) are the constrained and unconstrained gate arrival times for flight i. +V SIMULATION RESULTSThis section presents and discusses the metric comparison results between sectorizations.Each of the three algorithms used flight tracks generated from an unconstrained simulation of a current day traffic schedule (1X) to create unique sectorizations subject to the guidelines discussed in section III.B. Fig. 3 shows the resulting sectorizations for high altitude Fort Worth Center in black overlaying flight tracks in grey.The most notable difference between the sectorizations are the number of sectors.This will be discussed in section V.A below.Another notable difference is in boundary smoothness.The rough boundaries in the MIP sectorization are a result of the hex cell clustering method.All sectorizations tend to direct the major axis of their sectors toward the middle of the center.This follows the major flow patterns in and out of this center's largest airport located in the middle of the center.The same set of unconstrained 1X flight tracks were used to estimate sector capacities for each of these sectorizations and the modified current day sectorization.These sector capacities were then used to simulate constrained 1X flight tracks through each sectorization.In order to test the sectorizations with more futuristic flight traffic levels, a flight demand generation tool called AvDemand [16] was used to homogeneously grow the 1X flight traffic schedule to 1.5X.Another unconstrained simulation was used to generate unconstrained 1.5X flight tracks for this traffic schedule.No new sectorizations or capacity estimations were generated using the unconstrained 1.5X flight tracks.Instead, the 1X generated sectorizations and capacity estimations were used to simulate constrained 1.5X flight tracks.The resulting 16 sets of flight track data were used to calculate the metrics defined in section IV.They consisted of 4 (sectorizations) by 2 (constrained and unconstrained) by 2 (1X and 1.5X).The following subsections discuss the compared results of these 16 sets of metrics. +A Number of SectorsNo design restriction for number of sectors for each region was placed on the sectorization algorithms.Therefore, some sectorizations resulted in very different numbers of sectors from the current day sectorization.altitude stratum.The NAS includes all 20 centers.For the algorithm generated sectorizations, n s (NASall) is the sum of n s (NAShigh) and n s (NASsuper).This relationship does not hold for the current day sectorization becasue there were many sectors classified as both high and super-high altitude due to the inconsistent altitude structure.The MIP algorithm was given a goal number of sectors per center similar to the average number of sectors at any one altitude for that center.The MIP number of sectors is the same for both high and super-high altitude stratums.The number of sectors for the Flight Clustering and Voronoi Genetic sectorizations were influenced by other algo- rithm parameters as described in sections II.A and II.B.Both resulted in fewer super-high sectors than high sectors.All the algorithm generated sectorizations have more total sectors than the current day sectorization, perhaps because of the NAS-wide altitude stratification design guideline.The Flight Clustering produced more than twice as many total sectors.If number-offlight-tracks-per-cluster parameter were increased, the resulting number of sectors would be decreased.A sensitivity analysis has not yet been completed to determine the optimal parameter setting.The relative numbers of sectors per center for each sectorization follow similar patters as the NAS.There are several exceptions for high altitude centers where the Voronoi Genetic algorithm produces close to or more than the number of sectors as the Flight Clustering Algorithm.These centers are ZDC, ZID, ZNY, and ZOB, all busy centers. +B CapacityLike number of sectors, capacity is something that is unique to each sectorization.Although 1X unconstrained track data was used to estimate the capacities of each sectioization, the same set of capacities was used in both the 1X and 1.5X constrained simulations.In general, average capacities, C(r), are lower for the algorithm generated sectorizations than the current day sectorization.However, because algorithm sectorization resulted in more sectors per region, the sum capacities, Σ c (NASall), are higher than for the Current Cay sectorization, especially in the case of the Flight Clustering sectorization.Fig. 5 shows Σ c (NASall) for each sectorization.The Flight Clustering and Voronoi Genetic algorithms show general increases in σ c (r) over Current Day, whereas the MIP algorithm shows general decreases in σ c (r) over Current Day.Consistency in capacity between sectors is not as important as constancy in capacity ratios or complexity discussed in later sections. +C Demand and ThroughputThe demand and throughput metrics are important metrics for assessing how well the sectorization accommodates traffic demand.These metrics compare track data from unconstrained and constrained simulations within the same 1X or 1.5X traffic level.Fig. 6 and7 show the NAS-wide average through- put/demand percentages for each altitude stratum for 1X and 1.5X traffic, respectively.Higher percentages indicate that the traffic is less constrained and throughput is being allowed to reach demand levels. +Altitude StratumThroughput/Demand As seen in Fig. 6 for 1X traffic, the algorithm generated sectorization increases average NAS throughput/demand percentages, but there is very little room to improve because the Current Day sectorization 1X throughput/demand percentage is already above 99%.In some cases, the percentage is slightly above 100%.This is possible due to minor traffic schedule shifting, which causes more aircraft to occupy an airspace at the same time than was originally demanded.The 1.5X traffic strains the system to more than the 1X traffic.It more effectively evaluates the algorithm generated sectorizations improvements over the Current Day sectorization.Fig. 7 clearly shows that all three algorithm generated sectorizations produce higher throughput/demand percentages than the Current Day sectorization.Flight Clustering and MIP sectorizations produce larger improvements at the high altitude stratum and the Voronoi Genetic algorithm produces the largest improvement at the super-high altitude stratum.Fig. 8 shows the recovered throughput, R t (r), for both stratums combined at each region.For each algorithm generated sectorization, there are one to three centers with negative recovered throughout.These are instances where the throughput is lower than in the Current Day sectorization.This can happen in less busy centers when the m d0 (r) -m t0 (r) term from Eqn. 4 is less than m d (r) -m t (r).In each case, the sectorizations with negative throughput/demand percentage are the ones with the lowest number of sectors for the center. +D Capacity RatiosThe standard deviations of demand/capacity and throughput/capacity ratios between sectors in a region are more important that the ratios themselves.The standard deviations, especially for ρ d as opposed to ρ t , indicate whether capacity is distributed appropriately to accommodate demand.A smaller σ ρ (r) means that the sectorization placed more capacity where demand needed it, and less capacity where it wasn't needed.Because, each sectorization was design for a single center at a single stratum at a time, it makes the most sense to evaluate σ ρ (r) at the single center and stratum level.act same trend between sectorizations and altitude stratum as in Fig. 9.The the average σ ρ d (r)s increased by 70% and the average σ ρt (r)s increased by 46% between 1X and 1.5X traffic levels.The percent increase in average σ ρ d (r) is the same as the percent increase in P d (r).This indicates that the 1.5X flight tracks are a good representation of homogeneous 1.5X growth from the 1X flight tracks.The percent increase in average σ ρt (r) is a little lower than the percent increase in P t (r) due to the smoothing effect of delaying flights to meet the capacity constraints. +E ComplexityComplexity is computed to serve as a more realistic measure of controller workload than just occupancy count.All of the sectorization algorithms utilize metrics that relate to complexity components other than occupancy count.Although, occupancy count is still a driving factor in the design.Fig. 10 shows the average NAS-wide complexity within each altitude stratum for 1X and 1.5X, and unconstrained and constrained simulations.Because occupancy count is a heavily weighted component of complexity, 1.5X has larger X(NAS)s than 1X, and X t (NAS) is lower than X d (NAS) for every sectorization as occupancy counts are controlled to be below capacity contraints.X t (NAS) is only slightly lower than X d (NAS) for 1X because the throughput did not deviate from demand much at 1X.The Flight Clustering sectorization reduces complexity the most due to it's high number of sectors and consequently low number of flights per sector.Other complexity components in- crease more for Flight Clustering due to the reduced average size of more sectors, but reduced occupancy count makes up for increases in other components.The Voronoi Genetic sectorization has the most similar X(NAS) between altitude stratums because number of sectors designed for each center and altitude stratum were based on the occupancy count.MIP is the only algorithm generated sectorization with higher X(NAS) than Current Day for the high-altitdue stratum.MIP also had the least number of sectors at this stratum.Fig. 11 shows the percentage each algorithm generated sectorization reduces X d (r) from the Current Day sectorization at 1X broken out by center regions.The Flight Clustering sectorization is the only one that reduces X d (r) for every center.Many of the centers for which the Voronoi Genetic or MIP sectorization increase complexity are the same centers for which R t (r) was negative in Fig. 8.These were the centers with the lowest number of sectors.Fig. 12 shows the average σ χ (r) for all centers within each altitude stratum for 1X and 1.5X and unconstrained and constrained simulations.Average σ χ (r) follows the same trend as X(NAS) between 1X and 1.5X and between unconstrained and constrained simulations.Super-high altitude average σ χ (r)s are consistantly lower than for High altitude.Average σ χ (r)s are significantly lower for the algorithm generated sectorizations than the Current Day sectorization.The Flight Clustering and Voronoi Genetic sectorizations have very similar average σ χ (r).The average σ χ (r) for MIP is much lower in each case.This means that the MIP algorithm did the best job of distributing workload, as measured by SDD complexity, evenly between sectors. +F DelayDelay is the ultimate measure of cost to the airspace system customers.Table II shows the average total delay simulated for all flights, (NAS), at 1X and 1.5X traffic for each sectorization.All of the algorithm generated sectorizations significantly reduce delay by similar amounts.While delays are still reduced at 1.5X, the amount varies more by sectorization.The Voronoi Genetic algorithm shows the most robust partitioning with respect to delay for increasing traffic demand.The Voronoi Genetic objective function directly affected delay by minimizing the possibility of traffic demand exceeding capacity constraints.Fig. 13 and 14 show the percent recovered delay, R (r), for each center region.There were more instances of negative R (r), indicating increased delay over the Current Day sectorization.This is because at many centers, the 1X traffic did not stress the system enough to incur significant delay, especially Cleveland Center (ZOB).ZOB is a very busy center for which the airspace was more recently designed than other centers to accommodate it's high demand.Only the MIP sectorization shows negative R (r) at 1.5X for the same three centers with significant negative R (r) at 1X. +Region Percent recovered delay +G Results SummaryThe benefits of new airspace configurations are expected to include improved system efficiency and more balanced controller workload.A few of the metrics discussed above, R t and R reflect system efficiency.σ χ d and σ ρ d reflect controller workload balance.These metrics, along with number of sectors, portray a summary of new sectorization benefits.Table III shows the total number of sectors, system efficiency, and workload balancing metrics computed for the entire NAS for each sectorization with 1.5X traffic.The best result for each metric is underlined.Some of these metrics portray the sector benefits a little differently with respect to one another when altitude stratum are combined vs. when they are computed separately.All three algorithm generated sectorizations show improvements in all summary benefit metrics except number of sectors.Flight Clustering more than doubles the number of sectors.Voronoi Genetic and MIP methods have more comparable numbers of sectors to current day but still increase the total number.This may be due to the NAS-wide altitude stratification guideline.There are many areas of the NAS where a single stratum of sectors is used from above flight level 240, whereas algorithm generated sectorizations assumed two stratum.Flight Clustering and Voronoi Genetic show comparable benefits in both system efficiency and workload balancing metrics, but the Voronoi Genetic sectorization has a much more comparable number of sectors to current day than the Flight Clustering sectorization.MIP shows lower system efficiency gains than the other two sectorizations, but it is very good at balancing complexity and demand/capacity between sectors, while maintaining a comparable number of sectors to current day.Due to it's low number of sectors, most robust increases in system efficiency, and second best workload balancing, the Voronoi Genetic sectorization appears to be the best sectorization compared in this experiment. +VI CONCLUSIONSThis experiment is the first US nationwide, side-by-side comparison of different algorithm-generated sectorizations.Simulating traffic through each of these sectorizations for a 1.5X traffic schedule improved the comparison by stressing the system enough to evaluate the algorithms strengths and weaknesses.Each algorithm shows its strengths and weaknesses through the different metrics.The Flight Clustering sectorization significantly increased throughput, while reducing complexity and delay, but only at the cost of doubling the number of sectors that exist in today's system.The Voronoi Genetic sectorization had a more comparable number of sectors to today's system, while increasing throughput and reducing delay similar to the Flight Clustering algorithms.The Voronoi Genetic sectorization had a more modest reduced complexity over the Current Day sectorization with respect to Flight Clustering.The MIP sectorization also had comparable numbers of sectors to Current Day with similar increases in throughput to Flight Clustering and Voronoi Genetic.However, the MIP algorithm's greatest strength was in balancing capacity and complexity.Overall, the Voronoi Genetic sectorization performed the best in this experiment.Two major realizations from this experiment were that the number of sectors designed for each region and the altitudes at which airspace is stratified are not trivial airspace design factors.These design factors need to be better incorporated into the algorithms.The number of sectors has been well integrated in to the Flight Clustering design but sensitivity analyses still need to be performed to evaluate how the rest of the metrics are affected as the number-of-flight-tracks-per-cluster parameter is increased and the number of sectors decreases.All three algorithm sectorization showed improved system efficiency and workload balancing over current day.This indicates that these sectorization methods have merit and are worth continued development and more detailed analysis.r = region comprised of a group of sectors t e (s, i) = flight i egress time from sector s t i (s, i) = flight i ingress time into sector s t u (i) = flight i unconstrained gate arrival time t c (i) = flight i constrained gate arrival time N f (s) = the set of all flights within sector s between 7AM and 7PM local time N f (r) = the set of all flights flying through region r N s (r) = the set of all sectors in region r n f (s) = total number of flights in set N f (s) n f (r) = total number of flights in set N f (r) n s (r) = total number of sectors in set N s (r) c(s) = capacity (MAP value) for sector s δ(s) = average time a flight spends in sector s m(s, k) = average flight count in sector s for quarter-hour k m(s) = average flight count in sector s between 4PM and 12AM UTC M (r) = average m(s)∀s ∈ N s (r) σ m (r) = standard deviation of m(s)∀s ∈ N s (r) Σ c (r) = sum of c(s)∀s ∈ N s (r) C(r) = average c(s)∀s ∈ N s (r) σ c (r) = standard deviation of c(s)∀s ∈ N s (r) ρ(s) = flight count/capacity ratio for sector s P (r) +x 7 ] = array of 7 complexity components [w 1 ...w 7 ] = array of 7 complexity weight coefficients χ(s, k) = sector s complexity for quarter-hour k χ(s) = average χ(s, k) for 4PM-12AM UTC X(r) = average χ(s)∀s ∈ N s (r) σ χ (r) = standard deviation of χ(s)∀s ∈ N s (r) subscripts d = demand, unconstrained t = throughput, constrained I INTRODUCTION +Figure 1 :1Figure 1: Experiment process. +Figure 2 :2Figure 2: Continental United States airsapce centers. +1 )1Occupancy Count (x 1 ) and Aircraft per Sector Volume (x 5 ): The occupancy count component for SDD metrics is the average instantaneous sector flight count for the given quarterhour.Therefore, x 1 (s, k) = m(r, k) where N s (r) = {s}.Aircraft per sector volume is simply x 1 divided by the sector volume in cubic kilometers. +Fig. 4 Figure 3 :43Fig.4shows each sectorization's total number of sectors for the National Airspace System (NAS), n s (NAS), at each +Figure 4 :4Figure 4: Total number of sectors in the National Airspace System, ns(NAS), for each altitude stratum. +Figure 5 :5Figure 5: Sum capacity in the National Airspace System, Σc(NASall), for each sectorization. +Figure 6 :6Figure 6: Average 1X Throughput/Demand in the National Airspace System, m t (NAS) m d (NAS) , for each altitude stratum. +Figure 8 :8Figure 8: Percent recovered throughput by region, Rt(r), for 1.5X traffic. +Figure 7 :7Figure 7: Average 1.5X Throughput/Demand in the National Airspace System, m t (NAS) m d (NAS) , for each altitude stratum. +Fig. 9 Figure 9 :99Fig.9shows the average σ ρ d (r) across all centers in each altitude stratum for each sectorization.All three algorithm generated sectorizations show a reduced average σ ρ d (r) over the Current Day sectorization.The MIP algorithm does an especially good job of distributing capacity appropriately to match demand.It has half the average σ ρ d (r) for high altitude and a quarter the average σ ρ d (r) for super-high altitude.The 1X average σ ρt (r)s are almost identical to the 1X average σ ρ d (r)s .This is not surprising, considering how little throughput deviated from demand for all of the 1X simulations.Both the average σ ρ d (r)s and σ ρt (r)s for 1.5X show the ex- +Figure 11 :Figure 10 :1110Figure 11: Percent that X d (r) is reduced from the Current Day to each algorithm generated sectorization by region for 1X traffic. +Figure 12 :12Figure 12: Average standard deviation of Complexity for all centers, σχ(r), within each altitude stratum. +Figure 13 :13Figure 13: Percent recovered delay by region, R (r), for 1X traffic. +Figure 14 :14Figure 14: Percent recovered delay by region, R (r), for 1.5X traffic. +TABLE I :IPROXIMITY SEVERITY LEVEL CRITERIASeverity LevelVertical Sep. Horizontal Sep.1<1000 ft<5 nmi2<1000 ft5 to 7.5 nmi3<1000 ft7.5 to 10 nmi4≥1000 ft<5 nmi +TABLE II :II(NAS) FOR EACH CONSTRAINED SIMULATION.Sectorization(NAS) for 1X(NAS) for 1.5XCurrent Day3.99 min45.02 minFlight Clustering 0.39 min18.68 minVoronoi Genetic0.47 min16.50 minMIP0.55 min30.11 min +TABLE III :IIINAS-WIDE SUMMARY METRICS FOR EACH SIMULATION WITH 1.5X TRAFFIC.SectorizationRtRσχ dσρ dnsCurrent Day--16.80.22470Flight Clustering 55%59% 13.40.161031Voronoi Genetic50%63% 14.00.17565MIP31%33% 12.50.18593 + + + + +ACKNOWLEDGMENTMany thanks to Charlene Cayabyab for her above-andbeyond excellent support on this research in the ACES lab at NASA Ames Research Center. + + + + + + + + + + + Dynamic Airspace Configuration Management Based on Computational Geometry Techniques + + JoeMitchell + + + GirishkumarSabhnani + + + RobertHoffman + + + JimmyKrozel + + + ArashYousefi + + 10.2514/6.2008-7225 + AIAA-2008-7225 + + + AIAA Guidance, Navigation and Control Conference and Exhibit + Honolulu, Hawaii + + American Institute of Aeronautics and Astronautics + 2008. August + + + Mitchell, J., Sabhnani, G., Krozel, J., Hoffman, B., and Yousefi, A. 2008. Dynamic Airspace Configuration management Based on Computational Geometry Techniques. In AIAA Guidance, Navigation and Control Con- ference and Exhibit, August, Honolulu, Hawaii. AIAA-2008-7225. + + + + + Airspace Sector Redesign Based on Voronoi Diagrams + + MinXue + + 10.2514/6.2008-7223 + AIAA-2008-7223 + + + AIAA Guidance, Navigation and Control Conference and Exhibit + Honolulu, Hawaii + + American Institute of Aeronautics and Astronautics + 2008 + + + Xue, M. 2008. Airspace Sector Redesign Based on Voronoi Diagrams. In AIAA Guidance, Navigation and Control Conference and Exhibit, Au- gust, Honolulu, Hawaii. AIAA-2008-7223. + + + + + Airspace partitioning using flight clustering and computational geometry + + ChrisRBrinton + + + StephenPledgie + + 10.1109/dasc.2008.4702800 + + + 2008 IEEE/AIAA 27th Digital Avionics Systems Conference + St. Paul, Minnesota + + IEEE + 2008 + + + Brinton, C. and Pledgie, S. 2008. Airspace Partitioning Using Flight Clustering and Computational Geometry. In 27th Digital Avionics Sys- tem Conference, St. Paul, Minnesota. + + + + + Trigger Metrics for Dynamic Airspace Configuration + + ArashYousefi + + + RobertHoffman + + + MarcusLowther + + + BabakKhorrami + + + HerbertHackney + + 10.2514/6.2009-7103 + No. 34N1207-001-R0 + + + 9th AIAA Aviation Technology, Integration, and Operations Conference (ATIO) + + American Institute of Aeronautics and Astronautics + December 2007 + + + Tech. Rep. Report + Yousefi, A., Khorrami, B., Hoffman, R., and Hackney, B. "Enhanced Dynamic Airspace Configuration Algorithms and Concepts, Metron Avi- ation Inc., Tech. Rep. Report No. 34N1207-001-R0, December 2007. + + + + + Analysis of an optimal sector design method + + MichaelDrew + + 10.1109/dasc.2008.4702801 + + + 2008 IEEE/AIAA 27th Digital Avionics Systems Conference + St. Paul, Minnesota + + IEEE + 2008 + + + Drew, M. 2008. Analysis of an Optimal Sector Design Method. In 27th Digital Avionics System Conference, St. Paul, Minnesota. + + + + + Build 4 of the Airspace Concept Evaluation System + + LarryMeyn + + + RobertWindhorst + + + KarlinRoth + + + DonaldVan Drei + + + GregKubat + + + VikramManikonda + + + SharleneRoney + + + GeorgeHunter + + + AlexHuang + + + GeorgeCouluris + + 10.2514/6.2006-6110 + AIAA-2006- 6110 + + + AIAA Modeling and Simulation Technologies Conference and Exhibit + Keystone, CO + + American Institute of Aeronautics and Astronautics + 2006 + + + Meyn, L., Windhorst, R., Roth, K., Van Drei, D., Kubat, G., Manikonda, V., Roney, S., Hunter, G., Huang, A., and Couluris, G. 2006. Build 4 of the Airspace Concept Evaluation System. In AIAA Modeling and Simu- lation Technologies Conference an Exhibit, Keystone, CO. AIAA-2006- 6110. + + + + + Applications of a Macroscopic Model for En Route Sector Capacity + + JerryWelch + + + JohnAndrews + + + BrianMartin + + + EricShank + + 10.2514/6.2008-7221 + + + AIAA Guidance, Navigation and Control Conference and Exhibit + Barcelona, Spain + + American Institute of Aeronautics and Astronautics + 2007 + + + Welch, J. 2007. Macroscopic Workload Model for Estimating En Route Sector Capacity. In 7th USA/Europe ATM R&D Seminar, Barcelona, Spain. + + + + + Predicting Sector Capacity for TFM Decision Support + + LixiaSong + + + CraigWanke + + + DanielGreenbaum + + 10.2514/6.2006-7827 + + + 6th AIAA Aviation Technology, Integration and Operations Conference (ATIO) + Barcelona, Spain + + American Institute of Aeronautics and Astronautics + 2007 + + + Song, L., Wanke, C., and Greenbaum, D. 2007. Predicting Sector Capac- ity for TFM. In 7th USA/Europe ATM R&D Seminar, Barcelona, Spain. + + + + + FAA Air Traffic Activity, Fiscal Year 1976. Federal Aviation Administration, U.S. Department of Transportation, 800 Independence Avenue, S.W., Washington, D.C. 20591. June 30, 1976. 293p + 10.1177/004728757701600140 + JO 7210.3V + + + + Journal of Travel Research + Journal of Travel Research + 0047-2875 + 1552-6763 + + 16 + 1 + + + SAGE Publications + + + cited 26 November 2008 + Federal Aviation Administration, Order JO 7210.3V, "Sec- tion 7. Monitor Alert Parameter," [online publication] URL: http://www.faa.gov/airports airtraffic/air traffic/publications/ ATpubs/FAC/Ch17/s1707.html [cited 26 November 2008]. + + + + + Dynamic Density and Complexity Metrics for Realtime Traffic Flow Management + + AMasalonis + + + MCallaham + + + CWanke + + + + 5th USA/Europe ATM R&D Seminar + Budapest, Hungary + + 2003 + + + Masalonis, A., Callaham, M., and Wanke, C. 2003. Dynamic Density and Complexity Metrics for Realtime Traffic Flow Management. In 5th USA/Europe ATM R&D Seminar, Budapest, Hungary. + + + + + Airspace Complexity Measurement: An Air traffic Control Simulation Analysis + + PKopardekar + + + + 7th USA/Europe ATM R&D Seminar + Barcelona, Spain + + 2007 + + + Kopardekar, P. 2007. Airspace Complexity Measurement: An Air traffic Control Simulation Analysis. In 7th USA/Europe ATM R&D Seminar, Barcelona, Spain. + + + + + Air Traffic Complexity: An Input-Output Approach + + KeumjinLee + + + EricFeron + + + AmyPritchett + + 10.1109/acc.2007.4282989 + + + 2007 American Control Conference + New York + + IEEE + 2007. July 11-13 + + + Lee, K., Feron, E., and Pritchett, A. 2007. Air Traffic Complexity: An Input-Output Approach. In Proceedings of the 2007 American Control Conference, July 11-13, New Youk City, New York. + + + + + Simplified dynamic density: A metric for dynamic airspace configuration and NextGen analysis + + AlexanderKlein + + + MarkDRodgers + + + KennethLeiden + + 10.1109/dasc.2009.5347539 + + + 2009 IEEE/AIAA 28th Digital Avionics Systems Conference + + IEEE + 2008 + + + unpublished + Klein, A., Rodgers, M., Lucic, P., and Leiden, K. 2008. "Airspace Com- plexity Metric for Dynamic Airspace Configuration: Simplified Dy- namic Density (SDD)," unpublished. + + + + + Dynamic Airspace Configuration Algorithms + + MRodgers + + + PLucic + + + AKlein + + + KLeiden + + + 2008 + + + unpublished + Rodgers, M., Lucic, P., Klein, A., and Leiden, K. 2008. "Dynamic Airspace Configuration Algorithms," unpublished. + + + + + Simplified dynamic density: A metric for dynamic airspace configuration and NextGen analysis + + AlexanderKlein + + + MarkDRodgers + + + KennethLeiden + + 10.1109/dasc.2009.5347539 + + + 2009 IEEE/AIAA 28th Digital Avionics Systems Conference + + IEEE + 2009 + + + unpublished + Klein, A., Rodgers, M., and Leiden, K. 2009. "Simplified Dynamic Density: a Metric for Dynamic Airspace Configuration and NEXTGEN Analysis," unpublished. + + + + + Futuristic US Flight Demand Generation Approach Incorporating Fleet Mix Evolution + + AlexHuang + + + DavidSchleicher + + 10.2514/6.2008-6678 + AIAA 2008-6678 + + + AIAA Modeling and Simulation Technologies Conference and Exhibit + Honolulu, Hawaii + + American Institute of Aeronautics and Astronautics + 2008. August 2008 + + + + Huang, A. and Schleicher, D. 2008. Futuristic US Flight Demand Gener- ation Approach Incorporating Fleet Mix Assumptions. In AIAA Model- ing and Simulation Technologies Conference an Exhibit, 18 -21 August 2008, Honolulu, Hawaii. AIAA 2008-6678. + + + + + + diff --git a/file826.txt b/file826.txt new file mode 100644 index 0000000000000000000000000000000000000000..c9bf72d96fb3b3ebd9ba7256018937f01055de0a --- /dev/null +++ b/file826.txt @@ -0,0 +1,1809 @@ + + + + +IntroductionOne of the greatest sets of choke points in the National Airspace System is the combined terminal area and surface.Arrival, departure, and surface operations are nodes within one great fluid network, however scheduling research has been largely segregated between these domains.Whereas the segregation simplifies each scheduler, the following resource utilization inefficiencies in the current system may be alleviated more by integrating.First, uncertainties passed between domains and between service provider and users are treated as hard constraints.The segregated domains use firstcome-first-served traffic management to cope with the uncertainty, which can add inefficiencies and can even increase the uncertainty.For example, the firstcome-first-served traffic management policy incentivizes front-loading arrival meter fixes, arrival runways and departure queues.The current system uses this technique to maintain high throughput in the midst of high uncertainty by providing queues close to constrained resources.However, queue congestion adds delay to trajectories, reducing fuel efficiency.Airlines not only burn extra fuel to meet their schedules, but they pad their schedules, which further escalates uncertainty.Detailed information passed between domains and more intelligent management or negotiation of the uncertainties could alleviate some of the constraints and allow more precise queuing, resulting in more efficiency.Second, terminal areas with constrained resources often favor one set of operations over another (e.g.arrivals over departures or one airport over another within the same terminal area), and they typically do not consider user preferences when scheduling to these resources.This is because each set of operations is scheduled separately to gain control of resources on a first-come-first-served basis.Often the operations that have the earliest opportunity to reserve the resources get the resources, even if it causes another part of the system to gridlock.A more integrated approach could more equitably schedule the constrained resources to benefit the system as a whole.Finally, the segregated schedulers make the system slow to adjust to changes or recover from disturbances.For example, if convective weather forces a stream of arrivals to a different arrival meter fix, they may be assigned more path distance than necessary to conform to the runway configuration.By the time the runway configuration changes to adapt to the new dominant flow direction, the weather may have cleared, returning the flow to normal.Integrated scheduling may enable downstream operations to adjust to upstream disturbances proactively rather than reactively, enhancing the system's robustness and resilience.This paper proposes a framework for integrating scheduling between arrival, departure, and surface operations to address the potential drawbacks of segregated scheduling.The goals of integrated scheduling are:1. Reduce and manage uncertainty to simultaneously maximize throughput, efficiency and schedule integrity.2. Equitably manage operations competing for resources and incorporate user preferences.3. Enhance system robustness and resilience to change or disturbances.Although integrated scheduling can improve operations efficiency, it may also increase complexity related implementation cost and operator workload.Therefore, integrated scheduling development should also consider the degree of integration necessary to achieve the above goals while minimizing complexity.The following section discusses the current state-of-the-art both in the field and in NASA development for each of the segregated scheduling domains and the limitations of segregating these domains.Then an integrated scheduling hierarchy is proposed that organizes scheduling tasks by time horizon rather than domain.Finally, this paper identifies current NASA research gaps and proposes key areas where future research should focus to facilitate scheduler integration. +Segregated SchedulingThus far, terminal area schedulers have been individually developed to solve very domain-specific problems with different time horizons.This section describes individual flight schedulers decomposed into the arrival, departure, and surface domains shown in figure 1.The configuration scheduler is discussed as a fourth domain, which allots time windows when given resources (meter fixes, routes, runways, taxiways, etc.) are available to the individual flight schedulers. +Arrival SchedulingArrival scheduling focuses on the phase of flight just prior to top-of-descent to landing.Flights that depart from airports within the arrival scheduling horizon are also included.Arrival schedulers primarily aim to maximize arrival throughput and fuel efficiency.This is achieved by scheduling arrivals to precise time slots at coordination points (i.e.meter fixes, merge points, and runways) along an assigned fixed route from meter fix to runway.The accuracy with which flights can conform to their scheduled slots determines the amount of scheduling buffer required between slots.The more accurate the schedule conformance, the smaller the scheduling buffers required, the greater the potential throughput.The most mature example of arrival scheduling NASA technology is the Traffic Management Advisor with Terminal Metering (TMA-TM 1 ) [1] which uses a staged first-come-first-served arrival scheduler with a freeze horizon ~40 minutes prior to the estimated time of arrival (ETA) at the runway.For jet aircraft, this freeze horizon extends into Center airspace, ~150 nmi radius from the Terminal Radar Approach Control (TRACON) boundary and ~200 nmi radius from airport.This large freeze horizon enables the scheduler to push back all delay not anticipated to be absorbed with speed control within the TRACON to the Center.This minimizes vectoring in the TRACON, which serves to increase schedule conformance, leading to the primary goal of increased throughput.This also achieves a secondary goal of increased fuel efficiency by minimizing level segments in the TRACON so that continuous lowthrust descents can be executed without interruption [2].Flights with frozen arrival schedules are monitored and controlled precisely to meet the schedule at each coordination point along the fixed path either by on-board avionics for aircraft with highly advanced equipage or by TRACON controllers.It is assumed that any deviation from this schedule can be kept within the bounds of the scheduling buffer.If not, it is the controller's responsibility to tactically fit the flight within a natural gap in the schedule.If the schedule is sufficiently saturated due to high demand, there are few gaps to absorb deviating flights and the disturbance can extend all the way to Center airspace.In such cases, when the schedule is deemed unrecoverable, all schedules are unfrozen and rescheduled, referred to as "list rippling".A recent study attempted to automate this form of rescheduling to accommodate go-around procedures [3].Currently, an advanced tactical component of the arrival scheduler is being developed at NASA Ames to mitigate the impact of scheduling disturbances called Method to Enhance Scheduled Arrival Robustness (MESAR) 2 .This component monitors the system to catch emerging scheduling disturbances early and selectively re-schedules the a subset of flights necessary to recover the schedule.In a packed schedule situation, the tactical scheduler can temporarily reduce the scheduling buffers for the subset of flights being re-scheduled due to the higher accuracy of arrival time estimates once flights are within the TRACON. +Surface SchedulingSurface scheduling focuses on the taxi phase of flight from the arrival runway to the gate, and the gate to the departure runway.One surface scheduler objective is to minimize total delay or maximize runway throughput.Another objective is to minimize fuel burn by minimizing taxi-time on the active movement area such that most of the taxi-time is spent in active taxi rather than waiting at taxi intersections or runway crossings.Minimal taxi-time is achieved by holding flights at the gate for as long as possible to reduce congestion.However, due to large taxi-time uncertainties and arrival constraints, gate holding for just-in-time departure can reduce throughput.Due to the large uncertainties, surface schedulers have placed less focus on meeting specific departure times and more focus on managing spot and runway queue size to simultaneously maximize efficiency and throughput.Unfortunately, the greatest uncertainties do not lie within the active movement area (i.e.taxiways and runways), but with gate pushback time, and ramp movement time from gate pushback to spot (entrance to active movement area), both of which are most often controlled by airlines and not the Air Navigation Service Provider (ANSP).For this reason, not only do surface schedulers typically start at the spot, but they continuously reoptimize the schedule on a short planning horizon and even shorter freeze horizon to mitigate the uncertainty.Recently surface metering has been field tested at several airports [4,5].When gate-departure demand exceeds runway departure capacity, individual flights are held at the gate or pre-assigned holding pad with engines off to shorten the queue lines.In addition to reducing taxi-out time and therefore fuel consumption, surface metering reduces takeoff delay.The most mature example of surface scheduling NASA technology is the Spot and Runway Departure Advisor (SARDA) [6].The most recent human-inthe-loop (HITL) simulation of SARDA at Dallas Fort Worth International used a 15-minute planning horizon that reschedules every 10 seconds, and only the first three flights in both spot release and runway sequence from the last scheduling iteration were frozen [7].SARDA uses a two-stage scheduler.The first stage considers the departures and arrivals expected at the runway within a 15-minute planning window along with wake vortex and traffic management initiative constraints, and generates expected takeoff times for departures.The second stage of the scheduler generates departure spot release times by subtracting nominal taxi times with an uncertainty buffer from the runway schedule.More recently, the algorithm has been extended to include gate pushback times in the ramp area by subtracting nominal transit time between gates and spots from spot release times.A candidate concept for an integrated system of both strategic and tactical components of SARDA advisory tools, via a Collaborative Decision Making (CDM) mechanism between airlines and ATC, has been developed [8].The strategic advisory component assigns gate pushback times with a planning window that can range from 30 minutes to 2 hours.SARDA-CDM guarantees a spot release time window to flights that pushback no later than their assigned pushback time.Flights that do not meet their assigned latest gate pushback time are scheduled tactically subject to availability. +Departure SchedulingDeparture scheduling focuses on the phase of flight from takeoff to cruise.Departure schedulers aim to meter and de-conflict flights at departure meter fixes or fill specific slots in en route or arrival streams.Currently fielded departure scheduling is a function of TMA called tactical departure scheduling [9].TMA's en-route departure capability schedules outbound departures to metering arcs that join enroute streams.Arrival TMA schedules inbound departures that originate within the arrival freeze horizon of the destination airport.The planning horizon of tactical departure scheduling is similar to that of arrival scheduling in order for inbound departures to compete fairly for arrival runway slots.However, the accuracy of departure scheduling is severely limited by the current state of departure trajectory prediction and uncertain departure wheels-off times [9].Departure trajectory prediction suffers from a wide variation of departure path (even along Area Navigation RNAV routing) and inaccurate aircraft weight estimates [10,11].Departure wheels off time uncertainties are due to all the same uncertainties that plague surface scheduling.Departure scheduling is especially challenging when weather blocks departure fixes or gates causing departure fix compression or fix swapping.Not only are fewer meter fixes available, but miles-in-trail restrictions are often imposed on the remaining fixes to account for increased uncertainty and spacing requirements associated with vectoring near the fixes [12]. +Integrated Departure Route Planning (IDRP) is a Traffic Management Coordinator (TMC) Decision +SupportTool (DST) planned for 2017 implementation in the field [13,14].It integrates information about weather and congestion impacts on departure routes into constraints and recommends how to reroute not-yet-airborne departures to avoid the constraints with a 30-to 60-minute planning horizon.Expedited Departure Path (EDP) was a tactical Air Traffic Controller (ATC) DST designed to offer advisories for optimal sequencing and merging of departures to departure meter fixes as scheduled [15].A secondary goal of departure schedulers is to enable continuous climb by opportunistically shooting gaps in arrival streams.EDP offered tactical advisories for continuous climb when no lateral conflicts with crossing traffic are found.Research in departure control DSTs continued with Sharing of Airspace Resources (SOAR) [16].SOAR has been developing communication procedures and DSTs to enable schedule-based continuous climb of departure through crossing arrival streams, and efficient departure stream merging at departure meter fixes.The most mature example of NASA technology facilitating departure scheduling is Precision Departure Release Capability (PDRC) recently transferred to the FAA [17].PDRC schedules departures constrained by Call For Release procedures, which require the Tower to get Center approval prior to releasing departures to specific destinations.PDRC improves the accuracy of the OFF time predictions used by tactical departure scheduling by enabling the surface scheduler to share it's predicted OFF times with the departure scheduler.For the current surface management system, this pushes back the tactical departure scheduling horizon to the spot.In site tests, PDRC also made more specific horizontal profile TRACON departure routing and departure runway assignment available to the departure scheduler to reduce TRACON transit time error.NASA is currently developing a departure scheduler that extends the tactical component of surface scheduling to consider terminal airspace constraints [18].Departures are continuously resequenced and rescheduled across multiple airports on a 5-second scheduling cycle to produce controller OFF times ensuring that minimum separation is maintained at both the runway threshold and departure fix.This enables departure scheduling coordination between multiple airports sharing departure fixes. +Configuration SchedulingConfiguration scheduling does not operate on individual flights, but rather it determines time windows for which specific terminal area resources (routing, runways, meter fixes, taxi-ways etc.) are available.The configuration or available resources for a given time period narrows down the routing options the arrival, departure, or surface scheduler may assign to an individual flight.The simplest form of configuration scheduling is when a group of route segments and fixes are blocked by convective weather rendering the routing inaccessible.If convective weather can be predicted in advance of the scheduling horizon, the scheduler can avoid assigning blocked routes.Currently fielded Route Availability Planning Tool (RAPT) assigns convective weather blockage status to departure routes up to 30 minutes in advance to help traffic managers determine if and when specific published routes are available for use [19].Dynamic routing research has developed algorithms to create dynamic meter fixes or routes around blocked airspace [20][21][22][23][24].These algorithms could be used to design a larger set of pre-defined weather contingency routes than currently exists.A more advanced solution to blocked routing is to use these algorithms to dynamically generate temporary routes around the weather in real-time.Recently NASA experimented with how dynamic routing could be used in extended terminal airspace to funnel arrivals around convective weather to an arrival meter fix [25].Dynamic routes were designed with a 45-minute freeze horizon and updated every 15 minutes.From a controller perspective, several different routing structures may be active within the airspace at the same time.However, a given individual flight will enter a single route structure predicted to be unblocked and stable for the entire time the flight traverses the airspace to the meter fix.The arrival scheduler will then use the route structure associated with a given flight to schedule its meter fix, merge point, and runway crossing times.Other resources frequently associated with configuration scheduling are runways.The runway configuration (which runways may be used for arrival, departure, or both) determines the available TRACON routing as well.A runway configuration change may be scheduled to adapt to a change in winds, visibility, noise or emission level curfews, runway obstruction, traffic volume, dominant traffic direction, or an arrival/departure push.Several algorithms have been developed to generate optimal runway configuration schedules [26][27][28][29][30][31].The most mature example of configuration scheduling NASA technology is Tactical Runway Configuration Management (TRCM) designed to select runway configuration plans that maximize throughput and minimize delays associated with transitioning between configurations [31].As with dynamic routing, TRCM is expected to use a ~45 minute freeze horizon large enough to inform individual flight schedulers of upcoming configuration changes prior to generating their individual flight schedules.TRCM also limits the frequency of major directional shifts in configuration to no more than 1-in-30 minutes, and minor runway assignment policy changes to no more than 1-in-15 minutes. +Segregated Scheduling LimitationsArrival scheduling development has not considered departures as constraints to minimize some of the uncertainty problems surface and departure schedulers faced.For aircraft nearing their destination, excessive delay burns extra fuel, which would soon be exhausted, making landing aircraft as quickly as possible a top safety priority.Therefore, scheduled arrival times were passed to the surface scheduler as hard constraints.As the precision of arrival operations improves and uncertainty diminishes, the resulting precision of arrival constraints is a huge benefit to the surface scheduler, enabling it to find gaps for inserting takeoffs on mixed usage runways and efficiently sequence takeoffs and departure runway crossings.But precise arrival constraints do little to help departure scheduling in peak arrival conditions if they are so tightly packed that there are no gaps for departure schedulers to use.This is why mixed usage runways have either alternating arrival/departure pushes, or they handle overflow arrival operations that do not saturate the runway.For the most part, arrival routes are procedurally segregated from departure routes.This is done laterally where possible to create more efficient continuous descent and climb vertical profiles.However, this segregation technique can extend the length of both arrival and departure routes.Where this kind of segregation is not possible, arrivals are typically given the more efficient path, leaving departures to tunnel underneath arrival streams or fly extra path miles to loop above them.In some cases, departures are left to opportunistically shoot gaps in arrival streams to fly more efficient vertical profiles.Such procedures require a lot of extra controller attention and so are not often used at busier or complex TRACONs or in higher traffic volume conditions, where they could have the greatest benefit.Opportunistic changes to departure trajectories also affect departure transition times to meter fixes or en-route slots.Most of the disadvantage given to surface and departure schedulers comes from large uncertainties, the largest of which is gate pushback time.This uncertainty induces a large tradeoff between precision and scheduling horizon, influencing these schedulers to focus on very tactical, short time horizon solutions.These short time horizon solutions make it very difficult for departures to compete with arrivals for precisely scheduled resources. +Integrated Scheduling FrameworkWhereas there is a clear functional distinction between configuration scheduling and the other domains, the boundaries are less distinct between arrival, departure, and surface scheduling.These domains are making progress towards the first integrated scheduling goal of reducing and managing uncertainty and third goal of enhancing system robustness and resilience, but not the second goal of equitably managing resources between them to maximize system efficiency.Individual scheduler performance is limited by hard constraints imposed by segregation and they do not address scheduler imbalance.Integrated procedures have been disregarded due to lack of precision and robustness, but recent individual scheduler advances make this next step more viable.Refocusing research toward arrival/departure/surface scheduler integration strategies will not only free scheduling research from accustomed hard constraints, but will begin to address scheduler imbalance.The solution is by no means one mega, global optimization engine.Even the individual domain schedulers are broken down into functional elements, some with different time horizons, which share information.The arrival scheduler development, while initially less tactical, has discovered that this leaves the solutions vulnerable to disturbances too large to be rectified by it's narrow range of control authority.Current research is adding a more tactical component (MESAR's selective re-scheduling) to the arrival scheduler to cope with large disturbances.On the other hand, the surface scheduler is no stranger to large disturbances (i.e.uncertainty).Surface scheduler development has discovered that its highly tactical and reactive approach to optimizing the solution may be the best way to manage uncertainty.However, recent research added a more strategic component (SARDA-CDM) to the surface scheduler to manage its largest disturbance, gate pushback uncertainty, making it possible to generate a candidate solution prior to tactical optimization.As the arrival scheduler becomes more tactically capable and the surface scheduler becomes more strategically capable, these domains are poised for integration.In order for these segregated scheduler domains to integrate, they must be broken down into common hierarchical components to ensure that information is passed between domains at every level.Figure 2 shows a flow chart of information exchange between four hierarchy levels: 1) configuration schedule, 2) flight schedule, 3) flight schedule update, and 4) schedule conformance.Each flow chart element represents arrival, departure, and surface domains. +Level 1: Configuration ScheduleThe configuration schedule occupies the top level of the integrated scheduling hierarchy with a planning horizon on the order of hours and a freeze horizon of at least 45 minutes to provide the subset of resources available to the lower levels.Inputs used to generate a configuration schedule solution include a library of resources and configurations that utilize predefined sets of resources, nonnegotiable and negotiable constraints, and demand forecast.Nonnegotiable constraints, such as weather forecast, pre-filter the solution space and static negotiable constraints are used to refine the solution space and generate a configuration schedule tailored to the demand forecast.Nonnegotiable constraints are constraints that automatically prohibit the use of a particular set of resources and any configurations that make use of them.Weather-related nonnegotiable constraints are the most numerous and least predictable.Convective weather can prohibit the use of specific route segments, fixes and runways.Wind direction and magnitude can prohibit the use of specific runways in one or both directions.Some weather conditions only prohibit a particular configuration and not the resources themselves.For example, Instrument Meteorological Conditions may prohibit the use of both parallel runways as arrivals, but one may be used for arrivals and one for departures.Other nonnegotiable constraints are policy driven.For example, some arrival or departure routes over heavily populated areas may be prohibited during certain hours of the day. +Figure 2. Integrated Scheduling HierarchyNegotiable constraints prohibit the use of some resources or configurations at the same time as others.For example runways and route segments may not be used (or it would be extremely inefficient to schedule their use) in both directions at the same time.Herein lies the choice of which set of coexistent configurations to make available to lower hierarchy schedulers.At this point, traffic demand input is evaluated to maximize efficiency or throughput, and minimize the cost of transitioning from one configuration to another.In general, the transition cost is less, the more similar the new configuration is to the old, or the earlier the configuration change is planed before the change occurs.Frequency of a With a planning horizon on the order of hours, resource demand for preferred routes may be evaluated as a rate (e.g. per 15 minutes).Short notice resource related disturbances that necessitate a frozen configuration schedule to update, may require evaluating demand at the individual flight level and coordinating directly with the flight schedule update level.Due to the potential cascading disruption a short notice configuration schedule update may have, these should only be triggered by nonnegotiable constraints.Examples include unpredicted pop-up weather, sudden change in wind direction or magnitude, visibility, or runway/taxiway closures due to aircraft mechanical failure or other obstruction.Solutions to configuration schedule updates may call upon resources reserved specifically for solving short-notice configuration changes such as transition routes or holding areas.For example, the runway itself can be used as a taxiway when queued aircraft have to move to the other end of the runway due to a configuration change.Evaluating the efficiency or throughput benefit of a configuration can be quite complex and requires detailed domain-specific traffic information as demonstrated by TRCM research.TRCM attempts to globalize configuration scheduling, but the process can be modularized when the negotiable constraints governing the configurability of one set of resources are independent from another.For example, arrivals from a similar direction generally enter a TRACON in the vicinity of the same arrival fix regardless of the runway configuration in effect.Therefore, dynamic routing from a given arc direction to a given meter fix could remain separate from runway-configuration scheduling.The meter fix becomes the coupling point between the two configuration schedulers.If the runway configuration change requires that the meter fix location be moved, this information must be passed to the dynamic routing scheduler with enough lead time to satisfy the freeze horizon. +Level 2: Flight ScheduleThe flight schedule hierarchy level provides the initial optimal flight-specific schedule.As flights enter the planning horizon, the flight scheduler determines the available route options for each flight given the active configuration, and calculates feasible scheduled time of arrival (STA) ranges at coordination points along each route.Available route options may include several different routes between multiple meter fix/runway pairs.Route availability may be subject to individual aircraft performance capabilities and equipage.The scheduler then tries to find the best route and schedule within this solution space that minimizes delay, maximizes throughput and efficiency, and may incorporate airline preference input as well.Ideally, arrivals and departures would be scheduled together at the same planning horizon.Distributed parallel processes could prioritize route options based on cost functions incorporating user preferences and calculate feasible STA ranges for each domain or even each individual flight.But this information should be fed to a centralized scheduler to organize the constraints and costs into an optimal schedule.However, if some of this information (e.g.gate pushback time) is not available at a comparable level of uncertainty within the same time horizon, then a multi-stage scheduler is needed.Figure 3 diagrams information flow for a twostage coordinated flight scheduling approach between arrival and surface/departure schedulers.In an attempt to balance arrival/departure demand and capacity, the first stage arrival scheduler uses flight plan departure times to schedule some gaps in the arrival schedule for departures at shared arrival/departure coordination points which can then be translated to target takeoff times and departure routes.The second stage consists of the more tactical surface and departure scheduling to fill the gaps and meet the target times provided by the first stage.At this low level of precision, the arrival and surface/departure schedulers are coordinated rather than fully integrated.The surface/departure scheduler still works opportunistically at the smaller time horizon, but the arrival scheduler works with the information it has at the larger planning horizon to maximize the opportunity.The resulting coordinated schedule is the combination of arrival and departure schedules.MITRE proposed a more near-term two-stage coordinated scheduling concept called High Density Area Departure/Arrival Management (HDDAM) to manage metroplex arrival and departure meter fixes [32].In HDDAM both arrival and departure flight scheduling occur independently within the second stage.The first stage consists of an arrival/departure slot-negotiator function, which takes as inputs demand requests and resource capacities from the entities that own them, and assigns generic arrival and departure slots per airport at each shared resource.The independent arrival and departure flight schedulers then schedule individual flights to fill their respective allotted slots. +Level 3: Flight Schedule UpdateThe flight schedule update is the more tactical reactive level of the terminal scheduling hierarchy addressing the third integrated scheduling goal of enhancing system robustness and resilience.It is a schedule-based method to quickly contain and recover from schedule disruptions due to nonconformance.For arrivals, scheduling disturbances include late flights unable to meet their scheduled time, emergencies requiring a flight to land as soon as possible, or missed approaches causing flights to go-around and fit back into the arrival queue.In the case of late or go-round flights, one solution is to vector the flight into a holding pattern until a natural gap opens up to fit the flight back in.This is similar to the approach taken by NASA's new CDM surface scheduler concept [8].If a flight does not pushback from the gate by the agreed upon latest gate pushback time, it's spot release time window can no longer be guaranteed and it must wait for the tactical surface scheduler to opportunistically fit the flight in.However, in high traffic volume, the flight could be waiting a long time to reenter the queue.Emergencies are more complicated as they force the issue of creating a gap where none may exist and may affect flights that were in conformance.The extreme schedule-based solution to these disturbances would be to unfreeze and reschedule all flights (list rippling) to rectify the nonconformance.This is highly disruptive in a segregated terminal environment.It could be even more disruptive in an integrated terminal environment depending on the cascading dependencies of the schedule.A more surgical approach would be to reschedule a subset of flights to contain and rectify the disturbance.In addition, the earlier a developing disturbance can be detected and mitigation initiated, the more efficient the mitigation can be.a disturbance is predicted, the next function is to identify the resolution method and subset of flights that could be affected by the disturbance.Finally, the schedule is updated for the minimum cost subset of affected flights.In an integrated scheduling environment, the rescheduled flight subset may include aircraft within in all three domains.In order to keep the rescheduled flight subset from growing too large in a high volume traffic situation with very few natural gaps, the schedule update must be allowed to temporarily relax constraints imposed on the initial flight schedule.For example, temporal scheduling buffers may be reduced or originally unavailable resources (route options, taxi-ways, etc.) reserved for such situations may be made available to the schedule update. +Level 4: Schedule ConformanceThe lowest level of the terminal scheduling hierarchy includes the control techniques used to achieve schedule conformance.The control techniques used by each domain may be very different and may not pass any information between domains.Whether flight-deck or ground based, the function of schedule conformance control techniques is to efficiently resolve any deviations from the schedule.The performance of this lowest level bounds the solution space of the higher levels.The precision of a technique will determine the scheduling buffer required to dampen the system, affecting throughput.The flexibility and fast effectiveness of a technique will temper the use of flight schedule updates to respond to disturbances. +Research GapsTerminal area and surface scheduling research is already strongly aligned to the proposed integration framework.However research efforts are not yet synchronized.Research directed specifically toward integrating domains are at different levels of maturity targeting different implementation time frames and TRACONs.To inform a more balanced integrated terminal scheduling research portfolio, this section identifies current research gaps and proposes key areas where future research should focus to facilitate scheduler integration. +Level 1: Configuration Schedule GapsA stand-alone concept was developed to dynamically reroute arrivals around weather as they were funneled through extended terminal airspace to their meter fix.Rerouting algorithms were developed [20] and HITL simulations evaluated the operational feasibility of dynamically changing route structure without specific scheduling constraints [25].Recently, en-route Dynamic Weather Rerouting (DWR) has been successfully tested in the field [33] and is exploring the possibility of extending this technology to the extended terminal area to support arrival scheduling.Whereas reference [25] updated an entire route structure on a synchronized update rate, terminal DWR may update arrival routes to meter fixes on an individual basis.Moving forward, extended terminal-area weather re-routes (whether they are route structure based or individual flight based) should be integrated with scheduling by feeding more accurate ETAs for weather route options to the flight schedule prior to the freeze horizon.These above efforts are primarily concerned with processing non-negotiable weather constraints and providing more accurate information to the arrival scheduler and not with generating an integrated schedule of terminal-area and surface configurations.NASA configuration scheduling development under TRCM supported integrated arrival, departure, and surface operations from the start.However, TRCM technologies still need to address uncertainty.Evaluations of these technologies assume perfect forecasts, analyzing historical data for potential benefits.Integrated configuration scheduling development should move to higher fidelity fast-time simulations that incorporate uncertainty associated with weather and demand forecast. +Level 2: Flight Schedule GapsCurrently arrival scheduling dominates this level with ~40-minute freeze horizons, but very little reliable information is passed from surface and departure scheduling.PDRC has made the greatest progress in correcting this imbalance by improving departure takeoff time estimates enough to compete for arrival slots.This gives these flights more equitability as arrivals in their destination TRACON, but not as departures competing with arrivals for resources within their original TRACON.The key areas where flight scheduling should focus to facilitate integration are to expand route options, mitigate uncertainty with stochastic and collaborative scheduling techniques, and incorporate user preference in the schedule. +Expand Route OptionsCurrently, TMA-TM typically assumes there is only one schedulable route between any meter-fix and runway pair for a given performance based navigation equipage level.A flight is scheduled to a meter-fix, one or two merge points, and the runway, totaling to three or four coordination points.Departure crossings will only add more coordination points.The problem with adding more coordination points into the system is that it adds more constraints, which limit the solution space such that the coupled solution may incur more delay than the uncoupled solution.This can be alleviated by adding more route options, thereby introducing an extra degree of freedom to the scheduler, and expanding the solution space.Many surface scheduling algorithms consider multiple route options along taxiways between spot and runway [34][35][36][37][38]. Fewer airspace schedulers consider multiple route options for a given aircraft between the same meter fix/runway pair [39][40][41].Multiple terminal airspace route options have also been considered in lower level scheduling hierarchies [42][43][44].Rather than segregating all arrival and departure routes, several optimal departure routes should be designed even if they must temporarily occupy arrival airspace.Likewise, multiple routes should be available to arrivals for any meter-fix-to-runway pair for a given equipage level.When the demand is too high for a shared resource, alternate less efficient routes must be scheduled, but they may be scheduled to arrivals as well as departures depending on which option is more optimal for the system.Research has already shown that a combination of route segregation and temporal separation at a coordination point is a more effective method of arrival/departure scheduling than either method alone [45].Arrival scheduling may also consider departure demands on surface resources as well.This is already done in current practice by necessity at airports with dual use or crossing arrival and departure runways.But the current practice is to meter the arrivals such that there are gaps in which to opportunistically depart aircraft.These gaps could be more optimally sequenced with more flight-specific demand information passed [46][47][48]. +Mitigate UncertaintyThe main deterrent to arrival scheduling considering departures is the level of uncertainty in departure trajectory prediction at the necessary freeze horizon.In addition to improving departure trajectory prediction with CDM gate pushback times and RNAV departure routing, arrival schedulers may use stochastic techniques to mitigate the uncertainty.Stochastic scheduling incorporates flight arrival-time uncertainty at the coordination point by including the probability of separation in the cost function [49][50][51][52].If the flight scheduler cannot get an accurate 40minute advance gate pushback schedule from surface CDM, some system benefits may still be possible using flight plan departure times, even though surface scheduler assigned departure times would be much better. +Incorporate User PreferenceThe benefit metrics that have driven terminalarea scheduling are largely system oriented.It is assumed that all flights wish to fly shorter distances and use less fuel.However, airlines may have other preferences that could influence scheduling.Strategic schedule integrity is a large concern, especially for flights in and out of large hubs.An airline may not wish one flight to arrive early when another flight feeding multiple connections is running late.SARDA-CDM [8] and other surface CDM schedulers [53,54] begin to address airline preference by allowing them to negotiate scheduled gate pushback time.Arrival scheduling should also develop techniques other than separation-constraint-modified first-come-first-served to address airline preferences [55]. +Level 3: Flight Schedule Update GapsSARDA evolved with tactical updates as an integral part of the flight scheduler.The schedule is constantly updated with only the sequence of a few flights at the front of the queue frozen.Because of this short freeze horizon, any propagating delays incurred can quickly be incorporated into the spot scheduler.When paired with CDM, the SARDA tactical scheduler acts as the flight schedule update hierarchy level.Airspace scheduling has only recently begun to develop automated tactical schedule updates with MESAR, which is being applied to arrivals only in initial development.Simulations of arrival operations to LaGuardia's crossing runways exposed the need for tactical schedule update in this situation but the updates were performed manually [56].The possibility of arrival/departure schedule negotiation to improve departure runway utilization at LaGuardia is currently being explored.Future simulations may incorporate MESAR-like automation to synch operations to crossing runways.Before MESAR is extended to a fully integrated arrival/departure flight scheduler, a similar technique may be applied to the coordinated flight scheduler.First an arrival schedule is developed to include some number of gaps for departures based on departure demand extrapolated from flight plan departure times.The departure scheduler attempts to use the gaps created in the arrival schedule as given, but may also request the arrival scheduler to update the schedule for a small set of arrivals to modify gap size or temporal location.In this way, tactical schedule update is used to negotiate a refined coordinated schedule between arrival and departures.If these negotiations will be nominal, they may occur more often than the occasional off-nominal nonconformance triggering a MESAR tactical arrival schedule update.Research is needed to determine how often such a negotiation would be attempted or can be accommodated.In the departure phase of flight, path control (i.e.vectoring and direct-to) and altitude clearances are the primary methods of control.There has been little DST development for precisely controlling departures in terminal airspace.The SOAR effort has recently developed DSTs to enable early altitude clearances [16] and just begun development of some precision path-based control [44]. +Level 4: Schedule Conformance GapsOn the surface, ANSP control mainly consists of taxi clearances at the spot, takeoff clearances at the runway, and runway crossing clearances.Pilots control speed to provide visual separation along the 2D fixed-path taxiways.Previous development of DSTs enabled pilots to follow 4D trajectory clearances on the surface [59,60].More recently, the SARDA surface scheduler was integrated with the flight-deck 4D taxi capability and tested in a humanin-the-loop simulation [61].In the simulation, both takeoff sequence and departure times generated by SARDA were displayed to the pilots and an errornulling algorithm provided speed advisories to meet the runway RTA.In order to support schedule conformance, control methods will need to be more trajectorybased and employ more flexibility to resolve errors due to uncertainty.Of the control techniques described above, arrival speed control is the most precise but also the least responsive.More precise path control methods should be developed to add responsiveness to the arrival domain and precision to the departure and surface domains.Because the system performance is dependent on the schedule conformance precision, control methods and scheduling concepts should incentivize flight-deck trajectory precision when possible.For example, the SARDA-CDM method of handling gate pushback nonconformance is an incentive for meeting the scheduled gate pushback time.Nonconforming aircraft are sent to a separate queue that does not affect the integrity of the original schedule of conforming aircraft.In the arrival phase, appropriately equipped aircraft can fly shorter Required Navigation Performance (RNP) routes with radius-to-fix turns onto a short final approach leg.When controllers are given the option to delegate separation to FIM equipped aircraft, they are encouraged to let the FIM aircraft follow the scheduled trajectory and instead apply more invasive control to less precise aircraft.In places where departure routes tunnel under arrival streams, controllers can give early altitude clearances to departures when they see a sufficiently large arrival gap.In a more trajectory-based schedule-driven environment where these arrival gaps are scheduled, the departure would have incentive to meet the scheduled gap rather than tunneling.Current scheduling techniques use time buffers between time slots to mitigate uncertainty and ensure minimum required separation.Aircraft equipped to achieve higher precision may be scheduled with smaller buffers.However, this approach benefits the system almost as much if not more than the equipped aircraft, making the advantage of equipping early less attractive.Path-based buffering (reserving path shortcuts rather than time to resolve nonconformance) may allow precision-equipped aircraft to be scheduled to a shorter path than less precise aircraft requiring larger path buffers [40]. +Degree of IntegrationAnother key area of research that must be addressed is the necessary degree of integration.The need for integration is driven by resource utilization inefficiencies.Where and when no such inefficiencies exist, flight schedules may not need to be integrated or flights may not need to be scheduled at all.A mechanism is needed to watch for resource competition leading to inefficiencies, and manage the degree of scheduling and integration accordingly.Real-time metrics are needed to justify the tradeoff between complexity and efficiency that integration may bring. +ConclusionThis paper presented a framework to integrate arrival, departure, and surface operations scheduling.The framework consists of a hierarchy of scheduling functions organized by scheduling horizon rather than the traditional arrival, departure, and surface domains.Configuration schedule determines the schedule of configurations (i.e.set of airspace and surface resources) available to lower level schedulers.Future research in this area can expand route options, mitigate uncertainty with stochastic and collaborative scheduling techniques, and incorporate user preference in the schedule.Flight schedule update may be used to negotiate integrated arrival/departure schedules by giving the arrival schedule the flexibility to adapt to surface uncertainties.Schedule conformance research should develop more precision path-based control methods to add more flexibility to the arrival domain and more precision to the departure and surface domains.Finally, real-time metrics need to be defined to watch for resource competition and manage the degree of scheduling and integration required to justify the added complexity.Figure 1 .1Figure 1.Flight Scheduling Domains. +change can also influence the transition cost.Ideally, the configuration schedule should be frozen prior to the planning horizon of the highest-level individual flight scheduler. +Figure 3 .3Figure 3. Coordinated Arrival/Departure Flight Schedule +Figure 44Figure 4 diagrams how arrival schedule updates can be incorporated into coordinated arrival departure scheduling. +Figure 4 .4Figure 4. Coordinated Flight Schedule with Arrival Updates +Each domain has been developing DSTs to aid the pilot or controller in achieving the scheduler goals.The domain specific DSTs cater to the preferred control methods of the domain.In the arrival phase of flight, speed control with as little path deviation as possible is preferred.Flight-deck speed control technologies include Required Time of Arrival (RTA) and Flight-deck Interval Management (FIM).Ground based DST development for arrivals are called Controller Managed Spacing (CMS) and Ground Interval Management (GIM).Recent evaluations of these technologies integrated FIM with CMS in a mixed-equipage environment[56,57]. +Flight schedule determines initial individual flight trajectories (i.e.route and schedule).Flight schedule update monitors flight ETAs for disturbances and updates the flight schedule when required.Finally, schedule conformance includes the control techniques flights use to meet the schedule.Research gaps within this framework were discussed and key areas of research focus recommended.Configuration schedule research should consider uncertainty more during its development.Otherwise, conformance scheduling should start to integrate with the flight scheduling hierarchy.Flight schedule research is having difficulty bridging the gap between arrival scheduling horizon and departure and surface uncertainties. +1Known as Time Based Flow Management (TBFM) by the FAA.Also known as Terminal Area Precision Scheduling and Spacing System (TAPSS) by NASA or Terminal Sequencing and Spacing (TSS) by the FAA, when scheduling is coupled with TRACON controller tools. + This research is currently unpublished.For more information contact the MESAR scheduling lead at Jaewoo.Jung@nasa.gov. + + + +Appendix I: List of Acronyms + + + + + + + Design and Evaluation of the Terminal Area Precision Scheduling and Spacing System + + HSwenson + + + JThipphavong + + + ASadovsky + + + LChen + + + CSullivan + + + LMartin + + + 2011 + Berlin, Germany + + + Ninth USA/Europe Air Traffic Management Research and Development Seminar + + + Swenson, H., J. Thipphavong, A. Sadovsky, L. Chen, C. Sullivan, and L. Martin, 2011, Design and Evaluation of the Terminal Area Precision Scheduling and Spacing System, Ninth USA/Europe Air Traffic Management Research and Development Seminar, Berlin, Germany. + + + + + Efficiency Benefits Using the Terminal Area Precision Scheduling and Spacing System + + JaneThipphavong + + + HarrySwenson + + + PaulLin + + + AnthonySeo + + + LeonardBagasol + + 10.2514/6.2011-6971 + + + 11th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Virginia Beach, Virginia + + American Institute of Aeronautics and Astronautics + 2011 + + + Thipphavong, J., H. Swenson, P. Lin, A. Seo, and L. Bagasol, 2011, Efficiency Benefits Using the Terminal Area Precision Scheduling ans Spacing System, 11 th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, Virginia Beach, Virginia. + + + + + Development and evaluation of the terminal precision scheduling and spacing system for off-nominal condition operations + + HNSwenson + + + Jaewoo Jung + + + JThipphavong + + + Liang Chen + + + LMartin + + + JNguyen + + 10.1109/dasc.2012.6382303 + + + 2012 IEEE/AIAA 31st Digital Avionics Systems Conference (DASC) + Williamsburge, Virginia + + IEEE + 2012 + + + Swenson, H., J. Jung, J. Thipphavong, L. Chen, L. Martin, and J. Nguyen, 2012, Development and Evaluation of the Terminal Precision Scheduling and Spacing System for Off-Nominal Condition Operations, 31 st Digital Avionics Systems Conference, Williamsburge, Virginia. + + + + + Demonstration of reduced airport congestion through pushback rate control + + IoannisSimaiakis + + + HarshadKhadilkar + + + HamsaBalakrishnan + + + TomGReynolds + + + RJohnHansman + + 10.1016/j.tra.2014.05.014 + + + Transportation Research Part A: Policy and Practice + Transportation Research Part A: Policy and Practice + 0965-8564 + + 66 + + 2011 + Elsevier BV + Berlin, Germany + + + Ninth USA/Europe Air Traffic Management Research and Development Seminar + + + Simaiakis, I., H. Khadilkar, H. Balakrishnan, T. Reynolds, R. Hansmann, B. Reilly, and S. Urlass, 2011, Demonstration of Reduced Airport Congestion Through Pushback Rate Control, Ninth USA/Europe Air Traffic Management Research and Development Seminar, Berlin, Germany. + + + + + Assessing the impacts of the JFK ground management program + + StevenStroiney + + + BenjaminLevy + + + HarshadKhadilkar + + + HamsaBalakrishnan + + 10.1109/dasc.2013.6712660 + + + 2013 IEEE/AIAA 32nd Digital Avionics Systems Conference (DASC) + Syracuse, New York + + IEEE + 2013 + + + Stoiney, S., B. Levy, H. Khadilkar, and H. Balakrishnan, 2013, Assessing the Impacts of the JFK Ground Management Program, 32 nd Digital Avionics Systems Conference, Syracuse, New York. + + + + + A Concept and Implementation of Optimized Operations of Airport Surface Traffic + + YoonJung + + + TyHoang + + + JustinMontoya + + + GautamGupta + + + WaqarMalik + + + LeonardTobias + + 10.2514/6.2010-9213 + + + 10th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Fort Worth, Texas + + American Institute of Aeronautics and Astronautics + 2010 + + + Jung, Y., T. Huang, J. Montoya, G. Gupta, W. Malik, and L. Tobias, 2010, A Concept and Implementation of Optimized Operations of Airport Surface Traffic, 10 th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, Fort Worth, Texas. + + + + + Performance Evaluation of Individual Aircraft Advisory Concept for Surface Managemenet + + GGupta + + + WMalik + + + LTobias + + + YJung + + + THuang + + + MHayashi + + + 2013 + Chicago, Illinois + + + Tenth USA/Europe Air Traffic Management Research and Development Seminar + + + Gupta, G., W. Malik, L. Tobias, Y. Jung, T. Huang, and M. Hayashi, 2013, Performance Evaluation of Individual Aircraft Advisory Concept for Surface Managemenet, Tenth USA/Europe Air Traffic Management Research and Development Seminar, Chicago, Illinois. + + + + + An Integrated Collaborative Decision Making and Tactical Advisory Concept for Airport Surface Operations Management + + GautamGupta + + + WaqarMalik + + + YoonJung + + 10.2514/6.2012-5651 + + + 12th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference and 14th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference + Indianapolis, Indiana + + American Institute of Aeronautics and Astronautics + 2012 + + + Gupta, G., W. Malik, and Y. Jung, 2012, An Integrated Collaborative Decision Making and Tactical Concept for Airport Surface Operations Management, 12 th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, Indianapolis, Indiana. + + + + + Characterization of Tactical Departure Scheduling in the National Airspace System + + RichardCapps + + + ShawnEngelland + + 10.2514/6.2011-6835 + + + 11th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Virginia Beach, Virginia + + American Institute of Aeronautics and Astronautics + 2011 + + + Capps, A. and S. Engelland, 2011, Characterization of Tactical Departure Scheduling in the National Airspace System, 11 th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, Virginia Beach, Virginia. + + + + + Analysis of divergences from area navigation departure routes at DFW airport + + PaulFBorchers + + + KevinDay + + 10.1109/dasc.2009.5347532 + + + 2009 IEEE/AIAA 28th Digital Avionics Systems Conference + Orlando, Florida + + IEEE + 2009 + + + Borchers, P. and K. Day, 2009, Analysis of Divergences from Area Navigation Departure Routes at DFW Airport, 28 th Digital Avionics Systems Conference, Orlando, Florida. + + + + + Challenges to Modeling Vectored Area Navigation Departures at Dallas/Fort Worth International Airport + + PaulBorchers + + + OusmaneDiallo + + + KevinDay + + 10.2514/6.2011-6836 + + + 11th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Virginia Beach, Virginia + + American Institute of Aeronautics and Astronautics + 2011 + + + Borchers, P., O. Diallo, and K. Day, 2011, Challenges to Modeling Vectored Area Navigation Departures at DFW Airport, 11 th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, Virginia Beach, Virginia. + + + + + Characterization of Nationwide TRACON Departure Operations + + MatthewSKistler + + + AlanCapps + + + ShawnAEngelland + + 10.2514/6.2014-2019 + + + 14th AIAA Aviation Technology, Integration, and Operations Conference + Atlanta, Georgia + + American Institute of Aeronautics and Astronautics + 2014 + + + Kistler, M., A. Capps, and S. Engelland, 2014, Characterization of Nationwide TRACON Departure Operations, 14 th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, Atlanta, Georgia. + + + + + + RDelaura + + + NUnderhill + + + LHal + + + YRodriguez + + Evaluation of the Integrated Departure Route Planning (IDRP) Tool + + 2012. 2011 + + + DeLaura, R., N. Underhill, L. Hal, and Y. Rodriguez, 2012, Evaluation of the Integrated Departure Route Planning (IDRP) Tool 2011 + + + + + + + Prototype + + ATC-388 + + Lincold Laboratory, MIT, Lexigton, Massachusetts + + + Project Report + Prototype, Project Report ATC-388, Lincold Laboratory, MIT, Lexigton, Massachusetts. + + + + + + JDearmon + + + NTaber + + + HBateman + + + LSong + + + TMasek + + + DGilani + + Benefits Analysis of a Departure Management Prototype for the + New York Area; Berlin, Germany + + 2011 + + + Ninth USA/Europe Air Traffic Management Research and Development Seminar + + + DeArmon, J., N. Taber, H. Bateman, L. Song, T. Masek, and D. Gilani, 2011, Benefits Analysis of a Departure Management Prototype for the New York Area, Ninth USA/Europe Air Traffic Management Research and Development Seminar, Berlin, Germany. + + + + + Development of Conflict-Free, Unrestricted Climbs for a Terminal Area Departure Tool + + YoonJung + + + DouglasIsaacson + + 10.2514/6.2003-6794 + + + AIAA's 3rd Annual Aviation Technology, Integration, and Operations (ATIO) Forum + Denver, Colorado + + American Institute of Aeronautics and Astronautics + 2003 + + + Jung, Y. and D. Isaacson, 2003, Development of Conflict-Free, Unrestricted Climbs for a Terminal Area Departure Tool, 3 rd AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, Denver, Colorado. + + + + + Scheduling and separating departures crossing arrival flows in shared airspace + + EricChevalley + + + BonnyParke + + + PaulLee + + + FaisalOmar + + + HwasooLee + + + NancyBienert + + + JoshuaKraut + + + EverettPalmer + + 10.1109/dasc.2013.6712522 + + + 2013 IEEE/AIAA 32nd Digital Avionics Systems Conference (DASC) + Syracuse, New York + + IEEE + 2013 + + + Chevalley, E., B. Parke, P. Lee, F. Omar, H. Lee, N. Bienert, J. Kraut, and E. Palmer, 2013, 32 nd Digital Avionics Systems Conference, Syracuse, New York. + + + + + Trajectory-Based Takeoff Time Predictions Applied to Tactical Departure Scheduling: Concept Description, System Design, and Initial Observations + + ShawnEngelland + + + RichardCapps + + 10.2514/6.2011-6875 + + + 11th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Virginia Beach, Virginia + + American Institute of Aeronautics and Astronautics + 2011 + + + Engelland, S. and A. Capps, 2011, Trajectory- Based Takeoff Time Predictions Applied to Tactical Departure Scheduling: Concept Development, System Design, and Initial Observations, 11 th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, Virginia Beach, Virginia. + + + + + Design Characteristics of a Terminal Departure Scheduler + + AlanCapps + + + MathewSKistler + + + ShawnAEngelland + + 10.2514/6.2014-2020 + + + 14th AIAA Aviation Technology, Integration, and Operations Conference + Atlanta, Georgia + + American Institute of Aeronautics and Astronautics + 2014 + + + Capps, A., M. Kistler, and S. Engelland, 2014, Design Characteristics for a Terminal Departure Scheduler, 14 th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, Atlanta, Georgia. + + + + + The Route Availibility Planning Tool (RAPT): Evaluation of Departure Management Decision Support in New York During the 2008 Convective Weather Season + + MRobinson + + + RDelaura + + + NUnderhill + + + 2009 + Napa Valley, California + + + Eighth USA/Europe Air Traffic Management Research and Development Seminar + + + Robinson, M., R. DeLaura, and N. Underhill, 2009, The Route Availibility Planning Tool (RAPT): Evaluation of Departure Management Decision Support in New York During the 2008 Convective Weather Season, Eighth USA/Europe Air Traffic Management Research and Development Seminar, Napa Valley, California. + + + + + Flexible, Performance-Based Route Planning for Super-Dense Operations + + JosephPrete + + + JimmyKrozel + + + JosephMitchell + + + JoondongKim + + + JasonZou + + 10.2514/6.2008-6825 + + + AIAA Guidance, Navigation and Control Conference and Exhibit + Honolulu, Hawaii + + American Institute of Aeronautics and Astronautics + 2008 + + + Prete, J., J. Krozel, J. Mitchell, J. Kim, and J. Zou, 2008, Flexible, Performance-based Route Planning for Super-Dense Operations, AIAA Guidance, Navigation, and Control Conference, Honolulu, Hawaii. + + + + + Dynamic Metroplex Airspace + + WilliamHall + + + StephenAtkins + + 10.2514/6.2011-7064 + + + 11th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Virginia Beach, Virginia + + American Institute of Aeronautics and Astronautics + 2011 + + + Hall, W. and S. Atkins, 2011, Dynamic Metroplex Airspace, 11 th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, Virginia Beach, Virginia. + + + + + Weather Avoidance Optimal Routing for Extended Terminal Airspace in Support of Dynamic Airspace Configuration, 31 st Digital Avionics Systems Conference + + JChen + + + AYousefi + + + SKrishna + + + BSliney + + + PSmith + + + 2012 + Williamsburge, Virginia + + + Chen, J., A. Yousefi, S. Krishna, B. Sliney, and P. Smith, 2012, Weather Avoidance Optimal Routing for Extended Terminal Airspace in Support of Dynamic Airspace Configuration, 31 st Digital Avionics Systems Conference, Williamsburge, Virginia. + + + + + Identification of Robust Terminal-Area Routes in Convective Weather + + DMichalek-Pfeil + + + HBalakrishnan + + + + Transpotation Science + + 46 + 1 + + 2012 + + + Michalek-Pfeil, D. and H. Balakrishnan, 2012, Identification of Robust Terminal-Area Routes in Convective Weather, Transpotation Science, Vol. 46, No. 1, pp. 65-73. + + + + + Modeling and Optimization of Terminal Area Utilization by Assigning Arrival and Departure Fixes + + BosungKim + + + John-PaulClarke + + 10.2514/6.2013-5256 + + + AIAA Guidance, Navigation, and Control (GNC) Conference + + American Institute of Aeronautics and Astronautics + 2013 + + + Kim, B. and J. P. Clarke, 2013, Modeling and Optimization of Terminal Area Utilization by Assigning Arrival and Departure Fixes, AIAA Guidance, Navigation, and Control (GNC) + + + + + + + Conference + + + Boston, Masachusetts + + + Conference, Boston, Masachusetts. + + + + + An examination of selected datacom options for the near-term implementation of trajectory based operations + + WalterWJohnson + + + SummerLBrandt + + + JoelLachter + + + VernolBattiste + + + VeranikaLim + + + RobertKoteskey + + + Arik-Quang V.Dao + + + SarahVLigda + + + Shu-ChiehWu + + 10.1109/dasc.2011.6096233 + + + 2011 IEEE/AIAA 30th Digital Avionics Systems Conference + Villasimius, Sardinia + + IEEE + 2012 + + + Johnson, W., H. Latcher, S. Brandt, R. Koteskey, A. Dao, J. Kraut, S. Ligda, and V. Battiste, 2012, Evaluation of Controller and Pilot Performance, Workload, and Acceptability Under A NextGen Concept for Dynamic Weather Adapted Arrival Routing, 30 th European Association for Aviation Psychology (EAAP) Conference, Villasimius, Sardinia. + + + + + + RZhang + + + RKincaid + + Robust Optimization Model for Runway Configurations Management, 11 th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Virginia Beach, Virginia + + 2011 + + + Zhang, R. and R. Kincaid, 2011, Robust Optimization Model for Runway Configurations Management, 11 th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, Virginia Beach, Virginia. + + + + + Tactical airport configuration management + + ChristopherAProvan + + + StephenCAtkins + + 10.1109/icnsurv.2011.5935346 + + + 2011 Integrated Communications, Navigation, and Surveillance Conference Proceedings + + IEEE + 2011 + + + Integrated + Provan, C. and S. Atkins, 2011, Tactical Airport Configuration Management, Integrated + + + + + Integrated communications navigation and surveillance (ICNS) conference + + RichardJehlen + + 10.1109/icnsurv.2013.6548683 + + + 2013 Integrated Communications, Navigation and Surveillance Conference (ICNS) + Herndon, Virginia + + IEEE + + + + Communications, Navigation, and Surveilance (ICNS) Conference, Herndon, Virginia. + + + + + Optimal Selection of Airport Runway Configurations + + DimitrisBertsimas + + + MichaelFrankovich + + + AmedeoOdoni + + 10.1287/opre.1110.0956 + + + Operations Research + Operations Research + 0030-364X + 1526-5463 + + 59 + 6 + + 2011 + Institute for Operations Research and the Management Sciences (INFORMS) + + + Bertsimas, D., M. Frankovich, and A. Odioni, 2011, Optimal Selection of Airport Runway Configurations, Operations Research, Vol. 59, No. 6, pp. 1407-1419. + + + + + Heuristic Search for Tactical Runway Configuration Management + + JenniferThorne + + + RexKincaid + + 10.2514/6.2012-5499 + + + 12th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference and 14th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference + Indianapolis, Indiana + + American Institute of Aeronautics and Astronautics + 2012 + + + Thorne, J. and R. Kincaid, 2012, Heuristic Search for Tactical Runway Configuration Management, 12 th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, Indianapolis, Indiana. + + + + + Decision Support for Optimal Runway Reconfiguration + + XiaoliBai + + + PadmanabhanKMenon + + 10.2514/6.2013-4397 + + + 2013 Aviation Technology, Integration, and Operations Conference + Los Angeles, California + + American Institute of Aeronautics and Astronautics + 2013 + + + Bai, X. and P. K. Menon, 2013, Decision Support for Optimal Runway Reconfiguration, 13 th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, Los Angeles, California. + + + + + Benefits Assessment for Tactical Runway Configuration Management Tool + + RosaMOseguera-Lohr + + + NipaPhojanamongkolkij + + + GaryWLohr + + + JamesWFenbert + + 10.2514/6.2013-4395 + + + 2013 Aviation Technology, Integration, and Operations Conference + Los Angeles, California + + American Institute of Aeronautics and Astronautics + 2013 + + + Oseguera-Lohr, R., N. Phojanamongkolkij, and G. Lohr, 2013, Benefits Assessment for Tactical Runway Configuration Management Tool, 13 th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, Los Angeles, California. + + + + + Framework for High-Density-Area Departure and Arrival Traffic Management + + ChristineTaylor + + + TudorMasek + + + HiltonBateman + + 10.2514/1.57273 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 36 + 4 + + 2013 + American Institute of Aeronautics and Astronautics (AIAA) + + + Taylor, C., T. Masek, and H. Bateman, 2013, Framewok for High-Density-Area Departure and Arrival Traffic Management, Journal of Guidance, Control, and Dynamics, Vol. 36, No. 4, pp. 1134- 1149, DOI: 10.2514/1.57273. + + + + + Dynamic Weather Routes: Two Years of Operational Testing at American Airlines + + DavidMcnally + + + KapilSheth + + + ChesterGong + + + MikeSterenchuk + + + ScottSahlman + + + SusanHinton + + + ChuHanLee + + + Fu-TaiShih + + 10.2514/atcq.23.1.55 + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 23 + 1 + + 2013 + American Institute of Aeronautics and Astronautics (AIAA) + Chicago, Illinois + + + Tenth USA/Europe Air Traffic Management Research and Development Seminar + + + McNally, D., K. Sheth, C. Gong, P. Borchers, J. Osborne, D. Keany, B. Scott, S. Smith, S. Sahlman, C. Lee, and J. Cheng, 2013, Operational Evaluation of Dynamic Weather Routes at American Airlines, Tenth USA/Europe Air Traffic Management Research and Development Seminar, Chicago, Illinois. + + + + + A Mixed Integer Linear Program for Solving a Multiple Route Taxi Scheduling Problem + + JustinMontoya + + + ZacharyWood + + + SivakumarRathinam + + + WaqarMalik + + 10.2514/6.2010-7692 + + + AIAA Guidance, Navigation, and Control Conference + Boston, Masachusetts + + American Institute of Aeronautics and Astronautics + 2010 + + + Montoya, J., Z. Wood, S. Rathinam, and W. Malik, 2010, A Mixed Integer Linear Program for Solving a Multiple Route Taxi Scheduling Problem, AIAA Guidance, Navigation, and Control (GNC) Conference, Boston, Masachusetts. + + + + + Generation of Optimized Routes and Schedules for Surface Movement of Aircraft on Taxiways + + PrateekGupta + + + HarishwarSubramanian + + + RajkumarPant + + 10.2514/6.2010-9211 + + + 10th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Fort Worth, Texas + + American Institute of Aeronautics and Astronautics + 2010 + + + Gupta, P., H. Subramanian, and R. Pant, 2010, Generation of Optimized Route and Schedules for Surface Movement of Aircraft on Taxiways, 10 th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, Fort Worth, Texas. + + + + + Optimization of Taxiway Routing and Runway Scheduling + + Clare + + + G + + + ARichards + + 10.1109/TITS.2011.2131650 + + + IEEE Transactions on Intelligent Transportation Systems + + 12 + 4 + + 2011 + + + Clare, G. and A. Richards, 2011, Optimization of Taxiway Routing and Runway Scheduling, IEEE Transactions on Intelligent Transportation Systems, Vol. 12, No. 4, pp. 1000-1013, DOI: 10.1109/TITS.2011.2131650. + + + + + A Collaborative Optimization Model for Ground Taxi Based on Aircraft Priority + + YuJiang + + + ZhihuaLiao + 0000-0001-9618-3208 + + + HonghaiZhang + + 10.1155/2013/854364 + + + Mathematical Problems in Engineering + Mathematical Problems in Engineering + 1024-123X + 1563-5147 + + 2013 + + 2013 + Hindawi Limited + + + Jiang, Y., Z. Liao, and H. Zhang, 2013, A Collaborative Optimization Model for Ground Taxi Based on Aircraft Priority, Mathematical Problems in Engineering, DOI: 10.1155/2013/854364. + + + + + An Algorithm for Managing Aircraft Movement on an Airport Surface + + UrbanoTancredi + + + DomenicoAccardo + + + GiancarmineFasano + + + AlfredoRenga + + + GiancarloRufino + + + GiuseppeMaresca + + 10.3390/a6030494 + + + Algorithms + Algorithms + 1999-4893 + + 6 + 3 + + 2013 + MDPI AG + + + Tancredi, U., D. Accardo, G. Fasano, A. Renga, G. Rufino, and G. Maresca, 2013, An Algorithm for Managing Aircraft Movement on an Airport Surface, Algorithms, No. 6, pp. 494-511, DOI: 10.3390/a6030494. + + + + + Arrival Management Architecture and Performance Analysis with Advanced Automation and Avionics Capabilities + + AslaugHaraldsdottir + + + JulienScharl + + + JanetKing + + + MatthewBerge + + 10.2514/6.2009-7006 + + + 9th AIAA Aviation Technology, Integration, and Operations Conference (ATIO) + Hilton Head, South Carolina + + American Institute of Aeronautics and Astronautics + 2009 + + + Haraldsdottir, A., J. Scharl, J. King, and M. Berge, 2009, Arrival Management Architecture management Architecture and Performance Analysis with Advanced Automation and Avionics Capabilities, 9 th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, Hilton Head, South Carolina. + + + + + Design considerations for shortcut path-based time recovery + + ShannonZelinski + + 10.1109/dasc.2013.6719601 + + + 2013 IEEE/AIAA 32nd Digital Avionics Systems Conference (DASC) + Syracuse, New York + + IEEE + 2013 + + + Zelinski, S., 2013, Design Considerations for Shortcut Path-Based Time Recovery, 32 st Digital Avionics Systems Conference, Syracuse, New York. + + + + + Precision Arrival Scheduling for Tactical Reconfiguration + + ShannonZelinski + + 10.2514/6.2014-3152 + + + 14th AIAA Aviation Technology, Integration, and Operations Conference + Atlanta, Georgia + + American Institute of Aeronautics and Astronautics + 2014 + + + Zelinski, S., 2014, Precision Arrival Scheduling for Tactical Reconfiguration, 14 th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, Atlanta, Georgia. + + + + + Investigating the impact of off-nominal events on high-density ‘green’ arrivals + + ToddJCallantine + + + ChristopherCabrall + + + MichaelKupfer + + + LynneMartin + + + JoeyMercer + + + EverettAPalmer + + 10.1109/dasc.2011.6096195 + + + 2011 IEEE/AIAA 30th Digital Avionics Systems Conference + Seattle, Washingon + + IEEE + 2011 + + + Callantine, T., C. Cabrall, M. Kupfer, L. Martin, J. Mercer, and E. Palmer, 2011, Investigating the Impact of off-Nominal Events on High-Density 'Green' Arrivals, 30 th Digital Avionics Systems Conference, Seattle, Washingon. + + + + + Autonomous System for Air Traffic Control in Terminal Airspace + + AnastasiosNikoleris + + + HeinzErzberger + + 10.2514/6.2014-2861 + + + 14th AIAA Aviation Technology, Integration, and Operations Conference + Atlanta, Georgia + + American Institute of Aeronautics and Astronautics + 2014 + + + Nikoleris, T., H. Erzberger, R. Paielli, and Y. Chu, 2014, Autonomous System for Air Traffic Control in Terminal Airspace, 14th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, Atlanta, Georgia. + + + + + Development of a route crossing tool for shared airspace environments + + DaphneRein-Weston + + + EricChevalley + + 10.1109/dasc.2014.6979655 + + + 2014 IEEE/AIAA 33rd Digital Avionics Systems Conference (DASC) + Colorado Springs, Colorado + + IEEE + 2014 + + + Rein-Weston, D., E. Chevalley, A. Globus, R. Jacoby, and E. Palmer, 2014, Development of Route Crossing Tool for Shared Airspace Environments, 33 rd Digital Avionics Systems Conference, Colorado Springs, Colorado. + + + + + Optimal Integration of Departures and Arrivals in Terminal Airspace + + MinXue + + + ShannonZelinski + + 10.2514/1.60489 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 37 + 1 + + 2014 + American Institute of Aeronautics and Astronautics (AIAA) + + + Xue, M. and S. Zelinski, 2014, Optimal Integration of Departures and Arrivals in Terminal Airspace, Journal of Guidance, Control, and Dynamics, Vol. 37, No. 1, pp. 207-213, DOI: 10.2514/1.60489. + + + + + The Design and Optimization of a Combined Arrival-Departure Scheduler + + AnkitTyagi + + + FrederickWieland + + 10.2514/6.2012-5697 + + + 12th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference and 14th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference + Indianapolis, Indiana + + American Institute of Aeronautics and Astronautics + 2012 + + + Tyagi, A. and F. Wieland, 2012, The Design and Optimization of a Combined Arrival-Departure Scheduler, 12 th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, Indianapolis, Indiana. + + + + + + SChandrasekar + + + IHwang + + Optimal Arrival and Departure Sequencing on a Runway System, AIAA Guidance, Navigation, and Control (GNC) Conference + Boston, Masachusetts + + 2013 + + + Chandrasekar, S. and I. Hwang, 2013, Optimal Arrival and Departure Sequencing on a Runway System, AIAA Guidance, Navigation, and Control (GNC) Conference, Boston, Masachusetts. + + + + + Calculating capacity of dependent runway configurations: A discrete-event simulation approach for analysing the effect of aircraft sequencing + + JoeyKlugt + + + PaulCRoling + + + RobTenHove + + + RichardCurran + + 10.2514/6.2013-4353 + + + 2013 Aviation Technology, Integration, and Operations Conference + Los Angeles, California + + American Institute of Aeronautics and Astronautics + 2013 + + + Klugt, J., P. Roling, R. Curran, R. Hove, 2013, Calculating Capacity of Dependent Runway Configurations: A Discrete-Event Simulation Approach for Analysing the Effect of Aircraft Sequencing, 13 th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, Los Angeles, California. + + + + + Runway Operations Optimization in the Presence of Uncertainties + + GustafSolveling + + + SenaySolak + + + John-PaulClarke + + + EllisJohnson + + 10.2514/1.52481 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 34 + 5 + + 2010 + American Institute of Aeronautics and Astronautics (AIAA) + + + Solveling, G., S. Solak, J. P. Clarke, and E. Johnson, 2010, Runway Operations Optimization in the Presence of Uncertainties, Journal of Guidance, Control, and Dynamics, Vol. 34, No. 5, pp. 1373- 1382, DOI: 10.2514/1.52481. + + + + + An approach to optimization of airport taxiway scheduling and traversal under uncertainty + + RossAnderson + + + DejanMilutinović + + 10.1177/0954410011433238 + + + Proceedings of the Institution of Mechanical Engineers, Part G: Journal of Aerospace Engineering + Proceedings of the Institution of Mechanical Engineers, Part G: Journal of Aerospace Engineering + 0954-4100 + 2041-3025 + + 227 + 2 + + 2011 + SAGE Publications + + + Anderson, R. and D. Milutinovic, 2011, An Approch to Optimization of Airport Taxiway Scheduling and Traversal Under Uncertainty, Proceedings of the Institution of Mechanical Engineers, part G: Journal of Aerospace Engineering, Vol. 227, No. 2, pp. 273-284. + + + + + Optimizing integrated terminal airspace operations under uncertainty + + ChristabelleBosson + + + MinXue + + + ShannonZelinski + + 10.1109/dasc.2014.6979545 + + + 2014 IEEE/AIAA 33rd Digital Avionics Systems Conference (DASC) + Colorado Springs, Colorado + + IEEE + 2014 + + + Bosson, C., M. Xue, and S. Zelinski, 2014, Optimizing Integrated Terminal Airspace Operations Under Uncertainty, 33 rd Digital Avionics Systems Conference, Colorado Springs, Colorado. + + + + + Dynamic stochastic scheduler for integrated arrivals and departures + + MinXue + + + ShannonZelinski + + 10.1109/dasc.2014.6979399 + + + 2014 IEEE/AIAA 33rd Digital Avionics Systems Conference (DASC) + Colorado Springs, Colorado + + IEEE + 2014 + + + Xue, M. and S. Zelinski, 2014, Dynamic Stochastic Scheduler for Integrated Arrivals and Departures, 33 rd Digital Avionics Systems Conference, Colorado Springs, Colorado. + + + + + Ration by Schedule for airport arrival and departure planning and scheduling + + ChrisBrinton + + + StephenAtkins + + + LaraCook + + + StevenLent + + + TomPrevost + + 10.1109/icnsurv.2010.5503239 + + + 2010 Integrated Communications, Navigation, and Surveillance Conference Proceedings + Herndon, Virginia + + IEEE + 2010 + + + Brinton, C., S. Atkins, L. Cook, S. Lent, and T. Prevost, 2010, Ration by Schedule for Airport Arrival and Departure Planning and Scheduling, Integrated Communications Navigation and Surveillance (ICNS) Conference, Herndon, Virginia. + + + + + A Collaborative Optimization Model for Ground Taxi Based on Aircraft Priority + + YuJiang + + + ZhihuaLiao + 0000-0001-9618-3208 + + + HonghaiZhang + + 10.1155/2013/854364 + + + Mathematical Problems in Engineering + Mathematical Problems in Engineering + 1024-123X + 1563-5147 + + 2013 + + 2013 + Hindawi Limited + + + Jiang, Y., Z. Liao, and H. Zhang, 2013, A Collaborative Optimization Model for Ground Taxi Based on Aircraft Priority, Mathematical Problems in Engineering, DOI: 10.1155/2013/854364. + + + + + Scheduling aircraft landings using airlines’ preferences + + MJSoomer + + + GJFranx + + 10.1016/j.ejor.2007.06.017 + + + European Journal of Operational Research + European Journal of Operational Research + 0377-2217 + + 190 + 1 + + 2008 + Elsevier BV + + + Soomer, M. and G. Franx, 2008, Scheduling Aircraft Landings Using Airlines' Preferences, European Journal of Operational Research, Vol. 190, No. 1, pp.277-291. + + + + + Nextgen operations in a simulated NY area airspace + + NancyMSmith + + + BonnyParke + + + PaulLee + + + JeffHomola + + + ConnieBrasil + + + NathanBuckley + + + ChrisCabrall + + + EricChevalley + + + CindyLin + + + SusanMorey + + + FaisalOmar + + + DaphneRein-Weston + + + Hyo-Sang Yoo + + 10.1109/dasc.2013.6712607 + + + 2013 IEEE/AIAA 32nd Digital Avionics Systems Conference (DASC) + Syracuse, New York + + IEEE + 2013 + + + Smith, N., B. Parke, P. Lee, J. Homola, C. Brasil, N. Buckley, C. Cabrall, S. Morey, F. Omar, D. Rein-Weston, and H. Yoo, 2013, NextGen Operations in a Simulated NY Area Airspace, 32 nd Digital Avionics Systems Conference, Syracuse, New York. + + + + + Considerations for Interval Management Operations in a Mixed-Equipage Environment + + LesleyWeitz + + + JonathanHammer + + + RandallBone + + + WilliamPenhallegon + + + PeterMoertl + + 10.2514/6.2012-5616 + + + 12th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference and 14th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference + Indianapolis, Indiana + + American Institute of Aeronautics and Astronautics + 2012 + + + Weitz, L., R. Katkin, P. Moertl, W. Penhallegon, J. Hammer, R. Bone, and T. Peterson, 2012, Considerations for Interval Management Operations in a Mixed-Equipage Environment, 12 th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, Indianapolis, Indiana. + + + + + Initial Investigations of Controller Tools and Procedures for Schedule-Based Arrival Operations with Mixed Flight-Deck Interval Management Equipage + + ToddCallantine + + + ChristopherCabrall + + + MichaelKupfer + + + JoeyMercer + + + ThomasPrevot + + 10.2514/6.2012-5673 + + + 12th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference and 14th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference + Chicago, Illinois + + American Institute of Aeronautics and Astronautics + 2013 + + + Tenth USA/Europe Air Traffic Management Research and Development Seminar + + + Callantine, T., M. Kupfer, L. Martin, and T. Prevot, 2013, Simulations of Continuous Decent Operations with Arrival-Management Automation and Mixed Flight-Deck Interval Management Equipage, Tenth USA/Europe Air Traffic Management Research and Development Seminar, Chicago, Illinois. + + + + + Flight Deck Surface Trajectory-Based Operations (STBO): Simulation Results and ConOps Implications + + DFoyle + + + BHooey + + + DBakowski + + + JWilliams + + + CKunkle + + + 2011 + Berlin, Germany + + + Ninth USA/Europe Air Traffic Management Research and Development Seminar + + + Foyle, D., B. Hooey, D. Bakowski, J. Williams, and C. Kunkle, 2011, Flight Deck Surface Trajectory-Based Operations (STBO): Simulation Results and ConOps Implications, Ninth USA/Europe Air Traffic Management Research and Development Seminar, Berlin, Germany. + + + + + Air-Ground Integrated Concept for Surface Conflict Detection and Resolution + + VeeraVaddi + + + VictorCheng + + + JasonKwan + + + SandyWiraatmadja + + + SandraLozito + + + YoonJung + + 10.2514/6.2012-5645 + + + 12th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference and 14th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference + Indianapolis, Indiana + + American Institute of Aeronautics and Astronautics + 2012 + + + Vaddi, S., V. Cheng, J. Kwan, and S. Wiraadmatja, 2012, Air-Ground Integrated Concept for Surface Conflict Detection and Resolution, 12 th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, Indianapolis, Indiana. + + + + + Flight deck robustness/conformance testing with a surface management system + + DavidCFoyle + + + DeborahLBakowski + + + BeckyLHooey + + + LaraW SCheng + + + CynthiaAWolter + + 10.1145/2669592.2669687 + + + + Proceedings of the International Conference on Human-Computer Interaction in Aerospace + the International Conference on Human-Computer Interaction in AerospaceSanta Clara, California + + ACM + 2014 + + + Foyle. D, D. Bakowski, B. Hooey, L. Cheng, and C. Wolter, 2014, Flight Deck Robustness/Conformance Testing with a Surface Managmeent System: An Integrated Pilot-Controller Human-in-the-loop Surface Operations Simulation, International Conference on Human-Computer Interaction in Aerospace (HCI-Aero), Santa Clara, California. Email Address Shannon.j.zelinski@nasa.gov + + + + + + diff --git a/file827.txt b/file827.txt new file mode 100644 index 0000000000000000000000000000000000000000..b9f58478c7ce7714c7ae7711970f5680536fc250 --- /dev/null +++ b/file827.txt @@ -0,0 +1,596 @@ + + + + +IntroductionThere are many concepts being developed to address capacity and safety challenges in high density terminal airspace.Several of these require nominal arrival and departure paths to be defined to/from runways to take advantage of a more predictable trajectory based environment.The Terminal-Tactical Separation Assured Flight Environment [1] concept uses nominal terminal routing to approximate horizontal flight intent and aid it's conflict detection algorithms.Automation concepts utilizing precision schedulers [2,3] use terminal arrival routes defined all the way to the runway to precisely schedule arrivals to not only the runway, but to upstream merge points.Several concepts provide visual controller aids relative to terminal arrival routes to facilitate merging and spacing [4,5].A number of simulation environments have been developed to assess these and similar trajectory-based terminal area concepts.These range from time-based queuing [6][7][8], to trajectory-based fast-time [9,10], to human-in-the-loop [11] simulation environments.All of these simulation environments require models of existing real-world or researcher-designed routes to accurately assess concepts.Many of today's published routes do not contain details between the meter fixes and runways.Because aircraft generally follow the same nominal paths or flows, historical track data can be used to design research models of arrival and departure trajectories.Route design methods are needed to identify not only distinct nominal terminal area arrival and departure flows while ignoring outliers, but standard path control techniques controllers use to manage the traffic.Ideally, these methods should also require low manual effort to allow the assessment scope to expand as the concept requires.Terminal route models have been developed largely manually [12], sometimes with the aid of trajectory visualization and design software [13][14][15].But this is still a lengthy process that has limited concept assessment scope.For example, most Trajectory Analysis and Modeling Environment [10] assessments have been scoped to Houston Intercontinental and most terminal area Multi Aircraft Control System [11] assessments have been scoped to Los Angeles International.Leiden and Atkins [16] presented two algorithmic trajectory bundling methods to define nominal terminal area routing: a ridge detection method and a k-means method.The ridge detection method could easily identify route segments of high density but there were gaps between segments where no ridge could be detected.The k-means method could identify fully connected routes without any gaps but its heuristic search approach required it to run many times with different seeds before converging to a global minima and the number of desired routes was required as an input.Leiden and Atkins addressed these difficulties by clustering hierarchically to achieve quicker convergence, and by 4B2-2 using the number of ridges found with the ridge detection method to estimate the number of desired routes.However, this method does not filter outliers, thus many of the subtleties of terminal area routing may be lost due to blending dominant flows with those that are distinct yet less dominant.This paper presents another algorithmic method for defining nominal terminal routing using graph representations of historical flight tracks.Unlike the k-means method, this graph-based method is deterministic and self defines the number of nominal routes identified.When the graph is searched for dominant flows, outliers are ignored, which avoids blending multiple distinct flows.This method is used to identify not only the nominal routes, but commonly used path control techniques such as path stretches and shortcuts, making it well suited to aid prototyping of terminal area route models for concept assessment. +MethodPrevious work defined route structure representing en-route traffic patterns by clustering individual flight trajectory intersections into nodes and then defining links between nodes [17].Unique properties of terminal airspace, such as frequent turns and higher degree of precision required, made it difficult to directly apply the en-route technique to the terminal area.This section describes a node-link graph routing method designed specifically for terminal area traffic. +Determining Flight Runway AssignmentBefore generating nominal routes, flight track data were filtered to arrivals or departures of a single airport runway and engine type.This was done because paths to different runways are often segregated some distance from the runway thresholds, and performance variation often precludes using the same path for different engine types.The flight track data included aircraft type used to determine engine type, but it did not contain accurate runway assignment information.Therefore, arrival and departure runway assignments were determined by runway relative position and course of tracks just prior to arrival or after departure.Figure 1 shows the metrics used to assign arrival and departure runways.Each track point was rotated to a coordinate system with origin at the runway threshold and the positive x-axis oriented along the runway.The black dot in Figure 1 represents the runway threshold and the thick black arrow points along the runway.The dashed black lines represent the x-and y-axes with origin at the runway threshold.The large orange dot represents the track point position where x and y are the track point coordinates relative to the runway.The track point course relative to the runway is labeled as ω.For arrivals, the track point course is calculated relative to the previous track point.Departure track point course is calculated relative to the next track point. +Figure 1. Runway Assignment MetricsThe track point must be within the gray shaded area to meet the criteria for a given runway assignment.For arrivals, this region is a rectangle defined by y min , y max , x min , and x max shown in Figure 1.In addition, abs(ω) must be less than a given ω max not shown in the Figure .The last three track points prior to crossing the runway threshold must meet these criteria to be assigned the arrival runway.If a flight met the criteria for more than one runway, as was often the case with closely spaced parallel runways, the runway that produced the smallest abs(y) for the last track point was assigned.Departure runway assignment was more complicated due to common sharp turns just after departure.For departures, the gray rectangle's y dimension must be larger than for arrivals to accommodate these turns.Also, the rectangle corners closest to the runway threshold are cut off by defining a θ max relative to x min .Assuming x min is a 4B2-3 likely minimum wheels-off point, even sharp turning departures are not likely to enter theses cut off corners.Finally, it was assumed that tracks with abs(y) more than a given y ω were turning and their course should be in a similar direction requiring that y and ω have the same sign.Only the first departure track must meet these criteria to be assigned the departure runway.As with arrival assignments, if more than one runway met the criteria, the runway producing the smallest abs(y) was assigned.Table 1 shows the criteria values used to assign arrival (Arr) and departure (Dep) runways.Values for x and y are in nmi and values for ω and θ are in degrees.Arr x and y values are based on runway assignment criteria used by Robinson and Kamgarpour [18].Arr ω max and all Dep values were determined based on trial and error such that reasonable runway assignment results could be achieved at different airports using the same values.Reasonable runway assignments were verified visually for at least one randomly chosen day for each airport.Figure 2 shows an example visual verification for Los Angeles International (LAX) arrivals to 24R and 25L and departures from 24L and 25R on February 4, 2010.The three vectors prior to crossing the arrival runway threshold are shown for arrivals and the first vector after departure is shown for departures.This method does not account for missed approaches, nor will it assign all flights to a runway.For the example shown in Figure 2, 13% of arrivals and 3% of departures were unassigned.Because the purpose of these assignments is to identify nominal routes, unassigned flights only serve to filter possible outliers. +Figure 2. Runway Assignment Verification +Generating the Node-Link GraphsFor each runway assignment made, the runway threshold point was added to the end of arrival trajectories or to the beginning of departure trajectories.The sequence of all departure tracks was reversed so that they could be processed like arrivals.Each flight trajectory was then simplified using the Douglas-Peucker algorithm [19] with a distance dimension of 0.5 nmi in x-y space.This tended to leave vertices at turning points with straight-line segments in between.An agglomerative method was used to cluster vertices of all the simplified flight trajectories of the same engine type assigned to the same runway.Even though departure trajectories were reversed, they were clustered separately from arrivals.Initial vertex clusters were generated by defining a 0.2 nmi resolution x-y density map of the vertices.Each cluster position was given by the average position of it's vertices and cluster weight was given by the number of vertices.Then, in order of decreasing weight, each cluster was merged with all other clusters within a distance threshold of 1.5 nmi.This step iterated until all clusters were separated by at least the threshold distance.The resulting clusters were used as nodes to generate a directed graph.Each adjacent pair of vertices along a simplified flight trajectory that belonged to different nodes contributed to a link between those nodes.Link weight was given by the number of unique flights contributing to the link.For any set of flight track data to a single airport runway, there should be multiple source nodes identifying where flights entered the terminal airspace, and a single sink node identifying the runway threshold. +4B2-4A nominal route from each source node to the runway sink node was found using Dijkstra's algorithm [20].Each link was assigned a cost equal to its Euclidean distance in x-y space (d) divided by its weight (w).The shortest path found using this cost definition tended to minimize the total path distance while maximizing the weights of chosen links.In order to control relative emphasis placed on minimizing path length or maximizing weight, the cost function became J=d a /w b , where a and b were used to weight d and w.The shortest path graph links were reweighted according to the number of flights whose path from a source to sink traveled the link. +Filtering the Shortest Path GraphsThe shortest path graph still had many low weight paths that needed filtering.This was partially caused by some truncated flight track data generating source nodes well into terminal airspace.Leaves of the tree were pruned to a desired minimum link weight w min .In addition, all leaves closer than a desired minimum Euclidean distance from the runway d min were pruned.Figure 4 shows the shortest path graph from Figure 1 filtered using w min = 100 and d min = 30 nmi.Note that the only difference between Figure 3 and 4 is that many nodes and low weight links have been removed in Figure 4 so that only the most dominant routes remain. +Refining the Shortest Path GraphsFinally, the filtered shortest path graph was refined to a minimal set of nodes and links to represent nominal routes.This included creating nodes at intersections, consolidating links that partially overlapped, and removing unnecessary nodes along straight paths.In between each step, the shortest path from all source nodes to sink, and shortest path graph filtering was repeated.Figure 5 illustrates the refinement steps on a portion of the filtered shortest path graph from Figure 4.Figure 5A shows the original filtered shortest path graph.Because links, (6,5) and (7,1) intersect, a new node is created at the intersection labeled 9 in Figure 5B.The new shortest path from node 6 to 1 now travels straight from node 9 to 1 rather than going through 5, so link (9,5) is removed.Link (5,1) remains in Figure 5B because node 5 served as a source node for more than w min = 100 flights.The next step, from Figure 5B to 5C, locates and merges all nodes with their closest neighboring link 4B2-5 within 0.5 nmi.Nodes 5, 6, and 8 all merge with their closest neighboring link.This consolidates partially overlapping links and more appropriately positions merge points.Finally, between Figure 5C and 5D, unnecessary nodes along fairly straight paths are removed.Unnecessary nodes were identified using the Douglas-Peucker algorithm with a distance dimension of 0.5 nmi in x-y space, as was used to simplify original flight trajectories.Nodes 5, 2, and 3 are removed because they deviate less than 0.5 nmi from a straight line between nodes 4 and 1.Similarly, node 9 is removed because it deviates less than 0.5 nmi from a straight line between nodes 7 and 1. +Figure 5. Graph Refinement +ResultsNominal routes in Southern California TRACON (SCT) were generated using the graphbased method.The data analyzed consisted of 52 24hour Center/TRACON Automation System [21] recordings of TRACON flight plan and track data between February 4 and May 6, 2010.Flights were assigned to runways with no distinction made between instrument and visual operations.Observations were made between arrival and departure routes for different engine types.Engine types were analyzed separately because performance variation often precludes using the same path for different engine types.The graph-based routes were also compared with routes generated manually [12], and with routes generated using the k-means method [16]. +Route Comparison by Engine TypeFor the 52 days of SCT track data analyzed, the most frequently assigned runways were LAX 24R and 25L for arrivals, and 24L and 25R for departures.LAX flights mainly consist of jets and turboprops.One of LAX's smaller satellite airports Orange County (SNA) 19R, was the most assigned SCT arrival and departure runway for pistons.Together, these runways represent all 3 engine types, arrivals and departures, and large and small airports for graph-based route analysis.Table 2 shows an engine type break down of numbers of flights mapped to each of these LAX and SNA runways over the 52 days of track data. +LAX Departure Routes by Engine TypeThere are significant differences between jet and turboprop routes.Turboprops tend to use more direct routes somewhat segregated from jets.Departure and arrival routes appear to be well segregated as well, crossing more orthogonally closer to the runway.Figure 8 shows SNA 19R graph-based arrival and departure routes together with magnified runway area in the lower left.Because of the lower volume of flights, SNA routes were generated for a minimum link weight of w min = 25 versus w min = 100 for LAX.Contrary to LAX, jet and turboprop routes appear very similar for both arrivals and departures.It is the piston routes that are segregated from the rest.For the most part, arrival and departure routes are segregated.However, there are a few segments where arrivals and departures follow the same path, presumably within different altitude ranges.Most noticeably, one piston departure route turns North to follow the same path as the only major piston arrival route generated. +Figure 8. SNA Arrival and Departure Routes by Engine TypeIn general, departure routes tend to display less clean branching that occur much closer to the airport than arrival routes.This is because, other than to avoid arrival routes, departure fanning to their various destinations requires less precision than arrivals merging to a single runway. +Route Comparison by MethodThe graph-based routes from Figures 678were compared to results from two other route generation methods, manual and k-means.The manual routes [12] were developed for use in high-fidelity human-in-the-loop simulations of LAX.They included arrival and departure procedures (including altitude restrictions) between en-route transition waypoints in center airspace and runway 4B2-7 thresholds for six SCT airports including LAX and SNA.Three stages of terminal routes were considered, en-route transitions, common routes, and runway transitions.Published STARs (Standard Terminal Arrival Routes) and SIDs (Standard Instrument Departures) and four days of track data were analyzed to manually build runway transitions between the end of STARs or the beginning of SIDs and runway thresholds where none existed, and to modify STARs and SIDs where the dominant flow pattern appeared to deviate.In addition, routes were created where dominant flows did not follow any published STARs or SIDs, and any unused STARs and SIDs were deleted.The k-means routes [16] were developed for use in medium-fidelity fast-time simulations of congested multiple-airport terminal areas, also know as metroplexes.Three-dimensional routes were developed for several metroplex terminal areas including SCT.A separate ridge method was used to determine the desired k number of routes to represent a given set of trajectories.Then each trajectory was sampled to have the same number of waypoints.The average Euclidian distance of trajectory waypoints from their counterpart waypoints along candidate routes served as the distance function for k-means clustering.A final processing step merged the resulting individual routes to a connected network, with a method similar to the step depicted from Figure 5B to 5C, to ensure that all network points were either merge/diverge points or more than a distance threshold from any other network segment.K-means routes were generated for a single day's worth of either arrivals or departures to/from a single airport.All three methods display similar en-route transitions and common routes with differences mainly in routes that were filtered due to an absence of critical traffic volume.k-means routes up to the last point before the runways.The final segment would connect directly to some mid point over the airport.The insufficient resolution is most apparent in Figure 9 where arrivals from the West appear to bypass their base turn onto the final approach.In addition, inappropriate choice of k can cause blending of distinct flows to create a flow where none exists.This can be seen best in SNA arrivals.Figure 13 shows a close-up of Figure 11 including flight tracks in gray. +Figure 13. K-Means Blending ExampleBoth manual and graph-based routes show arrivals from the North-North-West taking two distinct routes to the runway.The k-means route shown is a blending of these two routes where there is no observable flow pattern from track data.Leiden and Atkins solved this problem with hierarchical clustering and demonstrate it on New York TRACON [16] where runway transitions diverging from a single common route are both large enough flows.However, as seen in Figure 8, the Northern runway transition is primarily used by pistons.There may be four times as many jets and turboprops on the other transition route, prompting the k-means method to ignore the diversion and treat it as a single route.This nuance may have a larger impact on both a scheduling concept assessment as the k-means route is more than three nmi shorter than in reality, and a conflict detection concept assessment as the k-means is displaced more than three nmi.The manual and graph-based LAX arrival routes in Figure 9 are quite similar close to the runway, turning onto final approach in similar locations.However, LAX departure routes in Figure 10 are quite different close to the runway.Manual routes show a departure loop from each runway that flights may take to gain altitude over the runways and head North East.The two-dimensional graph-based routes bypass the loop to shorten the path and head North East directly from the runway.The runways are in West flow configuration, so the direct North East departure routes would not be flyable.The k-means North East departure route actually captures the looping behavior in lower resolution because the departure start point shown is to the West of the runways.This suggests that the graph-based method should be adapted to three-dimensions. +4B2-9Even without the loop departures, the manual method produced more LAX departure runway transition paths than either of the algorithmic methods.These included published and new runway transitions along with common bypasses to the same common route.The graph-based method produces route networks with a tree structure such that there may be many en-route transition paths merging into fewer common routes, merging in to even fewer runway transition paths.This tends to produce more en-route transition paths and fewer runway transition paths than the manual method.The k-means method groups entire trajectories together, not just the en-route or runway transitions or common routes.Even with higher resolution, the strongest clusters would minimize great distances between en-route transitions rather than the smaller distances between runway transitions.Distinct runway transition would only be captured if all the flights using the same en-route transition also used the same runway transition.Other than dimensionality, the graph-based routes differ from the other two methods in that they represent specific engine types.As seen in the previous section, nominal flight paths between the different engine types can differ significantly to segregate slower from faster aircraft until they are very close to the runway.Many of the manual routes that graph-based captured and k-means did not are LAX turboprop or SNA piston routes.This would be a simple matter for most route generation methods to fix by adding engine type, runway and arrival/departure as trajectory filtering parameters. +Increasing Route OptionsThe defining characteristic of graph-based method is its selection of a single dominant distinct traffic flow path.The strength of this method is that it filters less common paths such as vectored path stretches and shortcuts.On the other hand, in busy TRACONS, the dominant path may include some nominal delay.Also, if multiple runway transitions exist from a single common route, the method will select only the shortest path as defined by its cost function.The segregation of routes by engine type may appear to defy this structure when different engine types use similar common routes but very different runway transitions.But this still will not capture multiple route options available to a single flight.One way to develop more route options is to vary the exponents a and b in the shortest path cost function J=d a /w b .However, this only captures a variation of routes where path distance is at odds with weight, i.e.where longer routes are more commonly used and shorter routes are less commonly used.This is more often the case for arrivals than departures.Therefore departure routes tend to change very little as a and b are varied.Here, path stretching is so common that the a=b route from the North that then curves West appears to be a path stretch.Two longer and even more common path stretch routes labeled as ab routes bypass the path stretch and merge with another stream from the East.Another way to develop more route options is to expand the shortest path search beyond source to sink pairs.Figure 15 shows LAX 24R jet arrival shortest paths to the runway from all original graph links representing at least 100 flights (w≥100).Compare this to Figure 3, which shows shortest paths from all 4B2-10 sources to the runway for the same flights.The shortest paths from w≥100 links capture several additional common path stretch, direct, and base extension options. +Figure 15. LAX 24R Jet Arrival OptionsFigure 16 shows all shortest path routes from w≥100 links to the runway for ORD 09R jet arrivals.Shortest path graph source nodes are highlighted in green identifying a few unconnected runway transition paths.This is solved by finding shortest path routes from all source nodes farther than 30 nmi from the runway to all source nodes within 30 nmi of the runway.Figure 17 shows these additional transitions resulting in a fully connected set of path stretch, direct, and base extension options for ORD 09R jet arrivals. +Figure 16. ORD 09R Jet Incomplete RoutesThis same method was used to find all shortest path routes and additional transitions from w≥200 links for LAX jet departures shown in Figure 18.The manual routes for both runways are overlaid in magenta.The multiple shortest path route options now display the same looping behavior as the manually constructed routes.Thus, by expanding the shortest path search, the strength of the graph-based methods in identifying a distinct dominant path can be preserved and expanded to identify multiple dominant paths. +ConclusionsThis paper presented a graph-based method for algorithmically defining nominal terminal routing.First a fully connected directed graph was created to represent a given set of flight tracks.Then Dijkstra's algorithm is used to find shortest paths between graph nodes.Shortest path routes were generated between source nodes and a runway sink node for each of three engine types.Routes of different engine types could differ significantly.Graph-based routes were compared with routes generated from a k-means algorithm and with manually generated routes for two airports in SCT.For the most part, routes from these three methods compared well.The manual routes were assumed to be of sufficient accuracy for humanin-the-loop simulation but take a lot of manual effort to generate.The k-means routes are lower fidelity 4B2-11 due to the resolution used.But even if resolution were increased, the desired k number of routes must be carefully selected to avoid blending distinct routes together.The graph-based single path method ignored looping departures in favor of unrealistically direct paths due to the algorithm's current twodimensionality.This paper also presented a method to identify multiple route options.This is done by finding shortest paths between all links of sufficient weight to the runway.This method captured common path-stretch and direct route options as well as the departure loops ignored by the single shortest path.The overall strength of the graph-based method is that it is able to capture distinct flows and nuances of highly complex terminal airspace.This method is well suited to aid rapid prototyping of terminal area route models for concept assessment.Route modeling was once a lengthy process of trial and error requiring subject matter expertise.This method can significantly reduce modeling time, enabling assessment scope to expand as the concept requires.Figure 33Figure 3 shows the shortest path graph results for 14,316 Los Angeles International Airport (LAX) 24R jet arrivals taken from 52 24-hour traffic samples between February 4 and May 6, 2010.The cost function weights, a and b, were both set to 1. Darker thicker blue lines represent higher weight links.Nodes are shown as orange dots.The graph overlays gray dashed lines representing original simplified flight tracks used to create the graph.The shortest path graph has a tree structure where branches of the tree merge and increase in weight until all paths lead to a single sink. +Figure 3 .3Figure 3. Shortest Path Graph +Figure 4 .4Figure 4. Filtered Shortest Path Graph +Figures 6 and 7 6 Figure 6 .Figure 7 .7667Figures 6 and 7 show LAX graph-based arrival and departure routes, respectively.Boxes at the upper right of each Figure magnify routes near the runways shown in orange.Only jet and turboprop routes are shown because there were too few piston flights to generate routes. +Figures 9 ,9Figures 9, 10, 11, and 12 compare manual, graph-based, and k-means routes for LAX arrivals, LAX departure, SNA arrivals, and SNA departures, respectively.K-means routes prior to the merging step are shown because the 3 nmi distance threshold distorted the result far more than the 0.5 nmi threshold used for the graph-based routes.Only manual and graph-based route waypoints are shown because they represent turn and merge points, and published fix locations in the case of manual routes.Whereas k-means waypoint locations were the result of sampling the trajectories every 8 to 12 nmi. +Figure 9 .Figure 10 . 8 Figure 11 .Figure 12 .91081112Figure 9. LAX Arrival Routes by Method +show +Figure 1414Figure 14 shows arrival routes for a range of a and b.For constant b = 1.0, a was varied between 0.0 and 1.0 in increments of 0.1.Also, for constant a = 1.0, b was varied between 0.1 and 1.0 in increments of 0.1.Results for b=0.0 are not shown because the routes are unrealistic and obstruct the other routes. +Figure 14 .14Figure 14.Effects of Varying Cost Function Dark blue lines represent the original a=b=1.0routes.All additional routes are shown in a lighter blue.Reducing a did not have an effect on LAX 24R routes.Therefore, all lighter blue routes shown represent less commonly used shorter routes when b was reduced.For contrast, Chicago O'Hare (ORD) 09R arrival routes are shown on the right of Figure 14.Here, path stretching is so common that the a=b route from the North that then curves West appears to be a path stretch.Two longer and even more common path stretch routes labeled as ab routes bypass the path stretch and merge with another stream from the East. +Figure 17 .Figure 18 .1718Figure 17.ORD 09R Jet Arrival Options + + +Table 1 . Runway Assignment Criteria x min x max y min y max ω max θ max y ω1Arr -10 0-.25 .2510Dep .754-.5.56045.1 +Table 2 . Number of Flights by Engine Type2RunwayJetTurboPistonArrivalLAX 24R LAX 25L SNA 19R14,316 13,653 4,4601,728 1,269 32311 12 317DepartureLAX 24L LAX 25R SNA 19R12,068 14,493 7,2081,446 883 5435 7 1,582 + + + +Email AddressesShannon.J.Zelinski@nasa.gov31st Digital Avionics Systems Conference October [14][15][16][17][18]2012 + + + + + + + Tactical Conflict Detection in Terminal Airspace + + HuabinTang + + + JohnERobinson + + + DallasGDenery + + 10.2514/1.51898 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 34 + 2 + + 2011 + American Institute of Aeronautics and Astronautics (AIAA) + + + Journal of Guidance, Control, and Dynamics + Tang, H., J. Robinson III, D. G. Denery, 2011, Tactical Conflict Detection in Terminal Airspace, Vol. 34, No. 2, Journal of Guidance, Control, and Dynamics. + + + + + Design and Evaluation of the Terminal Area Precision Scheduling and Spacing System, 9 th USA + + HNSwenson + + + JThipphavong + + + ASadovsky + + + LChen + + + CSullivan + + + LMartin + + + + Europe Air Traffic Management R&D Seminar + + 2011 + Berlin, Germany + + + Swenson, H. N., J. Thipphavong, A. Sadovsky, L. Chen, C. Sullivan, L. Martin, 2011, Design and Evaluation of the Terminal Area Precision Scheduling and Spacing System, 9 th USA/Europe Air Traffic Management R&D Seminar, Berlin, Germany. + + + + + Arrival Management Architecture and Performance Analysis with Advanced Automation and Avionics Capabilities + + AslaugHaraldsdottir + + + JulienScharl + + + JanetKing + + + MatthewBerge + + 10.2514/6.2009-7006 + + + 9th AIAA Aviation Technology, Integration, and Operations Conference (ATIO) + Hilton Head, South Carolina + + American Institute of Aeronautics and Astronautics + 2009. September 21-23 + + + Haraldsdottir, A., J. Scharl, J. King, M. E. Berge, 2009, Arrival Management Architecture and Performance Analysis with Advanced Automation and Avionics Capabilities, AIAA 9 th Aviation Technology, integration and Operations (ATIO) Conference, September 21-23, Hilton Head, South Carolina. + + + + + RNP RNAV Arrival Route Coordination + + PaulMacwilliams + + + ArthurSmith + + + ThomasBecher + + 10.1109/dasc.2006.313682 + + + 2006 ieee/aiaa 25TH Digital Avionics Systems Conference + Portland, Oregon + + IEEE + 2006 + + + MacWilliams, P. V., A. P. Smith, T. A. Becher, 2006, RNP NRAV Arrival Route Coordination, 25 th Digital Avionics Systems Conference, Portland, Oregon. + + + + + Controller-Managed Spacing - A Human-In-The-Loop Simulation of Terminal-Area Operations + + MichaelKupfer + + + ToddCallantine + + + JoeyMercer + + + LynneMartin + + + EverettPalmer + + 10.2514/6.2010-7545 + + + AIAA Guidance, Navigation, and Control Conference + Berlin, Germany + + American Institute of Aeronautics and Astronautics + 2011 + + + Kupfer, M., T. Callantine, L. Martin, J. Mercer, E. Palmer, 2011, Controller Support Tools for Schedule- Based Terminal-Area Operations, 9 th USA/Europe Air Traffic Management R&D Seminar, Berlin, Germany. + + + + + Modeling System-wide Predictability and Associated Air Carrier Benefits + + SeliAgbolosu-Amison + + + StephaneMondoloni + + 10.2514/6.2011-6810 + + + 11th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Virginia Beach, Virginia + + American Institute of Aeronautics and Astronautics + 2011 + + + Agbolosu-Amison, S., S. Mondoloni, 2011, Modeling System-Wide Predictability and Associated Air Carrier Benefits, 11 th AIAA Aviation Technology, Integration and Operations Conference, Virginia Beach, Virginia. + + + + + Aviation Simulation with SIMMOD + + DorothyBrady + + 10.1061/40530(303)14 + + + + The 2020 Vision of Air Transportation + + American Society of Civil Engineers + 2011 + + + SIMMOD Reference Manual + Federal Aviation Administration, 2011, SIMMOD Reference Manual, http://www.airporttools.com/ + + + + + Design and Evaluation of a Stochastic Time-Based Arrival Scheduling Simulation System + + DanielMulfinger + + + AlexanderSadovsky + + 10.2514/6.2011-6874 + + + + 11th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Virginia Beach, Virginia + + American Institute of Aeronautics and Astronautics + 2011 + + + Mulfinger, D., A. Sadovsky, 2011, Design and Evaluation of a Stochastic Time-Based Arrival Scheduling Simulation System, 11 th AIAA Aviation Technology, Integration and Operations Conference, Virginia Beach, Virginia. apt/html/docu mentation.php + + + + + The Airspace Concepts Evaluation System Terminal Area Plant Model + + RobertWindhorst + + + LarryMeyn + + 10.2514/6.2007-6555 + + + AIAA Modeling and Simulation Technologies Conference and Exhibit + Hilton Head, South Carolina + + American Institute of Aeronautics and Astronautics + 2007 + + + Windhorst, R. D., L. A. Meyn, 2007, The Airspace Concepts Evaluation System Terminal Area Plant Model, AIAA Modeling and Simulation Technologies Conference, Hilton Head, South Carolina. + + + + + A Trajectory Modeling Environment for the Study of Arrival Traffic Delivery Accuracy + + JulienScharl + + + AslaugHaraldsdottir + + + EwaldSchoemig + + 10.2514/6.2006-6612 + + + AIAA Modeling and Simulation Technologies Conference and Exhibit + Keystone, Colorado + + American Institute of Aeronautics and Astronautics + 2006 + + + Scharl, J., A. Haraldsdottir, E. G. Schoemig, 2006, A Trajectory Modeling Environment for the Study of Arrival Traffic Delivery Accuracy, AIAA Modeling and Simulation Technologies Conference, Keystone, Colorado. + + + + + Human-In-the-Loop Evaluation of NextGen Concepts in the Airspace Operations Laboratory + + ThomasPrevot + + + PaulLee + + + ToddCallantine + + + JoeyMercer + + + JeffreyHomola + + + NancySmith + + + EverettPalmer + + 10.2514/6.2010-7609 + + + AIAA Modeling and Simulation Technologies Conference + Toronto, Canada + + American Institute of Aeronautics and Astronautics + 2010 + + + Prevôt, T., P. Lee, T. Callantine, J. Mercer, J. Homola, N. Smith, et al., 2010, Human-In-The-Loop Evaluation of NextGen Concepts in the Airspace Operations Laboratory, AIAA Modeling and Simulation Technologies Conference, Toronto, Canada. + + + + + Super Density Operations Airspace Modeling for the Southern California Metroplex + + SebastianTimar + + + GauravNagle + + + AdityaSaraf + + + PeterYu + + + PeterHunt + + + AndrewTrapani + + + NickJohnson + + 10.2514/6.2011-6535 + + + AIAA Modeling and Simulation Technologies Conference + Portland, Oregon + + American Institute of Aeronautics and Astronautics + 2011 + + + Timar, S. D., G. Nagle, A. Saraf, P. Yu, P. Hunt, A. Trapani, N. Johnson, 2011, Super Density Operations Airspace Modeling of the Southern California Metroplex, AIAA Modeling and Simulation Technologies Conference, Portland, Oregon. + + + + + Health hazard evaluation report: HHE-81-042-832, Federal Aviation Administration, New York Air Route Traffic Control Center, Ronkonkoma, New York. + 10.26616/nioshhhe81042832 + + + + Terminal Area Route Generation, Evaluation, and Traffic Simulation + + U.S. Department of Health and Human Services, Public Health Service, Centers for Disease Control, National Institute for Occupational Safety and Health + 2011 + + + MITRE Center for Advanced Aviation System Development, 2011, Terminal Area Route Generation, Evaluation, and Traffic Simulation, http://www.mitrecaasd.org/work/project_ details.cfm?item_id=152 4B2-12 + + + + + New brochure targets marine industry + 10.1016/s1350-4789(08)70293-9 + + + + Sealing Technology + Sealing Technology + 1350-4789 + + 2008 + 5 + 4 + 2011 + Mark Allen Group + + + CSSI, Inc., 2011, Terminal Area Route Generation, Evaluation and Traffic Simulation Software Brochure, http://targets.cssiinc.com/external/downloa ds/documents/TARGETS-Brochure.pdf + + + + + An Integrated Tool for NextGen Concept Design, Fast-Time Simulation, and Analysis + + ToddCallantine + + 10.2514/6.2008-6343 + + + AIAA Modeling and Simulation Technologies Conference and Exhibit + Honolulu, Hawaii + + American Institute of Aeronautics and Astronautics + 2008 + + + Callantine, T., 2008, An Integrated Tool for NextGen Concept Design, Fast-Time Simulation, and Analysis, AIAA Modeling and Simulation Technologies Conference, Honolulu, Hawaii. + + + + + Trajectory Clustering for Metroplex Operations + + KennethLeiden + + + StephenAtkins + + 10.2514/6.2011-7066 + + + 11th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Virginia Beach, Virginia + + American Institute of Aeronautics and Astronautics + 2011 + + + Leiden, K., S. Atkins, 2011, Trajectory Clustering for Metroplex Operations. 11 th AIAA Aviation Technology, Integration and Operations Conference, Virginia Beach, Virginia. + + + + + Defining dynamic route structure for airspace configuration + + SZelinski + + + MJastrzebski + + 10.1177/0954410011417354 + + + Proceedings of the Institution of Mechanical Engineers, Part G: Journal of Aerospace Engineering + Proceedings of the Institution of Mechanical Engineers, Part G: Journal of Aerospace Engineering + 0954-4100 + 2041-3025 + + 226 + 9 + + 2012 + SAGE Publications + + + Zelinski, S., M. Jastrzebski, 2012, Defining Dynamic Route Structure for Airspace Configuration, Vol. 226, No. 9, Proceedings of the Institution of Mechanical Engineers, Part G, Journal of Aerospace Engineering, pp. 1161-1170. + + + + + Benefits of Continuous Descent Operations in High-Density Terminal Airspace Considering Scheduling Constraints + + JohnRobinson Iii + + + MaryamKamgarpour + + 10.2514/6.2010-9115 + + + 10th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Fort Worth, Texas + + American Institute of Aeronautics and Astronautics + 2010 + + + Robinson III, J.E., M. Kamgarpour, 2010, Benefits of Continuous Descent Operations in High- Density Terminal Airspace Under Scheduling Constraints, 10th AIAA Aviation Technology, Integration and Operations Conference, Fort Worth, Texas. + + + + + ALGORITHMS FOR THE REDUCTION OF THE NUMBER OF POINTS REQUIRED TO REPRESENT A DIGITIZED LINE OR ITS CARICATURE + + DavidHDouglas + + + ThomasKPeucker + + 10.3138/fm57-6770-u75u-7727 + + + Cartographica: The International Journal for Geographic Information and Geovisualization + Cartographica + 0317-7173 + 1911-9925 + + 10 + 2 + + 1973 + University of Toronto Press Inc. (UTPress) + + + Digitized Line or It's Caricature + Douglas, D. H. and T. K. Peuker, 1973, Algorithms for the Reduction of the Number of Points Required to Represent a Digitized Line or It's Caricature, Vol. 10, No. 2, Cartographica: The International Journal for Geographic Information and Geovisualization, pp. 112-122. + + + + + Algorithmen - Eine Einführung + + ThomasHCormen + + + CharlesELeiserson + + + RonaldRivest + + + CliffordStein + + 10.1515/9783110522013 + + 2001 + De Gruyter + + + + Second Edition + Cormen, T. H., C. E. Leiserson, R. L. Rivest, C. Stein, 2001, Section 24.3: Dijkstra's algorithm, Introduction to Algorithms (Second Edition), MIT Press and McGraw-Hill, pp. 595-601. + + + + + Research Foundations Main, NASA Ames Aviation Systems Division + + SantosDe Los + + + + NASA Ames Research + + 2008 + + + De Los Santos, V., 2008, Research Foundations Main, NASA Ames Aviation Systems Division, NASA Ames Research + + + + + Initial Characterization of the 30 kW Miniature Arc Jet (mARC II) at NASA Ames Research Center + + Center + + 10.2514/6.2020-3108.vid + + + + American Institute of Aeronautics and Astronautics (AIAA) + + + Center, http://www.aviationsystemsdivision.arc.nasa. gov/ research/foundations/index.shtml + + + + + + diff --git a/file828.txt b/file828.txt new file mode 100644 index 0000000000000000000000000000000000000000..d9772cd70b14e53ed02571f3f5c78369b112b0a1 --- /dev/null +++ b/file828.txt @@ -0,0 +1,559 @@ + + + + +Mulfinger [1] showed that delay and throughput benefits are more sensitive to these spacing buffers than the uncertainty itself.But the appropriate spacing buffer to achieve a desired controller intervention rate does decrease significantly with the uncertainty, allowing more benefit.To this end, benefit analyses have focused on showing the arrival time accuracy achievable for a specific technology or concept of operations employing a set of technologies [2][3][4][5][6][7].The larger benefit of the increased accuracy is unclear.Some studies have extrapolated or measured the benefit of scheduling and spacing precision on delay and throughput as compared to current day operations.The Terminal Area Precision Scheduling and Spacing System [8][9] verified in human-in-the-loop simulations that precision scheduling and spacing tools can allow spacing buffers of 0.3-0.4nmi to accommodate 10-15% more traffic than current operations.However, the results are limited to a handful of simulated 100-minute traffic scenarios for a single airport.Ballin and Erzberger [10] analyzed the arrival A spacing precision for 30 Dallas Fort Worth (DFW) rush periods and estimated that a 0.25 nmi separation buffer should allow for at least 15% more capacity than was observed.Again, this analysis focused on a single airport and since then, DFW operations have changed considerably.In addition to allowing reduced spacing buffers, precision scheduling and spacing concepts use defined fixed arrival paths such as Area Navigation (RNAV) and Required Navigation Performance (RNP) routing to achieve full benefit.Economic and environmental benefits (fuel usage and noise) of RNP have been quantified in case studies [11][12].The broader impact of fixed arrival routing and reduced separation buffers and the variability of this impact between different terminal areas are unclear.This paper characterizes the observed arrival spacing behavior of 29 runways belonging to 15 airports within 8 of the busiest terminal areas across the US for 32-60 days worth of traffic.Potential capacity increases due to reducing the observed 0.5 to 1.5 nmi spacing buffers to 0.3 nmi are estimated.Then fixed arrival paths are designed for each runway.The impact of these fixed paths and reduced buffers on average flight arrival time is then estimated. +II. MethodArrival flights to multiple airport runways from multiple terminal areas were analyzed to encompass a wide range of traffic arrival environments, from metroplexes to super-hubs, across the National Airspace System (NAS).The traffic analysis was performed using recorded Terminal Radar Approach Control (TRACON) flight plan and track data.First, separation behavior was characterized for each airport runway.Theoretical runway arrival capacities were estimated based on observed current spacing behavior and expected spacing behavior afforded by precision scheduling and spacing concepts.This study also estimated the effect precision scheduling and spacing along fixed arrival routes would have on terminal area flight time.RNAV arrival routes were designed to each runway and new scheduled arrival times were compared with historical arrival times using historical aircraft arrival paths. +A. Scope of StudyFlight plan and track data from the eight busiest TRACON facilities in the NAS (Atlanta, Chicago, Denver, Dallas/Fort Worth, New York, Potomac, Northern California, and Southern California) were examined.The data covered between 32 and 60 24-hour traffic samples distributed between February 2010 and early May 2010.Traffic samples were required to have uninterrupted TRACON track data between 0600 and 2200 local time.These same data were used in two previous studies where data statistics are described in more detail [13][14].This analysis focused on airport runways with relatively large numbers of tightly spaced instrument arrival operations.These included 29 runways from 14 of the 29 NPIAS (National Plan of Integrated Airport Systems [15]) identified large hubs and one satellite airport (TEB) diagramed in Fig. 1.Locations of satellite airports for which track data was available but were not included in this analysis are also shown. +B. Separation Behavior from Track DataTo characterize the unique separation behavior of each runway, Ballin and Erzberger's [10] method was modified to accommodate samples outside rush periods and differentiate between behavior observed during Visual Approach Conditions (VAC) and Instrument Approach Conditions (IAC).Robinson and Kamgarpour [13] describe the method used to determine each aircraft's landing runway and threshold crossing time from the track data.For each sequential pair of aircraft landing at the same runway, the position of the trailing aircraft at the time the leading aircraft crossed the runway threshold was extrapolated from the two track points on either side of this time.The observed separation was then calculated as the trailer's along-path distance from this point to the runway threshold.Sequential aircraft pairs to the same runway were segregated by their minimum in-trail separation requirements (standard radar separation [16][17]) determined by leader/trailer weight class.Table 1 shows the leader/trailer weight class to required separation mapping in nmi.For pairs requiring 3 nmi separation, in-trail separation at the runway threshold can be reduced to 2.5 nmi under certain conditions [16].However, these conditions were not considered in this analysis.Then a +/-3-sigma Gaussian kernel smoother with a 2 nmi span was used to smooth each histogram.The maximum bin of the smoothed histogram within a -1 to +5 nmi range of the required separation standard was assumed to be the observed target separation s.The curves generally followed the two-part separation model convolving a normal distribution of the positioning accuracy of the controller/pilot team with a Poisson distribution of the arrival gaps.The Poisson distribution can be large due to low demand periods, which can shift the mean far to the right of the maximum.Therefore, the standard deviation σ from s was calculated using only observed separations (not the smoothed curve) less than or equal to s.This is shown as an arrow pointing to the left of the target separation in Fig. 2. Let r be the minimum required separation.The separation buffer b was then s-r.These separation behavior metrics were calculated the same way for VAC and IAC.The subscripts VAC and IAC differentiate metrics by weather condition in the results section. +Fig. 2 Example separation histogram.A final behavior metric, safety margin, was calculated as x=b/σ.It was designed as a simple proxy for separation conformance.Let n s be the number of separations less than or equal to s and let n r be the number of separations less than r.The empirical separation conformance is given by c e =1-(n r /n s ), that is the ratio of observed separations to the left of s that conform to the required separation minimum r.The analytical probability of separation conformance is given by the Gauss error function c a =erf(x/√2).Assuming that the curve to the left of s has a normal distribution, c a was expected to be similar to c e .With this simplification, if 95% conformance was required to be considered safe, then x would need to be at least 2. Values of x below 2 would quickly fall out of acceptable conformance.The above metrics require a significant number of high arrival rate operations to generate a meaningful characterization.The 29 runway subset for this analysis was chosen for their relatively high values of n s in IAC. Figure 3 shows the IAC n s at each of these runways for required separations of 3, 4, and 5 nmi.The numbers of aircraft pairs are shown on a logarithmic scale. +Fig. 3 Number of instrument arrival separations ≤ s (n s ).The most frequent required separation at any runway is 3 nmi, with much fewer instances as the separation requirement increases.This is due to a preponderance of aircraft of large weight class.The number of instances for required separations greater than 5 nmi were less than 10 at all but two runways at r=6 nmi and so are not shown in Fig. 3.Of the runways shown, 8 have fewer than 9 instances for r=5 nmi and the separation histograms were too sparse to represent a normal distribution.For the remaining 21 runways for r=5 nmi, and all 29 runways for r=3 or 4 nmi, the arrival separations less than or equal to s have a fairly normal distribution as demonstrated by how similar the analytical separation conformance is to the empirical separation conformance in Fig. 4. Figure 4 shows the differences between empirical c e and analytical c a separation conformance by required separation. +Fig. 4 Separation conformance comparisonOn average the empirical ratio of separations conforming to the required minimum is higher than the analytical probability by just 0.02 (on a 0 to 1 scale).Therefore, the distributions are fairly normal and the safety metric x is a good proxy for separation conformance.These separation behavior metrics were calculated the same way for VAC and IAC.The subscripts VAC and IAC are used to differentiate metrics by weather condition in the results section. +C. Runway Arrival CapacityRunway arrival capacity is the estimated number of arrival slots that can be scheduled to a runway within a given amount of time.Average slot size can vary depending on the typical landing speeds and mix of aircraft weight class at each runway, as well as the spacing buffer used.For each aircraft, the desired spacing with its leading aircraft is computed as the required separation plus a given buffer.The time at which the aircraft reaches this along path distance from the runway threshold is calculated by interpolating between track points on either side of this along-path distance.This desired spacing time is then subtracted from the runway threshold crossing time to get the scheduling slot size for the flight.The average slot size for each runway is then divided into 15 minutes to get a quarter hourly arrival capacity. +D. Fixed Routing and Path Stretch DelayPrecision scheduling concepts require fixed arrival routes such as RNAV and RNP but very few arrival routes have been defined all the way to the runway in today's system.Therefore, fixed arrival routes to each runway studied were designed for the purposes of this analysis.Most RNAV/RNP procedures designed in practice overlay existing routes to avoid lengthy environmental reviews [18].Fixed arrival routes for this study were designed to follow the nominal flow of arrival traffic.A graph-based trajectory bundling algorithm [19] was used to identify the most consistently used paths to each runway.Separate routes were generated for turboprops and jets where it was clear their nominal paths deviated significantly.In some cases the resulting routes curved unnecessarily in places due to consistently used path stretch maneuvers and these route segments were smoothed out manually.A flight was assigned on a fixed route to its arrival runway at the track point that first came within 3 nmi of the fixed route.From this point, the flight was directed to the next downstream waypoint along the route and then followed the route to the runway.The flight was expected to follow a deceleration profile similar to that of its original trajectory as it followed the fixed route.To model this assumption, first the original flight track was segmented using the closest track point to each fixed route waypoint.Original flight track segments were then mapped to fixed route segments using these closest points.Figure 5 illustrates the path stretch delay calculation for a sequence of mapped closest flight track and fixed route points labeled 1 through 5. Let p ij and q ij be mapped original flight track (gray) and fixed route segments (black) respectively.Let t(p ij ) and t(q ij ) be the respective times to travel each segment, and let l(p ij ) and l(q ij ) be the respective lengths of each segment.Then t(q ij )=t(p ij )l(q ij )/l(p ij ).The sum of the differences in flight time (t(p ij )-t(q ij )) across each mapped segment is the path stretch delay d.Note that path stretch delay may be negative if the original flight path is sufficiently shorter than the fixed route. +E. Fixed Route Scheduling and Time SavedTo determine the effect of fixed routing and scheduling buffers on terminal area delay, flights were rescheduled according to their fixed route Estimated Time of Arrival (ETA).The fixed route ETA was calculated as Observed Time of Arrival (OTA) minus the path stretch delay.The flights were then re-sequenced according to their fixed route ETAs, and delay was applied as Scheduled Time of Arrivals (STA) assigned to place flights within arrival slots appropriately sized to meet the required separation with respect to the leading aircraft plus a given buffer.The time saved or reduced path stretch delay by using the fixed path and given buffer is then OTA-STA.Note that time saved cannot exceed path stretch delay.The time saved is somewhat conservative because flights with path stretch delay may have been slowed as well and the path stretch delay calculation does not compensate for this, but assumes a similar speed profile to the original flight trajectory. +III. Results +A. Separation Behavior by Required SeparationQuartiles for IAC and VAC separation buffers, standard deviations, and safety margins by The mean IAC separation buffer is approximately 1 nmi with an increasing trend as required separation increases.This additional buffer compensates for the increasing IAC standard deviation.It even tends to overcompensate as can be seen by the increasing IAC safety margin.VAC standard deviation also increases with required separation but it tends to be higher than IAC standard deviation.This means that a single target separation is less often accurately achieved under VAC.Unlike the IAC, VAC buffers remain fairly stable as standard deviation increases so there is no compensation.Under IAC, buffers are never below 0.4 mi, whereas the VAC buffers can be negative.This means that the VAC target separation can be below the IAC required separation.Even though the VAC safety margin is quite low and decreases as required separation increases, this does not mean that these operations are unsafe.The VAC safety margin is calculated with respect to IAC requirements and so this metric merely shows how different the separation behavior is between IAC and VAC. +B. Separation Behavior by RunwayIndividual separation behavior by runway is compared for 3 nmi required separation as most aircraft pairs fall in this category.Figure 7 shows separation behavior metrics for each individual runway for 3 mi required separation.The dark and light gray columns represent IAC and VAC target separation (s) respectively, with single tail standard deviation (σ) whiskers.The 3 nmi required separation (r) is highlighted in black so that the part of the target separation columns protruding above can be visualized as the buffer (b).The IAC safety margins (x) are shown as white diamonds to visualize runways' need to reduce standard deviation or freedom to reduce buffers relative to each other.In general, runways within the same TRACON tend to behave similarly.DEN and SFO runways have the largest x IAC because they have the smallest IAC standard deviations but among the largest separation buffers.They may benefit from advanced scheduling that reduces the buffers.DFW also has fairly high spacing buffers but they are scaled appropriately to their standard deviations.DFW's x values between 1.7 and 2.0 equate to c a probabilities between 0.92 and 0.95, which is quite reasonable.More precision would enable DFW to further reduce its separation buffers.Airports in the remaining TRACONs, particularly LGA and IAD are not using separation buffers that sufficiently compensate for their standard deviations.The significant benefit of increased precision at these heavily strained airports may be to reduce controller workload and separation violations.By contrast, the two runways with the largest IAC buffers and standard deviations, TEB 06 and SAN 27, have the lowest arrival traffic volume among the runways analyzed within their TRACONs.They may be neglected due to higher priority airports nearby. +Fig. 7 Separation metrics by runway.In general, IAC target separations are larger than VAC.The difference between IAC and VAC is negligible for DEN, DFW, JFK, DCA, IAD, and LAX, suggesting that advanced scheduling would provide similar benefits under both IAC and VAC.In contrast, ATL has much lower target separations in VAC than IAC and so very little if any benefit would be gained during VAC using IAC separation requirements.All four ATL runways analyzed can use 2.5 nmi separation which was not considered, so their actual buffer may be 0.5 nmi larger.Only three runways analyzed (ORD 09R, SFO 19L and 28R) have VAC target separations that are larger than IAC.This may be due to procedural constraints.ORD 09R handles arrivals and departures most of the time.However, it may be used mostly for arrivals in IAC. Figure 8 shows smoothed VAC and IAC spacing histograms for ORD 09R.Assuming the shape of the The histograms for 19R and 19L show a similar but less pronounced effect.Given that the observed IAC buffers are between 0.4 and 2 nmi as seen in Fig. 7, it is not surprising that 0.3 nmi buffers are estimated to increase IAC capacity up to 30%.ATL VAC buffers are so low that using 0.3 nmi buffers would actually reduce capacity based on 3 nmi rather than 2.5 nmi required separation.Although VAC buffers and capacity estimates tend to be lower than IAC, many of the VAC buffers are above 0.3 nmi, resulting in estimated capacity increases around 10-20%.Many of these runways such as TEB may lack sufficient demand to achieve these theoretical capacities or they are affected by other constraints such as shared arrival/departure operations or crossing runway restrictions.Other airports such as EWR, LGA, and IAD show an increase over IAC capacity with little to no change in VAC capacity, which suggests that demand is not lacking and that with precision spacing, it may be possible to achieve VAC throughput under IAC. +D. Path Stretch Delay and Time Saved by RunwayEach flight was assigned to the first fixed route that came within 3 nmi of its original track, which was used to calculate the flight's path stretch delay and fixed route ETA.Flights were sequenced according to fixed route ETA and assigned scheduled arrival slots using spacing buffers ranging between 0 and 1.0 nmi in 0.1 nmi increments.Figure 12 shows the 5% trimmed mean path stretch delays and time saved for the range of scheduling buffers.The 5% trimmed mean is used rather than the pure mean to filter the effects of a few large outliers such as undetected go-arounds and other processing errors.The white rectangles mark the 5% trimmed mean path stretch delays for each runway.The 5% trimmed mean time saved results are represented by gray rectangles with width proportional to the buffer used.The 0.3 nmi buffer results for each runway are highlighted in black as this buffer size has been successfully tested in simulations of DAL and PHX arrival operations.As expected, time saved is inversely proportional to the buffer size.Some runways tend to be more sensitive to buffer size than others as can be seen in Fig. 12 by the rapidly decreasing time saved results as the buffer size increases from 0 to 1.0 nmi.Runways at ATL, ORD, EWR, andLGA are particularly sensitive to buffer size.These runways have tightly packed schedules with visual arrivals often spaced closer than the minimum required spacing for instrument arrivals.Therefore, any increase in spacing buffer quickly compounds delay as each delay is shared by each flight in the tightly packed stream.Runways TEB 06 and SAN 27 are least sensitive to buffer size because both their IAC and VAC observed buffers were larger than 1.0 nmi, the upper range of buffer size tested.Figure 13 shows smoothed histograms of the time saved results for two runways with different sensitivity to buffer size.LGA 22 is very sensitive to buffer size, which can be seen in Fig. 13a as the histograms shift to the left with increasing buffer size.In contrast, DEN 35R in As can be seen in Fig. 12 most runways have a positive path stretch delay.That is, in general, more flights are given a path stretch than a short cut relative to the fixed arrival routing identified.Figure 14 shows the flight tracks and fixed routing for the runways with lowest (MDW 04R) and highest (ORD 09L) path stretch delay.The dotted lines represent the lateral paths of historical tracks color-coded by path stretch delay relative to the fixed routing represented by the black nodes and links.The dark gray tracks are at least 4 minutes longer than the fixed route.Even though the fixed routing for MDW 04R is consistent with published RNAV instrument approach procedures for that runway, most flights from the East take shorter paths, perhaps due to visual approaches.Relatively few flights with more than 4 minutes of path stretch delay can be seen holding from the West or performing a base turn from the East.However, ORD 09L has a large amount of path stretch delay.Holding patterns and S-turns can be seen near all the major entry points.But the main contributor to path stretch delay is a large extended base turn to the West of the runway.Traffic from all directions is affected by this inefficiency, even traffic from the West, which doubles back on itself to merge with the extended base turn.In general, there is more scheduled delay in IAC than VAC (light gray bars tend to be longer than dark gray bars).Fast-time simulations of DEN [7] estimated the amount of delay that could be absorbed by speed control to be on the order of 2-3 minutes depending on the length of the route.Most runways in Fig. 15 have scheduled delay under 2 minutes.With the exception of ATL 10, and ORD 09R and 10, which all have delays over 3 minutes, relatively little delay should need to be absorbed in Center airspace in order to stay on fixed routes in the TRACON. +F. Runway Interdependency AnalysisThe above results are based on in-trail spacing restrictions alone.However, many runways may have other restrictions due to parallel arrival operations, shared arrival/departure operations, or crossing runways.In addition to arrival weather conditions, ASPM [18] denotes which runways were configured for arrival and departure operations per quarter hour.These data were analyzed for operations between 0600 and 2200 during the months of Feb -May 2010.As can be seen in Fig. 1, many of the runways analyzed cross other runways or are very close (2500 ft) to a parallel runway.Active crossing or parallel runways require coordinated operations and impose additional restrictions that affect spacing behavior.The stacked columns of Fig. 17 Only SFO and LAX runways configure both parallels for arrivals at the same time and LAX's configures both parallels for arrivals mostly in VAC.Even though SFO runways are configured for parallel arrivals most of the time in VAC and IAC, Fig. 9 shows that SFO 28L is rarely ever used for arrivals in IAC.This inconsistency illustrates that the above runway configuration analysis only identifies possible (not actual) sources of procedural constraints that could affect spacing behavior.This analysis also identifies runways that are likely NOT affected by these kinds of procedural constraints because they are configured for dedicated arrival operations without active crossing or closely spaced parallel runway most of the time.These include all the runways analyzed at ATL, DEN, DFW, EWR, JFK, and IAD in VAC and IAC, as well as LAX runways in IAC only. +IV. ConclusionsThis paper discussed an analysis of spacing behavior across 29 runways from 8 top TRACONs and estimated the benefits and impacts of precision scheduling and spacing along fixed arrival routes.Likely candidate airports and runways that would benefit from precision scheduling and spacing concepts were identified.The spacing behavior and estimated capacity increases by runway suggest that all runways analyzed would benefit from a concept that would safely allow buffers to be reduced to 0.3 nmi (at least in IAC).The analysis reinforced how different spacing behaviors can be between TRACONs, airports, and even individual runways, and that evaluations of concepts adapted to a specific site cannot be arbitrarily extended to another site.In general, increasing separation buffers for increased required separation tend to overcompensate for the increasing standard deviations.Precision scheduling and spacing concepts would remove this unnecessary extra spacing from aircraft already widely spaced due to wake hazard and make it easier for controllers to manage these less frequently occurring aircraft pairs.Possible procedural constraints affecting spacing due to shared arrival/departure operations, crossing runways, and closely spaced parallel approaches must be considered.Of the runways that do not appear to be affected by such procedures, reducing buffers to 0.3 nmi could increase capacity 10-20% at ATL, DEN, DFW, EWR, JFK, and LAX in IAC, and at DEN, DFW, and JFK in VAC.With the exception of LAX, all these airports have at least one runway with significant path stretch delay and could benefit from adhering to fixed arrival routing.For half of ATL and DEN runways and all DFW, and JFK runways studied, some of this delay can be reduced through precision scheduling as scheduling buffers are reduced and throughput increases.DFW and JFK runways have less than 2 minutes of scheduled delay, which may be absorbed in the TRACON with speed control.ATL 10 and DEN 35R have higher scheduled delays and so they would likely need to pass 1-2 minutes of this delay to the center.Having identified likely candidate airports and runways that would benefit from precision scheduling and spacing concepts, the next step should be to conduct more detailed analyses and simulations on these runways.A significant number of runways with possible procedural constraints such as shared arrival/departure operations and active crossing or closely spaced parallel runways showed potential benefits.This suggests future research is needed for integration and evaluation of precision arrival scheduling and spacing concepts with these procedures.Nomenclature b = spacing buffer (nmi) c a = analytical separation conformance (ratio of flights) c e = empirical separation conformance (ratio of flights) d = path stretch delay (sec) l = length of a track or route segment (nmi) n r = number of separations less than the required separation n s = number of separations less than or equal to the target separation p ij = flight track segment q ij = fixed route segment r = required separation (nmi) s = target separation (nmi) t = time to travel a track or route segment (sec) x = safety margin (nmi/nmi) σ = separation standard deviation (nmi) I. Introduction DVANCES in arrival scheduling and spacing have potential benefits of reducing aircraft delays and increasing airport arrival throughput.Schedulers use fixed arrival paths to estimate aircraft time-to-fly and assign them arrival slots.The size of an arrival slot is based on the required separation between leading and trailing aircraft plus a buffer used to mitigate arrival time uncertainty and reduce the probability of a separation violation.Thipphavong and +Fig. 11Fig. 1 Runway diagrams. +Fig. 55Fig. 5 Example path stretch delay calculation. +required separation are shown in Fig. 6 as box and whisker plots.From top to bottom the box and whisker divisions represent max, 75%, 50%, 25%, and min values for each set of data.The quartiles for the 3 and 4 nmi required separations represent all 29 runways.The quartiles for 5 nmi required separation represent only the 21 runways for which IAC n s was at least 9. +Fig. 66Fig. 6 Separation metrics by required separation. +VAC curve is due to normal shared arrival/departure operations, the IAC curve appears to be a convolution of dedicated arrival operations and shared arrival/departure operations.The dashed curves show a possible deconvolution of IAC where 0.35VAC (the VAC curve multiplied by 0.35) represents the shared arrival/departure behavior and the remainder (IAC-(0.35VAC))represents the dedicated arrival behavior. +Fig. 8 ORD8Fig. 8 ORD 09R spacing. +Fig. 9 SFO9Fig. 9 SFO 28L and 28R spacing. +Figure 10 Fig. 101010Figure 10 shows estimated quarter hour capacity for each runway using scheduling buffers of +Fig. 1111Fig. 11 Estimated percent change in capacity. +Fig. 1212Fig. 12 Path stretch delay and time saved by runway. +Fig. 13b isFig. 13b is less sensitive to buffer size with time saved results very similar to the path stretch +Fig. 1313Fig. 13 Time saved histograms a) LGA 22 b) DEN 35R. +Fig. 14 Fig. 151415Fig. 14 Sample flight tracks and fixed routing a) MDW 04R b) ORD 09L. +Figure 16 shows16Figure 16 shows stacked columns for IAC and VAC (right and left respectively) representing +Fig. 1616Fig. 16 Shared arrival/departure runway configurations. +show the relative percentage of time when the given runway is configured for arrivals and one or more of its crossing runways is configured for arrivals, departures, or no crossing runways are active.JFK 04L and DCA 01 are relatively free of crossing runway activity.The remaining runways shown have one or more crossing runways configured for departures most of the time.In general, crossing runway activity is more common in VAC. +Fig. 1717Fig. 17 Crossing runway configurations. +Fig. 1818Fig. 18 Parallel runway configurations. + + + + +Table 1 Minimum required separation (nmi) Trailer1Super Heavy B757 Large SmallSuper 668810Heavy 44556LeaderB757 4 Large 34 34 34 35 4Small 33333Aircraft pairs with the same separation requirement were further segregated between VACand IAC according to the airport's weather conditions during the quarter hour in which the trailerlanded. The FAA's Aviation System Performance Metrics (ASPM † ) data provide VAC and IACstate in quarter hour intervals. VAC and IAC separation histograms were generated with a 0.1nmi separation bin size. Figure 2 shows an example separation histogram for characterizingseparation behavior. + † Federal Aviation Administration, Aviation System Performance Metrics (ASPM) Airport Efficiency, http://aspm.faa.gov. + + + + +AcknowledgmentsThis work was funded by the Concept and Technology Development Project within NASA's Airspace Systems Program. + + + + + + + + + Design Considerations for a New Terminal Area Arrival Scheduler + + JaneThipphavong + + + DanielMulfinger + + + AlexanderSadovsky + + 10.2514/6.2010-9290 + + + 10th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Fort Worth, TX + + American Institute of Aeronautics and Astronautics + 2010 + + + Thipphavong, J. and Mulfinger, D., "Design Considerations for a New Terminal Area Arrival Scheduler," 10 th AIAA Aviation Technology, Integration and Operations Conference, Fort Worth, TX, 2010. + + + + + Simulation Results for Airborne Precision Spacing along Continuous Descent Arrivals + + BryanBarmore + + + WilliamCapron + + + TerenceAbbott + + + BrianBaxley + + 10.2514/6.2008-8931 + + + The 26th Congress of ICAS and 8th AIAA ATIO + Anchorage, AK + + American Institute of Aeronautics and Astronautics + 2008 + + + Barmore, B. E., Abbott, T. S., Capron, W. R., and Baxley, B. T., "Simulation Results for Airborne Precision Spacing along Continuous Descent Arrivals," 8 th AIAA Aviation Technology, Integration and Operations Conference, Anchorage, AK, 2008. + + + + + Performance of Airborne Precision Spacing under realistic weather conditions + + FrederickWieland + + + MichelSantos + + + WilliamKrueger + + + VincentEHouston + + 10.1109/dasc.2011.6095981 + + + 2011 IEEE/AIAA 30th Digital Avionics Systems Conference + Seattle, WA + + IEEE + 2011 + + + Wieland, F., Santos, M., Krueger, W., and Houston, V. E., "Performance of Airborne Precision Spacing Under Realistic Weather Conditions," 30th Digital Avionics System Conference, Seattle, WA, 2011. + + + + + Arrival Management with Required Navigation Performance and 3D Paths + + AHaraldsdottir + + + JScharl + + + MBerge + + + ESchoemig + + + MCoats + + + + Europe Air Traffic Management R&D Seminar + + 2007 + Barcelona, Spain + + + Haraldsdottir, A., Scharl, J., Berge, M., Schoemig, E., and Coats, M., "Arrival Management with Required Navigation Performance and 3D Paths," 7th USA/Europe Air Traffic Management R&D Seminar, Barcelona, Spain, 2007. + + + + + Analysis of Arrival Management Performance with Aircraft Required Time of Arrival Capabilities + + AslaugHaraldsdottir + + + JulienScharl + + + JanetKing + + + EwaldSchoemig + + + MatthewBerge + + 10.2514/6.2008-8932 + + + The 26th Congress of ICAS and 8th AIAA ATIO + Anchorage, AK + + American Institute of Aeronautics and Astronautics + 2008 + + + Haraldsdottir, A., Scharl, J., King, J., Schoemig, E. G., and Berge, M., "Analysis of Arrival Management Performance with Aircraft Required Time of Arrival Capabilities," 26th International Congress of the Aeronautical Sciences, Anchorage, AK, 2008. + + + + + Arrival Management Architecture and Performance Analysis with Advanced Automation and Avionics Capabilities + + AslaugHaraldsdottir + + + JulienScharl + + + JanetKing + + + MatthewBerge + + 10.2514/6.2009-7006 + + + 9th AIAA Aviation Technology, Integration, and Operations Conference (ATIO) + Hilton Head, SC + + American Institute of Aeronautics and Astronautics + 2009 + + + Haraldsdottir, A., Scharl, J., King, J., and Berge, M., "Arrival Management Architecture and Performance Analysis with Advanced Automation and Avionics Capabilities," 9th AIAA Aviation Technology, Integration and Operations Conference, Hilton Head, SC, 2009. + + + + + Terminal area arrival management concepts using tactical merge management techniques + + AslaugHaraldsdottir + + + JanetKing + + + JulienScharl + + 10.1109/dasc.2010.5655508 + + + 29th Digital Avionics Systems Conference + Salt Lake City, UT + + IEEE + 2010 + + + Haraldsdottir, A., King, J., and Scharl, J., "Terminal Area Arrival Management Concepts Using Tactical Merge Management Techniques," 29th Digital Avionics System Conference, Salt Lake City, UT, 2010. + + + + + Design and Evaluation of the Terminal Area Precision Scheduling and Spacing System + + HNSwenson + + + JThipphavong + + + ASadovsky + + + LChen + + + CSullivan + + + LMartin + + + 2011 + Berlin, Germany + + + USA/Europe Air Traffic Management R&D Seminar + + + Swenson, H. N., Thipphavong, J., Sadovsky, A., Chen, L., Sullivan, C., and Martin, L., "Design and Evaluation of the Terminal Area Precision Scheduling and Spacing System," 9th USA/Europe Air Traffic Management R&D Seminar, Berlin, Germany, 2011. + + + + + Evaluation of the terminal area precision scheduling and spacing system for near-term NAS application + + JThipphavong + + + HSwenson + + + LMartin + + + PLin + + + JNguyen + + 10.1201/b12321-12 + + + Advances in Human Aspects of Aviation + San Francisco, CA + + CRC Press + 2012 + + + + Thipphavong, J., Swenson, H., Martin, L., Lin, P., and Nguyen, J., "Evaluation of the Terminal Precision Scheduling and Spacing System for Near-Term NAS Application," 1st International Conference on Human Factors in Transportation, San Francisco, CA, 2012. + + + + + Potential Benefits of Terminal Airspace Traffic Automation for Arrivals + + MarkGBallin + + + HeinzErzberger + + 10.2514/2.4324 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 21 + 6 + + 1998 + American Institute of Aeronautics and Astronautics (AIAA) + + + Ballin, M. and Erzberger, H., "Potential Benefits of Terminal Airspace Traffic Automation for Arrivals," Journal of Guidance, Control, and Dynamics, Vol. 21, No. 6, 1998. + + + + + The impact of advanced technology on Nex Generation Air Transportation: A case study of Required Navigation Performance (RNP) + + YoonjinYoon + + + MarkHansen + + 10.2514/6.2007-6697 + + + AIAA Modeling and Simulation Technologies Conference and Exhibit + Hilton Head, SC + + American Institute of Aeronautics and Astronautics + 2007 + + + Yoon, Y. and Hansen, M., "The Impact of Advanced Technology on Next Generation Air Transportation: A Case Study of Required Navigation Performance (RNP)," AIAA Modeling and Simulation Technologies Conference and Exhibit, Hilton Head, SC, 2007. + + + + + Evaluation of the Potential Environmental Benefits of RNAV/RNP Arrival Procedures + + DelriMuller + + + PayunaUday + + + KarenMarais + + 10.2514/6.2011-6932 + + + 11th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Virginia Beach, VA + + American Institute of Aeronautics and Astronautics + 2011 + + + Muller, D., Uday, P., and Marais, K. B., "Evaluation of the Potential Environmental Benefits of RNAV/RNP Arrival Procedures," 11th AIAA Aviation Technology, Integration and Operations Conference, Virginia Beach, VA, 2011. + + + + + Benefits of Continuous Descent Operations in High-Density Terminal Airspace Considering Scheduling Constraints + + JohnRobinson Iii + + + MaryamKamgarpour + + 10.2514/6.2010-9115 + + + 10th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Fort Worth, TX + + American Institute of Aeronautics and Astronautics + 2010 + + + Robinson III, J.E. and Kamgarpour, M., "Benefits of Continuous Descent Operations in High-Density Terminal Airspace Under Scheduling Constraints," 10th AIAA Aviation Technology, Integration and Operations Conference, Fort Worth, TX, 2010. + + + + + Comparison of Trajectory Synthesis Algorithms for Monitoring Final Approach Compression + + JohnRobinson Iii + + + OusmaneDiallo + + + RonaldReisman + + 10.2514/6.2011-6900 + + + 11th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Virginia Beach, VA + + American Institute of Aeronautics and Astronautics + 2011 + + + Robinson III, J.E., Diallo, O., and Reisman, R.J., "Comparison of Trajectory Synthesis Algorithms for Monitoring Final Approach Compression," 11th AIAA Aviation Technology, Integration and Operations Conference, Virginia Beach, VA, 2011. + + + + + Estimating Airport Operations at General Aviation Airports Using the FAA NPIAS Airport Categories + + MaryJohnson + + + YueGu + + 10.15394/ijaaa.2017.1151 + + + + International Journal of Aviation, Aeronautics, and Aerospace + IJAAA + 2374-6793 + + 27 September 2010 + ERAU Hunt Library - DIGITAL COMMONS JOURNALS + + + 2011-2015 National Plan of Integrated Airport Systems Report + Federal Aviation Administration, "2011-2015 National Plan of Integrated Airport Systems Report," National Plan of Integrated Airport Systems Reports [online database], URL: http://www.faa.gov/airports/planning_capacity/npias/reports [cited 27 September 2010]. + + + + + Practice for Application of Federal Aviation Administration (FAA) Federal Aviation Regulations Part 21 Requirements to Unmanned Aircraft Systems (UAS) + 10.1520/f2505 + + + + Chapter 5. Section 5. Radar Separation, FAA Orders & Notices + + ASTM International + null + + + Federal Aviation Administration. online database. cited 9 February 2012 + Federal Aviation Administration, "Air Traffic Control FAA Order JO 7110.65U," Chapter 5. Section 5. Radar Separation, FAA Orders & Notices, [online database], URL: http://www.faa.gov/regulations_policies/orders_notices/ [cited 9 February 2012]. + + + + + Practice for Application of Federal Aviation Administration (FAA) Federal Aviation Regulations Part 21 Requirements to Unmanned Aircraft Systems (UAS) + 10.1520/f2505 + + + + Federal Aviation Administration + + ASTM International + 18 June 2012 + + + FAA Orders & Notices. online database + Federal Aviation Administration, "Air Traffic Control FAA Order JO 7110.582 Procedures for Airbus A380-800 (A388) Flights," FAA Orders & Notices, [online database], URL: http://www.faa.gov/regulations_policies/orders_notices/ [cited 18 June 2012]. + + + + + (2021) Volume 2, Issue 4 Cultural Implications of China Pakistan Economic Corridor (CPEC Authors: Dr. Unsa Jamshed Amar Jahangir Anbrin Khawaja Abstract: This study is an attempt to highlight the cultural implication of CPEC on Pak-China relations, how it will align two nations culturally, and what steps were taken by the governments of two states to bring the people closer. After the establishment of diplomatic relations between Pakistan and China, the cultural aspect of relations between the two states also moved forward. The flow of cultural delegations intensified after the 2010, because this year was celebrated as the ‘Pak-China Friendship Year’. This dimension of relations further cemented between the two states with the signing of CPEC in April 2015. CPEC will not only bring economic prosperity in Pakistan but it will also bring two states culturally closer. The roads and other communication link under this project will become source of cultural flow between the two states. Keyswords: China, CPEC, Culture, Exhibitions Pages: 01-11 Article: 1 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)01 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)01 Download Pdf: download pdf view article Creative Commons License Political Persona on Twittersphere: Comparing the Stardom of Prime Minister(s) of Pakistan, UK and India Authors: Maryam Waqas Mudassar Hussain Shah Saima Kausar Abstract: Political setup demands to use Twittersphere for preserving its reputation because of significant twitter audience, which follows celebrities and political figures. In this perspective, political figures frequently use twitter to highlight their political as well as personal lives worldwide. However, political figures take the stardom status among the twitter audience that follow, retweet and comment by their fans. The purpose of this study is, to analyze what kind of language, level of interest is made by political figures while communicating via twitter, text, phrases and languages used by political figures, and do their tweets contribute in their reputation. The qualitative content analysis is used for evaluation of the interests shared by PM Imran Khan, PM Boris John Son and PM Narendra Modi with the key words of tweets. A well-established coding sheet is developed for the analysis of text, phrases and words in the frames of negative, positive and neutral from March 2020 to May 2020. The results are demonstrating on the basis of content shared by Prime Ministers of three countries i.e., From Pakistan, Imran Khan, United Kingdom, Johnson Boris and India, Narendra Modi on twitter. The findings also reveal that varied issues discussed in tweets, significantly positive and neutral words are selected by these political figures. PM Imran tweeted more negative tweets than PM Boris Johnson and PM Narendra Modi. However, PM Boris Johnson and PM Narendra Modi make significant positive and neutral tweets. It is observed that political figures are conscious about their personal reputation while tweeting. It also revealed that the issues and tweets shared by these leaders contribute to their personal reputation. Keyswords: Imran Khan, Johnson Boris, Narendra Modi, Political Persona, Stardom, Twittersphere Pages: 12-23 Article: 2 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)02 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)02 Download Pdf: download pdf view article Creative Commons License An Empirical Relationship between Government Size and Economic Growth of Pakistan in the Presence of Different Budget Uncertainty Measures Authors: Sunila Jabeen Dr. Wasim Shahid Malik Abstract: Relationship between government size and economic growth has always been a debated issue all over the world since the formative work of Barro (1990). However, this relationship becomes more questionable when policy uncertainty is added in it. Hence, this paper presents evidence on the effect of government size on economic growth in the presence of budget uncertainty measured through three different approaches. Rather than relying on the traditional and complicated measures of uncertainty, a new method of measuring uncertainty based on government budget revisions of total spending is introduced and compared with the other competing approaches. Using time series annual data from 1973-2018, the short run and long run coefficients from Autoregressive Distributed Lag (ARDL) framework validate the negative effect of budget uncertainty and government size on economic growth of Pakistan regardless of the uncertainty measure used. Therefore, to attain the long run economic growth, along with the control on the share of government spending in total GDP, government should keep the revisions in the budget as close to the initial announcements as it can so that uncertainty can be reduced. Further, the uncertainty in fiscal spending calculated through the deviation method raises a big question on the credibility of fiscal policy in Pakistan. Higher will be the deviation higher will be the uncertainty and lower the fiscal policy credibility hence making fiscal policy less effective in the long run. Keyswords: Budget Uncertainty, Economic Growth, Government Size, Policy Credibility Pages: 24-38 Article: 3 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)03 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)03 Download Pdf: download pdf view article Creative Commons License Despair in The Alchemist by Ben Jonson Authors: Dr. Fatima Syeda Dr. Faiza Zaheer Numrah Mehmood Abstract: This research aims to challenge the assumption that The Alchemist by Ben Jonson is one of the greatest examples of the “explicit mirth and laughter” (Veneables 86). The paper argues that The Alchemist is a cynical and despairing play created in an atmosphere not suitable for a comedy. This is a qualitative study of the text and aims at an analysis of the theme, situations, characters, language, and the mood of the play to determine that Jonson is unable to retain the comic spirit in The Alchemist and in an attempt to “better men” (Prologue. 12) he becomes more satirical and less humorous or comic. This research is important for it contends that the play, termed as a comedy, may be read as a bitter satire on the cynical, stinky, and despairing world of the Elizabethan times. Keyswords: Comedy, Despair, Reformation Pages: 39-47 Article: 4 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)04 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)04 Download Pdf: download pdf view article Creative Commons License Analysis of Principles of Coordinated Border Management (CBM) in articulation of War-Control Strategies: An Account of Implementation Range on Pakistan and Afghanistan Authors: Dr. Sehrish Qayyum Dr. Umbreen Javaid Abstract: Currently, Border Management is crucial issue not only for Pakistan but for the entire world due to increased technological developments and security circumstances. Pakistan and Afghanistan being immediate states have inter-connected future with socio-economic and security prospects. Principles of Coordinated Border Management (CBM) approach have been extracted on the basis of in-depth interviews with security agencies and policymakers to understand the real time needs. The current research employs mixed method approach. Process Tracing is employed in this research to comprehend the causal mechanism behind the contemporary issue of border management system. A detailed statistical analysis of prospect outcomes has been given to validate the implication of CBM. Implication range of CBM has been discussed with positive and probably negative impacts due to its wide range of significance. This research gives an analysis of feasibility support to exercise CBM in best interest of the state and secure future of the region. Keyswords: Afghanistan, Coordinated Border Management, Fencing, Pakistan, Security Pages: 48-62 Article: 5 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)05 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)05 Download Pdf: download pdf view article Creative Commons License The Belt and Road Initiative (BRI) vs. Quadrilateral Security Dialogue (the Quad): A Perspective of a Game Theory Authors: Muhammad Atif Prof. Dr. Muqarrab Akbar Abstract: Containment is the central part of the U.S.'s foreign policy during the cold war. With the application of containment Policy, the U.S. achieved much success in international politics. Over time China has become more powerful and sees great power in international politics. China wants to expand and launched the Belt and Road Initiative (BRI). The primary purpose of The Belt and Road Initiative (BRI) is to achieve support from regional countries and save their interests from the U.S. In 2017, the American administration launched its Containment policy through Quadrilateral Security Dialogue (the Quad) to keep their interest from China. The Quadrilateral Security Dialogue (Quad) is comprising of Australia, the United States, Japan, and India. This Study is based on Qualitative research with theoretical application of Game theory. This research investigates both plans of China (BRI) and the U.S. (the Quad) through a Game Theory. In this study, China and the U.S. both like to act as gamers in international politics. This study recommends that Game theory can predict all developments in the long term. Keyswords: Containment, Expansionism, Quadrilateral Security Dialogue, The Belt and Road Initiative (BRI) Pages: 63-75 Article: 6 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)06 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)06 Download Pdf: download pdf view article Creative Commons License Narendra Modi a Machiavellian Prince: An Appraisal Authors: Dr. Imran Khan Dr. Karim Haider Syed Muhammad Yousaf Abstract: The comparison of Narendra Modi and Machiavellian Prince is very important as policies of Modi are creating problems within India and beyond the borders. The Prince is the book of Niccolo Machiavelli a great philosopher of his time. If Indian Prime Minister Narendra Modi qualifies as a Prince of Machiavelli is a very important question. This is answered in the light of his policies and strategies to become the undisputed political leader of India. Much of the Machiavellian Prince deals with the problem of how a layman can raise himself from abject and obscure origins to such a position that Narendra Modi has been holding in India since 2014. The basic theme of this article is revolving around the question that is following: Can Modi’s success be attributed to techniques of The Prince in important respects? This article analyzed Narendra Modi's policies and strategies to develop an analogy between Machiavellian Prince and Modi in terms of characteristics and political strategies. This research work examines, how Narendra Modi became the strongest person in India. Keyswords: Comparison, India, Machiavelli, Modus Operandi, Narendra Modi Pages: 76-84 Article: 7 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)07 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)07 Download Pdf: download pdf view article Creative Commons License Analyzing Beckett's Waiting for Godot as a Political Comedy Authors: Muhammad Umer Azim Dr. Muhammad Saleem Nargis Saleem Abstract: This study was devised to analyze Samuel Beckett’s play Waiting for Godot in the light of Jean-Francois Lyotard’s theory of postmodernism given in his book The Postmodern Condition (1984). This Lyotardian paradigm extends a subversive challenge to all the grand narratives that have been enjoying the status of an enviable complete code of life in the world for a long time. Even a cursory scan over the play under analysis creates a strong feel that Beckett very smartly, comprehensively and successfully questioned the relevance of the totalizing metanarratives to the present times. Being an imaginative writer, he was well aware of the fact that ridicule is a much more useful weapon than satire in the context of political literature. There are so many foundationalist ideologies that he ridicules in his dramatic writing. Christianity as a religion is well exposed; the gravity of philosophy is devalued; the traditional luxury that the humans get from the art of poetry is ruptured and the great ideals of struggle are punctured. He achieves his artistic and ideologically evolved authorial intentions with a ringing success. It is interesting to note that he maintains a healthy balance between art and message. Keyswords: Beckett, Lyotard, The Postmodern Condition, Waiting for Godot Pages: 85-94 Article: 8 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)08 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)08 Download Pdf: download pdf view article Creative Commons License Effect of Parenting Styles on Students’ Academic Achievement at Elementary Level Authors: Hafsa Noreen Mushtaq Ahmad Uzma Shahzadi Abstract: The study intended to find out the effect of parenting styles on students’ academic achievement. Current study was quantitative in nature. All elementary level enrolled students at government schools in the province of the Punjab made the population of the study. Multistage sampling was used to select the sample from four districts of one division (Sargodha) of the Punjab province i.e., Sargodha. A sample size i.e., n=960; students and their parents were participated in this study. Research scales i.e. Parenting Styles Dimension Questionnaire (PSDQ) was adapted to analyze and measure parents’ parenting styles and an achievement test was developed to measure the academic achievement of the elementary students. After pilot testing, reliability coefficient Cronbach Alpha values for PSDQ and achievement test were 0.67 and 0.71 Data was collected and analyzed using frequencies count, percentages, mean scores and one way ANOVA. Major findings of the study were; Majority of the parents had authoritative parental style, a handsome number of parents keep connection of warmth and support with their children, show intimacy, focus on discipline, do not grant autonomy to their children, do not indulge with their children and as well as a handsome number of students were confident during their studies and study, further, found that parental style had positive relationship with academic achievement. Recommendations were made on the basis of findings and conclusion such as arrangement of Parents Teachers Meetings (PTM‘s), parents’ training, provision of incentives and facilities to motivate families might be an inclusive component of elementary education program. Keyswords: Academic Achievement, Elementary Education, Parenting Styles Pages: 95-110 Article: 9 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)09 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)09 Download Pdf: download pdf view article Creative Commons License Kashmir Conflict and the Question of Self-Determination Authors: Izzat Raazia Saqib Ur Rehman Abstract: The objective of this paper is to explore relations between Pakistan and India since their inception in the perspective of Kashmir conundrum and its impact on the regional security. Kashmir is the unfinished agenda of partition and a stumbling block in the bilateral relations between Pakistan and India. After the partition of sub-continent in 1947, Pakistan and India got their sovereign status. Kashmir conflict, a disputed status state, is the byproduct of partition. Pakistan and India are traditional arch-foes. Any clash between Pakistan and India can bring the two nuclear states toe-to-toe and accelerate into nuclear warfare. Due to the revulsion, hostility and lack of trust between the two, the peaceful resolution of the Kashmir issue has been long overdue. Ever-increasing border spats, arms race and threat of terrorism between the two have augmented anxiety in the subcontinent along with the halt of talks between India and Pakistan at several times. Additionally, it hampers the economic and trade ties between the two. India, time and again, backtracked on Kashmir issue despite UN efforts to resolve the issue. Recently, Indian government has responded heavy-handedly to the Kashmiri agitators’ demand for sovereignty and revocation of ‘Special Status’ of Kashmir impacting the stability of the region in future. Keyswords: India, Kashmir Conundrum, Pakistan, Regional Security, Sovereignty Pages: 111-119 Article: 10 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)10 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)10 Download Pdf: download pdf view article Creative Commons License Exploring Image of China in the Diplomatic Discourse: A Critical Discourse Analysis Authors: Muhammad Afzaal Muhammad Ilyas Chishti Abstract: The present study hinges on the major objective of analyzing Pakistani and Indian diplomatic discourses employed in portrayal of image of China. Data comprises the official discourse which is used in diplomatic affairs of both the states. The extensive investigation seeks insights from the fundamentals of Critical Discourse Analysis propounded by van Dijk, Fairclough and Wodak with a special focus on Bhatia’s (2006) work. The study reveals that the image of China has always been accorded priority within Indian and Pakistani diplomatic discourse even though nature of bilateral relations among China, India and Pakistan is based on entirely different dynamics; Indian and Pakistani diplomatic discourses are reflective of sensitivities involved within the bilateral relations. Through employment of linguistic techniques of ‘positivity’, ‘evasion’ and ‘influence and power’, Indian diplomats have managed not to compromise over the fundamentals in bilateral relations with China despite Pakistan’s already strengthened and deep-rooted relations with China. While Pakistani diplomatic fronts have been equally successful in further deepening their already strengthened relations in the midst of surging controversies on CPEC, BRI and OBOR. Hence, diplomatic fronts of both the counties, through employment of ideologically loaded linguistic choices, leave no stone unturned in consolidation of the diplomatic relations with China. Keyswords: CDA, China Image, Corpus, Language of Diplomacy, Political Discourse Analysis Pages: 120-133 Article: 11 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)11 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)11 Download Pdf: download pdf view article Creative Commons License Students’ Perception about Academic Advising Satisfaction at Higher Education Level Authors: Rukhsana Sardar Zarina Akhtar Shamsa Aziz Abstract: The purpose of the study was to examine the students’ perception about academic advising satisfaction at higher education level. All the students from two years master (M.A) degree programme and four years (BS) degree programme of eight departments from International Islamic University Islamabad (IIUI), Faculty of Social Sciences were taken as a population of the study. 475 students were randomly selected as a sample of the study. The Academic Advising Inventory (AAI) was used to assess Academic Advising Style. For measuring level of the satisfaction, descriptive statistics was used. To compare the mean difference department-wise and gender-wise about academic advising satisfaction t.test was applied. It was concluded that from the major findings of the study those students who received departmental academic advising style are more satisfied as compared to those students who provided prescriptive academic advising style. Female students seemed more satisfied as compared to male students regarding the academic advising style provided to them. Students who satisfied from developmental academic advising style and they were also highly satisfied from the advising provided to them at Personalizing Education (PE) and this is the subscale of developmental academic advising whereas students who received prescriptive academic advising they were also satisfied from the advising provided to them regarding personalizing education and academic decision making but their percentage is less. It is recommended to Universities Administration to focus on Developmental Academic Advising Style and establish centers at universities/department level and nominate staff who may be responsible to provide developmental academic advising. Keyswords: Academic Advising, Higher Level, Students’ Perception Pages: 134-144 Article: 12 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)12 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)12 Download Pdf: download pdf view article Creative Commons License Perceptions of Sexual Harassment in Higher Education Institutions: A Gender Analysis Authors: Ruhina Ghassan Dr. Subha Malik Nayab Javed Abstract: Sexual harassment is a social issue which is present in every society, globally, which interferes in an individual’s social and professional life. It happens almost everywhere i.e. at workplaces, public places or institutes as well. The focus of the present study was to explore the differences of male and female students’ perception of sexual harassment. This study was a quantitative research. Sample of the study included of 400 students (200 males and 200 females) from two government and two private universities. In the present study, Sexual Harassment Perception Questionnaire (SHPQ) was used to find out these differences in perceptions as every person has his own view for different situations. The study revealed the significant differences in perception of students. Study showed that both genders perceived that female students get more harassed than male students. The factors that affect the perception frequently were gender and age. The findings recommended that regulations for sexual harassment should be implemented in universities; laws should be made for sexual harassment in higher education institutes. Students should be aware of sexual harassment through seminars, self-defense classes and awareness campaigns. And every institute should have a counseling center for the better mental health of students. Keyswords: Gender Differences, Higher Educational Institutions, Sexual Harassment Pages: 145-158 Article: 13 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)13 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)13 Download Pdf: download pdf view article Creative Commons License Role of IMF Over the Governance Structure and Economic Development of Pakistan Authors: Ali Qamar Sheikh Dr. Muhammad Imran Pasha Muhammad Shakeel Ahmad Siddiqui Abstract: Developing countries like Pakistan seeks for financial assistance in order to fulfil their deficits. IMF is one of the largest financial institution who give loans to countries who need it. This research has studied the IMF role and the effects of IMF conditions on the economy of Pakistan. To carry out this research, both quantitative data from primary sources has been gathered and qualitative analysis has been made to signify whither this borrowing creating and maintaining dependency of Pakistan on West and financial and governance structure constructed to curtail Countries like Pakistan. The results concluded that there is negative and insignificant relationship between GDP and IMF loans in the long run. The short-term dynamic shows that weak economic and Political Institutions in Pakistan. The Development dilemma constitutes dependency even today. The Current Budget Deficit Pakistan's fiscal deficit climbs to Rs 3.403 trillion in 2020-21 needs to be readdressed in such a manner that Pakistan can counter Balance of Payments and import/export imbalance. Keyswords: Dependency, Development, IMF, Loans, Debt, Pakistan, Governance structure Pages: 159-172 Article: 14 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)14 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)14 Download Pdf: download pdf view article Creative Commons License Climate Change and the Indus Basin: Prospects of Cooperation between India and Pakistan Authors: Sarah Saeed Prof. Dr. Rana Eijaz Ahmad Abstract: Climate change is transforming the global societies. The shift in average temperature is putting negative impacts on human health, food production and the natural resources. In the wake of the altered climate, water flow in the river systems is experiencing variability and uncertainty. This paper aims at studying the negative impacts of climate change on the water resources of the Indus Basin and investigate the prospects of cooperation between India and Pakistan; two major riparian nations sharing the basin. Adopting the case study approach, a theoretical framework has been built on the ‘Theory of the International Regimes’. It has been argued that institutional capacity and the dispute resolution mechanism provided in any water sharing agreement determine the extent of cooperation among the member states. Since India and Pakistan are bound by the provisions of the Indus Waters Treaty, this study tries to assess the effectiveness of this agreement in managing the negative consequences of the climate change. Keyswords: Climate Change, Cooperation, Dispute Resolution Mechanism, Institutional Capacity Pages: 173-185 Article: 15 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)15 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)15 Download Pdf: download pdf view article Creative Commons License Translation, Cultural Adaptation and Validation of Behavioral-Emotional Reactivity Index for Adolescents Authors: Saima Saeed Farah Malik Suzanne Bartle Haring Abstract: Measuring differentiation of self in terms of behavioral/emotional reactivity towards parents is important because of the complex parent-child connection. This needs a valid and reliable measure to assess the differentiation of self particularly in a relationship with parents. Behavior\Emotional Reactivity Index is such a tool that fulfills this purpose. The present study was carried out to culturaly adapt and translate BERI into the Urdu language and establish the psychometric properties of Urdu version. A sample of 303 adolescents of age (M = 16.07, SD = 1.77) was taken from different schools and colleges. Scale was split into Mother and father forms for the convenience of respondents. Findings supported the original factor structure of the BERI-original version. Higher-order factor analysis showed good fit indices with excellent alpha ranges (α= .91 to α=.80). BERI scores were compared for the adolescents who were securely attached with parents and insecurely attached with parents which showed a significant difference between the groups. BERI-Urdu version was found to be a valid and reliable measure in the Pakistani cultural context which gives researchers new directions to work with adolescents. Keyswords: Adolescence, Differentiation of Self, Behavioral, Emotional Reactivit, Index, Parental Attachment Pages: 186-200 Article: 16 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)16 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)16 Download Pdf: download pdf view article Creative Commons License Notion of Repression in Modern Society: A Comparative Analysis of Sigmund Freud and Herbert Marcuse Authors: Khadija Naz Abstract: One of the fundamental issues for modern civilized man is how to adapt a modern society without losing his individual status. Is it possible for an individual to adjust in a society where he/she loses his/her individuality and becomes part of collectivity? One point of view is that for society to flourish, man needs to be repressed. But to what extent is repression necessary for societies to rise and survive? This paper shall examine the above given questions from the standpoint of two thinkers who greatly influenced twentieth-century thought: Sigmund Freud and Herbert Marcuse. To undertake this task, first the term Repression shall be examined and then the notions of Freud and Marcuse will be discussed to determine the degree of repression required for the development of modern society. Keyswords: Modern Society, Performance Principle, Repression, Surplus-Repression, The Pleasure Principle, The Reality Principle Pages: 201-214 Article: 17 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)17 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)17 Download Pdf: download pdf view article Creative Commons License Perceptions of Teacher Educators about Integration of (ESD) in Elementary Teachers Education Program Authors: Dr. Rukhsana Durrani Dr. Fazal ur Rahman Dr. Shaista Anjum Abstract: Education and sustainable development have a close relationship as education provides sustainability to society. This study explored the perceptions of teacher educators for integration of Education for Sustainable Development (ESD) in B.Ed. 4 years’ elementary program. Four major components of ESD i.e., Education, Social & Culture, Economic and Environment were included in study. 127 teacher educators from departments of education were randomly selected from public universities of Pakistan who were offering B.Ed. 4 years’ elementary program. Data was collected through questionnaires from teacher educators. The findings recommended the inclusion of the components of Education for Sustainable Development (ESD) in curriculum of B.Ed. 4 years’ elementary program. Keyswords: B.Ed. 4 Years Elementary Curriculum, Sustainable Development, Integration, Teacher Education Pages: 215-225 Article: 18 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)18 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)18 Download Pdf: download pdf view article Creative Commons License Exploring TPACK skills of prospective teachers and challenges faced in digital technology integration in Pakistan Authors: Tariq Saleem Ghayyur Dr. Nargis Abbas Mirza Abstract: The current study was aimed to explore TPACK skills of prospective teachers and challenges faced in digital technology integration in Pakistan. The study was qualitative in nature and semi structured interview schedule was developed to collect data from prospective teachers. Purposive sampling technique was employed to collect data from 20 prospective teachers of 7 public sector universities. It was concluded that majority of the prospective teachers used general technological and pedagogical practices (GTPP), technological knowledge practices (TKP), Technological Pedagogical Knowledge practices (TPKP), Technological Content Knowledge practices (TCKP). Majority of prospective teachers reported multiple challenges in integration of digital technology in teacher education programs including lack of teacher training as one of the largest hurdle in digital technology integration, lack of digital technology resources or outdated digital technology resources, inadequate computer lab, lack of learning apps (courseware), financial constraints, lack of teachers’ motivation to use digital technology, slow computers available at computer labs, and unavailability of technical support. It was recommended that digital technology infrastructure should be improved across all teacher education institution and it was further recommended that TPACK model of digital technology integration should serve digital technology integration in teacher education programs in Pakistan. Keyswords: Challenges, Digital Technology Integration, Digital Technology Resources, Digital Technology, TPACK Pages: 226-241 Article: 19 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)19 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)19 Download Pdf: download pdf view article Creative Commons License Revisiting the Linkage between Money Supply and Income: A Simultaneous Equation Model for Pakistan Authors: Zenab Faizullah Dr. Shahid Ali Muhammad Imad Khan Abstract: A reliable estimate of the money supply is an important sign of the Gross Domestic Product (GDP) and many other macroeconomic indicators. It is widely discussed that over a long period of time, there is a strong link between GDP and money supply. This link is significantly important for formation of monetary policy. The main aim of this study is to estimate the income-money supply model for Pakistan. This study estimates the income-money supply model for Pakistan over the period of 2009 to 2019. The study uses Two Stage Least Square (2SLS) econometric technique due to the presence of endogeneity problem in the model under consideration. The existence of simultaneity between money supply (M2) and income (GDP) is also clear from the results of Hausman Specification test for simultaneity between M2 and GDP. The results further show that there exists a strong money-income relationship in case of Pakistan. Keyswords: Money Supply, Income, Simultaneous Equations Pages: 242-247 Article: 20 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)20 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)20 Download Pdf: download pdf view article Creative Commons License Analyzing the Mechanism of Language Learning Process by the Use of Language Learning Strategies Authors: Shafiq Ahmad Farooqi Dr. Muhammad Shakir Sher Muhammad Awan Abstract: This analytical research study involves the use of learning strategies to know the mechanism of learning a second language. People acquire their native language (L1) without any conscious effort and they have a complete knowledge of L1 and are competent in their native language even without going to school. It is believed that language learning is a process as well as an outcome and the focus of current study is to understand the process of learning a second language. The population in this study comprised of 182 boys and Girls Govt. Higher Secondary Schools studying at intermediate level in the 11 Districts of the Southern Punjab. The sample was selected through random probability sampling and consisted of 40 subject specialists teaching the subject of English in Govt. higher secondary schools with 400 students studying English at Intermediate level. A questionnaire comprising some common and easily accessible learning strategies was designed to determine the frequency of these strategies used in the classrooms by the language learners through the specialists of the subject. The data was collected from the selected sample through the subject specialists teaching in these schools. The data was collected quantitatively and was analyzed in the statistical package for social sciences (SPSS) version 20. The most common 27 language learning strategies (LLS) were applied to analyze the process of language learning. In the light of the results of the study, it was concluded that application of the learning strategies according to the nature of the text is helpful in understanding the language functions and its application. Keyswords: Language Acquisition, Learning Strategies, Mechanism of Language Learning Pages: 249-258 Article: 21 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)21 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)21 Download Pdf: download pdf view article Creative Commons License Secondary School Science Teachers’ Practices for the Development of Critical Thinking Skills: An Observational Study Authors: Dr. Muhammad Jamil Dr. Yaar Muhammad Dr. Naima Qureshi Abstract: In the National curriculum policy documents, to produce rationale and independent critical thinkers, different pedagogical practices have been recommended like cooperative learning, questioning, discussion, etc. This qualitative case study aimed at analyzing secondary school science teachers’ practices for the development of critical thinking skills in secondary school students. There were twelve classrooms (four from each subject of Physics, Chemistry and Biology) selected as cases. Video recording was used for the observations for six lessons in each classroom. In this way, a total of 72 observations were conducted lasting for approximately 35 minutes. Qualitative content analysis was used for data analysis through Nvivo 12. The findings of the observations revealed that all the teachers used the lecture method. They used this to cover the content at a given specific time. There was not much focus on the development of critical thinking. In a few of the classrooms, the students were engaged and active during learning different specific topics. Whiteboard was used as a visual aid by most of the teachers. Furthermore, to some extent, discussion, questioning, and daily life examples were used in different classrooms. It is recommended that teachers’ professional development should be conducted to focus on the development of critical thinking skills through pedagogical practices which have been recommended by the national education policy documents. Keyswords: Analysis, Critical Thinking, Curriculum Policy, Pedagogy, Secondary Level Pages: 259-265 Article: 22 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)22 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)22 Download Pdf: download pdf view article Creative Commons License Historical Development of Clinical Psychology in Pakistan: A Critical Review-based Study Authors: Muhammad Nawaz Shahzad Dr. Mushtaq Ahmad Dr. Muhammad Waseem Tufail Abstract: Clinical Psychology is clinical and curing psychological practices in Pakistan. The present research study endeavors to examine the contemporary status of Clinical Psychology in the country and descriptively analyzes the significant contribution of various psychologists in its development. The study also elaborates the emergence of Clinical Psychology and its treatment aspects in the country. The experimental approach of the treatment psychology has also been defined. The role of different scholars to set and promote the Clinical Psychology as discipline and dealing about treatment of Human mind has also been discussed here. The study also presented the scenario of the issues of legislative acknowledgment, qualifications mandatory for practice, communal awareness of cerebral treatment, the tradition of ethnic and native practices about the clinical psychological treatments has also been discussed. Keyswords: Approaches, Clinical Psychology, Psychologist, Therapist Pages: 266-272 Article: 23 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)23 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)23 Download Pdf: download pdf view article Creative Commons License Impact of Devolution of Power on School Education Performance in Sindh after 18th Constitutional Amendment Authors: Abdul Hafeez Dr. Saima Iqbal Muhammad Imran Abstract: Devolution of the authority from central units of empowering authorities to the local level to develop and exercise policies at local or organizational level is under debate in various countries of the world. The legation in with the name of 18th constitutional amendment in constitution of 1973 of Pakistan ensures more autonomy to federal units. The difference between province and federation mostly creates misunderstanding in the belief of cooperation and universalism of education standards, expenditures and service delivery. Very currently the ministry of education and local government encoring principles and headmasters to adopt self-management skills to be updated to accept the spin of power from higher authorities to lower authorities’ pedagogical and local schools. In this qualitative research semi structured questioner were incorporated as data collection tool equally, the data was analyzed by usage of NVivo software. In this regard Government of Sindh has introduced various reforms and new trends like objectives and policy pillars, better government schools, improved learning outcomes and increased and improved funding in the education sector Sindh government has so far been unable to effectively use its resources to implement effective governance system which provides quality and sustained education in the province. To achieve this basic universal education, equally fourth objective of Sustainable Development Goal (SDG) the educational leaders must develop a comparative education setup that help to educate planers to plan and design standards for school leaders, instruction, appropriate professional development of teachers, ways to support school leaders to change in mission. Parallel, develop new program for early childhood, school and class size and ensure school enrollment. Keyswords: 18th Constitutional Amendment, Devolution of Power, Sindh Education Performance Pages: 273-285 Article: 24 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)24 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)24 Download Pdf: download pdf view article Creative Commons License Legal Aspects of Evidence Collected by Modern Devices: A Case Study Authors: Muhammad Hassan Zia Alvina Ali Abstract: This paper is a qualitative research of different case laws dealing with modern technological evidence. Courts were required to adopt new methods, techniques and devices obtained through advancement of science without affecting the original intention of law. Because of modern technology, a benefit could be taken from said technology to preserve evidences and to assist proceedings of the Court in the dispensation of justice in modern times. Owing to the scientific and technological advancements the admissibility of audio and visual proofs has grown doubtful. No doubt modern evidence assist the court in reaching out to the just decision but at the same time certain criteria need to be laid down which must be satisfied to consider such evidence admissible. Different Case laws are discussed here to show how the cases were resolved on the basis of technological evidence and when and why such evidence have been rejected by the court, if it did. Moreover, legal practices developed in various countries allow our Courts to record evidence through video conferencing. The Honorable Supreme Court of Pakistan directed that in appropriate cases statement of juvenile rape victims and other cases of sensitive nature must be recorded through video conferencing to avoid inconvenience for them to come to the Court. Nevertheless, it has some problems. The most important among them is the identification of the witness and an assurance that he is not being prompted when his statement is recorded. In this paper protocols that are necessary to follow while examining witness through video link are discussed Keyswords: DNA Profiling, Finger Prints, , Telephone Calls, Video Tape Pages: 286-297 Article: 25 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)25 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)25 Download Pdf: download pdf view article Creative Commons License The Political Economy of Terrorisms: Economic Cost of War on Terror for Pakistan Authors: Muhammad Shakeel Ahmad Siddiqui Dr. Muhammad Imran Pasha Saira Akram Abstract: Terrorism and its effect on contemporary society is one of the core and vital subjects of International Political Economy (IPE) during the last years. Despite the fact that this is not a new phenomenon, special attention has been given to this issue, specifically after the terrorist attacks of 9/11, 2001. The objective of this paper analyzes to what dimensions terrorism affects the global economy mainly the two predominant actors of the conflict i.e. Pakistan and the United States. For this purpose, this article will take a look at the financial cost of War for Pakistan and how Pakistan’s decision to become frontline State has affected its Economy, its effect on agriculture, manufacturing, tourism, FDI, increased defense costs The normative and qualitative methodology shows a significant disadvantage between terrorist activities and economic growth, social progress, and political development. The results shows that Pakistan has bear slow economic growth while facing terrorist activities more than US. In this last section, the paper suggests ways and means to satisfy people around the world not to go in the hands of fundamentals and terrorists. Keyswords: Cost of War, Economic Growth, Frontline States, Pak Us Relations, Terrorism Pages: 297-309 Article: 26 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)26 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)26 Download Pdf: download pdf view article Creative Commons License A Comparative Study of Grade 10 English Textbooks of Sindh Textbook Board and Cambridge “O Level” in the perspective of Revised Bloom’s Taxonomy Authors: Mahnoor Shaikh Dr. Shumaila Memon Abstract: The present study evaluated the cognitive levels of reading comprehension questions present in grade 10 English Textbooks namely English Textbook for grade 10 by Sindh Textbook Board and compared it to Oxford Progressive English book 10 used in Cambridge “O Level” in the perspective of Revised Bloom’s Taxonomy. Qualitative content analysis was used as a methodology to carry out the study. To collect the data, a checklist based on Revised Bloom’s taxonomy was used as an instrument. A total of 260 reading comprehension questions from both the textbooks were evaluated. The findings of the study revealed that reading comprehension questions in English textbook for grade 10 were solely based on remembering level (100%) whereas the questions in Oxford Progressive English 10 were mainly based on understanding level (75.5%) with a small percentage of remembering (12.5%), analyzing (11.1%) and evaluating level (0.74%). This suggests that the reading comprehension questions in both the textbooks are dominantly based on lower-order thinking skills. Keyswords: Bloom’s Taxonomy, Content Analysis, Reading Comprehension, Textbook Evaluation Pages: 310-320 Article: 27 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)27 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)27 Download Pdf: download pdf view article Creative Commons License Assessing the Preparedness of Government Hospitals: A Case of Quetta City, Balochiatan Authors: Sahar Arshad Syed Ainuddin Jamal ud din Abstract: Earthquake with high magnitude is often resulting in massive destruction with more causalities and high mortality rate. Timely providence of critical healthcare facilities to affected people during an emergency response is the core principle of disaster resilient communities. The main objective of this paper is assessing the hospital preparedness of government hospitals in Quetta. Primary data was collected through questionnaire survey. Total of 165 sample size chosen via simple random sampling. Relative important index (RII) is used to analyze the overall situation of hospitals preparedness in term of earthquake disaster. Findings of the study showed that the preparedness level of government hospitals in Quetta is weak to moderate level. Based on the findings this study recommends the necessary measures to minimize the risk of earthquake disaster including training and exercise programs for the staff of hospital, proper resource management to efficiently use the existing machinery and equipment in the meeting of disaster to enhance employee’s performance and preparedness of government hospitals in Quetta to deal with earthquake disaster. Keyswords: Earthquake, Preparedness, Relative Important Index Pages: 321-329 Article: 28 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)28 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)28 Download Pdf: download pdf view article Creative Commons License Development of Reasoning Skills among Prospective Teachers through Cognitive Acceleration Approach Authors: Memoona Bibi Dr. Shamsa Aziz Abstract: The main objectives of this study were to; investigate the effects of the Cognitive Acceleration approach on the reasoning skills of the prospective teachers at the university level and compare the effects of the Cognitive Acceleration approach and traditional approach concerning reasoning skills of prospective teachers’ at the university level. The study was experimental and followed a pre-test post-test control group experimental design. The sample of the study included the experimental group and control group from the BS Education program in the Department of Education at International Islamic University Islamabad. A simple random sampling technique was used to select the sample after pre-test and pairing of prospective teachers. CTSR (classroom test for scientific reasoning) developed by A.E. Lawson (2000) was used to collect the data through pre-tests and post-tests. The experimental group’s perception about different activities of the experiment was taken through a self-made rating scale. Collected data were analyzed by calculating mean scores and t-test for hypothesis testing by using SPSS. The main findings of the study revealed that the Cognitive Acceleration teaching approach has a significant positive effect on the reasoning skills development of prospective teachers at the university level. Findings also showed that participants found this teaching approach effective and learned many new concepts and skills with the help of thinking activities. Based on findings it has been concluded that the Cognitive Acceleration teaching approach might be encouraged for training prospective teachers at the university level and training sessions about the use of the Cognitive Acceleration approach must be arranged by teacher education programs and institutions. Keyswords: Cognitive Acceleration Approach, Prospective Teachers, Reasoning Skills, Traditional Approach Pages: 330-342 Article: 29 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)29 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)29 Download Pdf: download pdf view article Creative Commons License Spatial Injustice in Shamsie’s Kartography Authors: Syeda Hibba Zainab Zaidi Dr. Ali Usman Saleem Sadia Waheed Abstract: Social space under postmodernism and wave of globalization have suffered in and its idealistic representations are lost and deteriorated which ultimately led to discursiveness in the lives of postmodern man, especially Karachiites. The boundaries of geographies play a significant role in shaping fates, biographies, social superstructures and shared collective histories of its residents. Considering this, Henri Lefebvre and Edward William Soja, argue that space is something which determines the living circumstances within the particular social framework and instigates and controls various societal happenings. City space of Karachi suffers from appalling distortions as a part of postmodern, globalized and capitalist world. By employing Lefebvre’s idea of spatial triad and Soja’s views of the trialectrics of spaciality, this paper foregrounds how social space enforces spatial injustice and serves for the inculcation of spatial cleansing in the lives of inhabitants of urban space. Using Shamsie’s Kartography as an interpretive tool for contemporary urban environment, this paper inquires the engrafting of spatial cleansing in the lives of Karachiites resulting in multiple standardization and segregation on the basis of living standards among different social strata. This research substantiates how in Kartography, Materialism nibbles the roots of social values and norms while sequentially administering Spatial Injustice in the lives of Karachiites. This paper proclaims the scarcity of execution of Spatial Justice in the lives of common people in this postmodern globalized capitalist era. This paper urges the possibility of a utopian urban space with enforced spatial justice where people can be saved from dilemmas of injustice and segregation, especially Karachiites. Keyswords: Capitalistic Hegemony, City Space, Globalization, Spatial Cleansing, Spatial Injustice Pages: 343-352 Article: 30 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)30 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)30 Download Pdf: download pdf view article Creative Commons License A Quasi-Experimental Study on the Performance and Attitudes of Pakistani Undergraduate Students towards Hello English Language Learning Application Authors: Wafa Pirzada Dr. Shumaila Memon Dr. Habibullah Pathan Abstract: With the advancement of technology, more and more avenues of bringing creativity and innovation in language learning have opened up. These exciting advances have given rise to a new field of study within linguistics, termed Mobile Assisted Language Learning (MALL). This paper aims to fill the gap of MALL research in the area of grammar teaching in the Pakistan. Two BS Part 1 classes from University of Sindh, Jamshoro, were chosen for this quasi-experimental study. In total, 62 out of 101 students volunteered to use the Hello English application for 2 months, making up the experiment group, and the remaining 39 students were put in a control group. Paired Samples T-Test was run on pretest and posttest results which revealed no significant difference in both groups’ performances, proving that Hello English application could not significantly improve students’ grammar performance. However, in spite of the lack of a significant difference between the test results, the data gathered through the attitudinal survey showed that students still found mobile application very easy to use and effective in language learning. Keyswords: Attitudes, Grammar Learning, Hello English, Mobile Language Learning, Technology In Language Learning Pages: 353-367 Article: 31 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)31 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)31 Download Pdf: download pdf view article Creative Commons License Impact of Determinants on the Profile Elevation of Secondary School Teachers in Pakistan Authors: Zahida Aziz Sial Dr. Farah Latif Naz Humaira Saadia Abstract: The foremost purpose of this research paper was to interrogate the effects of determinants on the educational and social profile of secondary school teachers in Pakistan. The key question taken was related to determinants that affect teachers’ profile. The Population of the study was secondary school teachers of Punjab province. A questionnaire was used as research instrument. The researcher personally visited the schools to administer the questionnaire. E-Views software was used for data analysis. Moreover, OLS regression model and LOGIT regression model were carried out. It was found that the variable years of teaching experience (EXPYR) (*** 0.03) can have a vital concrete effect upon the societal figuration of teachers as the experience of teachers grows, so does their social interactions with officials, colleagues, students and friends increases. The said variable is significant at 10 percent level. The variable, Residence (RESIDE) (** 0.53) have a significant impact upon civic links. This obviously associated with less community connection of country side teachers than the teachers residing in urban areas. Keyswords: Determinants, Elevation, Educational Profile, Social Profile, Secondary School Teacher Pages: 368-372 Article: 32 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)32 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)32 Download Pdf: download pdf view article Creative Commons License Impact of War on Terror on the Tourism Industry in Swat, Pakistan Authors: Sabir Ihsan Prof. Dr. Anwar Alam Aman Ullah Abstract: The present study was designed to ascertain the status of tourism before insurgency, during insurgency and after insurgency in District Swat-KP Pakistan. The study is quantitative and descriptive in nature. A diverse sample size of 370 out of 9014 was selected through convenient sampling strategy. Notwithstanding, the objectives of the study was achieved through structured questionnaire. Data was analysed through chi-square at Bi Variate level. Findings of the study revealed that earning livelihood in swat was significantly associated (P=0.016), (P=0.003) with tourism industry prior 2009 and present time respective, but the same statement was observed non-significant (P=0.075) at the time of insurgency. Arranging different festivals in the study area and establishment of different showrooms for local handcrafts, artificial jewellery and woollen shawl are some of the recommendations of the study. Keyswords: Business, Insurgency, Swat, Tourism Pages: 373-385 Article: 33 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)33 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)33 Download Pdf: download pdf view article Creative Commons License Challenges and Prospects of Pak-China Economic Corridor Authors: Muhammad Mudabbir Malik Prof. Dr. Muqarrab Akbar Abstract: Pak-China has historic relationships from the emergence of both states, and were proved long-lasting in every thick and thin times. In initial times they supported each other in foreign policies and regional issues. Pakistan and China have border disputes with India, which forced them to come close to counter India, letter on the economic interests strengthened these relations. In order to maximize the economic benefits, China announced economic corridor with the name China Pakistan Economic Corridor (CEPC). It was thought it will boost the economic growth of China, and as a prime partner Pakistan will also get economic benefits. In order to completely understand how Pakistan and China came on the same page and decided to put CPEC into reality we have to understand the Geo-political Importance of Pakistan, Strategic and economic importance of CPEC for China and Pakistan, Influence and concerns of West and neighboring countries including India. Domestic limitations and all the possible benefits and risks involved in this project for both Pakistan and China, this research acknowledges all these questions. Keyswords: Challenges, China, CPEC, Domestic Limitations Economic Growth, Pakistan, Western and Regional Concerns Pages: 386-404 Article: 34 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)34 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)34 Download Pdf: download pdf view article Creative Commons License An Analysis of Learning Practices and Habits of Children at Early Childhood Education: Students’ Perspective Authors: Masood Ahmad Sabiha Iqbal Shaista Noreen Abstract: The study was designed to analysis learning practices and habits of children at early childhood education. The major objective of the study was to find out the learning practices and habits of children. Problem was related to current situation, so survey method was exercised, 220 students were selected with the help of convenient sampling technique. Self-constructed questionnaire were exercised. The collected data was analyzed and calculate frequency, percentage, mean score, standard deviation and t-test of independent variable. The major findings of the study were; students learn from the pictures, cartoons and funny face; student’s eyes get tired of reading. When student read context continuously then they feel that their eyes get tired. There was a significance difference between male and female student about learning practices and habits of children. Keyswords: Early Childhood Education, Learning Practices and Habits, Pre-School Students Pages: 405-416 Article: 35 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)35 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)35 Download Pdf: download pdf view article Creative Commons License Gender Identity Construction in Akhtar’s Melody of a Tear Authors: Dr. Amna Saeed Hina Quddus Abstract: This study aims to discuss the notion of gender in terms of performativity and social construction. It also draws upon the idea of gender identity construction and how it relates to the society, performativity and biology. As its theoretical framework, the study relies upon the Performative Theory of Gender and Sex (1990) presented by Judith Butler and studies the gender identity construction in the female protagonist of Akhtar’s Melody of a Tear. Zara is a girl who is raised as a boy from his father and there is a kind of dilemma in Zara’s personality related to being masculine and feminine. The cultural norms of a particular gender are also a cause of this dilemma. Throughout the novel, she is in a conflicting state whether she should behave feminine or masculine. She is being depicted as an incomplete person until she finds and resolves this issue of gender identity. The paper discusses the gender performativity, social construction, cultural norms and identity as these are all contributing to the confusion and construction of the protagonist’s identity. Character analysis is used as the methodology of analysis. Keyswords: Cultural Norms, Femininity And Identity Confusion, Gender, Performativity, Masculinity, Social Construction Pages: 417-427 Article: 36 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)36 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)36 Download Pdf: download pdf view article Creative Commons License The Level of Impulsivity and Aggression among Crystal Meth and Cannabis Users Authors: Dr. Umbreen Khizar Muhammad Shafique Sana Nawab Abstract: Cannabis and crystal meth use is pervading in our society. Present study was conducted to explore the relationship between level of impulsivity and aggression among crystal meth and cannabis users. The sample of the present study was comprised of 100 participants. There were 50 cannabis and 50 crystal meth users who were diagnosed on the basis of DSM-V without any comorbidity. The sample were taken from all age range of population. The minimum education level was primary and maximum education level was graduation and above. The sample was selected from different drug rehabilitation centers of Rawalpindi and Islamabad, Pakistan. Demographic Performa was used to collect the initial important information, The “Barratt Impulsiveness Scale was used to measure the impulsivity and “Aggression Questionnaire” were used to measure the level of aggression. Finding of the study showed that there are significant differences among crystal meth and cannabis users on level of aggression. The calculated mean value for crystal meth user and for cannabis users indicates that crystal meth users have higher level of aggression as compared to the cannabis user. Over all analysis indicates a significant positive correlation of impulsivity with the variable aggression. The alpha coefficient value for all scale is acceptable. Keyswords: Aggression, Cannabis Users, Crystal Meth, Impulsivity Pages: 428-439 Article: 37 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)37 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)37 Download Pdf: download pdf view article Creative Commons License Impact of Social Factors on the Status of Tribal Women: A Case Study of the (Erstwhile) Mohmand Agency Authors: Sadia Jabeen Prof. Dr. Anwar Alam Muhammad Jawad Abstract: This study investigates the impact of socio-economic and cultural factors on the status of tribal women in the erstwhile Mohmand agency of the Ex-Federally Administered Tribal Area (FATA), Pakistan. Cultural practices and illiteracy impede the role of women in socio-economic development. The respondents were randomly selected from tehsil Ekka Ghund and Pindialai with a sample size of 370, through stratified random sampling. Data collected through structured interview schedule, FGD and observation technique. The study reveals that tribal practices early marriages, joint family system, tradition of forced marriages, compensation/Swara, exchange, purchase marriages, hampers women’s socioeconomic status. The illiteracy rate is high among the tribal women and it further undermines their role and negatively affects their socio-economic status. However, improvement in women status needs peace and stability, reforms in the constitution for women empowerment and active participation, improvement in the quality and quantity of education, women employability, skills development and women entrepreneurship Keyswords: Empowerment and Education, Marriage Types, Tribal Women Role, Tribal Women Status, Violence against Women Pages: 440-455 Article: 38 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)38 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)38 Download Pdf: download pdf view article Creative Commons License Effects of Heavy School Bags on Students’ Health at Primary Level in District Haveli (Kahutta) Azad Jammu and Kashmir Authors: Dr. Muhammad Mushtaq Shamsa Rathore Mishbah Saba Abstract: Heavy school bags is a very serious issue for the health of the primary level students throughout the world particularly in Azad Jammu and Kashmir. This study intends to explore the effect of heavy school bags on students’ health at primary level in district Kahuta. Naturally the study was descriptive and survey method was used, the population consists of one hundred ninety teachers and a sample of one hundred twenty seven teachers was selected using non probability sampling technique. A likert scale questionnaire was developed validated and distributed among the sampled respondents. The researcher personally visited the schools and collected the filled questionnaire. The data was coded and fed to the SPSS to analyze and interpret. The Chi Square test was applied to see the effect of heavy school bags on student’s health and academic achievement. The study found that heavy bags have negative effect on their health as well as their academic achievement. Students were found complaining their sickness, body and back pain. They were also found improper in their gait and their body postures. The researcher recommended the policy makers to take and develop strategies to decrease the heavy school bags. The school administration needs to make alternate days’ time tables of the subjects. Keyswords: Health, Primary Level, School, Bags, Students Heavy Pages: 456-466 Article: 39 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)39 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)39 Download Pdf: download pdf view article Creative Commons License Exploring the ‘Civil Repair’ Function of Media: A Case Study of The Christchurch Mosques Shootings Authors: Ayaz Khan Dr. Muhammad Junaid Ghauri Riffat Alam Abstract: This research endeavor is an attempt to explore and analyze the discourse produced by The New Zealand Herald; a newspaper from New Zealand and by The News International; a Pakistani newspaper. The researchers intend to determine whether and to what extent both the newspapers have the role of ‘civil repair’ played after the Christchurch mosques shootings. The researchers have incorporated the ‘lexicalization’ and the ‘ideological square’ techniques proposed by Tuen A. van Dijk within the scope of Critical Discourse Analysis. The findings of this study show that both the selected newspapers assuming the social status of ‘vital center’ performed the role of ‘civil repair’ in the aftermath of the shootings by producing the ‘solidarity discourse’. The ‘solidarity discourse’ has been produced in terms of the ‘we-ness’, harmony, understanding, and by mitigating the conflicting opinions. Keyswords: Christchurch Mosque Shootings, Civil Repair, Civil Sphere Theory, Lexicalization, Solidarity Discourse Pages: 467-484 Article: 40 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)40 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)40 Download Pdf: download pdf view article Creative Commons License China Pakistan Economic Corridor: Regional Dominance into Peace and Economic Development Authors: Tayba Anwar Asia Saif Alvi Abstract: The purpose of this qualitative study was to investigate the true motivations behind CPEC idea and the advantages it delivers to Pakistan and China. It also recognizes the Corridor's potential for mixing regional economies while dissolving geographical borders. The study is deductive in character, since it examines financial, political, and military elements of Pakistan and China's positions and situations. Enhancing geographical linkages through improved road, train, and air transport systems with regular and free exchanges of development and individual’s interaction, boosting through educational, social, and regional civilization and wisdom, activity of larger quantity of investment and commerce flow, generating and moving energy to provide more optimal businesses for the region. Keyswords: Geographical Linkages, Globalized World, Landlocked, Regional Connectivity, Regionalization Pages: 485-497 Article: 41 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)41 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)41 Download Pdf: download pdf view article Creative Commons License China’s New Great Game in Central Asia: Its Interest and Development Authors: Bushra Fatima Rana Eijaz Ahmad Abstract: Central Asia is rich in hydrocarbon resources. It’s geostrategic, geopolitical, and geo-economic significance has grasped the attention of multiple actors such as China, the USA, Russia, Turkey, the European Union, Pakistan, Afghanistan, and India. Due to its location, the Central Asian region appeared as a strategic hub. In the present scenario, China’s strategy is massive economic development, energy interest, peace, and stability. This article highlights China’s interest, political and economic development, and its role as a major player in the New Great Game in Central Asia. Shanghai Cooperation Organization (SCO) which presents as a platform where China is playing an active role in political, economic, and security concerns for achieving its objectives in Central Asia. The new step of the Belt and Road Initiative (BRI) sheds light on China’s progressive move in this region via land and sea routes, which creates opportunities for globalization. Keyswords: Belt and Road Initiative, Central Asia, China, New Great Game Pages: 498-509 Article: 42 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)42 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)42 Download Pdf: download pdf view article Creative Commons License Personality Traits as Predictors of Self-Esteem and Death Anxiety among Drug Addicts Authors: Umbreen Khizar Saira Irfan Iram Ramzan Abstract: This study seeks to investigate whether personality traits predict self-esteem and death anxiety among drug addicts. The sample consisted of 100 drug addicts taken from the two hospitals in Multan city. Only men between the ages of 20 and 65 were included in the study. Data was collected through reliable and valid questionnaires. Results revealed positive relationship between conscientiousness, openness to experience and self-esteem. Moreover, findings showed positive relationship between extraversion and death anxiety, and negative correlation between neuroticism and death anxiety. Findings also showed that self-esteem and death anxiety are significantly and negatively correlated. Additionally, findings revealed that conscientiousness positively predicted self-esteem and neuroticism negatively predicted death anxiety. Furthermore, significant differences were observed in self-esteem, and death anxiety based on age. Significant differences were also found in extraversion, agreeableness, openness to experience, and death anxiety based on location. Understanding how personality traits affect behavior can help drug addicts get the support they need to live a better life and reduce their risk of death anxiety and premature death. Keyswords: Death Anxiety, Drug Users, Personality Traits, Self- Esteem Pages: 510-524 Article: 43 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)43 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)43 Download Pdf: download pdf view article Creative Commons License Middle East: A Regional Instability Prototype Provoking Third Party Interventions Authors: Waseem Din Prof. Dr. Iram Khalid Abstract: Third party interventions always prolong the interstate or civil wars with unending sufferings and devastations. The entire Middle East region is fraught with tensions, conflicts, civil wars and rivalries. From strategic interests to power grabbing, sectarian divisions, flaws in the civil and social structure of the state and society, ethnic insurrections, and many other shapes of instability syndromes can be diagnosed in this region. In the post-Arab Spring, 2011, the emerging new regional hierarchical order for power/dominance, in addition to the weakening/declining dominant US power in the region, changed the entire shape of already conflict-ridden region. New weak or collapsing states and bifurcation of the ‘status quo’ and ‘counter-hegemonic’ states along with their respective allies, made this region a prototype of instability in the regional security complex of the Middle East, as a direct result of these developments. The perpetuation of these abnormalities would not recede this instability conundrum from the region, provoking third party intervention, if not contained. Keyswords: Conflicts/Civil Wars, Dominant Power, Instability, Intervention, Middle East, Middle Powers, Regional Hierarchy, Regional Powers, Security Complex, Weak State Pages: 525-542 Article: 44 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)44 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)44 Download Pdf: download pdf view article Creative Commons License Impact of Classroom Environment on Second Language Learning Anxiety Authors: Zohaib Zahid Abstract: Second language learning anxiety has attained the attention of the researchers in almost every part of the world. Pakistan is a country where English is taught as a second language from the very beginning of school education. Second Language learning anxiety is a phenomenon which has been prominently found among the learners because of their less proficiency in learning English language. This study has been conducted to investigate the effect of anxiety in learning and using English language in classroom, university and outside the classroom. There are variables that affect language learning performance of the learners but this paper has solely investigated the effect of anxiety. The paper has concluded that anxiety is a variable which has a striking affect in second language learning and its use inside classrooms. Keyswords: Effect of Anxiety, Proficiency, Second Language Learning Anxiety, Striking Affect Pages: 485-497 Article: 45 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)45 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)45 Download Pdf: download pdf view article Creative Commons License Struggling for Democracy: A Case of Democratization in Pakistan Authors: Ammara Tariq Cheema Dr. Rehana Saeed Hashmi Abstract: The objective of this research paper is to review the challenges for democratization in Pakistan. The problem of democratization and consolidation refers to the structure of democracy following the collapse of non-democratic regime. Ten factors as given by Michael J. Sodaro are considered effective in helping a democratically unstable state to stabilize its system in other words helps in the democratic consolidation. It is argued in this research that the ten factors of democratization as given by Michael J. Sodaro have been absent in the political system of Pakistan and working on these factors can lead Pakistan to the road of democratization. This study uses qualitative method of research and proposes a novel framework for the deed of parliament, because the effectiveness of parliament can contribute positively to democratization/consolidated democracy. Keyswords: Electoral Politics, General Elections, Political Participation, Women Empowerment Pages: 554-562 Article: 46 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)46 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)46 Download Pdf: download pdf view article Creative Commons License Impact of Dependency Ratio on Economic Growth among Most Populated Asian Countries Authors: Dilshad Ahmad Salyha Zulfiqar Ali Shah Abstract: Demographic transition through different channels significantly influences economic growth. Malthusian view postulated as dependency ratio adversely affects economic growth while Julian Simon's view is quite different, highlighted the long-run benefits of the population in the range of 5 to15 years on economic growth. This study can be a valuable addition in research to analyzing the association of dependency ratio and economic growth of the five most populated Asian countries (Bangladesh, China, Indonesia, India, and Pakistan). Empirical findings of the study indicated that a total dependency and younger dependency ratio has a positive and significant influence on economic growth in both short-run and long-run scenarios while the old dependency ratio shows a negative influence on economic growth in the long run while short-run results are unpredictable. There is a need for state-based proper policy measures in focusing the higher financing in human capital development specifically in education and health. Keyswords: Economic Growth, Gross Saving, Old Dependency Ratio, Young Dependency Ratio Pages: 563-579 Article: 47 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)47 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)47 Download Pdf: download pdf view article Creative Commons License Chinese Geo-Strategic Objectives and Economic Interests in Afghanistan under President Xi Jinping Authors: Farooq Ahmed Prof. Dr. Iram Khalid Abstract: China has its own distinctive interests, concerns and strategies with respect to the changing security dynamics in Afghanistan. China has taken an active interest, though retaining a low profile and avoiding direct military interaction. China has exclusively relished on economic engagement actively and provided numerous financial aid and financial support in the rebuilding of Afghanistan's economy. The aim of this research study is to analyze the geo-strategic objectives and economic interests of China under the leadership of President Xi Jinping. This study looks at the actual diplomatic, economic and protection commitments of both countries as well as the basis of the geopolitical complexities – core variables that form China's current foreign policy to Afghanistan. Keyswords: Afghanistan, BRI, China, NATO Withdrawal Pages: 580-592 Article: 48 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)48 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)48 Download Pdf: download pdf view article Creative Commons License The Argument Structure of Intransitive Verbs in Pashto Authors: Abdul Hamid Nadeem Haider Bukhari Ghani Rehman Abstract: This study focuses on the description and categorization of intransitive verbs in terms of its argument structure. The study concludes that the unaccusative verbs only project an internal argument. It does not require the event argument. However, the said verb can be causativised by adding external argument and at the same time the event argument gets included in the valency of the derived causative of the unaccusative root. The unergative, on the other hand, requires an external argument as an obligatory argument while the internal argument is not the obligatory argument of the verb. The event argument is also a part of the valency of the verb. The APFs require one argument which is the internal argument of the verb. However, since the external argument is not available, the internal argument of the verb gets realized as the subject of the verb. The verb does not project event argument. The ergative predicates are derived by the suppression of the external argument and by the externalization of the internal argument. Keyswords: Argument Structure, Ergative Case, Event Argument, External Argument, Internal Argument, Valency Pages: 593-610 Article: 49 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)49 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)49 Download Pdf: download pdf view article Creative Commons License Positive, Negative and Criminal Orientation of Beggars in Okara: Perspective of Students Authors: Shahzad Farid Saif-Ur-Rehman Saif Abbasi Hassan Raza Abstract: This study aimed to measure the perspective of students about the criminal orientation of beggars. The sample size of the study (i.e., 100 students) was explored using Taro Yamane’ equation from the university of Okara, Punjab, Pakistan. The respondents were approached using simple random sampling and interviewed using face to face interview schedule. The data was collected using a structured questionnaire. The analysis was administered through SPSS-20.The study explored that parental illiteracy is associated with the high criminal and negative orientation of students towards beggars. It was also explored that females and respondents from rural background have low negative orientation towards beggars. However, males and respondents from urban background have medium criminal orientation and low positive orientation towards beggars, respectively. The study is useful for the government of Punjab, Pakistan campaign and policy for anti-begging. The study introduced the geometrical model of youth’s orientation toward begging. The study also contributed to the literature on begging by extending its domain from Law and Criminology to sociology as it incorporated social variables e.g., parents’ education, gender, etc., to explore their association with the youth’s socialization about begging. Keyswords: Begging, Crime, Education, Gender, Students Pages: 611-621 Article: 50 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)50 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)50 Download Pdf: download pdf view article Creative Commons License Relationship between Entrepreneurial Export Orientation and Export Entrepreneurship through Mediation of Entrepreneurial Capabilities Authors: Muhammad Saqib Nawaz Masood ul Hassan Abstract: Export led growth is prominent paradigm in developing world since decades. Exports play vital role in the economy by improving the level of balance of payments, economic growth and employment. Due to strategic importance of exports, organizational researchers focused on finding antecedents of export performance of the organizations. To line with this, current study aims to find the impact of entrepreneurial export orientation on export entrepreneurship through mediation of entrepreneurial capabilities in the Pakistani context. For this purpose, data was collected from 221 exporting firms of Pakistan by using questionnaire. Collected data was analyzed with the help of Smart PLS. In findings, measurement model confirmed the validity and reliability of measures of variables. Additionally, structural model provides the positive impact of entrepreneurial export orientation on export entrepreneurship. Similarly, entrepreneurial capabilities mediate the relationship between entrepreneurial export orientation on export entrepreneurship. The findings provide important implications for the managers of exporting firms to improve export performance. Keyswords: Entrepreneurial Capabilities, Entrepreneurial Export Orientation, Export Entrepreneurship Pages: 622-636 Article: 51 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)51 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)51 Download Pdf: download pdf view article Creative Commons License China Pakistan Economic Corridor: Explaining U.S-India Strategic Concerns Authors: Nasreen Akhtar Dilshad Bano Abstract: Regional and International political and economic landscape is being changed owing to China Pakistan Economic Corridor (CEPEC)-the new security paradigm has taken place-that has increased the strategic concerns of the U.S. and India. This research paper attempts to re-examine China-Pakistan relations in the new emerging geo-political compass. This paper has investigated the question that how regional, and global developments have impacted the China-Pakistan relationship? And why China – Pakistan have become partners of CPEC? In the global context, this paper assesses the emerging International Order, Indo-U. S strategic narrative vis-à-vis CPEC, and the containment of China through the new alliances and their impacts on China -Pakistan vis-à-vis the Belt Road Initiative (BRI). Quadrilateral (Quad) alliances is shaping the new strategic political and security paradigms in the world politics. Keyswords: BRI, China, CPEC, India, Pakistan, Silk Road, Strategic Concerns Pages: 637-649 Article: 52 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)52 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)52 Download Pdf: download pdf view article Creative Commons License The Structure of Domestic Politics and 1973 Constitution of Pakistan Authors: Dr. Fida Bazai Dr. Ruqia Rehman Amjad Rashid Abstract: Pakistan is located in a pivotal region. Its geo-strategic location affects its national identity as a nation state. Unlike Europe in South Asia security dilemma, proxy warfare and nuclear arms race are consistent features of the regional politics. The identity of Pakistan as security-centric state gives its army disproportional power, which created institutional imbalance that directly affected constitutionalism in the country. The constitution of Pakistan is based on principles of civilian supremacy and separation of power but in reality Pakistan’s army is the most powerful institution in country. This paper argues that the structure of Pakistani politics; created institutional imbalances by the disproportionate distribution of resources is the key variable in creating dichotomy. The structure of domestic politics is based upon the principles of hostility to India, use of Islam for national unity and strategic alliances with major powers to finance defense against the neighboring countries. Keyswords: Constitutionalism, Identity, Islam, South Asia Pages: 650-661 Article: 53 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)53 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)53 Download Pdf: download pdf view article Creative Commons License National Integration and Regionalism in Pakistan: Government’s Strategy and Response toward Regionalist Demands 1947-77 Authors: Najeeb ur Rehman Mohammad Dilshad Mohabbat Muhammad Wahid Abstract: The countries of South Asian region have pluralistic societies with different language, religious, and ethnic identities. Pakistan is no exception who is facing the challenge of regionalism since its inception. Different ethnic groups have been consistently raising their voices for separatism or autonomy within the frame work of an existing territorial state. The issues of provincialism, ethnicity, and regionalism is posing a serious challenge to the integrity of the country. This paper aims to explore the causes of the regionalism in Pakistan and intends to analyze the policies and strategies of different political governments which they launched to tackle this all important issue. The paper follows the historical method of research and analyzes different types of qualitative data to conclude the finding of the research. The paper develops the theory of “Regionalists Demand and Government Response” which shows how different regionalist forces put their demands and how the governments react on these demands. It recommends the grant of greater regional autonomy to the regionalists to enhance internal security and to protect the country from disintegration. Keyswords: Demands, Ethnicity, Government Strategy, National Integrity, Nationalism, Regionalism Pages: 662-678 Article: 54 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)54 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)54 Download Pdf: download pdf view article Creative Commons License Fostering Entrepreneurial Mindset through Entrepreneurial Education: A Qualitative Study Authors: Saira Maqbool Dr. Qaisara Parveen Dr. Muhammad Hanif Abstract: Research on entrepreneurial mindset has flourished in these recent years. Its significance lies in a critical suspicion and its matters for inventive behavior. Entrepreneurship joined with innovative abilities, seen as one of the most wanted in this day and age. This study aims to determine the perceptions about entrepreneurial mindset, its importance, and the role of entrepreneurship education and Training in developing the entrepreneurial mindset. This is a qualitative study based on interviews conducted by professors of Pakistan and Germany. The analysis was determined through content analysis. The results determine that 'Making Entrepreneurial Mindset' assists with seeing better all parts of business venture, which will undoubtedly influence their view of business venture, pioneering abilities, and mentalities. Keyswords: Entrepreneurship Education, Entrepreneurial Mindset Pages: 679-691 Article: 55 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)55 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)55 Download Pdf: download pdf view article Creative Commons License Benefits of Implementing Single National Curriculum in Special Schools of Lahore city for Children with Intellectual Disability: Teachers’ Perception Authors: Dr. Hina Fazil Khurram Rameez Sidra Ansar Abstract: Single national curriculum (SNC) is an important issue across the Punjab Province of Pakistan. Making and implementing SNC is not only focusing the education of normal pupils, but also focusing students with disabilities (SWD). The field of special education experienced an increased discussion of curriculum for students with intellectual disabilities (SID). The present research aimed to know the benefits to implement first stage of single national curriculum for students with Intellectual disability and to know the differences about the benefits between public and private schools regarding SNC for students with ID based on demographic characteristics. Likert type researchers-made questionnaire with reliability) Cronbach alpha .922) was used. 90 special educationists from public and private schools were chosen through random sampling technique. The findings raised some benefits such as: SNC will bridge the social and economic disparities which will increase the acceptance of ID students. It was recommended that SNC should include areas of adaptive skills, motor, and vocational skills to get involved in work activities. Keyswords: Benefits, Children with Intellectual Disability, Single National Curriculum Pages: 692-703 Article: 56 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)56 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)56 Download Pdf: download pdf view article Creative Commons License Last Rituals and Problems Faced by the Hindu Community in Punjab: A Case Study of Lahore Authors: Sabir Naz Abstract: Lahore is the provincial capital of Punjab, where a sizeable population of the Hindus has been residing there since the inception of Pakistan. There had been many crematoriums in the city but with the passage of time, one after another, disappeared from the land after partition of the Sub-continent. Those places were replaced by commercial or residential sites. There is also a graveyard in the city which is in the use of Hindu Valmik Sect. However, it was encroached by some Muslims due to very small size of population and indolence of the Hindus. Later on, the encroachments were removed by the District Government Lahore in compliance of order of the Supreme Court of Pakistan. Presently, there is a graveyard as well as a crematorium in the city. The community remained deprived of a place to dispose of a dead body according to their faith for a long period which is contravention with the guidelines of the Quaid-e-Azam, founder of the nation Keyswords: Crematorium, Graveyard, Hindu community, Last Rituals Pages: 704-713 Article: 57 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)57 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)57 Download Pdf: download pdf view article Creative Commons License Estimating Growth Model by Non-Nested Encompassing: A Cross Country Analysis Authors: Benish Rashid Dr. Shahid Razzaque Dr. Atiq ur Rehman Abstract: Whether models are nested or non-nested it is important to be able to compare them and evaluate their comparative results. In this study six growth models have been used for analyzing the main determinants of economic growth in case of cross countries, therefore by using these six models we have tested them for non-nested and nested encompassing through Cox test and F-test respectively. Data from 1980 to 2020 were used to analyze the cross country growth factors so therefore, the current study looked at about forty four countries with modelling these different comparative studies based on growth modelling. So, we can make these six individual models and we can estimate the General Unrestricted Model with the use of econometric technique of Non-Nested Encompassing. By evaluating the data using the Non-Nested Encompassing econometric technique, different sets of economic variables has been used to evaluate which sets of the economic variables are important to boost up the growth level of the country. And found that in case of nested model or full model it is concluded that model with lag value of GDP, trade openness, population, real export, and gross fix capital formation are the main and potential determinants to boost up the Economic Growth in most of the countries. Keyswords: Cross Country, Economic Growth, Encompassing, Nested, Non-nested Pages: 714-727 Article: 58 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)58 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)58 Download Pdf: download pdf view article Creative Commons License Assessment of Youth Buying Behaviour for Organic Food Products in Southern Punjab: Perceptions and Hindrances Authors: Ayousha Rahman Asif Yaseen Muhammad Arif Nawaz Abstract: This research examined the cognitive antecedental effects on organic food purchase behaviour for understanding the perceptions and hindrances associated with purchasing organic food products. Theory of Planned Behaviour (TPB) was adopted as a theoretical framework. A total of 250 young consumers in the two cities of Southern Punjab, Pakistan was randomly sampled and data were collected via a face-to-face survey method. Partial least square technique was employed to test the model. The results showed that attitude towards organic food purchasing motivated when moral norms were activated to consume organic food products. Further, environmental knowledge moderated the relationship of organic food purchase intentions and behaviour significantly. The findings highlighted the importance of moral norms as a meaningful antecedent that could increase the TP-based psychosocial processes if consumers have sufficient environmental knowledge. Therefore, farmers, organic products marketers, government administrators, and food retailers should take initiatives not only to highlight the norms and values but also when promoting organic food production and consumption. Keyswords: Environmental Knowledge, Organic Food Purchase Behaviour, Personal Attitude, PLS-SEM, Subjective & Moral Norms Pages: 728-748 Article: 59 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)59 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)59 Download Pdf: download pdf view article Creative Commons License An Analysis on Students Ideas about English and Urdu as Medium of Instructions in the Subjects of Social Sciences studying in the Colleges of the Punjab, Pakistan Authors: Ashiq Hussain Asma Amanat Abstract: The worth and usefulness of English education as a foreign language is of great concern to language rule and planning (LRP) researchers compared to teaching their native language globally in higher education. The study under research examines the perspectives of two similar groups of the final year students of at Higher Education Institutions of Pakistan. The first group consists of art students who received the Urdu medium of instruction (UMI), and the second group received the English medium of instruction (EMI). An empirical methodology was carried out in the present year, students answered questionnaires to find out the benefits and challenges of learning subject-based knowledge, what subject-based knowledge means to them, and their understanding of language as a teaching language. Interviews were conducted with the selected group of students who wished to participate in research. Additional information is available from the tests and results obtained in the two equivalent courses. Although many similarities have been identified between the two groups, the overall knowledge of disciplinary knowledge of English medium instruction students was not very effective, while that of UMI students was very effective. It explains the implications of the findings to continue the language rule as policy experience for teaching in higher education institutions. Keyswords: English as Medium of Instruction (EMI), Higher Education Institutions (HEIs), Urdu as Medium of Instruction (UMI) Pages: 749-760 Article: 60 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)60 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)60 Download Pdf: download pdf view article Creative Commons License Environment and Women in Kurt Vonnegut’s ‘Happy Birthday Wanda Juny’: An Eco- Critical and Feminist Analysis Authors: Dr. Muhammad Asif Safana Hashmat Khan Muhammad Afzal Khan Janjua Abstract: This is an Eco-feminist study of Vonnegut’s ‘Happy Birthday Wanda Juny’ and focuses on how both women and environment are exploited by patriarchy. Ecofeminism critiques masculine dominance highlighting its role in creating and perpetuating gender discrimination, social inequity and environmental degradation. Women suffer more because of power disparity in society. Environmental crises affect women more than men because of their already precarious existence and subaltern position. There is affinity between women and nature are victims of climate change and other environmental hazards. Cheryl Glotfelty introduced interdisciplinary approach to the study of literature and environment. Literary ecology as an emerging discipline explores the intriguing relationship between environment and literature. Ecofeminism draws on feminist critique of gender inequality showing how gender categories inscribed in power structure exploit both women and nature. Francoise d‘Eaubonne coined the term ecofeminism to critique the prevalent exploitation of both women and environment. Ecofeminism asserts that exploitation of women and degradation of the environment are the direct result of male dominance and capitalism. Ecofeminism argues for redressing the plight of women and protection of environment. Vonnegut’s play ‘Happy Birthday Wanda June’ was written at a time when the movement for the right of women and protection of environment were gaining momentum. The play shows how toxic masculinity rooted in power and capitalism exploit both women and environment. Keyswords: Eco-Feminism, Eco-Criticism, Ecology, Environment, Exploitation Pages: 761-773 Article: 61 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)61 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)61 Download Pdf: download pdf view article Creative Commons License Critical Analysis of Social Equity and Economic Opportunities in the Light of Quranic Message Authors: Prof. Dr. Muhammad Yousuf Sharjeel Mahnaz Aslam Zahida Shah Abstract: This study critically evaluated the key verses of Surah Al-Baqarah -the second chapter of Quran, a sacred scripture of Islam- which specifically relates to social equity opportunities and a code of conduct in the context of economics. The Quran claims that it is a book which explains every situation; therefore, the aim of this study remained to extract those verses of Surah Al-Baqarah which can guide us in Economics. The authentic and approved Islamic clerics and their translations were consulted for the interpretations of the Holy verses. The researchers chiefly focused and studied Surah Baqarah with regards to social equity and economic opportunities. The translations were primarily in the regional language Urdu so the interpretations must not be related exactly equitable in English. The study engaged the document analysis research strategy. This study is only an endeavour to decipher Holy Quran’s message from Allah for the mankind so it must not be considered as the full and complete solution to the all the economic issues, challenges and opportunities. Ahadees and the saying of the Holy prophet were referred to where ever required and available. The researcher also considered the Tafasir (detail intellectual interpretations) of the Quran done by the well-known scholars of Islam for the verses studied therein and any statements and/or material - such as ideas, studies, articles, documentation, data, reports, facts, statistics etc. For the study, data was collected and analyzed qualitatively. On the basis of the study, recommendations were also primed. Keyswords: Economic Issues and Challenges, Social Equity, Surah Al-Baqarah, Al Quran Pages: 774-790 Article: 62 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)62 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)62 Download Pdf: download pdf view article Creative Commons License A Critical Discourse Analysis of Dastak by Mirza Adeeb Authors: Muhammad Afzal Dr. Syed Kazim Shah Umar Hayat Abstract: The present research aims to explore ideology in Pakistani drama. The drama, “Dastak”, written by Mirza Adeeb, has been taken for exploration ideologically. Fairclough’s (1992) three-dimensional model has been used for analyzing the text of the above-mentioned drama which includes textual, discursive practice and social practice analyses. The linguistic and social analyses of the drama reveal the writer’s ideology about socio-cultural, conventional and professional aspects of life. The study has also explored the past and present states of mind of Dr. Zaidi, the central and principal character of the drama, Dastak. The text implies that the writer has conveyed personal as well as social aspects of his times through the drama of Dastak. Keyswords: Dastak, Drama, Ideology, Semiotics Pages: 791-807 Article: 63 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)63 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)63 Download Pdf: download pdf view article Creative Commons License Linking Job Satisfaction to Employee Performance: The Moderating Role of Islamic Work Ethics Authors: Dr. Shakira Huma Siddiqui Dr. Hira Salah ud din Khan Dr. Nabeel Younus Ansari Abstract: The most pervasive concern in public sector organizations is declining employee performance and workforce of these organizations are less satisfied with their jobs. The aim of this study is to investigate the impact of Job Satisfaction on employee’s performance and how Islamic work ethics moderates the above mentioned direct relationship in the public sector organizations of Pakistan. The data were collected from the sample of 193 permanent employees working in public sector organizations through stratified sampling technique. The results revealed that employees Job satisfaction is significantly related to higher performance. Further, the findings indicated that Islamic work ethics moderates the relationship between job satisfaction and employee performance. The present research has some theoretical and empirical implications for academicians, policymakers, especially of public sector organizations, for the improvement of performance of their workforce. Keyswords: Employee Performance, Islamic Work Ethics, Job Satisfaction, Person-Environment Fit Theory Pages: 808-821 Article: 64 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)64 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)64 Download Pdf: download pdf view article Creative Commons License Semantics of Qawwali: Poetry, Perception, and Cultural Consumption Authors: Rao Nadeem Alam Tayyaba Khalid Abstract: Semantics is about meanings and meanings are arbitrary and shared. Understanding qawwali context requires comprehension of semantics or process of meaning creation and meaning sharing among the qawwal party and the audience. This interactive activity might frequently be hindered when interrupted by subjective meanings creation during cultural consumption. Qawwali is a cultural tradition, its semantics are conditioned by axiological premises of poetry and perceptions which are transforming. The previous researches revealed that qawwali is associated with religion which provides the religious message by singing hamd and naat. It was a means to experience Divine; therefore, semantics are multi-layered and often crossroad with values and subjective experiences. It is novel due to its ritual of Sama. It has the therapeutic power that helps mentally disturbed people and they find refuge. This study is exploratory having a small sample size of twenty purposively selected audiences. This phenomenological inquiry used ethnographic method of conversational interviews at selected shrines and cultural spaces in Islamabad. The results indicate that qawwali is a strong refuge for people facing miseries of life and they attend Sama with a belief that attending and listening will consequently resolve their issues, either psychological or physiological. They participate in Sama which teaches them how to be optimistic in a negative situation; this paper brings forth this nodal phenomenon using the verbatim explanations by the interlocutors. Semantics of Qawwali are conditioned and some of these elements are highlighted including poetry and axiology based perceptions and cultural consumption of a cultural realm. Keyswords: Cognition, Culture, Poetry, Qawwal, Qawwali, Semantics Pages: 822-834 Article: 65 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)65 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)65 Download Pdf: download pdf view article Creative Commons License Political Economy of Smuggling: The Living Source for the Natives (A Case Study of Jiwani-Iran Border, Baluchistan) Authors: Abdul Raheem Dr. Ikram Badshah Wasia Arshed Abstract: This study explores the political economy of smuggling on Jiwani-Iran border. The natives are majorly involved in illegal transportation of goods and objects, therefore; the study sets to explain how significant smuggling for the local people is. It describes the kinship role in reciprocity of their trade and transportation. The qualitative methods such as purposive sampling and interview guide were employed for data collection. The research findings revealed that local people were satisfied with their illegal trading which is depended largely on their expertise and know-how of smuggling at borders. They disclosed that their total economy was predominantly based on smuggling of stuff like drugs, diesel, oil, gas, petrol, ration food from Iran, and human trafficking. They also enjoyed the privilege of possessing Sajjil (Iranian identity card), thus; the dual nationality helped them in their daily business and rahdari (border crossing agreement), enabling them to travel to Iran for multiple purposes. Keyswords: Drugs, Human, Navigation, Political Economy, Reciprocity, Smuggling, Trafficking Pages: 835-848 Article: 66 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)66 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)66 Download Pdf: download pdf view article Creative Commons License The Vicious Circles of System: A Kafkaesque Study of Kobo Abe’s The Woman in the Dunes Authors: Imran Aslam Kainat Azhar Abstract: This paper analyses the Kafkaesque/Kafkan features of Kobo Abe’s novel The Woman in the as formulated by Kundera in “Kafka’s World.” For Kundera, in a Kafkaesque work human existence is bleakly represented through intermingling of tragedy and comedy in an indifferent world dominated by hegemonic systems. The Kafkaesque is characterised by the following: World is a huge forking labyrinthine institution where the man has been thrown to suffer its complexities, confrontation with the labyrinth makes his existence meaningless because freedom is a taboo in no man’s land, he is punished for an unknown sin for which he seeks justification from the superior authorities, but his efforts are viewed as ludicrous or comic despite the underlying sense of tragedy. (5) The Kafkaesque tendency to present tragic situation comically is also explored in Abe’s novel. The paper studies the effect of higher authorities exercising their power over man and the inscrutability of cosmic structures continuously undermining human freedom in nightmarish conditions. The paper establishes Kobo Abe in the literary world as a writer who portrays the hollowness and futility of human lives with a Kafkaesque touch. Keyswords: Authority, Institutions, Kafka, Kafkaesque, Kafkan, Kobo Abe, Kundera, The Trial, The Woman in the Dune Pages: 849-861 Article: 67 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)67 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)67 Download Pdf: download pdf view article Creative Commons License Subjectivity and Ideological Interpellation: An Investigation of Omar Shahid Hamid’s The Spinner’s Tale Authors: Hina Iqbal Dr. Muhammad Asif Asia Saeed Abstract: Louis Althusser’s concept of interpellation is a process in which individuals internalize cultural values and ideology and becomes subject. Althusser believes that ideology is a belief system of a society in which ideological agencies establish hierarchies in society through reinforcement and discrimination for cultural conditioning. These agencies function through ideological state apparatuses. These ideological agencies help to construct individual identity in society. The undesirable ideologies promote repressive political agendas. The non-repressive ideologies are inhaled by the individuals as a natural way of looking at the culture and society. This research seeks to investigate Omar Shahid Hamid’s novel The Spinners Tales through the lens of Althusser’s ideology and interpellation. This study examines how the characters of Shahid’s novel inhaled ideology and became its subjects. This research also depicts the alarming effects of cultural hegemony that creates cultural infidelity and hierarchies between the bourgeoisie and proletariat classes. Keyswords: Cultural Hegemony, Ideological State Apparatus, Ideology, Interpellation, Repressive Factors Pages: 862-872 Article: 68 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)68 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)68 Download Pdf: download pdf view article Creative Commons License Blessing in Disguise: Recommendations of Indian Education Commission (1882) and Christian Missionaries’ Educational Policy in the Colonial Punjab Authors: Mohammad Dilshad Mohabbat Muhammad Hassan Muhammad Ayaz Rafi Abstract: Woods Education Despatch is considered to be the Magna Carta of Indian Education. It controlled the Indian education field till the establishment of Indian Education Commission, 1882. The Despatch provided space to Christian missionaries by promising government’s gradual withdrawal from the education in favour of missionaries. It also facilitated the missionaries by offering system of ‘grants on aid’ to the private bodies. Consequently, the missionaries fancied to replace the government institutions in the Punjab and initiated their efforts to increase the number of their educational institutions. They tried to occupy the educational field by establishing more and more educational institutions. But after the Recommendations of the Indian Education Commission 1882, a change in their policy of numeric increase of educational institutions is quite visible. With the turn of the century, they are found to be eager to establish a few institutions with good quality of education. This paper intends to analyse different factors behind the change of their policy of quantitative dominance to qualitative improvement. It also attempts to evaluate how their change of policy worked and what steps were taken to improve the quality of their educational institutions. Following the historical method qualitative data comprising educational reports, missionaries’ autobiographies, Reports of missionaries’ conferences, and the other relevant primary and secondary sources has been collected from different repositories. The analysis of the data suggests that the attitude of the administration of the education department and the recommendations of Indian Education Commission were the major driving forces behind the change of missionaries’ educational policy in the 20th century. The missionaries, after adopting the new policy, worked on the quality of education in their institutions and became successful. Keyswords: Christian Missionaries, Indian Education Commission, Missionary Schools, Numeric Increase, Quality of Education. The Punjab, Woods Education Despatch Pages: 873-887 Article: 69 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)69 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)69 Download Pdf: download pdf view article Creative Commons License Basic Life Values of Prospective Special Education Teachers Authors: Dr. Maria Sohaib Qureshi Dr. Syeda Samina Tahira Dr. Muhammad Irfan Arif Abstract: Future teachers' preconceived values about how to live their lives and how that affects the lives of their students were the focus of this study. Descriptive research was used by the researchers. The study was carried out by using Morris's Ways to Live Scale. Researchers used this scale to study prospective special education teachers' gender, social status, personal relationships, aesthetics and mental approach using purposive sampling method. Descriptive and inferential stats were used to analyse the data collected from those who participated in the study on basic life values of prospective teachers. Results indicated that being social and sympathetic are the most important values among prospective special education teachers. It was also found that male and female prospective special education teachers living in urban and rural areas had no significant differences in their basic life values. Keyswords: Special Education, Teacher, Values Pages: 888-896 Article: 70 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)70 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)70 Download Pdf: download pdf view article Creative Commons License Perception of Dowry: Effects on Women Rights in Punjab Authors: Dr. Bushra Yasmeen Dr. Muhammad Ramzan Dr. Asma Seemi Malik Abstract: Dowry is a common tradition in south Asian countries, especially in Pakistan and India. Daughters became curses and liability for parents causing serious consequences. For control, there are legal ban/restrictions (Dowry and Wedding Gifts (Restriction) Act, 1976; Amendment in Act, 1993) on its practice in Pakistan. Despite the legal cover, the custom has been extended. Dowry amount seems to be increasing due to changing lifestyle and trends of society. To understand males’ and females’ perceptions about dowry; impacts of dowry; why dowry is essential; and how it is affecting women’s rights and eventually affecting women’s autonomy. A qualitative study was conducted. Data was collected by using unstructured interviews from males and females including social activists, economists, and married couples about wedding expenses, demands, society pressure, men’s support, and perception against dowry especially with regards to women’s rights and autonomy. The study concluded heavy dowry especially in terms of furniture, electronics, kitchenware, car, furnished houses, and cash highly associated with women’s development and their rights. General people’s perception showed that dowry is no longer remained a custom or tradition in Asian countries. It is just a trend and people follow it as a symbol of respect for parents and women as well. Keyswords: Dowry, Effects, Impacts Of Dowry, Perceptions, Women Autonomy, Women Rights Pages: 897-909 Article: 71 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)71 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)71 Download Pdf: download pdf view article Creative Commons License NCOC-An Emblem of Effective Governance: An analysis of Pakistan’s Counter Strategy for Covid-19 as a Non-Traditional Security Challenge Authors: Dr. Iram Khalid Abstract: COVID -19 affected the world unprecedentedly. Lack of capacity and poor standards of governance caused nontraditional security challenges to Pakistan too. The NCOC is the central nerve center to guide the national response to COVID-19 by Pakistan and can be best analyzed in the light of the decision-making theory of Naturalist Decision Making (NDM). The study points out the effective role performed by NCOC at policy formation through a more prosaic combination of science, data, decision making and execution of decisions at the level of federalism. The study highlights the changing patterns of government’s approach during the pandemic at various levels. Pakistan faced economic, political and social crisis during this phase. This study uses a survey and key informant interviews as the source of analysis for qualitative data collection. By applying the decision- making theory, the paper extends that there is a need to use a model to balance the existing gap within the system, to meet challenges. The study suggests a coordinating approach among various units and center; that might raise the level of performance to meet the nontraditional security challenges with innovation, creativity and boldness. Keyswords: COVID-19, Decision Making Theory, Governance, Nontraditional Threats, Strategy Pages: 910-930 Article: 72 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)72 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)72 Download Pdf: download pdf view article Creative Commons License Comparative Implications of Wednesbury Principle in England and Pakistan Authors: Safarat Ahmad Ali Shah Dr. Sara Qayum Arzoo Farhad Abstract: Wednesbury principle is one of the most important and useful grounds of the Judicial Review. Judicial review is a remedy provided by the public law and is exercised by the superior and higher courts to supervise administrative authorities' powers and functions. The main objective of the judicial review is to ensure the fair and transparent treatment of individuals by public authorities. The ground of the judicial review, i.e., Unreasonableness or irrationality or popularly known as Wednesbury Unreasonableness was introduced by lord Greene in the Wednesbury Corporation case in 1948. Initially, the scope of this ground of judicial review was very narrow and was allowed only in rare cases. However, with the development of administrative law and Human rights, it also developed. Its development resulted in different controversies and issues about the application of this ground. The main issue is about its encroachment in the jurisdiction of other branches of the government i.e., the parliament and executive. The free and loose application of this principle results in confusion and conflict between different organs of the government. The present paper is based on the implications of the limitations on the ground of Wednesbury Unreasonableness both on the judicial and administrative bodies in Pakistan to avoid the chaos and confusion that results in the criticisms on this ground of judicial review. Keyswords: Administrative Authorities, Critical Analysis, Illegality, Judicial Review, Pakistan, Wednesbury Unreasonableness Pages: 931-946 Article: 73 , Volume 2 , Issue 4 DOI Number: 10.47205/jdss.2021(2-IV)73 DOI Link: http://doi.org/10.47205/jdss.2021(2-IV)73 Download Pdf: download pdf view article Creative Commons License Water Sharing Issues in Pakistan: Impacts on Inter-Provincial Relations + + RKoch + + + CTreakle + + + RDenmark + + + CEstada + + + AOlsen + + + ANossaman + + 10.47205/jdss.2021(2-iv)74 + AV-2011-025 + + + + Journal of Development and Social Sciences + JDSS + 2709-6254 + 2709-6262 + + 2 + IV + 10 December 2010 + Pakistan Social Sciences Research Institute (PSSRI) + + + OIG Report Number + online database + Koch, R., Treakle, C., Denmark, R., Estada, C., Olsen, A., and Nossaman, A., "FAA Needs to Implement More Efficient Performance-Based Navigation Procedures and Clarify the Role of the Third Parties," OIG Report Number AV-2011-025, [online database], URL: http://www.oig.dot.gov/library-item/5464 [cited 10 December 2010]. + + + + + A graph-based approach to nominal terminal routing + + SZelinski + + 10.1109/dasc.2012.6382333 + + + 2012 IEEE/AIAA 31st Digital Avionics Systems Conference (DASC) + Williamsburg, VA + + IEEE + 2012 + + + Zelinski, S., "A Graph-Based Approach to Defining Nominal Terminal Routing," 31st Digital Avionics System Conference, Williamsburg, VA, 2012. + + + + + + diff --git a/file829.txt b/file829.txt new file mode 100644 index 0000000000000000000000000000000000000000..a2201bacde830102811a41b9b1d19e7fec4ea9de --- /dev/null +++ b/file829.txt @@ -0,0 +1,578 @@ + + + + +IntroductionAdvances in arrival scheduling and spacing have potential benefits of reducing aircraft delays and increasing airport arrival throughput.Schedulers use fixed arrival paths to estimate aircraft time-to-fly and assign them arrival slots.The size of an arrival slot is based on the required separation between leading and trailing aircraft plus a buffer used to mitigate arrival time uncertainty and reduce the probability of a separation violation.Thipphavong and Mulfinger [1] showed that delay and throughput benefits are more sensitive to these spacing buffers than the uncertainty itself.But the appropriate spacing buffer to achieve a desired controller intervention rate does decrease significantly with the uncertainty, allowing more benefit.To this end, benefit analyses have focused on showing the arrival time accuracy achievable for a specific technology or concept of operations employing a set of technologies [2][3][4][5][6][7].The larger benefit of the increased accuracy is unclear.Some studies have extrapolated or measured the benefit of scheduling and spacing precision on delay and throughput as compared to current day operations.The Terminal Area Precision Scheduling and Spacing System [8][9] verified in human-in-theloop simulations that precision scheduling and spacing tools can allow spacing buffers of 0.3-0.4nmi to accommodate 10-15% more traffic than current operations.However, the results are limited to a handful of simulated 100-minute traffic scenarios for a single airport.Ballin and Erzberger [10] analyzed the arrival spacing precision for 30 Dallas Fort Worth (DFW) rush periods and estimated that a 0.25 nmi separation buffer should allow for at least 15% more capacity than was observed.Again, this analysis focused on a single airport and since then, DFW operations have changed considerably.In addition to allowing reduced spacing buffers, precision scheduling and spacing concepts use defined fixed arrival paths such as Area Navigation (RNAV) and Required Navigation Performance (RNP) routing to achieve full benefit.Economic and environmental benefits (fuel usage and noise) of RNP have been quantified in case studies [11][12].The broader impact of fixed arrival routing and reduced separation buffers and the variability of this impact between different terminal areas are unclear.This paper characterizes the observed arrival spacing behavior of 29 runways belonging to 15 airports within 8 of the busiest terminal areas across the US for 32-60 days worth of traffic.Potential capacity increases due to reducing the observed 0.5 to 1.5 nmi spacing buffers to 0.3 nmi are estimated.Then fixed arrival paths are designed for each runway.The impact of these fixed paths and reduced buffers on average flight arrival time is then estimated. +3A5-2 +MethodArrival flights to multiple airport runways from multiple terminal areas were analyzed to encompass a wide range of traffic arrival environments, from metroplexes to super-hubs, across the National Airspace System (NAS).The traffic analysis was performed using recorded Terminal Radar Approach Control (TRACON) flight plan and track data.First, separation behavior was characterized for each airport runway.Theoretical runway arrival capacities were estimated based on observed current spacing behavior and expected spacing behavior afforded by precision scheduling and spacing concepts.This study also estimated the effect precision scheduling and spacing along fixed arrival routes would have on terminal area flight time.RNAV arrival routes were designed to each runway and new scheduled arrival times were compared with historical arrival times using historical aircraft arrival paths. +Scope of StudyFlight plan and track data from the eight busiest TRACON facilities in the NAS (Atlanta, Chicago, Denver, Dallas/Fort Worth, New York, Potomac, Northern California, and Southern California) were examined.The data covered between 32 and 60 24hour traffic samples distributed between February 2010 and early May 2010.Traffic samples were required to have uninterrupted TRACON track data between 0600 and 2200 local time.These same data were used in two previous studies where data statistics are described in more detail [13][14].This analysis focused on airport runways with relatively large numbers of tightly spaced instrument arrival operations.These included 29 runways from 14 of the 29 NPIAS (National Plan of Integrated Airport Systems [15]) identified large hubs and one satellite airport (TEB) diagramed in Figure 1.Locations of satellite airports for which track data was available but were not included in this analysis are also shown. +Separation Behavior from Track DataTo characterize the unique separation behavior of each runway, Ballin and Erzberger's [10] method was modified to accommodate samples outside rush periods and differentiate between behavior observed 3A5-3 during Visual Approach Conditions (VAC) and Instrument Approach Conditions (IAC).Robinson and Kamgarpour [13] describe the method used to determine each aircraft's landing runway and threshold crossing time from the track data.For each sequential pair of aircraft landing at the same runway, the position of the trailing aircraft at the time the leading aircraft crossed the runway threshold was extrapolated from the two track points on either side of this time.The observed separation was then calculated as the trailer's along-path distance from this point to the runway threshold.Sequential aircraft pairs to the same runway were segregated by their minimum in-trail separation requirements (standard radar separation [16][17]) determined by leader/trailer weight class.Table 1 shows the leader/trailer weight class to required separation mapping in nmi.For pairs requiring 3 nmi separation, in-trail separation at the runway threshold can be reduced to 2.5 nmi under certain conditions [16].However, these conditions were not considered in this analysis.Aircraft pairs with the same separation requirement were further segregated between VAC and IAC according to the airport's weather conditions during the quarter hour in which the trailer landed.The FAA's Aviation System Performance Metrics (ASPM [18]) data provide VAC and IAC state in quarter hour intervals.VAC and IAC separation histograms were generated with a 0.1 nmi separation bin size.Figure 2 shows an example separation histogram for characterizing separation behavior.Then a +/-3-sigma Gaussian kernel smoother with a 2 nmi span was used to smooth each histogram.The maximum bin of the smoothed histogram within a -1 to +5 nmi range of the required separation standard was assumed to be the observed target separation s.The curves generally followed the two-part separation model convolving a normal distribution of the positioning accuracy of the controller/pilot team with a Poisson distribution of the arrival gaps.The Poisson distribution can be large due to low demand periods, which can shift the mean far to the right of the maximum.Therefore, the standard deviation σ from s was calculated using only observed separations (not the smoothed curve) less than or equal to s.This is shown as an arrow pointing to the left of the target separation in Figure 2. Let r be the minimum required separation.The separation buffer b was then s-r.These separation behavior metrics were calculated the same way for VAC and IAC.The subscripts VAC and IAC differentiate metrics by weather condition in the results section. +Figure 2. Example Separation HistogramA final behavior metric, safety margin, was calculated as x=b/σ.It was designed as a simple proxy for separation conformance.Let n s be the number of separations less than or equal to s and let n r be the number of separations less than r.The empirical separation conformance is given by c e =1-(n r /n s ), that is the ratio of observed separations to the left of s that conform to the required separation minimum r.The analytical probability of separation conformance is given by the Gauss error function c a =erf(x/√2).Assuming that the curve to the left of s has a normal distribution, c a was expected to be similar to c e .With this simplification, if 95% conformance was required to be considered safe, then 3A5-4x would need to be at least 2. Values of x below 2 would quickly fall out of acceptable conformance.The above metrics require a significant number of high arrival rate operations to generate a meaningful characterization.The 29 runway subset for this analysis was chosen for their relatively high values of n s in IAC. Figure 3 shows the IAC n s at each of these runways for required separations of 3, 4, and 5 nmi.The numbers of aircraft pairs are shown on a logarithmic scale. +Figure 3. Number of Instrument Arrival Separations ≤ s (n s )The most frequent required separation at any runway is 3 nmi, with much fewer instances as the separation requirement increases.This is due to a preponderance aircraft of large weight class.The number of instances for required separations greater than 5 nmi were less than 10 at all but two runways at r=6 nmi and so are not shown in Figure 3.Of the runways shown, 8 have fewer than 9 instances for r=5 nmi and the separation histograms were too sparse to represent a normal distribution.For the remaining 21 runways for r=5 nmi, and all 29 runways for r=3 or 4 nmi, the arrival separations less than or equal to s have a fairly normal distribution as demonstrated by how similar the analytical separation conformance is to the empirical separation conformance in Figure 4. Figure 4 shows the differences between empirical c e and analytical c a separation conformance by required separation.On average the empirical ratio of separations conforming to the required minimum is higher than the analytical probability by just 0.02 (on a 0 to 1 scale).Therefore, the distributions are fairly normal and the safety metric x is a good proxy for separation conformance.These separation behavior metrics were calculated the same way for VAC and IAC.The subscripts VAC and IAC are used to differentiate metrics by weather condition in the results section. +Figure 4. Separation Conformance Comparison +Runway Arrival CapacityRunway arrival capacity is the estimated number of arrival slots that can be scheduled to a runway within a given amount of time.Average slot size can vary depending on the typical landing speeds and mix of aircraft weight class at each runway, as well as the spacing buffer used.For each aircraft, the desired spacing with its leading aircraft is computed as the required separation plus a given buffer.The time at which the aircraft reaches this along path distance 3A5-5 from the runway threshold is calculated by interpolating between track points on either side of this along-path distance.This desired spacing time is then subtracted from the runway threshold crossing time to get the scheduling slot size for the flight.The average slot size for each runway is then divided into 15 minutes to get a quarter hourly arrival capacity. +Fixed Routing and Path Stretch DelayPrecision scheduling concepts require fixed arrival routes such as RNAV and RNP but very few arrival routes have been defined all the way to the runway in today's system.Therefore, fixed arrival routes to each runway studied were designed for the purposes of this analysis.Most RNAV/RNP procedures designed in practice overlay existing routes to avoid lengthy environmental reviews [19].Fixed arrival routes for this study were designed to follow the nominal flow of arrival traffic.A graphbased trajectory bundling algorithm [20] was used to identify the most consistently used paths to each runway.Separate routes were generated for turboprops and jets where it was clear their nominal paths deviated significantly.In some cases the resulting routes curved unnecessarily in places due to consistently used path stretch maneuvers and these route segments were smoothed out manually.A flight was assigned on a fixed route to its arrival runway at the track point that first came within 3 nmi of the fixed route.From this point, the flight was directed to the next downstream waypoint along the route and then followed the route to the runway.The flight was expected to follow a deceleration profile similar to that of its original trajectory as it followed the fixed route.To model this assumption, first the original flight track was segmented using the closest track point to each fixed route waypoint.Original flight track segments were then mapped to fixed route segments using these closest points.Figure 5 illustrates the path stretch delay calculation for a sequence of mapped closest flight track and fixed route points labeled 1 through 5. Let p ij and q ij be mapped original flight track (orange) and fixed route segments (green) respectively.Let t(p ij ) and t(q ij ) be the respective times to travel each segment, and let l(p ij ) and l(q ij ) be the respective lengths of each segment.Then t(q ij )=t(p ij )l(q ij )/l(p ij ).The sum of the differences in flight time (t(p ij )-t(q ij )) across each mapped segment is the path stretch delay d.Note that path stretch delay may be negative if the original flight path is sufficiently shorter than the fixed route. +Figure 5. Example Path Stretch Delay Calculation +Fixed Route Scheduling and Time SavedTo determine the effect of fixed routing and scheduling buffers on terminal area delay, flights were rescheduled according to their fixed route Estimated Time of Arrival (ETA).The fixed route ETA was calculated as Observed Time of Arrival (OTA) minus the path stretch delay.The flights were then re-sequenced according to their fixed route ETAs, and delay was applied as Scheduled Time of Arrival (STA) assigned to place flights within arrival slots appropriately sized to meet the required separation with respect to the leading aircraft plus a given buffer.The time saved or reduced path stretch delay by using the fixed path and given buffer is then OTA-STA.Note that time saved cannot exceed path stretch delay.The time saved is somewhat conservative because flights with path stretch delay may have been slowed as well and the path stretch delay calculation does not compensate for this, but assumes a similar speed profile to the original flight trajectory. +Results +Separation Behavior by Required SeparationQuartiles for IAC and VAC separation buffers, standard deviations, and safety margins by required separation are shown in Figure 6 as box and whisker plots.From top to bottom the box and whisker 3A5-6 divisions represent max, 75%, 50%, 25%, and min values for each set of data.The quartiles for the 3 and 4 nmi required separations represent all 29 runways.The quartiles for 5 nmi required separation represent only the 21 runways for which IAC n s was at least 9. +Figure 6. Separation Metrics by Required SeparationThe mean IAC separation buffer is approximately 1 nmi with an increasing trend as required separation increases.This additional buffer compensates for the increasing IAC standard deviation.It even tends to overcompensate as can be seen by the increasing IAC safety margin.VAC standard deviation also increases with required separation but it tends to be higher than IAC standard deviation.This means that a single target separation is less often accurately achieved under VAC.Unlike the IAC, VAC buffers remain fairly stable as standard deviation increases so there is no compensation.Under IAC, buffers are never below 0.4 mi, whereas the VAC buffers can be negative.This means that the VAC target separation can be below the IAC required separation.Even though the VAC safety margin is quite low and decreases as required separation increases, this does not mean that these operations are unsafe.The VAC safety margin is calculated with respect to IAC requirements and so this metric merely shows how different the separation behavior is between IAC and VAC. +Separation Behavior by RunwayIndividual separation behavior by runway is compared for 3 nmi required separation as most aircraft pairs fall in this category.Figure 7 shows separation behavior metrics for each individual runway for 3 mi required separation.The blue and purple columns represent IAC and VAC target separation (s) respectively, with single tail standard deviation (σ) whiskers.The 3 nmi required separation (r) is highlighted in yellow so that the part of the target separation columns protruding above can be visualized as the buffer.The IAC safety margins (x) are shown as red diamonds to visualize runways' need to reduce standard deviation or freedom to reduce buffers relative to each other.In general, runways within the same TRACON tend to behave similarly.DEN and SFO runways have the largest x IAC because they have the smallest IAC standard deviations but among the largest separation buffers.They may benefit from advanced scheduling that reduces the buffers.DFW also has fairly high spacing buffers but they are scaled appropriately to their standard deviations.DFW's x values between 1.7 and 2.0 equate to c a probabilities between 0.92 and 0.95, which is quite reasonable.More precision would enable DFW to further reduce its separation buffers.Airports in the remaining TRACONs, particularly LGA and IAD are not using separation buffers that sufficiently compensate for their standard deviations.The significant benefit of increased precision at these heavily strained airports may be to reduce controller workload and separation violations.By contrast, the two runways with the largest IAC buffers and standard deviations, TEB 06 and SAN 27, have the lowest arrival traffic volume among the runways analyzed within their TRACONs. +3A5-7They may be neglected due to higher priority airports nearby. +Figure 7. Separation Metrics by RunwayIn general, IAC target separations are larger than VAC.The difference between IAC and VAC is negligible for DEN, DFW, JFK, DCA, IAD, and LAX, suggesting that advanced scheduling would provide similar benefits under both IAC and VAC.In contrast, ATL has much lower target separations in VAC than IAC and so very little if any benefit would be gained during VAC using IAC separation requirements.All four ATL runways analyzed can use 2.5 nmi separation which was not considered, so their actual buffer may be 0.5 nmi larger.Only three runways analyzed (ORD 09R, SFO 19L and 28R) have VAC target separations that are larger than IAC.This may be due procedural constraints.ORD 09R handles arrivals and departures most of the time.However, it may be used mostly for arrivals in IAC. Figure 8 shows smoothed VAC and IAC spacing histograms for ORD 09R.Assuming the shape of the VAC curve is due to normal shared arrival/departure operations, the IAC curve appears to be a convolution of dedicated arrival operations and shared arrival/departure operations.The lighter blue curves show a possible deconvolution of IAC where 0.35VAC (the VAC curve multiplied by 0.35) represents the shared arrival/departure behavior and the remainder (IAC-(0.35VAC))represents the dedicated arrival behavior. +Figure 8. ORD 09R Spacing BehaviorSFO has two sets of closely spaced parallel runways that perform simultaneous arrivals in VAC but 19L and 28R perform single arrivals in IAC. Figure 9 shows smoothed arrival spacing histograms for SFO 28L and 28R separately and together (28LR).Individually, 28R and 28L receive a large number of arrivals in VAC.In IAC, 28R accepts the vast majority of arrivals.When the runways are analyzed together (28LR), a single peak appears for IAC but two peaks appear for VAC.The first peak represents coupled aircraft performing simultaneous operations and the second peak represents the spacing between sets of simultaneous operations.The histograms for 19R and 19L show a similar but less pronounced effect. +3A5-8 +Figure 9. SFO 28L and 28R Spacing Behavior +Runway Arrival Capacity EstimatesSwenson et al [8] successfully demonstrated safe IAC operations within a simulation of LAX using a 0.4 nmi scheduled buffer and recent tests of the same concept simulating Dallas Love Field (DAL) arrivals have used 0.3 nmi scheduling buffers.Figure 10 shows estimated quarter hour capacity for each runway using scheduling buffers of b VAC , b IAC , or 0.3 nmi.The percent change in 0.3 nmi buffer capacity from the estimated VAC and IAC capacity is also shown.Given that the observed IAC buffers are between 0.4 and 2 nmi as seen in Figure 7, it is not surprising that 0.3 nmi buffers are estimated to increase IAC capacity up to 30%.ATL VAC buffers are so low that using 0.3 nmi buffers would actually reduce capacity based on 3 nmi rather than 2.5 nmi required separation.Although VAC buffers and capacity estimates tend to be lower than IAC, many of the VAC buffers are above 0.3 nmi, resulting in estimated capacity increases around 20%.Many of these runways such as TEB may lack sufficient demand to achieve these theoretical capacities or they are affected by other constraints such as shared arrival/departure operations or crossing runway restrictions.Other airports such as EWR, LGA, and IAD show an increase over IAC capacity with little to no change in VAC capacity, which suggests that demand is not lacking and that with precision spacing, it may be possible to achieve VAC throughput under IAC. +Figure 10. Estimated Arrival Capacity +3A5-9 +Path Stretch Delay and Time Saved by RunwayEach flight was assigned to the first fixed route that came within 3 nmi of its original track, which was used to calculate the flight's path stretch delay and fixed route ETA.Flights were sequenced according to fixed route ETA and assigned scheduled arrival slots using spacing buffers ranging between 0 and 1.0 nmi in 0.1 nmi increments.Figure 11 shows the 5% trimmed mean path stretch delays and time saved for the range of scheduling buffers.The 5% trimmed mean is used rather than the pure mean to filter the effects of a few large outliers such as undetected go-arounds and other processing errors.The black rectangles mark the 5% trimmed mean path stretch delays for each runway.The 5% trimmed mean time saved results are represented by blue rectangles with width proportional to the buffer used.The 0.3 nmi buffer results for each runway are highlighted in green as this buffer size has been successfully tested in simulations of DAL arrival operations.As expected, time saved is inversely proportional to the buffer size.Some runways tend to be more sensitive to buffer size than others as can be seen in Figure 11 by the rapidly decreasing time saved results as the buffer size increases from 0 to 1.0 nmi.Runways at ATL, ORD, EWR, and LGA are particularly sensitive to buffer size.These runways have tightly packed schedules with visual arrivals often spaced closer than the minimum required spacing for instrument arrivals.Therefore, any increase in spacing buffer quickly compounds delay as each delay is shared by each flight in the tightly packed stream.Runways TEB 06 and SAN 27 are least sensitive to buffer size because both their IAC and VAC observed buffers were larger than 1.0 nmi, the upper range of buffer size tested.Figure 12 shows smoothed histograms of the time saved results for two runways with different sensitivity to buffer size.LGA 22 is very sensitive to buffer size, which can be seen in Figure 12 as the histograms shift to the left with increasing buffer size.In contrast, DEN 35R is less sensitive to buffer size with time saved results very similar to the path stretch delay.DEN 35R is also the runway with the most clearly visible histogram multi modality due to different amounts of typical path stretch time each flight can save, depending on which fixed path they follow.As can be seen in Figure 11 most runways have a positive path stretch delay.That is, in general, more flights are given a path stretch than a short cut relative to the fixed arrival routing identified.Figure 13 shows the flight tracks and fixed routing for the runways with lowest (MDW 04R) and highest (ORD 09L) path stretch delay.The dotted lines represent the lateral paths of historical tracks colorcoded by path stretch delay relative to the fixed routing represented by the blue nodes and links.The green tracks are at least 2 minutes shorter than the fixed route and the red tracks are at least 4 minutes longer than the fixed route.Even though the fixed routing for MDW 04R is consistent with published RNAV instrument approach procedures for that runway, most flights from the East take shorter paths, perhaps due to visual approaches.Relatively few flights with more than 4 minutes of path stretch delay can be seen holding from the West or performing a base turn from the East.However, ORD 09L has a large amount of path stretch delay.Holding patterns and S-turns can be seen near all the major entry points.But the main contributor to path stretch delay is a large extended base turn to the West of the runway.Traffic from all directions is affected by this inefficiency, even traffic from the West, which doubles back on itself to merge with the extended base turn. +Path Stretch Delay and Time Saved by Weather ConditionFigure 14 shows 5% trimmed mean path stretch delay and time saved for b=0.3 nmi separated by weather condition for each runway.In general, path stretch delay and time saved are higher for IAC than VAC.The difference between the path stretch delay and time saved is scheduled delay that must be absorbed by the TRACON through speed control or passed to the center in order to keep flight on their fixed routes.Figure 15 shows an estimation of 5% trimmed mean scheduled delay by subtracting time saved from path stretch delay by weather condition from Figure 14.In general, there is more scheduled delay in IAC than VAC.Fast-time simulations of DEN [7] estimated the amount of delay that could be absorbed by speed control to be on the order of 2-3 minutes depending on the length of the route.Most runways in Figure 15 have scheduled delay under 2 minutes.With the exception of ATL 10, ORD 09R and 10, and LGA 22, which all have delays over 3 minutes, relatively little delay should need to be absorbed in Center airspace in order to stay on fixed routes in the TRACON. +Runway Interdependency AnalysisThe above results are based on in-trail spacing restrictions alone.However, many runways may have other restrictions due to parallel arrival operations, shared arrival/departure operations, or crossing runways.In addition to arrival weather conditions, ASPM [18] denotes which runways were configured for arrival and departure operations per quarter hour.These data were analyzed for operations between 0600 and 2200 during the months of Feb -May 2010.Figure 16 shows stacked columns for IAC and VAC representing the percentage of time ASMP reported runways configured for arrivals, departures, or both in each weather state.DCA 01 and SAN 27 are always configured for shared arrival/departure operations.Other than a small percentage of time when MDW 04 is configured for dedicated departures, MDW 04 performs mostly shared arrival/departure operations as well.As can be seen in Figure 1, many of the runways analyzed cross other runways or are very close (2500 ft) to a parallel runway.Active crossing or parallel runways require coordinated operations and impose additional restrictions that affect spacing behavior.The stacked columns of Figure 17 show the relative percentage of time when the given runway is configured for arrivals and one or more of its crossing runways is configured for arrivals, departures, or no crossing runways are active.JFK 04L and DCA 01 are relatively free of crossing runway activity.The remaining runways shown have one or more crossing runways configured for departures most of the time.In general, crossing runway activity is more common in VAC. +Figure 17. Crossing Runway Configurations +3A5-13The stacked columns of Figure 18 show the relative percentage of time when the given runway is configured as a single arrival runway (it's parallel runway is inactive) or if both parallel runways are active arrival runways. +Figure 18. Parallel Runway ConfigurationsOnly SFO and LAX runways configure both parallels for arrivals at the same time and LAX's configures both parallels for arrivals mostly in VAC.Even though SFO runways are configured for parallel arrivals most of the time in VAC and IAC, Figure 9 shows that SFO 28L rarely ever used for arrivals in IAC.This inconsistency illustrates that the above runway configuration analysis only identifies possible (not actual) sources of procedural constraints that could affect spacing behavior.This analysis also identifies runways that are likely NOT affected by these kinds of procedural constraints because they are configured for dedicated arrival operations without active crossing or closely spaced parallel runway most of the time.These include all the runways analyzed at ATL, DEN, DFW, EWR, JFK, and IAD in VAC and IAC, as well as LAX runways in IAC only. +ConclusionsThis paper discussed an analysis of spacing behavior across 29 runways from 8 top TRACONs and estimated the benefits and impacts of precision scheduling and spacing along fixed arrival routes.Likely candidate airports and runways that would benefit from precision scheduling and spacing concepts were identified.The spacing behavior and estimated capacity increases by runway suggest that all runways analyzed would benefit from a concept that would safely allow buffers to be reduced to 0.3 nmi (at least in IAC).The analysis reinforced how different spacing behaviors can be between TRACONs, airports, and even individual runways, and that evaluations of concepts adapted to a specific site cannot be arbitrarily extended to another site.In general, increasing separation buffers for increased required separation tend to overcompensate for the increasing standard deviations.Precision scheduling and spacing concepts would remove this unnecessary extra spacing from aircraft already widely spaced due to wake hazard and make it easier for controllers to manage these less frequently occurring aircraft pairs.Possible procedural constraints affecting spacing due to shared arrival/departure operations, crossing runways, and closely spaced parallel approaches must be considered.Of the runways that do not appear to be affected by such procedures, reducing buffers to 0.3 nmi could increase capacity 10-20% at ATL, DEN, DFW, EWR, JFK, and LAX in IAC, and at DEN, DFW, and JFK in VAC.With the exception of LAX, all these airports have at least one runway with significant path stretch delay and could benefit from adhering to fixed arrival routing.For half of ATL and DEN runways and all DFW, and JFK runways studied, some of this delay can be reduced through precision scheduling as scheduling buffers are reduced and throughput increases.DFW and JFK runways have less than 2 minutes of scheduled delay, which may be absorbed in the TRACON with speed control.ATL 10 and DEN 35R have higher scheduled delays and so they would likely need to pass 1-2 minutes of this delay to the center.Having identified likely candidate airports and runways that would benefit from precision scheduling and spacing concepts, the next step should be to conduct more detailed analyses and simulations on these runways.A significant number of runways with possible procedural constraints such as shared arrival/departure operations and active crossing or closely spaced parallel runways showed potential benefits.This suggests future research is needed for integration and evaluation of precision arrival scheduling and spacing concepts with these procedures.Figure 1 .1Runway Diagrams +Figure 11 . 10 Figure 12 .111012Figure 11.Path Strech Delay and Time Saved by Runway +Figure 13 .13Figure 13.MDW 04R and ORD 09L Tracks +Figure 14 .14Figure 14.Path Strech Delay and Time Saved by Weather Condition +Figure 16 .16Figure 16.Shared Arrival/Departure Runway Configurations Runways ORD 09R and 10 are shared between arrivals and departures roughly twice as often as they are dedicated for arrivals in IAC.In VAC, ORD 10 continues this trend.However, ORD 09R is hardly ever used for dedicated arrival operations in VAC, accounting for its vast difference in spacing behavior between IAC and VAC.LAX 24R and 25L are configured for shared arrival/departure operations twice as often as dedicated arrivals in VAC only.In IAC, they are dominated by dedicated arrival operations.Runway LGA 04 is dominated by dedicated arrivals in IAC and dedicated departures in VAC with very little shared operations.All remaining runways analyzed are dominated by dedicated arrival operations in both VAC and IAC. + + + + + +Table 1 . Minimum Required Separation Trailer Super Heavy B757 Large Small1Super 668810Heavy 44556B757 44445LeaderLarge 3 Small 33 33 33 34 3 + + + +Email AddressShannon.J.Zelinski@nasa.gov31st Digital Avionics Systems Conference October [14][15][16][17][18]2012 + + + + + + + Design Considerations for a New Terminal Area Arrival Scheduler + + JaneThipphavong + + + DanielMulfinger + + + AlexanderSadovsky + + 10.2514/6.2010-9290 + + + 10th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Fort Worth, Texas + + American Institute of Aeronautics and Astronautics + 2010 + + + Thipphavong, J., D. Mulfinger, 2010, Design Considerations for a New Terminal Area Arrival Scheduler, 10 th AIAA Aviation Technology, Integration and Operations Conference, Fort Worth, Texas. + + + + + Simulation Results for Airborne Precision Spacing along Continuous Descent Arrivals + + BryanBarmore + + + WilliamCapron + + + TerenceAbbott + + + BrianBaxley + + 10.2514/6.2008-8931 + + + The 26th Congress of ICAS and 8th AIAA ATIO + Anchorage, Alaska + + American Institute of Aeronautics and Astronautics + 2008 + + + Barmore, B. E., T. S. Abbott, W. R. Capron, B. T. Baxley, 2008, Simulation Results for Airborne Precision Spacing along Continuous Descent Arrivals, 8 th AIAA Aviation Technology, Integration and Operations Conference, Anchorage, Alaska. + + + + + Performance of Airborne Precision Spacing under realistic weather conditions + + FrederickWieland + + + MichelSantos + + + WilliamKrueger + + + VincentEHouston + + 10.1109/dasc.2011.6095981 + + + 2011 IEEE/AIAA 30th Digital Avionics Systems Conference + Seattle, Washington + + IEEE + 2011 + + + Wieland, F., M. Santos, W. Krueger, V. E. Houston, 2011, Performance of Airborne Precision Spacing Under Realistic Weather Conditions, 30th Digital Avionics System Conference, Seattle, Washington. + + + + + Arrival Management with Required Navigation Performance and 3D Paths, 7th USA + + AHaraldsdottir + + + JScharl + + + MBerge + + + ESchoemig + + + MCoats + + + + Europe Air Traffic Management R&D Seminar + + 2007 + Barcelona, Spain + + + Haraldsdottir, A., J. Scharl, M. Berge, E. Schoemig, M. Coats, 2007, Arrival Management with Required Navigation Performance and 3D Paths, 7th USA/Europe Air Traffic Management R&D Seminar, Barcelona, Spain. + + + + + Analysis of Arrival Management Performance with Aircraft Required Time of Arrival Capabilities + + AslaugHaraldsdottir + + + JulienScharl + + + JanetKing + + + EwaldSchoemig + + + MatthewBerge + + 10.2514/6.2008-8932 + + + The 26th Congress of ICAS and 8th AIAA ATIO + Anchorage, Alaska + + American Institute of Aeronautics and Astronautics + 2008 + + + Haraldsdottir, A., J. Scharl, J. King, E. G. Schoemig, M. Berge, 2008, Analysis of Arrival Management Performance with Aircraft Required Time of Arrival Capabilities, 26th International Congress of the Aeronautical Sciences, Anchorage, Alaska. + + + + + Arrival Management Architecture and Performance Analysis with Advanced Automation and Avionics Capabilities + + AslaugHaraldsdottir + + + JulienScharl + + + JanetKing + + + MatthewBerge + + 10.2514/6.2009-7006 + + + 9th AIAA Aviation Technology, Integration, and Operations Conference (ATIO) + Hilton Head, South Carolina + + American Institute of Aeronautics and Astronautics + 2009 + + + Haraldsdottir, A., J. Scharl, J. King, M. Berge, 2009, Arrival Management Architecture and Performance Analysis with Advanced Automation and Avionics Capabilities, 9th AIAA Aviation Technology, Integration and Operations Conference, Hilton Head, South Carolina. + + + + + Terminal area arrival management concepts using tactical merge management techniques + + AslaugHaraldsdottir + + + JanetKing + + + JulienScharl + + 10.1109/dasc.2010.5655508 + + + 29th Digital Avionics Systems Conference + Salt Lake City, Utah + + IEEE + 2010 + + + Haraldsdottir, A., J. King, J. Scharl, 2010, Terminal Area Arrival Management Concepts Using Tactical Merge Management Techniques, 29th Digital Avionics System Conference, Salt Lake City, Utah. + + + + + Design and Evaluation of the Terminal Area Precision Scheduling and Spacing System, 9th USA + + HNSwenson + + + JThipphavong + + + ASadovsky + + + LChen + + + CSullivan + + + LMartin + + + + Europe Air Traffic Management R&D Seminar + + 2011 + Berlin, Germany + + + Swenson, H. N., J. Thipphavong, A. Sadovsky, L. Chen, C. Sullivan, L. Martin, 2011, Design and Evaluation of the Terminal Area Precision Scheduling and Spacing System, 9th USA/Europe Air Traffic Management R&D Seminar, Berlin, Germany. + + + + + Evaluation of the terminal area precision scheduling and spacing system for near-term NAS application + + JThipphavong + + + HSwenson + + + LMartin + + + PLin + + + JNguyen + + 10.1201/b12321-12 + + + Advances in Human Aspects of Aviation + San Francisco, California + + CRC Press + 2012 + + + + Thipphavong, J., H. Swenson, L. Martin, P. Lin, J. Nguyen, 2012, Evaluation of the Terminal Precision Scheduling and Spacing System for Near- Term NAS Application, 1st International Conference on Human Factors in Transportation, San Francisco, California. + + + + + Potential Benefits of Terminal Airspace Traffic Automation for Arrivals + + MarkGBallin + + + HeinzErzberger + + 10.2514/2.4324 + + + Journal of Guidance, Control, and Dynamics + Journal of Guidance, Control, and Dynamics + 0731-5090 + 1533-3884 + + 21 + 6 + + 1998 + American Institute of Aeronautics and Astronautics (AIAA) + + + Ballin, M., H. Erzberger, 1998, Potential Benefits of Terminal Airspace Traffic Automation for Arrivals, Vol. 21, No. 6, Journal of Guidance, Control, and Dynamics. + + + + + The impact of advanced technology on Nex Generation Air Transportation: A case study of Required Navigation Performance (RNP) + + YoonjinYoon + + + MarkHansen + + 10.2514/6.2007-6697 + + + AIAA Modeling and Simulation Technologies Conference and Exhibit + Hilton Head, South Carolina + + American Institute of Aeronautics and Astronautics + 2007 + + + Yoon, Y., M. Hansen, 2007, The Impact of Advanced Technology on Next Generation Air Transportation: A Case Study of Required Navigation Performance (RNP), AIAA Modeling and Simulation Technologies Conference and Exhibit, Hilton Head, South Carolina. + + + + + Evaluation of the Potential Environmental Benefits of RNAV/RNP Arrival Procedures + + DelriMuller + + + PayunaUday + + + KarenMarais + + 10.2514/6.2011-6932 + + + 11th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Virginia Beach, Virginia + + American Institute of Aeronautics and Astronautics + 2011 + + + Muller, D., P. Uday, K. B. Marais, 2011, Evaluation of the Potential Environmental Benefits of RNAV/RNP Arrival Procedures, 11th AIAA Aviation Technology, Integration and Operations Conference, Virginia Beach, Virginia. + + + + + Benefits of Continuous Descent Operations in High-Density Terminal Airspace Considering Scheduling Constraints + + JohnRobinson Iii + + + MaryamKamgarpour + + 10.2514/6.2010-9115 + + + 10th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Fort Worth, Texas + + American Institute of Aeronautics and Astronautics + 2010 + + + Robinson III, J.E., M. Kamgarpour, 2010, Benefits of Continuous Descent Operations in High- Density Terminal Airspace Under Scheduling Constraints, 10th AIAA Aviation Technology, Integration and Operations Conference, Fort Worth, Texas. + + + + + Comparison of Trajectory Synthesis Algorithms for Monitoring Final Approach Compression + + JohnRobinson Iii + + + OusmaneDiallo + + + RonaldReisman + + 10.2514/6.2011-6900 + + + 11th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Virginia Beach, Virginia + + American Institute of Aeronautics and Astronautics + 2011 + + + Robinson III, J.E., O. Diallo, R.J. Reisman, 2011, Comparison of Trajectory Synthesis Algorithms for Monitoring Final Approach Compression, 11th AIAA Aviation Technology, Integration and Operations Conference, Virginia Beach, Virginia. + + + + + U.S. Airport Infrastructure and Sustainable Aviation Fuel + + KristiMoriarty + + + AllisonKvien + + 10.2172/1768316 + + + + Federal Aviation Administration, 2010, 2011-2015 National Plan of Integrated Airport Systems Report + + Office of Scientific and Technical Information (OSTI) + + + + Federal Aviation Administration, 2010, 2011- 2015 National Plan of Integrated Airport Systems Report, http://aspm.faa.gov. + + + + + U.S. Airport Infrastructure and Sustainable Aviation Fuel + + KristiMoriarty + + + AllisonKvien + + 10.2172/1768316 + + + + Federal Aviation Administration + + Office of Scientific and Technical Information (OSTI) + 2012 + + + Section 5. Radar Separation + Federal Aviation Administration, 2012, Air Traffic Control FAA Order JO 7110.65U Chapter 5. Section 5. Radar Separation, http://www.faa.gov. + + + + + U.S. Airport Infrastructure and Sustainable Aviation Fuel + + KristiMoriarty + + + AllisonKvien + + 10.2172/1768316 + + + + Air Traffic Control FAA Order JO 7110.567 Interim Procedures for Airbus A388 Flights + + Office of Scientific and Technical Information (OSTI) + 2011 + + + Federal Aviation Administration + Federal Aviation Administration, 2011, Air Traffic Control FAA Order JO 7110.567 Interim Procedures for Airbus A388 Flights, http://www.faa.gov. 3A5-15 + + + + + U.S. Airport Infrastructure and Sustainable Aviation Fuel + + KristiMoriarty + + + AllisonKvien + + 10.2172/1768316 + + + + Aviation System Performance Metrics (ASPM) Airport Efficiency + + Office of Scientific and Technical Information (OSTI) + 2012 + + + Federal Aviation Administration + Federal Aviation Administration, 2012, Aviation System Performance Metrics (ASPM) Airport Efficiency, http://aspm.faa.gov. + + + + + FAA Needs to Implement More Efficient Performance-Based Navigation Procedures and Clarify the Role of the Third Parties + + RKoch + + + CTreakle + + + RDenmark + + + CEstada + + + AOlsen + + + Nossaman + + AV-2011- 025 + + + Federal Aviation Administration Report + + 2010 + + + Koch, R., C. Treakle, R. Denmark, C. Estada, A Olsen, A. Nossaman, 2010, FAA Needs to Implement More Efficient Performance-Based Navigation Procedures and Clarify the Role of the Third Parties, Federal Aviation Administration Report AV-2011- 025. + + + + + A graph-based approach to nominal terminal routing + + SZelinski + + 10.1109/dasc.2012.6382333 + + + 2012 IEEE/AIAA 31st Digital Avionics Systems Conference (DASC) + Williamsburg, Virginia + + IEEE + 2012 + + + Zelinski, S., 2012, A Graph-Based Approach to Defining Nominal Terminal Routing, 31st Digital Avionics System Conference, Williamsburg, Virginia. + + + + + + diff --git a/file830.txt b/file830.txt new file mode 100644 index 0000000000000000000000000000000000000000..a8cddf82267e4fd579afa9d0b31dfa31d1015b26 --- /dev/null +++ b/file830.txt @@ -0,0 +1,278 @@ + + + + +IntroductionAs the volume of en-route flight traffic increases, the problem of increasing or better utilizing airspace capacity is of growing concern.Dynamic Airspace Configuration research is fo-cused on creating methods and algorithms that increase airspace capacity by redesigning airspace boundaries to reduce or redistribute controller workload and airspace complexity [1].Human factors studies show that airspace structure, consisting of dominant traffic flows and the points where these flows merge and intersect, helps increase capacity [2,3].Air traffic controllers use airspace structure to lower cognitive complexity and enable them to control increasing numbers of flights at one time.Airspace boundaries are designed to accommodate these airspace structures to minimize controller workload and maximize airspace capacity.Occasionally, inefficient airspace boundaries are redesigned to accommodate standard flows that may change slowly over time.Flows are identified manually by looking at historical flight tracks in the airspace.Then airspace designers use expert judgment from past experience to define and analyze new airspace boundaries.This process takes several months.Therefore, only standard flows under "nominal" situations are accommodated, and the airspace structures and boundaries remain fairly static.Static airspace structure may mitigate complexity in "nominal" situations, but in an "offnominal" situation, such as bad weather or congestion, these same structures could actually add to complexity [4].Reference [5] shows that reconfiguring airspace boundaries to complement severe weather reroutes can produce more evenly distributed and lower peak sector loads, improving workload distribution.This concept follows the same airspace boundary design process as today, but alternate airspace boundaries are designed to accommodate a set of preconceived routing scenarios from the National Severe Weather Playbook.The National Severe Weather Playbook is a large set of rerouting scenarios designed to adapt to changes in the weather.The extent to which this concept can adapt to "off-nominal" situations is still fairly rigid.A method of defining dynamic airspace structure elements derived from forecast trajectories is needed to allow more flexible airspace boundary design.This paper focuses on the points at which flows cross and merge.These points will henceforth be referred to as Òcritical points.Ó A method is introduced for extracting critical points directly from flight trajectories.Critical points are identified by clustering individual crossing and merge points between a set of flight trajectories.These critical points along with other airspace constraints may be used to dynamically reconfigure airspace boundaries and improve controller workload distribution.An outline of the paper is as follows.Section 2 explains the method of extracting critical points from flight tracks.Section 3 presents an analysis of critical points extracted from current day traffic with respect to airway intersections.Today's standard flows tend to adhere the static airways that are used to define flight plans.Therefore, it is expected that critical points derived from current day flight tracks will tend to be at airway intersection points.In addition, an analysis of a critical point's relative position to current airspace boundaries is performed to establish baseline sector design criteria. +Extracting Critical Points From Flight TrajectoriesThis section describes a method for extracting critical points from flight trajectories.Once flight tracks are extracted from raw data, pairs of flight tracks are compared to locate individual merging and crossing intersections.These merging and crossing points are then clustered separately to form merging and crossing critical points. +Identifying Trajectory Intersection PointsEvery pair of flight tracks is compared to find individual crossing and merging intersection points.Figure 1 in path A, the closest perpendicular distance to a segment on path B is computed.If this distance is less than one mile, the point is a candidate crossing or merge point.If there are three or less adjacent candidates, the point closest to the compared path is considered a crossing point.In Figure 1, only point A3 is closer than a mile to path B and so it is considered a crossing point.If there are more than three adjacent intersection point candidates, the two flights are considered to be following the same route for some time.If both paths begin at least one point before they occupy the same route, the first candidate is considered a merging point.Figure 2 shows a few more examples of perpendicular distance graphs resulting in different crossing and merging intersection results.There are three adjacent intersection candidates in Figure 2(a) within the threshold distance shown in red.The point with the minimum perpendicular distance is considered to be the crossing point.In 2(b), there are more than three adjacent intersection candidates.There is a point before the first candidate which has a perpendicular distance greater than one mile.Therefore this first candidate point is considered a merging point.There are also more than three adjacent intersection candidates in Figure 2(c) but the first candidate is also the first track point.Flight tracks may be filtered by altitude, regional location, or time before comparing.The flight path may have been truncated at point A1 and so A1 may not be a true merging point.Diverging points are not analyzed in this paper as they are not considered critical points for controller workload.Point A4 in Figure 2(c) would be considered a diverging point.The comparison is such that, intersection points differ slightly when path A is compared to path B than when path B is compared to path A. Both intersection points are considered when identifying critical points.Figure 3 shows an example of how three flight tracks are compared to one another.Lines A (green), B (red), and C (blue) are three sets of flight tracks compared.The orange circles are the intersection candidate points within a mile of the compared track.Squares and diamonds are the points identified as crossing or merging respectively.Notice how one set of crossing points is chosen as the closer of two adjacent intersection candidates when A and C are compared.Also notice how all points downstream from the merging points of A and B are intersection point candidates.Note that when comparing flight A to flight B, direction is only important for flight A in locating merging points.For example if B in figure 3 flowed in the reverse direction, A would still find a merging point when comparing to B and produce a distance profile similar to figure 2(b).But when B is compared to A, it will produce a distance profile similar to 2(c) and so no intersection point will be found. +Merging Trajectory Intersections into Critical PointsOnce every pair of flight trajectories has been compared, the intersection points must be merged into critical points.This is done by iterating over the following two steps.1) A center of mass calculation is used to cluster points within a square region that sweeps across the space of original points and weight the resulting clustered points.2) Lower weight clustered points are filtered if they are near enough to a clustered point with higher weight.This two step algorithm is repeated, refining the results of the previous iteration.A steady state is achieved when the clustering results are identical to the results of the previous iteration.These steady state clustered points are considered critical points. +Clustering PointsThe center of mass calculation finds the average coordinate position of all points within a square region weighted by the point weight.The method of identifying trajectory intersections is such that there may be multiple intersections found at the same coordinates.The first iteration of the clustering algorithm is performed on these trajectory intersection coordinates weighted by the number of trajectory intersections they represent.LetΛ(c) = [λ 1 (c), λ 2 (c), ..., λ n (c)] be an array of longitudes and Φ(c) = [ϕ 1 (c), ϕ 2 (c), ..., ϕ n (c)] be an array of latitudes. p i (c) = (λ i (c), ϕ i (c))∀i ∈ {1, 2, ..., n} is the ith point of n points within a square region with center c.Let W (c) = [w 1 (c), w 2 (c), ..., w n (c)] be the array of weights associated with these points.The center of mass clustered point coordinates for the square region centered at c is given asp (c) = (λ (c), φ (c))(1)λ (c) = Λ(c) •W (c) ∑ n i=1 w i (c)(2)ϕ (c) = Φ(c) •W (c) ∑ n i=1 w i (c)(3)This calculation is repeated for other square regions, sweeping across the space of points to be clustered.The sweep rate determines how much the sqare region is shifted for each new center of mass calculation.Let r be the sweep rate given byr = s/ρ, ρ ∈ {2, 3, 4, ...}(4)where s is the length of a side of the square region and ρ is the sweep ratio.The ρ must be an integer greater than or equal to 2 so that every point is used in the same number of center of mass calculations and at least more than once.This enables the same point to be tested within multiple different s by s regions.In fact, each intersection point is included in a center of mass calculation exactly ρ 2 times.There are ρ 2 different s by s square regions that overlap each r by r square region.This is illustrated in Figure 4 for ρ = 2.A single r by r region is shown in red along with 4 different overlapping s by s regions shown in blue.Note how any point within the r by r region shown will be used, in part, to calculate 4 different p (c) coordinates and their weights.In order to normalize the weight, the assigned weight of p (c) is given asw (c) = ∑ n i=1 w i (c) ρ 2 .(5) +Filtering Clustered PointsThe point clustering process above will always yield one point for every s by s region covering the point space.Some of the lower weight points must be filtered to give more emphasis to higher weight points in subsequent iterations of the algorithm and to reduce the final number of critical points produced.In some cases, different s by s regions may produce clustered points with identical coordinates.For example, this happens when the points within a single r by r region are the only points within the ρ 2 s by s regions that share that r by r region.The result is ρ 2 clustered points, each with 1 ρ 2 the combined weight of the points in the r by r region.Therefore, before lower weight clustered points are filtered, all p (c)s that share the same coordinates are merged into one point (p) with summed weight (w(p)).The total coordinate weight for clustered point coordinates, p, is given byw(p) = ∑ c w (c) ∀c s.t. p (c) = p.(6)Once coordinate weights have been adjusted in this way, lower coordinate weight points that are close enough to higher coordinate weight points are filtered.Of the ρ 2 center of mass calculations performed on these ρ 2 s by s square regions, filter all but the clustered point with the greatest coordinate weight.Figure 5(a) shows the points in a 4r by 4r region and (b) shows the clustered points that resulted from a center of mass sweep with ρ = 2.The size of each blue point represents it's weight.In (a) the red box shows the s by s region associated with center location 1 in red.Numbers 2 through 9 are placed in the center of the remaining 8 s by s regions that span this example.Notice that the shifting distance between s by s region centers is r.The resulting clustered points After these points are filtered, the only remaining clustered point is 1. +Algorithm IterationsThe algorithm iterates by creating new Λ(c), Φ(c), and W (c) arrays populated with the surviving clustered points from the previous iteration and their associated coordinate weights.Equations 5 and 6 play an important role in allowing the algorithm to repeat until it reaches a steady state.Sufficiently isolated points and their associated weights are unaffected by the algorithm and after several iterations, the algorithm will reach a steady state.Consider the case where a single intersection point p with weight w exists in an r by r region and it is the only point within the ρ 2 s by s regions that share that r by r region.This will results in ρ 2 clustered points at the same coordinates, each with weight w ρ 2 by Equation 5. Equation 6 will then sum these weights to equal the original weight.If the clustered point is sufficiently isolated from other clustered points as not to be fil-tered, it is ultimately unaffected by the algorithm iteration.In this way, the size of s determines how isolated a point must be and ρ determines the accuracy to which the point is placed.For the simple example shown in Figure 5, the algorithm reaches a steady state after the first pass.Figure 6 shows a larger example of merging and filtering intersection points over several sweeps of the algorithm.In this example, crossing and merging points were processed independently.The blue and green points represent crossing and merging points respectively.The point size represents the relative weight of the point with respect to other points plotted.The routes that produced the original intersection points according to the method discussed in section 2.1 are also plotted as black lines.Figure 6 +Analysis of Current Day Critical PointsIn order evaluate the method proposed in this paper, a set of critical points identified from historical track data are compared with current airway intersection points and sector boundaries.Comparing critical points to the current day route Major airway intersection points of the current route structure have a direct relationship with current sector design.Some considerations for sector design that are analyzed in this section include the number of large critical points or airway intersections in one sector [6] and their location with respect to sector boundaries. +Analysis ScopeOne could used planed flight trajectories from flight plan data or actual trajectories captured by historical track data to produce critical points.Track data were used because flight plan data are too closely tied to the current airway structure.Track data present more accurate representations of flight paths actually flown.Research has shown that controllers faced with a high workload tend to adhere to the route structure more closely [7].Therefore, it is expected that the critical points from tracks during the busiest hours of the day on a low weather day should follow the current airspace structure.The track data used to generate critical points were taken from Aircraft Situation Display to In-dustry "TZ messages" [8] for 4/17/2005, a low weather day.The track data were filtered to nation-wide cruise tracks above flight level 240 during the busiest four hours of the day, between 18:00 and 22:00 GMT.The sectors analyzed include 188 sectors that at least include flight levels 370 through 490.About 87% of the flight tracks considered are within this altitude range.Critical points for these tracks were generated using s = 0.2 degrees and ρ = 2.The algorithm reached a steady state in three iterations.Current airway intersection points were extracted from airways published in the FAAs resource aeronautical data effective March 17th 2005 [9].Any navigational aid or fix that appears in more than one airway is considered an intersection point.Note that airways alone with out flight traffic, have no directionality and so crossing and merging airway intersections are indistinguishable.Each intersection point is given a weight equal to the number of airways in which it appears.This number ranges from 2 to 22 intersecting airways.The airway intersection weight is not affected by traffic density whereas the critical point weight is.The airway intersecting weight is multiplied by the surrounding flight density (number of flights within a r(2ρ -1) = 0.3 squared region surrounding the r = 0.1 squared region containing the intersection) to make it comparable with critical point weight.Similarly, the critical point weight can be normalized by dividing it by its surrounding flight density to create a weight that is more comparable with the airway intersection weight.But critical point and airway intersection results using this weight are less comparable.Therefore, the un-normalized weights reflecting traffic density are used in this analysis. +Merging vs. Crossing Critical PointsCritical points are generated separately for crossing and merging intersections.In general, critical points formed from crossings have much larger weights than critical points formed from merge points.This is because the method described in Section 2.1 detected far more crossing points than merge points.Recall that there is no minimum altitude difference when detecting crossing and merging points in this analysis.Merges are far more likely to occur at similar altitudes than crossings.Therefore, another analysis, detecting only intersections of flights within a minimum altitude difference of one another, may detect more merging critical points.The maximum weights for crossing critical points, merging critical points, and airway intersections are 724, 354, and 2,068, respectively.In order to more easily compare the critical points and airway intersection weights, they are rescaled to weights between 0 and 10.Points with higher weight have greater effect on controller workload.Figure 7 shows a histogram of numbers of critical points and airway intersections by weights.The figure is scaled to easily view the smaller numbers of points with weights of 2 and higher.Numbers with weights between 1 and 2 reach the low hundreds and numbers with weights between 0 and 1 are in the thousands.As seen in Figure 7, even with this scale adjustment, there are far fewer merging critical points with higher weight than crossing critical points.The crossing critical points are more similar in number to airway intersections.The few merging critical points that have substantial weight are very close to high weight crossing critical points but shifted slightly downstream of the general flow.This is due to flights cutting corners when turning from one airway to another, merging in reality farther downstream from the airway intersection.Only two merging critical points with weight greater than 1 could not be paired with a crossing critical point within 15 miles, in histogram bins 1 and 2.In general, merging critical points appear to get closer to their paired crossing critical points as weight increases.The merging critical point in histogram bin 7 is just 0.017 nautical miles away from it's paired crossing critical point.There is no correlation between the weights of crossing and merging critical point pairs.The three largest merging critical points with weights greater than 6 all appear in Jacksonville center.By contrast, the eleven largest crossing critical points with weights greater than 6 are spread throughout different East coast centers where traffic density is higher and more structured.Figure 8 tors have a high density of very structured routes with small intersection angles.Routes with small intersection angles are where large merge points are most likely to be detected because crossing routes at small angles may yield more than three adjacent points within one mile of each other, identifying the intersection as a merge point according to the method described in Section 2.1 . +Critical Points vs. Airway IntersectionsMerging and crossing critical points are paired separately with their closest airway intersection within 15 nautical miles.points with weight greater than 5 is just 1.2 miles.This shows that the method described in Section 2 does a good job of placing crossing critical points at airway intersections where the current day flows intersect.The exceptions seen in Figures 10 and 11 illustrate how major flows and critical points cannot always be inferred from todays static route structure alone, even during times when traffic is most likely to conform to this route structure.There is a 75% correlation between the weights of paired crossing critical points and airway intersections.But this high correlation is heavily influenced by the surrounding traffic density factor in airway intersection weights.The number of intersecting airways alone correlates only 35% with crossing critical point weights.Even though an airway intersection may intersect a large number of airways, the traffic flow around the intersection may not be close enough to form a critical point or not all of the intersecting airways might be used.Figure 12 shows the flight tracks and airways around a point with large number of intersecting airways.Notice how there are far more airways intersecting at this point than flight tracks utilize.Merging critical points do not show any trend between weight bin and distance to paired airway intersections and they have only a 51% correlation with paired airway intersection weights.Because of the stronger relationship between crossing critical points and airway intersections, merging critical points are not considered in the remaining analyses. +Sector AnalysisThis section analyzes the geographic relationship between critical point and airway intersection density within current sector boundaries.Three metrics that may be useful in designing sector boundaries around critical points are investigated.These metrics are the number of significantly weighted critical points in each sector, the closest distance of a critical point to the sector boundary, and the average distance between critical points within the same sector.Sector metrics are averaged over the subset of sectors that contain both critical points and airway intersections to make them comparable.Figure 13 shows the number of sectors containing crossing critical points, airway intersections, and both crossing critical points and airway intersections with a weight greater than a given threshold.The number of sectors compared quickly diminishes as the weight threshold is increased.Figure 14 The number of critical points in a sector may make a big impact on controller workload because each new critical point splits controller focus.Figure 15 shows the average and maximum number of crossing critical points and airway intersections in a sector.Only sectors containing both crossing critical points and airway intersections were included.The figure has been scaled to better show average numbers of points.Both the average and maximum number of crossing critical points and airway intersections per sector are very similar.This indicates that the weight scaling matches well between crossing critical points and airway intersections.These data help evaluate at what weight threshold the critical point can be considered a controller focal point within the sector.Typically, there should not be more than one or two major focal points in a sector.Both average and maximum numbers of crossing critical points and airway intersections per sector are way too high for a weight threshold of 0. The average numbers for a weight threshold of 1 are more reasonable but not for maximum numbers.Ignoring points with weights less than 2 reduces the numbers of points within todays sectorization enough to be considered controller focal points.The correlation between the numbers of crossing critical points and airway intersections in the same sector range from 54% to 73% for weight thresholds between 2 and 6.At higher weight thresholds, there are too few sectors to compare well.The distance of critical points from the sector boundary affects the time a controller has to assess and control a flight's impact on the intersection traffic.For each sector, let α be the average closest critical point or airway intersection distance (miles) to the sector boundary weighted by w(p).Let A be the average α for all sectors that had both critical points and airway intersections above a given weight threshold.Figure 16 for crossing critical points and airway intersections are very similar to each other.The highest distances are for points with weight thresholds 2 through 5 averaging around 19 miles.The main reason for placing airway intersections at least a minimum distance from a sector boundary is to allow controllers enough time to access and control a flight's impact on the intersection traffic.If the minimum distance to the sector boundary is not along a significant flow path, the metric is less meaningfull.Like the distance of critical points from the sector boundary, the average distance between critical points within the same sector may impact how a controller deals with a critical point.The closer critical points are together, the more likely control actions involving those critical points will be lumped together.The distance between critical points in this analysis are already affected by the choice of s in Section 2.2.Let β be the average distance between critical points within the same sector weighted by w(p).Let B be the average β for all sectors containing multiple points above a given weight threshold.Figure 17 shows crossing critical point and airway intersection Bs for weight thresholds between 0 and 5. Once again Bs are very similar between crossing critical points and airway intersections.All Bs are over 45 miles. +ConclusionsComparing critical points to the current day airway intersections helps calibrate their weights so they can be used in sector design with different flight trajectories.Then critical points based on forecast trajectories and constraints may be used to define dynamic airspace boundaries that change with the air traffic.This would minimize the need to impose traffic management restrictions and reduce controller workload.This analysis shows that critical points (where dominant flows cross and merge) generated from current day flight paths are comparable to airway intersections.This shows promise for the method of defining critical points presented in this paper.Crossing critical points far outnumbered and outweighed merging critical points.When critical points were scaled to a maximum weight of 10, the critical points with weight greater than 2 paired best with airway intersections and resulted in the most reasonable number of critical points within each sector.This analysis found average distances of critical point and airway intersection from sector boundaries around 19 miles and average distances between large weight points over 45 miles. +Copyright StatementThe authors confirm that they, and/or their company or institution, hold copyright on all of the original material included in their paper.They also confirm they have obtained permission, from the copyright holder of any third party material included in their paper, to publish it as part of their paper.The authors grant full permission for the publication and distribution of their paper as part of the ICAS2008 proceedings or as individual off-prints from the proceedings.Fig. 11Fig. 1 Example comparison between two tracks. +Fig. 22Fig. 2 Example comparison between two tracks. +Fig. 33Fig. 3 Example of how multiple crossing and merging intersection points are detected when three flight tracks are compared to one another. +Fig. 44Fig. 4 For ρ = 2, there are ρ 2 = 4 different s by s square regions overlapping a single r by r square region. +Fig. 55Fig. 5 Example of how intersection points are merged and filtered to form critical points. +(a) shows the original trajectory intersection points, and (b) and (c) show merged points after the first and second sweep of the algorithm respectively.Note in (a) how crossing points are far more numerous than merging points which results in much larger crossing points in (b) and (c) with respect to merging points.Also note how most of the point consolidation occurs in the first algorithm sweep. +Fig. 66Fig. 6 Example of how intersection points are merged and filtered over several center of mass sweeps to form critical points. +Fig. 77Fig. 7 Numbers of crossing and merging critical points and airway intersections by weight. +Fig. 88Fig.8The three largest merging critical points shown with crossing critical points and airway intersection points for three Jacksonville sectors. +Figure 9 Fig. 999Fig.9The percent of merging and crossing critical points paired with airway intersections within 15 miles. +Fig. 1010Fig. 10 Flight tracks and airways surrounding the unpaired crossing critical point near crossing airways without an intersection. +Fig. 1111Fig. 11 Flight tracks and airways surrounding the unpaired crossing critical point where the traffic flow does not follow an airway. +Fig. 1212Fig. 12 Flight tracks and airways in sector ZSE46. +Fig. 1313Fig. 13 Number of sectors containing crossing critical points, airway intersections, and both crossing critical points and airway intersections with weights greater than a given threshold. +Fig. 1515Fig.15The average and maximum number of crossing critical points and airway intersections for all sectors containing both types of points. +Fig. 1616Fig. 16 Crossing critical point and airway intersection As for different weight thresholds. +Fig. 17 Crossing critical point and airway intersection Bs for different weight thresholds. +Fig. 17 Crossing critical point and airway intersection Bs for different weight thresholds. + + + + + + + + + Initial Concepts for Dynamic Airspace Configuration + + ParimalKopardekar + + + KarlBilimoria + + + BanavarSridhar + + 10.2514/6.2007-7763 + + + 7th AIAA ATIO Conf, 2nd CEIAT Int'l Conf on Innov and Integr in Aero Sciences,17th LTA Systems Tech Conf; followed by 2nd TEOS Forum + Belfast, Northern Ireland + + American Institute of Aeronautics and Astronautics + 2007 + + + Kopardekar, P., Bilimoria, K., Sridhar, B. 2007. Initial concepts for dynamic airspace configura- tion. In 7th AIAA Aviation Technology, Integra- tion and Operations Conference, Belfast, North- ern Ireland. + + + + + Impact of operating context on the use of structure in air traffic controller cognitive process + + HJDavison + + + JMHiston + + + MDRagnarsdottir + + + LMMajor + + + RJHansman + + + + 5th USA/Europe ATM R&D Seminar + Budapest, Hungary + + 2003 + + + Davison, H. J., Histon J. M., Ragnarsdottir, M. D., Major, L. M. and Hansman, R. J. 2003. Im- pact of operating context on the use of struc- ture in air traffic controller cognitive process. In 5th USA/Europe ATM R&D Seminar, Budapest, Hungary. + + + + + Partitioning Complexity in Air Traffic Management Tasks + + MLCummings + + + ChrisTsonis + + 10.1207/s15327108ijap1603_3 + + + The International Journal of Aviation Psychology + The Int. J. of Aviation Psych. + 1050-8414 + + 16 + 3 + + 2005 + Informa UK Limited + Oklahoma City, Oklahoma + + + Cummings, M. L., Tsonis, C. G. and Cunha, D. C. 2005. Complexity mitigation through airspace structure. In 13th International Symposium on Aviation Psychology, Oklahoma City, Oklahoma. + + + + + ATC complexity and controller workload: trying to bridge the gap + + SAthenes + + + Averty + + + SPuechmorel + + + DDelahaye + + + CCollet + + + + International Conference on Human-Computer Interaction in Aeronautics, MIT + Cambridge, Massachusetts + + 2002 + + + Athenes, S., Averty., Puechmorel, S., Dela- haye, D., and Collet, C. 2002. ATC complex- ity and controller workload: trying to bridge the gap. In International Conference on Human- Computer Interaction in Aeronautics, MIT, Cam- bridge, Massachusetts. + + + + + "Airspace Playbook": Dynamic Airspace Reallocation Coordinated with the National Severe Weather Playbook + + AlexanderKlein + + + ParimalKopardekar + + + MarkRodgers + + + HongKaing + + 10.2514/6.2007-7764 + + + 7th AIAA ATIO Conf, 2nd CEIAT Int'l Conf on Innov and Integr in Aero Sciences,17th LTA Systems Tech Conf; followed by 2nd TEOS Forum + Belfast, Northern Ireland + + American Institute of Aeronautics and Astronautics + 2007 + + + Klein, A., Kopardekar, P., Rodgers, M. 2007. "Airspace playbook": dynamic airspace realloca- tion coordinated with the national severe weather playbook. In 7th AIAA Aviation Technology, In- tegration and Operations Conference, Belfast, Northern Ireland. + + + + + Introducing Structural Considerations into Complexity Metrics + + JonathanMHiston + + + RJohnHansman + + + GuillaumeAigoin + + + DanielDelahaye + + + StephanePuechmorel + + 10.2514/atcq.10.2.115 + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 10 + 2 + + 2002 + American Institute of Aeronautics and Astronautics (AIAA) + + + Histon, J. M., Hansman, R. J., Aigoin, G., Dela- haye, D. and Puechmorel, S. 2002. Introducing structural considerations into complexity met- rics". In Air Traffic Control Quarterly. + + + + + U.S. International Air Traffic Statistics. Information Management Branch, Transportation Systems Center, Kendall Square, Cambridge, Massachusetts 02142. 1982. Various paging + + HGBusing + + 10.1177/004728758302200138 + + + Journal of Travel Research + Journal of Travel Research + 0047-2875 + 1552-6763 + + 22 + 1 + + 2005 + SAGE Publications + + + Massachusetts Institute of Technology, Research Project under Prof. R. John Hansman + + + Busing, H. G. 2005. "Air Traffic Control Oper- ating Modes and the management of Complex- ity". International Center for Air Transportation, Massachusetts Institute of Technology, Research Project under Prof. R. John Hansman. + + + + + Best Practices: August 2000 + 10.1037/e331952004-001 + + August 2000 + Test accounts + Cambridge, MA + + + Report no. ASDI-FD-001 + Aircraft Situation Display To Industry: Func- tional Description and Interface Control Docu- ment. August 2000. Report no. ASDI-FD-001. Volpe Center Automation Applications Division, Cambridge, MA. + + + + + Warner, Edward Pearson (1894-1958), aeronautical engineer and aviation consultant + + RogerDLaunius + + 10.1093/anb/9780198606697.article.2001081 + + + Federal Aviation Administration. Aeronautical Information Services. National airspace system resource aeronautical data + + Oxford University Press + March 17, 2005 + + + Federal Aviation Administration. Aeronautical Information Services. National airspace system resource aeronautical data. Effective March 17, 2005. + + + + + + diff --git a/file831.txt b/file831.txt new file mode 100644 index 0000000000000000000000000000000000000000..a3cbe2296f2fb97baa7ab6c35f19c91676a5be74 --- /dev/null +++ b/file831.txt @@ -0,0 +1,238 @@ + + + + +I. INTRODUCTIONUrban Air Mobility (UAM) concepts and technologies are being developed to safely enable operations of small, electricpowered or hybrid, pilot-optional, vertical-takeoff-and-landing (VTOL) passenger and cargo aircraft at vertiport facilities in urban and suburban environments.To realize door-to-door time savings for trips as short as 20 miles, UAM concepts envision passengers flying VTOL aircraft that can be summoned "ondemand" between local vertiports close to their origin and destination [1].In order to be economically viable for operators, service providers and the travelling public, orders-of-magnitude more aircraft and vertiports than operate today would be required [2].Studies of potential UAM markets have shown that demand on popular destinations will require vertiports to have multiple touchdown and liftoff (TLOF) areas.A preliminary assessment of the number of TLOFs required to support a representative UAM demand network applied to Fort Worth, TX estimated a need for 98 TLOFs distributed among 25 locations during peak traffic [3].Although this estimate averages to four TLOFs per location, 75% of the total demand was captured by seven locations suggesting that the required TLOFs could be closer to ten for some locations.In order to maximize the throughput of TLOFs, aircraft will need to taxi (or be conveyed) to parking areas for turnaround (unloading, loading, battery charging etc.) to free up TLOFs for arrival and departure operations.Initial work in vertiport design [4] characterized relationships between numbers of TLOFs and parking areas and ranges of assumed time requirements for TLOF operations, taxi, and turnaround time at parking spaces.For example, five parking spaces per TLOF achieved maximum TLOF utilization based on 60 sec arrival or departure operations, 15 sec taxi time, and 300 sec turnaround time.It is likely that many of the highest demand locations will be in space constrained urban environments, requiring vertiport designs to maximize throughput within a compact surface footprint.This presents an interesting operational challenge for designing safe, efficient, and compact vertiport surface topologies.Vertiport designs must consider space requirements of surface features like TLOFs, parking gates, taxiways, and passenger accommodation areas.They must also consider airspace constraints which could affect arrival and departure paths to the vertiport, including surrounding tall buildings, obstructions, noise restrictions, neighboring traditional-controlled airspace, winds, zoning, etc.There has not been much work exploring vertiport design standards.For example, one study exploring landing accuracy considerations for UAM, showed that representative models of electric VTOL aircraft for UAM can land within a radius of 20-30 ft 95% of the time under general light turbulence conditions [5].This only scratches the surface of developing standards for vertiport TLOF space requirements.However, existing heliport design standards [6] provide an excellent starting point for exploring vertiport topology design.This paper presents several generic vertiport topology design approaches and evaluates their relative surface area utilization and operational efficiency while meeting safety driven spacing constraints derived from heliport design standards and subject matter expert interviews.Section II describes the assumed vertiport attribute and constraints as well as three surface topology design approaches based on how taxiways are used to connect TLOFs and parking pads.Section III presents the surface area utilization analysis of the three design approaches as the number of parking spaces per TLOF increases.Section IV presents a model for mapping wind heading and speed to feasible vertiport TLOF configurations for accommodating approach and departure operations.In Section V operational efficiency is analyzed under each feasible configuration from this model.Section VI discusses the relative advantages and disadvantages of the three design approaches, and the paper is concluded in Section VII. +II. VERTIPORT TOPOLOGY DESIGNS +A. Design AttributesThis section discusses assumptions and constraints driving the design attributes for this analysis.It is assumed that the density of operations at a single vertiport will be great enough to require multiple independent TLOFs.For this analysis, each vertiport design contains four TLOFs.Because many high demand vertiport locations will be within space constrained urban environments, it is assumed that a minimum surface footprint is desired.The most compact configuration of four TLOFs is a square.For this analysis, each vertiport design utilizes a square surface footprint with one TLOF at each corner.Placing TLOFs at the corners also maximizes the arcs of airspace available for independent approach and departure operations to/from each TLOF without constraining surface features.It is assumed that the area immediately surrounding the surface is free of airspace obstructions that would interfere with approach/departure paths to these corner TLOFs.In order to maximize the throughput of TLOFs, it is assumed that aircraft will taxi or be conveyed to parking areas for turnaround to free up TLOFs for approach and departure operations.Therefore, each vertiport design must include taxi paths from each TLOF to multiple parking spaces.Taxi paths may be placed between multiple TLOFs as well to give aircraft using a particular parking space the option of using different TLOFs depending on the operating conditions.Whether the vertiport is at ground level or raised, the surface will still require space for passengers to access the aircraft for loading and unloading.For each design approach considered, priority is given to maximizing parking spaces and accommodating the required taxi paths, with any remaining space given to passenger access.It is assumed that any centrally located passenger area may be accessible from below the vertiport surface. +B. Surface Feature ConstraintsIn the absence of vertiport design standards, existing heliport design standards are used to define surface feature spacing constraints as prescribed for general aviation VFR heliports in AC 150/5390-2C [6] as follows. +1) Approach/Departure PadsThere are three components of the approach/departure pad.The TLOF is the load-bearing surface on which the vehicle lands or takes off.For multiple rotor vehicles, the TLOF is recommended to have a minimum diameter of one tip-to-tip span (TTS) which is the distance between extreme edges of spinning rotors providing vertical lift.The Final Approach and Takeoff Area (FATO) surrounds the TLOF as the area over which the pilot completes the final phase approach to land or initiates takeoff.The FATO is recommended to have a minimum diameter of 1.5x the largest dimension of the vehicle.Furthermore, FATOs must be separated by at least 200 ft to conduct simultaneous operations.The FATO Safety Area (FSA) surrounds the FATO to provide an extra margin of error in case vehicles accidentally diverge from the FATO.The FSA is recommended to extend beyond the edge of a FATO for the larger of 20 ft or 1/3 TTS.The FSA can extend past the edge of raised surfaces.For this analysis, FATO edges are placed right at the edges of the total surface area, allowing FSAs to extend past the surface boundary. +2) TaxiwaysThis analysis assumes vehicles will ground taxi, whether under their own power or conveyed by tug or dolly, as opposed to hover taxi.For ground taxiways, a total taxi route width of 1.5x TTS is recommended.For this analysis, the taxi route width requirement is enforced only for major taxiways connecting TLOFs.Separation requirements for minor taxiways connecting parking to the major taxiways are assumed to be satisfied by parking separation requirements discussed below. +3) ParkingTo conserve space used for taxiways, this analysis assumes "back-out" or "turn-around" parking, for which a minimum diameter of the helicopter's tail rotor arc is recommended.For the purposes of this analysis, the vehicle's largest dimension is used as the parking pad diameter.The minimum distance between the tail rotor arcs of 10 ft for "back-out" or 1/3 rotor diameter for "turn-around" parking.For "turn-around" parking a minimum clearance of 10 ft is required between the tail rotor arc and fixed objects.This analysis assumes a minimum separation of 1/3 TTS between multiple parking spaces and between parking and an FSA, and a minimum separation of 10 ft between parking and adjacent passenger accommodation areas.It is assumed that parking spaces may access eVTOL battery charging stations without impacting space and separation requirements.Vascik and Hansman [4] include a review of seven proposed 2-6 passenger eVTOL aircraft with maximum dimensions of 45 ft, generating TLOF, FATO, and FSA diameters of 45, 68, and 88 ft, respectively.This analysis assumes an only slightly more conservative spacing by rounding up the 1.5x largest dimension parameter and TTS to 70 ft.Fig. 1 shows a diagram of surface features and separations considered. +C. Design ApproahesThree generic design approaches were considered based on how taxiways are used to connect TLOFs and parking pads.Surface topologies were designed for each method with up to eight parking spaces per TLOF (32 total).The minimum parking spaces per TLOF differed by design method depending on how many could fit within the minimum sized surface as defined by the FATO-FATO separation requirements for simultaneous operations.FATOs of 70 ft diameter placed at each corner of the surface separated by 200 ft requires a minimum surface side length of 340 ft. +1) PerimeterThe Perimeter approach places major taxiways connecting TLOFs at adjacent corners around the perimeter of the surface, with minor taxiways to parking lining the inside edge of the major taxiways.Fig. 2 shows the surface layout for the Perimeter design node-link model with five parking spots per TLOF (P5).Although not shown, models with different numbers of parking spaces per TLOF were also developed for analysis.The smaller blue circles represent parking spaces.Each complete green circle represents a FATO with FSA arc surrounding it to the edge of the surface.Green lines represent major taxiways connecting green nodes at the center of each TLOF.Similarly, blue lines represent minor taxiways connecting major taxiways to blue nodes at the center of each parking space.The light green shaded region marks the 70 ft wide area required by the major taxiways.The light blue shaded region marks the area used by parking.This includes 10 ft separation from the light orange shaded region marking remaining area available for passenger accommodation.Each Perimeter node-link model is developed by first placing parking nodes arranged in square shape spaced 60 ft (45 ft parking diameter + 15 ft separation) from each other.Major taxiways are then wrapped around the parking spaces.Due to the parking-FSA spacing requirements for the corner parking spots, the edge of the major taxiways is placed an extra 2 ft from the edge of the parking spaces such that the major taxiway centerline is 59.5 ft (70/2 + 45/2 + 2 ft) ft from the parking nodes.TLOF nodes are placed at the intersections of major taxiways.Minor taxiways head straight out from the parking node to stay clear of adjacent parking spaces until they reach the edge of major taxiway.The minor taxiways then angle toward one of the TLOF nodes at either end of the major taxiway such that they intersect the major taxiway centerline at the same point as an adjacent minor taxiway heading in the opposite direction.Only corner parking nodes have direct access to TLOF nodes without joining a major taxiway.Perimeter node-link models were developed with as many as eight parking spaces per TLOF (P8) down to only two parking spaces per TLOF (P2) for which the minimum surface width of 309 ft is less than the 340 ft width required for independent operations.P2 was redesigned from the outside in, placing parking the same relative distance from TLOFs and major taxiways as the other Perimeter designs but with parkingparking spacing larger than the minimum 15 ft. +2) CentralThe Central approach places major taxiways connecting TLOFs at opposite corners through the center of the surface, with minor taxiways to parking along the edges of the major taxiways.Figs. 3 and4 show the surface layout for the Central design node-link model with five and six parking spots per TLOF (C5 and C6), respectively.Note that the surface side length for C6 is larger than for C5 to accommodate the extra parking spaces.Although not shown, additional models with different parking spaces per TLOF were also created for analysis.Each Central node-link model is developed by first placing major taxiways that intersect in the center and extend out toward the corners.Parking nodes are then placed 60 ft apart along the edges of the major taxiways such that the parking space and taxiway edges touch.There is a subtle difference in the Central node-link design depending on whether the number of parking spaces per TLOF is odd or even.If the number of parking spaces per TLOF is odd as in Fig. 3, there will be four parking spaces closest to the middle of the surface that touch the edges of two major taxiways each close to the central intersection.If the number of parking spaces per TLOF is even as in Fig. 4, each parking space will touch the edge of only one major taxiway.Once all parking nodes have been placed, the TLOF nodes are placed in the corners such that they meet parking-FSA spacing requirements (50 ft FSA radius + 22.5 ft parking radius + 15 ft separation).Minor taxiways head straight out from the parking node until they reach the edge of major taxiway, and then angle toward major taxiway intersection points equidistant from adjacent parking.The eight parking spaces closest to TLOFs are given direct access to the TLOF without joining a major taxiway.The four central parking spaces in designs with odd number of parking per TLOF, are given the option of joining the major taxiways at their central intersection as seen in Fig. 3.Central node-link models were developed with as many as eight parking spaces per TLOF (C8) down to only three parking spaces per TLOF (C3) for which the minimum surface width of 329 ft is less than the 340 ft width required for independent operations.As with P2, C3 was redesigned from the outside in to fill out the 340 ft x 340 ft space, placing parking the same relative distance from TLOFs and major taxiways as the other Central designs but with parking-parking spacing larger than the minimum 15 ft. +3) DisconnectedThe Disconnected approach uses taxiways only to connect each TLOF to multiple parking pads, not to other TLOFs.Therefore, there are no major taxiways.Fig. 5 shows the surface layout for the Disconnected design node-link model with five parking spots per TLOF (D5).Although not shown, additional models with different parking spaces per TLOF were also created for analysis.Each Disconnected node-link model is developed by first developing one quadrant.The parking nodes were evenly spaced along an arc of fixed radius centered at the TLOF node such that parking nodes were separated by 60 ft and the two parking spaces on the edge of the arc touched the edge of the surface.Then a minor taxiway connected each parking node directly to the TLOF.This first quadrant of the design was duplicated three more times, rotated, and arranged such that the TLOFs were at the corners of the surface and parking nodes at the edges were spaced 60 ft from those in adjacent quadrants to meet the 15 ft parking-parking spacing requirement.Disconnected node-link models were developed with as many as eight parking spaces per TLOF (D8) down to only four parking spaces per TLOF (D4) for which the minimum surface width of 329 ft is less than the 340 ft width required for independent operations.The four quadrants of D4 were separated to reach a surface width of 340 ft so that the parkingparking spacing of the edge parking was a bit larger than the minimum 15 ft. +III. SURFACE AREA UTILIZATION ANALYSISSurface area utilization was analyzed by comparing the relative surface areas required for each node-link model developed.The total surface area is a square bounded by the FATO edges of the TLOF nodes placed in the corners.Total surface area is decomposed into TLOF/taxi, parking, and passenger area shown in Fig. 2-5 as light green, blue, and orange shaded regions, respectively.TLOF/taxi area includes the FSA of each TLOF node and 70 ft wide corridors along major taxi routes.Parking area includes the parking spaces as well as any additional area required to accommodate parking-parking, parking-FSA, and parking-passenger spacing requirements.Passenger area includes any large area free of taxiways that is sufficiently separated from parking.All areas are bounded by straight edges to simplify calculation.Fig. 6 compares space utilization results for each node-link model.The same results are grouped by design approach on the top and by parking per TLOF on the bottom to facilitate comparison.As expected, the total surface area required for each design approach increases with the number of parking spaces per TLOF.The Perimeter and Central design approaches maintain fairly consistent percentages of surface area devoted to parking with a greater percentage of area shifting from TLOF/taxi to passenger as the parking per TLOF increases.Because the Disconnected design approach does not include major taxiways, its TLOF/taxi area is related to TLOF only and is fixed.As the parking increases, most additional surface area is devoted to parking.This is because as the parking increases, the minor taxiways (e.g.blue lines in Fig. 5) get longer to accommodate parking-parking spacing constraints.The smallest total surface allowable by the FATO-FATO spacing requirements (340 ft x 340 ft) accommodates two, three, and four parking spaces per TLOF for the Perimeter, Central, and Disconnected design approaches with respective node-links models designated as P2, C3, and D4.When comparing models with the same number of parking spaces on the bottom of Fig. 6, the Perimeter design approach always requires the greatest total area.However, this design also offers the greatest centralized passenger area, which may be desirable.For five or fewer parking spaces per TLOF, the Disconnected design approach requires the least total area.However, as parking per TLOF increases above five, Disconnected, requires more area than Central.Although Central appears to have a space utilization advantage over the other design at higher parking numbers, the total passenger area is actually shared among four segregated regions as seen in Figs. 3 and4, which may be undesirable. +IV. VERTIPORT CONFIGURATION MODELVertiport are sensitive to wind conditions which may inhibit the use of one or more TLOFs for approach (apr), departure (dep) or both.A model is presented to map wind condition to vertiport configuration and then taxi distance and capacity metrics are presented for each configuration in the following section.Vertiport configuration refers to the availability of individual TLOFs for approach and/or departure operations.The configuration may depend on many factors affecting arrival and departure paths to the vertiport, including surrounding tall buildings, obstructions, noise restrictions, neighboring traditional-controlled airspace, winds, zoning, and air rights.This section presents a vertiport configuration model combining static and dynamic directional constraints.Rather than limit a TLOF to a finite number of fixed approach/departure paths, it is assumed that a TLOF is available for approach and/or departure operations if there exists a large enough arch of constraint free airspace surrounding the TLOF.This approach can include any constraint expressed as radial ranges of direction relative to the TLOF.For example, if there are tall buildings or other fixed obstructions near the vertiport that would penetrate a TLOF's approach/departure surface as defined by Advisory Circular 150/5390-2C [6] to/from a certain range of directions, these ranges are included as static constraints to approach/departure operations to/from the TLOF.For the purposes of this analysis, the only static configuration constraints considered are due to the vertiport surface.Based on heliport operations subject matter expert (SME) interviews, the approach/departure surface should not be placed over parked or taxiing aircraft or surface infrastructure like buildings or walls.To enable simultaneous operations to adjacent TLOFs, the arcs of their respective approach/departure directions should not overlap.Fig. 7 shows the resulting set of static approach/departure direction ranges to which each TLOF is constrained.The static feasible arc ranges for approaches and departures for an individual TLOF are equal.For example, for TLOF A, (A) = (A) = [180, 270].Once the static constraints are established, dynamic constraints may be incorporated that restrict the use of individual TLOFs for approach and/or departure operations under certain conditions to produce a set of possible vertiport configurations.Prevailing wind direction and speed are where , , and are the wind speed, wind direction, and vehicle direction, respectively.Equation ( 1) imposes the constraint that vehicles should not attempt approach or departure operations with a tailwind.A tailwind can cause the vehicle to fly into its own downwash and enter a highly unstable vortex ring state.Equation ( 2) imposes the constraint that vehicles should not attempt approach or departure operations with a crosswind component greater than 15 kt.Fig. 8 shows the resulting dynamic wind constraints for approach/departure direction ranges depending on wind direction and speed .Unlike the static constraints, the wind feasible arc ranges for approaches and departures for a given wind condition are not equal.From ( 1) and ( 2), for > 0 the wind feasible approach path arc must be within arcsin(15/ ) of and the wind feasible departure path arc must be within arcsin(15/ ) of + 180.Based on heliport operations SME interviews, 15 was selected as a reasonable minimum arc measurement required to allow approach or departure operations.Therefore, in the above example, TLOF A is configured to allow departure operations but not approach operations under the given wind conditions.Fig. 9 shows the wind dependent configuration map for TLOF A assuming a 15 minimum arc measurement threshold to allow operations.When the wind conditions fall within the blue, red, or purple regions, TLOF A is configured for approach only, departure only, or dual operations, respectively.For all other wind conditions, the TLOF is closed to approach/departure operations.Fig. 10 shows the wind dependent configuration map for the entire four-TLOF vertiport with static constraints as pictured in Fig. 7.Each Apr|Dep configuration lists the TLOFs available for approach and departure operations.For example, given = 45 and = 10 kt which falls within the dark teal shaded region of the map labeled BCD|ABD, TLOFs B, C, and D may conduct approach 10 due to space constraints is ABCD|ABCD which occurs only in the center of the map when = 0 kt.The operational analysis presented in the following sections will compare results under these different vertiport configurations. +V. OPERATIONAL EFFICIENCY ANALYSISDue to the quadrantal symmetry of the vertiport designs analyzed, the operational efficiency results for all configurations with the same numbers of vertiports available for each operation are identical.Therefore Set 1, 2, 3, and 4 will each refer to a representative set with identical results as follows.Set 1 includes ABCD|ABCD alone.Set 2 includes ABC|ACD, ABD|BCD, ACD|ABC, and BCD|ABD.Set 3 includes AB|CD, AD|BC, BC|AD, and CD|AB.Set 4 includes A|C, B|D, C|A, and D|B. +A. Average Total Taxi DistanceThis section compares average total taxi distance between each of the node-link models under each of the four different representative sets of vertiport configurations.Routes were created between each parking space and TLOF using Dijkstra's shortest path algorithm to explore the node-link model.Parking nodes were discouraged as intermediate nodes by adding a large value cost to links that connected to parking nodes.This ensured that the resulting routes did not try to cut corners by passing through a parking node.However, TLOF nodes were allowed as intermediate nodes, which was relevant for the Perimeter node-link models as some TLOF-parking pairs are only reachable through other TLOF nodes.The Disconnected node-link models are the only ones for which routes do not exist for many TLOF-parking pairs.Once shortest path routes were identified for each TLOFparking pair where possible, the shortest total taxi distance using each parking spot was determined for each configuration using (3)(4)(5).𝑖𝑛 𝑖 = min (𝑑𝑖𝑠𝑡(𝑃 𝑎𝑝𝑟 , 𝑝 𝑖 )) ∀ 𝑃 𝑎𝑝𝑟 ∈ 𝑐𝑜𝑛𝑓𝑖𝑔 () = min (( , )) ∀ ∈ ()𝑡𝑜𝑡𝑎𝑙 𝑖 = 𝑖𝑛 𝑖 + 𝑜𝑢𝑡 𝑖 ()To find the shortest total taxi distance totali via parking node pi in configuration config, first find the shortest taxi-in distance ini among the routes between pi and TLOF nodes available for approach operations Papr within configuration config.Then, find the shortest taxi-out distance outi among the routes between pi and TLOF nodes available for departure operations Pdep within config.Finally, add the shortest taxi-in distance ini to the shortest taxi-out distance outi to get the shortest total taxi distance totali.The average total taxi distance for config is then the average totali for all reachable parking nodes in configuration config.Fig. 11 shows the average total taxi distance results for all node-link models in each configuration set.For Perimeter and Central node-link models, all parking nodes are reachable in all configurations because there exists a route for all TLOF-parking pairs.Therefore, their results reflect total taxi distance averages across all parking nodes.However, for the Disconnected nodelink models, the TLOF node associated with a parking node must be available for both approach and departure operations for the parking node to be reachable within a given configuration.None of Set 3 or 4 configurations have a TLOF that is available for both approach and departure operations, and so Disconnected node-link models have no complete taxi routes (and no total taxi distance results) in these configurations.Set 2 configurations have exactly two out of four TLOFs that are available for both approach and departure operations, and so only half of the Disconnected parking nodes have complete taxi routes to average total taxi distance.Only Set 1 with all four TLOFs available for both operations has complete taxi routes for all Disconnected parking nodes.Several trends can be seen in Fig. 11.For all design methods, the average total taxi distance increases with the parking per TLOF.This is because the larger surface necessitates longer taxi distances to additional parking.The Perimeter and Central designs have increased average total taxi time for configuration sets with fewer TLOFs available for approach to departure operations.When a parking node's preferred closest TLOF is not available, it must use a longer route to get to a TLOF that is available.This effect is more pronounced for Disconnected.Whereas Disconnected designs have fewer reachable parking nodes in Set 2 than Set 1, the average total taxi distance remains constant between the two sets as these parking nodes have only one route option.In general, the average total taxi distance is longest for Perimeter, followed by Central, and then Disconnected.There are some interesting exceptions of note.As the Disconnected total taxi distance become longer as parking is added, the average total taxi distance starts to exceed Central at six parking spaces per TLOF, and exceeds Perimeter as well for parking per TLOF greater than six.In Set 2, the Perimeter and Central average total taxi times are almost identical. +B. Operational CapacityThis section analyzes the theoretical operational capacity of each node-link model in each configuration set.Guerreiro et al. [8] introduced a theoretical model for comparing surface capacity with TLOF capacity to determine if a vertiport's capacity was limited by its parking or TLOFs.This analysis adopts a similar method shown in (6-7) where is the number of surface operations that can be processed in time window and / is the number of balanced TLOF operations (pair of approach/departure operations) that can be processed in . is the number of reachable parking nodes.Parking nodes that have no complete taxi routes to operational TLOFs are excluded from this count. is the number of operational TLOFs, regardless if they are dedicated to either approach or departure operations or if they allow both.Only TLOFs unavailable for approach or departure operations are excluded from this count.Heliport operations SME interviews identified the rule of thumb for VTOL ground taxi speed to be equivalent to "a brisk walk" (~2.7 mph or 4 ft/s) regardless of whether the vehicle taxis under its own power or is conveyed.The total taxi time was calculated from the average total taxi distance results shown in Fig. 11 using an average taxi speed of 4 ft/s.The total time spent at the parking node was selected to be a constant value of 8 min (480 sec) based on assumed separate loading and unloading times of 60 sec per passenger (used in [7][8]) and an average of 4 passengers for the 2-to 6-passenger eVTOL vehicles envisioned.Vehicle charging times in excess of these 8 min are not considered in this analysis.Therefore, it is assumed that either vehicles charge for this short period of time or return to other vertiports for more charging.The TLOF utilization time for approach and departure operations is assumed to be 60 sec each also used in [7][8].A of 15 minutes is arbitrarily chosen to compare capacity results.Table II shows the resulting / for each vertiport design approach and configuration set.Fig. 12 shows the results for each node-link model and configuration set.The / results from Table II are also shown as black lines underlying the bars of relevant Fig. 12. Surface Operations Capacity configuration sets for easy comparison.The surface capacity increases with parking per TLOF as additional parking availability counteracts the effect that increasing average taxi distance has in increasing average taxi time.However, this effect can be seen as surface capacity tends to decrease for configuration sets with fewer TLOFs available for approach/departure operations and therefore longer taxi distances and times.The impact of reduced number of reachable parking nodes for Disconnected node-link models in Set 2 in reducing surface capacity is even more dramatic.Due to slightly shorter average total taxi distances, Central surface capacity is slightly higher than that of Perimeter for all configurations but Set 2 where they are similar just like their average total taxi distances.For a parking per TLOF of five or greater, the surface capacity is either very similar to or greater than approach/departure capacity for all configurations. +VI. DISCUSSIONThe operational efficiency results suggest that five parking spaces per TLOF may be sufficient given the assumed values for average taxi speed, , , and .The designs with five parking spaces per TLOF are pictured in Figs. 2, 3 and5.From an operational efficiency perspective, Disconnected is at a clear disadvantage when TLOFs are restricted from performing both approach and departure operations for any reason.Disconnected may have comparable capacity to other methods under configuration Set 1.However, keep in mind that Set 1 consists of only a single configuration made possible under completely calm wind conditions.The greatest advantage of Disconnected is in its surface area utilization by offering the greatest number of parking spaces (four per TLOF) in the smallest allowable surface area for four independent TLOFs.At five parking spaces per TLOF, the required total surface area for Disconnected becomes comparable with Central and this advantage is lost.Although not analyzed in this paper, Disconnected may have an advantage with respect to surface congestion.As all its routes between TLOFs and parking are completely independent of one another, the Disconnected approach should not be susceptible to surface congestion which could potentially reduce the operational efficiency of Perimeter and Central designs dependent on interconnected taxiways.A possible Disconnected design enhancement for future analysis would be to connect parking spaces between adjacent quadrants with minor taxiways where possible.For example, as seen in Fig. 5, the pairs of parking spaces at the edges of the surface directly across from one another could be connected to allow through access to each other's TLOF.This could provide some capacity improvement in configurations where TLOFs are restricted to only one type of operation.Even though Central tends to have shorter average total taxi distance, it is on fairly equal footing with Perimeter when it comes to capacity, and both are robust to configuration adjustment to wind condition.The primary advantage of Central over Perimeter is that it requires less total surface area for the same amount of parking.A possible disadvantage is that whereas Perimeter passenger area is centralized, the Central passenger area is segregated into four separate areas which may present different passenger access and accommodation considerations.This analysis focused on four-TLOF square-shaped vertiport designs.It would be interesting to extend this analysis to rectangular designs and designs incorporating more than four TLOFs to determine if any more capacity can be gained and what the impact would be to surface area requirements.As this paper presented a purely analytical evaluation of theoretical capacity limits, it would be interesting to see what efficiency is lost in simulation where surface congestion may become an issue. +VII. CONCLUSIONThis paper presented three generic methods for vertiport surface topology design based on how taxiways are used to connect TLOFs and parking.Analyses of surface area utilization and operational efficiency produced some interesting insights into the advantages and disadvantages of the different design approaches.The Disconnected approach focuses on maximizing parking per TLOF without providing taxi connectivity between multiple TLOFs.Whereas this approach does provide the greatest amount of parking in the smallest total surface area, it is not robust to conditions where TLOFs may be restricted from performing both approach and departure operations for any reason.The Perimeter and Central design methods provide connectivity between multiple TLOFs by placing major taxiways around the perimeter and through the center of the vertiport surface, respectively.These approaches are both robust to configuration adjustments due to wind conditions.Whereas the Central approach requires less surface area, Perimeter provides a centralized passenger area rather than segregating into four areas like Central.The operational capacity analysis suggests that five parking spaces per TLOF may be sufficient assuming 60 sec TLOF utilization time for approach and departure operations, 4 ft/s taxi speed, and 8 min turnaround time at parking spaces.If longer turnaround times are desired for charging purposes, more parking may be desired.This paper also introduced a method of combining static and dynamic approach/departure directional constraints into a vertiport configuration model.This model may be used to analyze other vertiport surface design robustness to configuration change or integrated into fast-time simulations of vertiport surface operations.It can also be utilized to look into the design of vertiports placed on large airports with traditional traffic arrival and departure paths added as constraints. +ACKNOWLEDGMENTThe author thanks Rex Alexander for invaluable support as a heliport operations subject matter expert.Fig. 2 .2Fig. 2. Perimeter design with 5 parking spaces per TLOF (P5) Fig. 3. Central design with 5 parking spaces per TLOF (C5) +Fig. 5 .5Fig. 5. Disconnected design with 5 parking spaces per TLOF (D5) Fig. 4. Central design with 6 parking spaces per TLOF (C6) +Fig. 6 .6Fig. 6.Space Utilization Results +For example, given = 0 and = 20 kt, arcsin(15/20) = 49.Therefore, (0, 20 kt) = [0 -49, 0 + 49] = [-49 or 311, 49] and (0, 20 kt) = [180 -49, 180 + 49] = [131, 229].The combined feasible approach and departure arc ranges are the intersection of their respective static and wind feasible arc ranges given by (TLOF, , ) = ∩ and (TLOF, , ) = ∩ .For TLOF A, recall from Fig.7 that (A) = (A) = [180, 270].Therefore, (A, 0, 20 kt) = [180, 270] ∩ [-49 or 311, 49] = ∅ and (A, 0, 20 kt) = [180, 270] ∩ [131, 229] = [180, 229].The resulting arc measurements = 0 and = 229 -180 = 49 leave no feasible arc range for approaches and 49 arc range for departures. +Fig. 8 .8Fig. 8. Dynamic Wind Constraints for Apr/Dep Direction +Fig. 11 .11Fig. 11.Average Total Taxi Distance +Table I lists the surface feature separation requirements assumed for this analysis. +TABLE I .ISURFACE FEATURE MINIMUM SEPARATION REQUIREMENTSFig. 1. Surface Features and SeparationsSurface FeatureRequirementTLOF diameter45 ftFATO diameter70 ftFSA diameter100 ftFATO-FATO edge separation200 ftSurface side length340 ftMajor taxiway route width70 ftParking diameter45 ftParking-Parking or Parking-FSA separation15 ftParking-Passenger Area separation10 ft +TABLE IIII.APPRAOCH/DEPARTURE OPERATIONS CAPACITYDesign/Configuration Set 1Set 2Set 3Set 4Perimeter30303015Central30303015Disconnected301500 + + + + + + + + + Fast-forwarding to a future of on-demand urban air transportation + + JHolden + + + NGoel + + 10.3030/774086 + + + + UBER Elevate white paper + + Publications Office of the European Union + October 27, 2016 + + + J. Holden and N. Goel, "Fast-forwarding to a future of on-demand urban air transportation," UBER Elevate white paper, October 27, 2016. www.uber.com/us/en/elevate/ + + + + + Enabling Airspace Integration for High-Density On-Demand Mobility Operations + + EricRMueller + + + ParmialHKopardekar + + + KennethHGoodrich + + 10.2514/6.2017-3086 + + + 17th AIAA Aviation Technology, Integration, and Operations Conference + Denver, Colorado + + American Institute of Aeronautics and Astronautics + June 2017 + + + + E. Mueller, P. Kopardekar, and K. Goodrich, "Enabling airspace integration for high-density on-demand mobility operations," 17th AIAA Aviation Technology, Integration, and Operations Conference, 5-9 June 2017, Denver, Colorado. + + + + + Exploratory Analysis of the Airspace Throughput and Sensitivities of an Urban Air Mobility System + + KennethHGoodrich + + + BryanBarmore + + 10.2514/6.2018-3364 + + + 2018 Aviation Technology, Integration, and Operations Conference + Atlanta, Georgia + + American Institute of Aeronautics and Astronautics + June 2018 + + + + K. H. Goodrich and B. E. Barmore, "Exploratory analysis of the airspace throughput and sensitivities of an urban air mobility system," 18th AIAA Aviation Technology, Integration, and Operations Conference, 25-29 June 2018, Atlanta, Georgia. + + + + + Development of Vertiport Capacity Envelopes and Analysis of Their Sensitivity to Topological and Operational Factors + + ParkerDVascik + + + RJohnHansman + + 10.2514/6.2019-0526 + + + AIAA Scitech 2019 Forum + San Diego, California + + American Institute of Aeronautics and Astronautics + 7-11 January, 2019 + + + P. D. Vascik and R. J. Hansman, "Development of vertiport capacity envelopes and analysis of their sensitivity to topological and operational factors," AIAA SciTech Forum, 7-11 January, 2019, San Diego, California. + + + + + Energy and Landing Accuracy Considerations for Urban Air Mobility Vertiport Approach Surfaces + + EmreYilmaz + + + MatthewWarren + + + BrianGerman + + 10.2514/6.2019-3122 + + + AIAA Aviation 2019 Forum + Dallas, TX + + American Institute of Aeronautics and Astronautics + 17-21 June 2019 + + + E. Yilmaz, M. M. Warren, and B. J. German, "Energy and landing accuracy considerations for urban air mobility vertiport approach surfaces," AIAA Aviation Forum, 17-21 June 2019, Dallas, TX. + + + + + Heliport design — the total package + + RaymondASyms + + 10.1016/s0740-8315(88)80062-7 + + + Hospital Aviation + Hospital Aviation + 0740-8315 + + 7 + 7 + + 2012 + Elsevier BV + + + AC 150/5390-2C: Heliport design + Federal Aviation Administration, "AC 150/5390-2C: Heliport design," 2012. + + + + + Mission Planner Algorithm for Urban Air Mobility – Initial Performance Characterization + + NelsonMGuerreiro + + + RickyWButler + + + JeffreyMMaddalon + + + GeorgeEHagen + + 10.2514/6.2019-3626 + + + AIAA Aviation 2019 Forum + Dallas, TX + + American Institute of Aeronautics and Astronautics + 17-21 June 2019 + + + N. M. Gurreiro, R. W. Butler, J. M. Maddalon, and G. E. Hagen, "Mission planner algorithm for urban air mobility -initial performance characterization," AIAA Aviation Forum, 17-21 June 2019, Dallas, TX. + + + + + Capacity and Throughput of Urban Air Mobility Vertiports with a First-Come, First-Served Vertiport Scheduling Algorithm + + NelsonMGuerreiro + + + GeorgeEHagen + + + JeffreyMMaddalon + + + RickyWButler + + 10.2514/6.2020-2903 + + + AIAA AVIATION 2020 FORUM + + American Institute of Aeronautics and Astronautics + June 2020 + + + N. M. Gurreiro, R. W. Butler, J. M. Maddalon, and G. E. Hagen, "Capacity and throughput of urban air mobility vertiports with a first- come, first-served vertiport scheduling algorithm," AIAA Aviation Forum, June 2020. + + + + + + diff --git a/file832.txt b/file832.txt new file mode 100644 index 0000000000000000000000000000000000000000..da5c65c7b5c91244d8bf6ec2c861a2af81c0fed7 --- /dev/null +++ b/file832.txt @@ -0,0 +1,355 @@ + + + + +I. IntroductionECONFIGURATION is a change in airport runway usage, usually referring to a change in airspace flow direction as well.There are many reasons why an airport may reconfigure: a change in noise restrictions between day and night, a shift in demand between arrivals and departure or between dominant flow directions, or a change in wind direction or magnitude making it difficult to land in the current configuration.A nation-wide analysis of reconfiguration events showed that reconfiguration can have a significant impact on airport capacity causing loss of throughput at larger busier airports. 1More detailed analyses of Dallas/Fort Worth and Chicago O'Hare (ORD) reconfiguration operations distinguished between strategic and tactical reconfigurations in how delay is absorbed. 2,3Strategic reconfigurations are planned and coordinated by the traffic managers at the Tower, Terminal Radar Approach Control (TRACON), and Center levels.They are proactive in nature and most of the delay is absorbed outside of the TRACON.By contrast, tactical flow changes occur with minimal warning to or planning by the traffic managers.They are reactive in nature and delay is absorbed both inside and outside the TRACON.Whereas strategic reconfigurations are generally more common than tactical, tactical reconfigurations are more disruptive, especially in high traffic load conditions.Ref. 3 studied seven cases of reconfiguration between the top two ORD configurations occurring in April 2012, three of which resulted in disruptive holding and vectoring both inside and outside the TRACON.5][6] Reference 7 briefly addresses the problem of scheduling individual flights given a planned reconfiguration time.However, flights are scheduled to runways assuming the reconfiguration time is known hours in advance, falling into the strategic reconfiguration category.No scheduling research could be found in the current literature specifically addressing tactical reconfiguration.This research addresses the tactical reconfiguration scheduling problem with the concept of precision arrival scheduling along fixed routing.In nominal conditions, precision arrival scheduling concepts can increase peak airport throughput 10%. 8These concepts use fixed routing defined from the meter fix all the way to the runway to more accurately estimate time-to-fly, enabling more precise scheduling with smaller buffers, yielding higher throughput.However, these concepts must be made robust to off-nominal disturbances.Whereas research has begun to explore and enhance precision scheduling robustness to disturbances such as wind errors, emergency landings, and missed-approaches, 9,10 these were all tested with static arrival fixed routing.In the case of airport reconfiguration, fixed routing may no longer be fixed as one set of routes is swapped out for another.This paper extends the concept of precision arrival scheduling to this dynamic reconfiguration environment.Integrated fixed arrival routing is modeled for the top two peak traffic configurations at ORD.A first-come-firstserved multi-point scheduler is adapted to the tactical reconfiguration problem by rescheduling aircraft within the TRACON upon reconfiguration notification.Then the effect of the rescheduling on throughput and delay metrics is studied for a range of arrival rates and reconfiguration notification lead-times. +II. Method +A. Route ModelsAn analysis of 2011 and 2012 ORD reconfiguration operations 3 determined that the most common reconfiguration changes occurred between the two most commonly used arrival configurations.These are Plan X Trip (PX) using arrival runways 10, 09R and 04R, and West Flow (WF) using arrival runways 27L, 27R and 28 † depicted in Figure 1.Arrival route models for these configurations were generated using published routes and historical flight track data.Published routes were extracted from all ORD Standard Terminal Arrival Routes (STARs) and Instrument Approach Procedures (IAPs) for the six runways used by the PX and WF configurations.* Daily operating conditions by the quarter-hour for ORD for the entire year of 2012 were collected from Aircraft System Performance Metrics (ASPM) Airport Efficiency reports. 11Among other data, these reports included a list of runways configured for arrivals, the arrival rate capacity, and the actual number of arrival operations at ORD every quarter-hour.ORD was assumed to be in PX or WF configuration for a given quarter hour if all of their respective arrival runways were listed as configured for arrivals in the ASPM report.A total of 34 24-hour historic traffic samples were selected from 2012 based on their relatively high number of actual arrival operations during PX and WF configurations.Historical track data were collected for all aircraft that arrived at one of the PX or WF configuration runways during a quarter-hour when ORD was in that configuration.A total of 29,660 flight tracks were collected, roughly half landing in each configuration.Figure 2 shows jet tracks, which comprised 99.3% of the tracks collected, and the published routes that most closely represent them.The concentric circles are centered on the ORD fix with radii increasing in 10 mile increments to 100 miles.Tracks of flights landing in PX and WF configurations are shown in gray and tan (tan overlaying gray), respectively.STAR common and transition routes are overlaid in blue and maroon respectively.IAPs for the six runways of interest are shown in green.Whereas STARs (common and transition) capture much of † In late 2013, new runways were added to ORD and some runway designations have changed.the track paths, there are no transition paths defined between STARs and IAPs.In addition, several flows from the North do not follow any published STAR.Fixes MSN and BJB identify likely coordination points for these flows.For the purposes of studying configuration changes, the route model should have entry points that both configurations use with similar frequency so that comparisons can be made using the same traffic scenarios for both configurations.For the Northeast, Northwest, and Southwest flows to meet this criterion, it would be sufficient for the model to extend roughly 45 miles from ORD at the beginning of the STAR common routes where the handoff between Center airspace and TRACON airspace typically occurs.However, the North flow undefined by STARs extends more than 80 miles from ORD before a common entry point (BJB) is found.The two STAR common routes serving the Southeast flows extend much farther from ORD (more than 90 miles) than the other STAR common routes.In addition, the more southern route (ROYKO) is more commonly used in PX and the more northern route (WATSN) is more commonly used in WF.Therefore, the route model was extended to entry points roughly 80-90 miles from ORD (roughly 40-50 miles from the TRACON).Because the Northwest and Northeast sets of flight tracks each clearly merge into a single common route near the Center/TRACON handoff point, a single transition route entry point was chosen to represent each set of merging flows.These were MYTCH in the Northwest and WLTER in the Northeast.This assumes that if a schedule is given 110-160 miles from the TRACON, then by the time flights are at these entry points roughly 40-50 miles from the TRACON, the Center controller has arranged flights in the correct sequence and spacing for the merge.configurations.Although a second common entry exists originating from General Mitchell International Airport 55 miles North of ORD, this entry was not modeled, as these flights are relatively infrequent and too far within the 150-200 mile arrival scheduling freeze horizon.Future research should address integration of arrival traffic from nearby satellite airports.Analysis of the Southeast clustered flight paths revealed that there are two distinct sets of merging flows that tend to feed the WATSN common route in WF and the ROYKO common route in PX.One flow merges near MZZ and then diverges to either VEECK or HAUPO.The other flow merges at WATSN and then either continues along the WATSN common route or diverges at MKITA.To model this crossover, WATSN and VOYUP were modeled as distinct Southeast entry points.VOYUP also served as the diverging point between VEECK and HAUPO for the MZZ flow.Analysis of the Southwest clustered flight paths revealed that whereas two distinct flows exist at BDF and PNT, they both feed primarily the BENKY common route.Even most PNT traffic landing on 04R, whose IAP overlaps the TRTLL common route, merge with the BENKY common route at NEWRK before turning back to the 04R IAP.Therefore, BDF was modeled as a single entry point representing all Southwest flows.To aid more detailed route modeling within the TRACON, track data were processed 12 to identify the most commonly used flight paths again, this time with a clustering parameter that separated nodes by at least 1.5 miles.Once again, Jets were processed separately for each arrival runway.Once all lateral paths had been designed, flight track altitudes along crossing paths were analyzed to determine if altitude separation is maintained.If nominal altitude ranges overlapped between crossing paths, a shared intersection node was added to ensure lateral separation through scheduling.Figure 3 shows the integrated route model including multiple commonly used path options from each of six entry nodes to each of the six runways modeled.Routes serving PX (blue) and WF (maroon) runways are shown overlaying historical flights tracks (gray and tan, respectively) to those runways.Thick black arcs identify the arc range and percentage of arriving traffic that each modeled entry point represents.Percentages are based on each flight's first track point within 90 miles of ORD. Figure 4 shows the integrated route model magnified to show detailed routing in the TRACON close to the runways.Routes are color coded by the runways they serve.Unlike current precision arrival schedulers, which use a single path option per arrival fix-runway pair for a given aircraft, notice that multiple path options to each runway are available to the arrival scheduler including base leg extensions and crossovers.These path options give the scheduler additional control authority beyond speed alone.In addition to the static configuration routes, transition routes between static configuration routings were generated to ensure a path existed from every node of one configuration to at least one runway of the other configuration.This included runway sink nodes to test the case of reconfiguration without pre-notification.Transition routes were designed to rejoin static routing as soon as reasonable with as few intersections and new nodes as possible.Figure 5 shows transition routes designed between 04R (blue) and 27R (maroon). +B. Arrival SchedulerA first-come-first-served multi-point scheduling technique 13 was applied to the route model.The scheduler was implemented in fast-time in the Python programming language.Flights were scheduled in order of their Estimated Time of Arrival (ETA) at their entry node to emulate distance-based first-come-first-served ordering.As each new flight is considered, the scheduling technique produces a sequence of feasible time windows for which the flight may be scheduled at each node along a given route.The time windows must satisfy feasible ranges of transit time along each route segment and avoid blocked time slots at each node reserved for the Scheduled Times of Arrival (STAs) of previously scheduled flights.Then the earliest arrival times available at each node, especially the runway, are considered the preferable schedule for that route.Because the route model includes not only multiple arrival runway options, but also multiple route options to each runway, a schedule was produced for each route option to each runway and the one providing the earliest runway arrival time was chosen.The assigned route option with the earliest runway STA may not necessarily be the one with the earliest runway ETA. +Transit time rangesTransit time ranges were used to implicitly model how much delay could be absorbed along a given route segment with speed control.Because historical flight tracks may include path stretch delay, the nominal transit time along a route segment between two nodes was calculated based on nominal speeds observed in historical track data at end nodes and segment path distance.For each historical flight, the closest track point within 3 miles of each node was collected and analyzed to assign nominal speeds to the node.The median speeds for all flights landing at the same runway tended to differ slightly for some nodes if the flight path distance from the node to the runway was significantly different.Therefore, nodes were assigned separate nominal speeds for PX and WF, and in some cases a separate 04R speed from the other PX runways was assigned as well.Assume a flight plans to travel a route segment from node n 1 to node n 2 and the nominal speeds at n 1 and n 2 given its planned arrival runway are v 1 and v 2 respectively.The nominal transit time from n 1 to n 2 is calculated as:t nom (n 1 , n 2 ) = d(n 1 , n 2 ) (v 1 + v 2 ) / 2 v 1 > v 2 d(n 1 , n 2 ) v 1 v 1 ≤ v 2 " # $ $ % $ $ ,(1)where d(n 1 ,n 2 ) is the distance between n 1 and n 2 .Note that if nominal speed is decreasing (v 1 > v 2 ), transit time is distance divided by average node speed.Because arrival flights are not expected to accelerate (only decelerate), if v 1 < v 2 , as can be the case along a reconfiguration transition route, the nominal transit time is distance divided by the slower nominal speed v 1 .Just as a node's nominal speed may differ by runway, the route segment nominal transit time may differ by runway as well.The nominal transit time was considered to be the minimum (i.e.fastest) schedulable transit time, reserving faster performance to manage uncertainty when trying to meet the schedule.Thus, the scheduler may not time advance flights, only delay.Because it is estimated that speed control alone can modify flight time up to 10%, the maximum (i.e.slowest) transit times were assumed to be 10% higher than the nominal transit times. +Blocked time slotsThe size of a blocked time slot was used to implicitly model the radar separation minimum at a given node.Per FAA regulation, 14 the separation minimum is 5 miles when the aircraft are 40 miles or more from the terminal radar antenna.The separation minimum is 3 miles when the aircraft are less than 40 miles from the antenna.Aircraft conducting an instrument approach must be separated using wake turbulence minima, which depends on the relative weight classes of aircraft pair being considered.Based on the weight class mix of historical track data collected, 93%, 5%, and 2% of aircraft pairs would require 3, 4, and 5 mile wake turbulence minima, respectively.Therefore, for simplicity in this study, the 3-mile minimum representing 93% of aircraft pairs was assumed for all aircraft pairs.All route model nodes less than 40 miles from ORD modeled 3-mile separation, and all nodes 40 miles or more from ORD modeled 5-mile separation.A 0.3-mile separation buffer was added to these separation minima as a scheduling buffer.The separation requirements were converted from distance to time based on the nominal speeds associated with each node.Time ranges on both sides of reserved STAs of previously scheduled flights were blocked from consideration for new flight scheduling.The amount of time blocked on either side of each preexisting STA was based on the nominal speed of the flight that would be the follower.For example, if a previously scheduled flight f 1 had been assigned time t at a given node, the blocked time range used for scheduling a new flight f 2 would be [t-(r/v 1 ), t+(r/v 2 )] where r is the required separation distance and v 1 and v 2 are the nominal speeds of f 1 and f 2 , respectively.This way, if f 2 is scheduled in front of f 1 , then f 1 is the follower and separation will be at least r/v 1 , and if f 2 is scheduled behind f 1 , then f 2 is the follower and separation will be at least r/v 2 . +Rescheduling due to tactical reconfigurationThe first-come-first-served scheduler serves new flights in order of their entry point ETAs.However, in the case of a reconfiguration, many flights already scheduled and traveling assigned routes may need to be rescheduled and rerouted simultaneously.The scheduler is notified at time t N that a reconfiguration will occur at time t R .All flights with originally scheduled runway arrival times greater than t R are rescheduled.Each flight is rescheduled by processing route options from the first originally scheduled node after t N to the new configuration runways, choosing the one with the earliest schedulable runway arrival time.Two methods of prioritizing the rescheduling of these flights were tested.Rescheduling in order of original scheduled runway arrival time prioritizes flights that were closer to landing and will likely fly extra distance at inefficiently low altitudes and speeds during the reconfiguration.Rescheduling in order of earliest estimated runway arrival time in the new configuration prioritizes packing the new schedule in an attempt to maintain throughput. +C. ScenariosArrival traffic scenarios for ORD were randomly generated in 15 minute chunks according to a given entry node distribution and quarter hourly rate.Each 15 minutes of arrival traffic contained the desired number of flights per quarter hour with uniform randomly distributed entry ETA across the 15 minutes.Random entry node assignment was weighted by the traffic representation percentages shown in Figure 2.Historical ASPM data was analyzed to determine the arrival capacities of PX and WF configurations in current operations.In current operations, the most commonly reported quarter hourly arrival capacities are 24 for PX and 25 or 26 for WF.The actual arrivals per quarter hour are very similar for both configurations at higher traffic volumes with a mode throughput of 18 and 90th percentile throughputs similar to the capacities.The theoretical capacity for 3.3 miles separation at three runways with 130-knot nominal landing speed is 29 aircraft per quarter hour, but that is highly unlikely to be sustainable for extended periods of time.Therefore arrival traffic scenarios were created with quarter hourly rates ranging from 20 to 28.A throughput of 28 falls within the 10% increase in throughput that precision scheduling could be expected to enable over the highest commonly observed capacity of 26.The total time a flight could be scheduled to spend within the route model (entry to runway) ranged between 18 and 30 minutes depending on the route option chosen and speed control delay applied.Therefore, to ensure that the route model was fully populated with flights at the time of reconfiguration notification, t N was set to 30 minutes from the start of the scenario.The reconfiguration time t R varied 30 and 60 minutes after the start of the scenario in 5 minutes increments.Scenarios where t R =t N were expected to be the most disruptive reconfigurations with no advanced notice.In today's operations, this would be similar to a missed approach triggering the decision to reconfigure.Scenarios where t R =60 allowed 30 minutes lead-time for all flights occupying the route model at t N to land as originally scheduled, rescheduling only flights that had not yet entered the route model.Thus, the 30-minute lead time scenarios enter the realm of strategic reconfiguration.A total of 100 traffic scenarios per arrival rate were generated.This number was chosen to generate fairly stable average results without making the simulation data collection time over the range of variables to large.For each traffic condition, reconfiguration scheduling was simulated for the range of reconfiguration lead-times for both PX to WF and WF to PX reconfiguration directions.Table 1 summarizes the range of scenario conditions simulated 100 times each. +D. MetricsWhen a tactical or short notice reconfiguration occurs, the expected system impacts are temporary changes in throughput, and delay.Throughput represents the performance of the system as a whole, whereas delay can identify where the impact to the system was absorbed.In some simulations, a feasible reschedule solution cannot be found and the reconfiguration fails, which is a complexity indicator.No other metrics are collected for such simulations.Throughput and delay metrics were averaged across all successful simulations within the same variable subset. +Success RateThe reschedule success rate is the percentage of 100 simulations within the same variable subset that succeeded and for which all other metrics were collected. +ThroughputThroughput is measured as the number of final STAs for a particular scheduling point per quarter hour.Airport arrival throughput was measured as a combination of each arrival runway throughput per quarter hour.Throughput differences per minute were computed between each reconfiguration simulation and the corresponding baseline simulation of the same traffic scenario without any reconfiguration. +DelayTotal delay is runway ETA minus STA.However delay can be absorbed prior to the entry point (Center delay) or between the entry point and runway (TRACON delay).Center delay is entry the point ETA minus the STA.TRACON delay can be further broken down into path delay or speed delay.Recall that the assigned path with the earliest runway STA may not be the path with the earliest runway ETA.This means that the nominal time-to-fly (TTF) of the assigned route is longer than the preferred shortest path TTF.Path delay is the difference between the assigned path TTF and the shortest path nominal TTF.Speed delay is the difference between the scheduled TTF and nominal TTF of the assigned path.For aircraft requiring rescheduling within the TRACON due to reconfiguration, the shortest path from the old configuration and assigned path from the new configuration are used in the delay calculations.In the case that the new configuration assigned path is shorter than the old configuration shortest path, path delay is negative.Delay metrics were averaged across all aircraft with runway STA in the same quarter hour.Delay differences per minute were computed between each reconfiguration simulation and the corresponding baseline simulation of the same traffic scenario without any reconfiguration. +III. Results +A. Success RateThe reschedule success rate was most dependent on the notification lead-time and arrival rate.No noteworthy difference was found between rescheduling prioritization methods or reconfiguration direction.Figure 6 shows the rescheduling success rate vs. arrival rate for each lead-time (t R -t N ).Each success rate represents 400 simulations (100 for each reconfiguration direction and rescheduling prioritization method combination).The lowest and highest few lead-times were less sensitive to arrival rate.Almost no rescheduling attempts with 0 lead-time succeeded, whereas most or all rescheduling attempts with 20 minutes or greater lead-time succeeded.Lead-times between 5 and 15 minutes had higher success rates at low arrival rates.The success rates then degraded as the arrival rate increased.A failed reschedule does not mean that no rescheduling exists.It only means that the heuristic firstcome-first-served rescheduling based on the given flight sequence according to priority method, reached a point where the next flight served had no fixed path or speed solution within 10% of nominal.In many cases, the same arrival rate scenario with the same leadtime succeeded for one prioritization method and failed for the other.The rescheduling prioritization by original runway STA accounted for vast majority of the few successful reschedules with 0 lead-time.On the other hand, prioritization by new runway ETA had more success with lead-time between 5 and 15 minutes.Rescheduling success rate could be increased by applying a near-optimal rather than heuristic scheduling method, increasing path options, or expanding TTF bounds along path segments.Based on the low rescheduling success rate of 0 lead-time simulations, metrics from the few 0 lead-time successes were not included in the remaining analyses. +B. ThroughputA quarter hourly final runway STA throughput difference was calculated every minute and then smoothed with a 15 minute running average.Figure 7 shows the throughput difference results for each lead-time for one set of parameters (WF to PX, arrival rate = 27, original STA rescheduling priority).This shows the typical throughput response to reconfiguration.The 25-and 30-minute lead-time results hardly deviate from 0 (the 30minute lead-time overlaps the 25-minute curve), indicating that the reconfiguration had little to no effect on throughput.The other results have a period where the reconfiguration throughput is lower than that of the stable configuration, after which the reconfiguration throughput may be higher to recover from the recent throughput reduction and built up center delay.As expected, there is a general increase in throughput stress as arrival rate increases and lead-time decreases.However, there are also differences in magnitude between reconfiguration direction and rescheduling prioritization methods.Figure 9 shows the throughput stress magnitude difference between reconfiguration direction, and Fig. 10 shows the difference between rescheduling prioritization methods.With the exception of prioritizing by original runway STA with 5-minute lead-time, changing from WF to PX tends to stress throughput more than changing from PX to WF.This could be due to the complexity of transitioning from a 3-paralell runway configuration to 2-paralell plus 1-diagonal runway configuration.Rescheduling by the original runway STA tends to stress throughput more than rescheduling by the new runway ETA, especially when changing from PX to WF with lower lead-time.The throughput results suggest that in most cases, rescheduling prioritizing by new ETA is preferred over prioritizing by original STA.If the new ETA prioritization is used, 15 minute or greater lead-times will have a minimal impact on throughput at any arrival rate acceptable for stable configuration operations.Even at 10 minutes lead-time, the impact on throughput is low if the arrival rate is at or below 25 aircraft per quarter hour. +C. DelayAs with throughput, quarter hourly average delay differences were calculated every minute and then smoothed with a 15 minute running average.Although at first glance the reconfiguration directions appear to have very different reactions to the reconfiguration, this is only because in stable configuration PX generates less average path and speed delay than WF.Therefore, the PX to WF reconfiguration transitions from a lower to higher average delay and the WF to PX reconfiguration transitions from a higher to lower average delay.In both cases, the peak delay difference magnitudes increase as reconfiguration lead-time decreases.The peaks appear in sequence: first speed, then path, and finally center delay.The negative speed delay peak results from shifting speed delay to path delay as flights rescheduled within the TRACON are assigned new paths.When path delay peaks and speed delay saturates, then additional delay is passed to the center until natural gaps in the arrival demand return delay levels to what is normal for the new configuration.Because PX is not able to absorb as much path and speed delay, more delay is passed to the Center during the WF to PX reconfiguration.However, the total delay peaks are still similar between configuration change directions.It is important to note that all 30 and 25-minute lead-time delay results achieved very smooth transitions with no transient response as they stabilized to the new configuration's nominal delay level.In this experiment these leadtimes resulted in strategic reconfigurations.The lead-time was great enough that flights needing rescheduling were always on a path common to both configurations, making the transition between configuration routing seamless.To focus on the more transient response associated with tactical rescheduling, the remaining lead-time results were compared and analyzed relative to the 30-minute lead-time results.Figure 12 shows the maximum deviations of 5 to 20-minute lead-time total delay from the 30-minute lead-time total delay.In figure 12, separate graphs are shown for each lead-time, and reconfiguration direction and rescheduling prioritization methods are compared within each graph.Keep in mind that results for high arrival rates at 5-minute lead-time may be skewed due to the very low success rate and therefore lack of data for these cases.For the 10 to 20-minute lead-times, WF to PX reconfigurations produce consistently higher peak average delay differences which is independent of arrival rate until 27 and 28 aircraft per quarter hour are reached.In general, there are higher peak average delays for original STA prioritization than new ETA prioritization.The difference becomes more pronounced as the system is stressed with higher arrival rate and shorter lead-time.Figure 13 shows the maximum deviations of delay from the 30-minute lead-time delay decomposed into center, path, and speed delay on the same scale.The 20-minute lead-time results are not included as they are too small to view in the same scale.The impact of shorter lead-time on the delay response to reconfiguration amplifies for each type of delay as arrival rate increases.There is no significant speed delay difference between reconfiguration direction and rescheduling prioritization.On the other hand, for path delay, there is a very consistent difference between reconfiguration directions and a more subtle difference between rescheduling prioritization that amplifies as arrival rate increases.This is due to the difference in transition path options available to each reconfiguration.The center delay responses between reconfiguration direction and rescheduling prioritization tend to deviate from one another as arrival rate increases and lead-time decreases.This is a reflection of both the reconfiguration direction's capacity for path and speed delay, and the total delay impact of the rescheduling prioritization method.The delay results support the throughput analysis conclusions in that prioritizing by new ETA is preferred over prioritizing by original STA.The delay results also suggest that the reason there is little impact to throughput with lead-times of 10 minutes or greater is that the path and speed delay capacity of the reconfiguration transition routing at this lead-time is large enough that very little delay is passed to the center.Once center delay starts to build, throughput starts to become heavily impacted.Even though throughput may be robust, the 10-20 minutes lead-times still have a negative impact on individual flight efficiency.The shorter the lead-time, the longer aircraft are forced to fly at low speeds due to the constraint that flights may only reduce speed and limitation on transition routing available. +IV. ConclusionThis work developed a model of fixed path routing for reconfiguration between ORD's top two peak traffic configurations.A first-come-first-served multi-point scheduler was adapted to the reconfiguration problem by prioritizing the rescheduling of aircraft within the TRACON at the time of reconfiguration notification.Results suggest that a rescheduling prioritization based on earliest ETA in the new configuration is preferred over prioritization based on the original STA assigned in the old configuration.Neither method was 100% successful in rescheduling below 25 minutes lead-time, especially at higher arrival rates.A relaxation of the separation constraints (temporarily smaller scheduling buffer) or increase in speed control range or path options could increase the success rate.From a system impact perspective, reconfigurations with lead-times as short as 10 minutes at a nominal stable configuration arrival rate (~25 arrivals per quarter hour) could be accommodated with little impact to throughput.As lead-time shortened below 25-minutes, individual aircraft efficiency quickly degraded due to extra flight time at lower speeds.The reconfiguration generally had less impact in one reconfiguration direction than another due to the inherently different and asymmetrical route designs.Changing from the 3-parallel-runway configuration to the 2-parallel-plus-1-diagonal-runway configuration was more difficult than visa versa.This work demonstrates that first-come-first-served arrival scheduling of fixed routing for tactical reconfiguration is promising if at least 10-15 minutes lead-time is given before the reconfiguration is in effect.More thought must be given to transition routing design if this scheduling approach is to accommodate shorter lead-times.Figure 2 .2Figure 2. Flight tracks and published STARs and IAPs. +Figure 3 .3Figure 3. Integrated route model. +Figure 4 .4Figure 4. Integrated route model TRACON detail. +Figure 5 .5Figure 5. Transition routes designed between 04R and 27R. +Figure 6 .6Figure 6.Rescheduling success rate vs. arrival rate per lead-time. +Figure 7 .7Figure 7. Sample throughput difference response per lead-time. +Figure 88Figure8compares the peak throughput stress magnitude, which is the throughput difference when the reconfiguration drops below the stable configuration the most. +Figure 8 .8Figure 8.Throughout stress peak magnitude. +Figure 9 .9Figure 9. Throughput stress peak magnitude difference between reconfiguration direction. +Figure 11 compares center, path, speed, and total delay difference results for each lead-time between reconfiguration direction for a single set of remaining parameters (arrival rate = 25 aircraft per quarter hour, original STA rescheduling priority). +Figure 10 .10Figure 10.Throughput stress peak magnitude difference between rescheduling prioritization method. +Figure 11 .11Figure 11.Delay comparison by reconfiguration direction. +Figure 12 .12Figure 12.Total delay max deviation from 30-min lead-time results. +Figure 13 .13Figure 13.Decomposed delay maximum deviation from 30-min lead-time results. +Figure 1. Top two ORD arrival runway configurations.Plan X Trip (PX)West Flow (WF)9L27R27L10284R* STARs and IAPs were extracted from En Route Automation Modernization (ERAM) adaptation data (http://www.faa.gov/air_traffic/technology/eram/)accessed in July, 2013.Downloaded by NASA AMES RESEARCH CENTER on June 20, 2014 | http://arc.aiaa.org| DOI: 10.2514/6.2014-3152 +Table 1 . Scenario variables1VariableMin valueMax valueIncrementTotalReconfiguration directionPX to WFWF to PX-2Reschedule PriorityOriginal STANew ETA-2Arrival Rate20 aircraft/qtr28 aircraft/qtr1 aircraft/qtr9Reconfiguration lead-time (t R -t N ) 0 min30 min5 min7 + Downloaded by NASA AMES RESEARCH CENTER on June 20, 2014 | http://arc.aiaa.org| DOI: 10.2514/6.2014-3152 + + + + +AcknowledgmentsThis work was funded by the Concept and Technology Development Project, which is part of NASA's Airspace Systems Program. + + + + + + + + + Empirical Evaluation of Impact of Reconfiguration Events on Airport Capacity + + GeorgeHunter + + 10.2514/6.2012-5534 + AIAA-2012-4434 + + + 12th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference and 14th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference + Indianapolis, Indiana + + American Institute of Aeronautics and Astronautics + 2012 + + + Hunter, G., "Empirical Evaluation of Impact of Reconfiguration Events on Airport Capacity," 12 th AIAA Aviation Technology, Integration, and optimization (ATIO) Conference and 14 th AIAA/ISSM, AIAA-2012-4434, Indianapolis, Indiana, 2012. + + + + + Natural gamma aeroradioactivity map of the Fort Worth-Dallas area, Texas + + KRoach + + 10.3133/gp696 + + 2011 + US Geological Survey + Fort Worth, Texas + + + NASA North Texas Research Station (NTX) + + + Unpublished, for access contact the author at keenan.r.roach@nasa.gov + Roach, K., "Procedures for Changing the Flow Direction at Dallas/Fort Worth International Airport," NASA North Texas Research Station (NTX), Fort Worth, Texas, 2011 (Unpublished, for access contact the author at keenan.r.roach@nasa.gov). + + + + + Procedures for Changing the Flow Direction at Chicago O'Hare International Airport + + KRoach + + + 2013 + Fort Worth, Texas + + + NASA North Texas Research Station (NTX) + + + Unpublished, for access contact the author at keenan.r.roach@nasa.gov + Roach, K., "Procedures for Changing the Flow Direction at Chicago O'Hare International Airport," NASA North Texas Research Station (NTX), Fort Worth, Texas, 2013 (Unpublished, for access contact the author at keenan.r.roach@nasa.gov). + + + + + Benefits Assessment for Tactical Runway Configuration Management Tool + + RosaMOseguera-Lohr + + + NipaPhojanamongkolkij + + + GaryWLohr + + + JamesWFenbert + + 10.2514/6.2013-4395 + AIAA- 2013-4395 + + + 2013 Aviation Technology, Integration, and Operations Conference + Los Angeles, California + + American Institute of Aeronautics and Astronautics + 2013 + + + Oseguera-Lohr, R., N. Phojanamongkolkij, G. Lohr, and J. Fenbert, "Benefits Assessment for Tactical Runway Configuration Management Tool," 13 th AIAA Aviation Technology, Integration, and optimization (ATIO) Conference, AIAA- 2013-4395, Los Angeles, California, 2013. + + + + + Decision Support for Optimal Runway Reconfiguration + + XiaoliBai + + + PadmanabhanKMenon + + 10.2514/6.2013-4397 + AIAA-2013-4397 + + + 2013 Aviation Technology, Integration, and Operations Conference + Los Angeles, California + + American Institute of Aeronautics and Astronautics + 2013 + + + Bai, X. and P.K. Menon, "Decision Support for Optimal Runway Reconfiguration," 13 th AIAA Aviation Technology, Integration, and optimization (ATIO) Conference, AIAA-2013-4397, Los Angeles, California, 2013. + + + + + Robust Optimization Model for Runway Configurations Management + + RZhang + + + RKincaid + + AIAA-2011-6922 + + + 11 th AIAA Aviation Technology, Integration, and optimization (ATIO) Conference + Virginia Beach, Virginia + + 2011 + + + Zhang, R. and R. Kincaid, "Robust Optimization Model for Runway Configurations Management," 11 th AIAA Aviation Technology, Integration, and optimization (ATIO) Conference, AIAA-2011-6922, Virginia Beach, Virginia, 2011. + + + + + A Queuing Framework for Terminal Area Operations + + MonishTandale + + + VeeraVaddi + + + SandyWiraatmadja + + + VictorCheng + + 10.2514/6.2011-6363 + AIAA-2011-6363 + + + AIAA Guidance, Navigation, and Control Conference + Portland, Oregon + + American Institute of Aeronautics and Astronautics + 2011 + + + Tandale, M., S. Vaddi, S. Wiraatmadja, and V. Cheng, "A Queuing Framework for Terminal Area Operations," AIAA Guidance, Navigation, and Control Conference, AIAA-2011-6363, Portland, Oregon, 2011. + + + + + Design and evaluation of the terminal area precision scheduling and spacing system + + HSwenson + + + JThipphavong + + + ASadovsky + + + LChen + + + CSullivan + + + LMartin + + + + 9th USA/Europe Air Traffic Management R&D Seminar, FAA and EUROCONTROL + Berlin, Germany + + 2011 + + + Swenson, H., J. Thipphavong, A. Sadovsky, L. Chen, C. Sullivan, and L. Martin, "Design and evaluation of the terminal area precision scheduling and spacing system," 9th USA/Europe Air Traffic Management R&D Seminar, FAA and EUROCONTROL, Berlin, Germany, 2011. + + + + + Investigating the impact of off-nominal events on high-density ‘green’ arrivals + + ToddJCallantine + + + ChristopherCabrall + + + MichaelKupfer + + + LynneMartin + + + JoeyMercer + + + EverettAPalmer + + 10.1109/dasc.2011.6096195 + + + 2011 IEEE/AIAA 30th Digital Avionics Systems Conference + Seattle, Washington + + IEEE + 2011 + + + Callantine, T. J., C. Cabrall, M. Kupfer, L. Martin, J. Mercer, E. A. Palmer, "Investigating the impact of off-nominal events on high-density 'green' arrivals," 30th Digital Avionics Systems Conference, IEEE, Seattle, Washington, 2011. 10 + + + + + Development and evaluation of the terminal precision scheduling and spacing system for off-nominal condition operations + + HNSwenson + + + Jaewoo Jung + + + JThipphavong + + + Liang Chen + + + LMartin + + + JNguyen + + 10.1109/dasc.2012.6382303 + + + + 2012 IEEE/AIAA 31st Digital Avionics Systems Conference (DASC) + Williamsburg, Virginia + + IEEE + 2012. January 2013 + + + 31st Digital Avionics Systems Conference + Swenson, H., J. Jung, J. Thipphavong, L. Chen, L. Martin, and J. Nguyen, "Development and evaluation of the terminal precision scheduling and spacing system for off-nominal condition operations," 31st Digital Avionics Systems Conference, IEEE, Williamsburg, Virginia, 2012. 11 Federal Aviation Administration, Aviation System Performance Metrics (ASPM) Airport Efficiency, URL: http://aspm.faa.gov [cited 7 January 2013]. + + + + + A graph-based approach to nominal terminal routing + + SZelinski + + 10.1109/dasc.2012.6382333 + + + 2012 IEEE/AIAA 31st Digital Avionics Systems Conference (DASC) + Williamsburg, Virginia + + IEEE + 2012 + + + 12 Zelinski, S., "A Graph-Based Approach to Defining Nominal Terminal Routing," 31st Digital Avionics System Conference, IEEE, Williamsburg, Virginia, 2012. + + + + + A Closed-Form Solution to Multi-Point Scheduling Problems + + LarryMeyn + + 10.2514/6.2010-7911 + + + + AIAA Modeling and Simulation Technologies Conference + Toronto, Ontario Canada + + American Institute of Aeronautics and Astronautics + 2010. 2012 + 7110 + + + 14 Federal Aviation Administration. 65U Chapter 5. Section 5. Radar Separation. cited 11 May 2012 + Meyn, L., "A Closed-Form Solution to Multi-Point Scheduling Problems," AIAA Modeling and Simulation Technologies Conference, AIAA, Toronto, Ontario Canada, 2010. 14 Federal Aviation Administration, 2012, Air Traffic Control FAA Order JO 7110.65U Chapter 5. Section 5. Radar Separation, URL: http://www.faa.gov [cited 11 May 2012]. + + + + + + diff --git a/file833.txt b/file833.txt new file mode 100644 index 0000000000000000000000000000000000000000..978b8930dbe244987da4ce85387f1e39db63d1fb --- /dev/null +++ b/file833.txt @@ -0,0 +1,188 @@ + + + + +I. Introduction +A. BackgroundThe Airspace Concept Evaluation System(ACES) is a National Airspace System(NAS)-wide fast-time simulation tool developed at NASA Ames Research Center. 1 ACES models and simulates the NAS using interacting agents representing center control, terminal flow management, airports, individual flights, and other NAS elements.These agents pass messages between one another similar to real world communications.This distributed agent based system is designed to emulate the highly unpredictable nature of the NAS, making it a suitable tool to evaluate current and envisioned airspace concepts.To ensure that ACES is producing the most realistic results, the system must be validated.There is no way to validate future concepts scenarios using real world historical data, but current day scenario validations increase confidence in the models and the validity of future scenario results.Each operational day has unique weather and traffic demand schedules.The more a simulation utilizes the unique characteristic of a specific day, the more realistic the results should be.ACES is able to simulate the full scale demand traffic necessary to perform a validation using real world data.Through direct comparison with the real world, models may continue to be improved and unusual trends and biases may be filtered out of the system or used to normalize the results of future concept simulations.There have been many domain specific attempts at simulation validation using real world data such as with noise impacts 2 and airport operations 3 models.Very few system-wide simulations being developed today have published simulation validation results.One system-wide scale simulation that has performed some validation using real world data is MITRE's Jet:Wise. 4Jet:Wise is an agent based simulation of airline activity designed to predict future airline trends.The inputs include observed fleet mix, schedules, and fares from historical data.Validation outputs consist of observed passenger counts tested against each of the various airlines, routes, and markets.Similar methods are used to validate ACES from an airspace operations standpoint.Weather and traffic demand schedules serve as ACES inputs.ACES outputs are then compared to real world operational metrics and delay statistics to conduct the system validation. +B. Summarizing The ProcessThe validation process involves extracting both ACES inputs and output comparison data from real world data.Figure 1 shows a diagram of the basic validation process.ACES inputs from real world data included flight plans and airport capacities.ACES outputs to be compared with real world data included airport throughput rates, delay statistics, and individual flight tracks.ACES requires various input data to simulate the unique weather and traffic demand schedules representing a given day in the NAS.A demand schedule and set of flight plans are compiled by combining information from various NAS messages for each flight.The NAS messages are available through Aircraft Situation Display to Industry (ASDI) data. 5Although ACES does not model weather explicitly, ACES is able to model the effects of weather in the form of varying airport capacities throughout the day.These airport capacities are extracted from Aviation System Performance Metrics (ASPM) data. 6CES models vary in fidelity and complexity.For an initial validation of the system, only the most basic functionality is utilized.This functionality includes lateral enroute delay maneuvers, minimum separation at departure meter fix crossings, airport and sector constraints, and using winds.As agents representing elements of the NAS interact, ACES collects every message passed between them such as aircraft state messages and flight time messages.The data extracted from these messages may be as detailed or high level as the research requires.Thus validation comparison is limited only by the fidelity of the ACES models being used and real world data availability.ASPM provides high level summary statistics at major US airports which may be compared to ACES scheduled and actual throughput levels at those airports.Throughput comparisons show that ACES throughput levels are comparable to that of ASPM.The Bureau of Transportation Statistics (BTS) provides delay statistics for individual flights for certain airlines.These flight delays may be compared with their ACES counterparts.The results show comparable delays with minor variations in phase of flight.Finally, individual flight NAS tracking messages from ASDI may be compared with ACES aircraft state information to validate individual flight trajectories.ACES models individual flight trajectories extremely well when the real flight follows the flight plan.Deviations occur when the real flight modifies it's original flight plan with direct-to segments. +II. Preprocessing ACES InputsACES requires various types of inputs to simulate a realistic day in the NAS.Static inputs include center and sector boundaries and capacities, airport locations, and unimpeded airport taxi times.These inputs do not change often and only need occasional updating.Other inputs such as the schedule of flight plans or flight data set, varying airport states and capacities, and winds are unique for each day to be simulated.Real world data sources for the flight data set and airport states and capacities require a degree of preprocessing before using them in ACES.This section discusses ACES input data requirements and the real world data sources and preprocessing required to prepare them.It should be kept in mind that real world data quality can itself be questionable making validation using such data difficult. +A. Flight Data SetAn ACES flight data set (FDS) is a set of flight plans including origin and destination airports, scheduled departure time, aircraft type, proposed cruise speed and altitude, and trajectory way points.ASDI is an excellent real world source for this data.ASDI is a subsystem of the Enhanced Traffic Management System distributed as a text feed of raw NAS messages so no text decoding is required.NAS messages are detailed messages about the flight plan, flight plan amendments, departure, trajectory tracking, boundary crossings, arrival, or cancellation.Collectively ASDI NAS messages provide all the information needed to create a flight data set for ACES.Some ASDI information may need manipulation or combining with other source data to be used as ACES inputs.ASDI routes are delivered as a series of fixes and jet ways 7 and ACES uses way point trajectories.Fix and jet way locations 8-10 are used to convert the last given route before takeoff to a series of latitude/longitude way points.Anticipating using BTS delay statistics for validation, flights extracted from ASDI may be paired and synchronized with BTS flights.When ASDI scheduled departure time varies from BTS scheduled departure time, the BTS time is used.Any BTS flight that is not paired with an ASDI flight may also be added to the data set.BTS data only provides the flight ID, origin and destination airports, and scheduled departure time required for the flight plan.Remaining flight plan information may be extrapolated using flights with similar airport pairs and airlines as guides.Febuary 19, 2004 was chosen for simulation.It was a high traffic and fair weather day.A total of 56850 flights were extracted from ASDI and BTS for Febuary 19, 2004.Only 3% of the flights required some extrapolation to complete the flight plan.ACES configuration filters some flights for various reasons.ACES's international flight capability is currently limited to flights either originating or terminating within domestic airspace.International overflights are filtered.The airport model used in this validation has a generic circular Terminal Radar Approach Control (TRACON) with 40 mile radius.The origin and destination airport of a flight may not have intersecting TRACONs.Therefore, each origin destination airport pair must be at least 80 miles apart.Overflights and intersecting TRACON flights consist of 16% of the original data set.Another 2.6% are filtered due to ACES identified trajectory anomalies.The total number of simulated flights is 46243, a subset of the original 56850. +B. Airport States and CapacitiesAdverse weather conditions are not directly modelled in ACES but the effects of weather on airport capacity can be modelled to a certain degree with airport states.Different airport capacities are used for visual flight rules (VFR) when visibility-distance and ceiling-altitudes are good and instrument flight rules (IFR) when adverse weather reduces ceiling and visibility.Airport capacity refers to the maximum number of takeoffs, landings, and total operations an airport is able to handle.The total operations capacity is usually less than the sum of the takeoff and landing capacities creating a Pareto curve of feasible takeoff and arrival rates.ASPM provides airport demand, actual throughput, operation rates and state (VFR or IFR) with quarter hour resolution for the top 53 continental US airports.The actual departure and arrival rates may be used as maximum capacity values with total operations equalling the sum of arrivals and departures.This would force ACES to use the exact operations rates per quarter hour reported by ASPM.However, airport throughput is often higher than the reported operational rate during peak efficiency periods.Figure 2 shows ASPM published departure statistics for Fort Lauderdale International Airport (KFLL) on 2/19/2004.Notice how the reported departure rate is more a reflection of average throughput than the actual departure rate the airport performs.ACES treats airport capacities as hard maximums.Peak throughput periods would not be modeled properly if reported operation rates were used.To model the most realistic throughput, these peak maximums may be used as ACES maximum operations capacities.Using maximum hourly throughput rates instead of quarter hourly throughput rates removes especially high throughput rates that cannot be sustained for more than a quarter hour.The hourly average departures for KFLL are shown in figure 2 for comparison with the quarter hourly departures.The curve in the Pareto curve is maintained by collecting maximum departure, arrival, and total throughput rates separately.The sum of departure and arrival capacities may be greater than the total capacity.Thus, ACES Traffic Flow Management (TFM) agents are free to alter the airport operation mode between departures and arrivals as the flight schedule demands.Figure 3(a) shows the hourly actual departures, arrivals, and total operations throuput rates for one day at KFLL.Each operation maximum may not occur over the same hour.Therefore, even though at any one hour, the total operations throughput may equal the sum of departures and arrivals, the maximum total operations may be less than the sum of the maximum departures and arrivals per hour.Figure 3(b) shows the resulting Pareto curve for this example.The maximum hourly throughput rates used in this validation were collected between 2/12/2004 and 2/26/2004.Throughput rates were also collected separately for VFR and IFR airport states.If any airports VFR throughput is lower than its IFR throughput, the VFR capacity is raised to at least equal the IFR maximum capacity.Using maximum throughput rates as airport capacities allows ACES' conservative airport Traffic Flow Management model to simulate realistic throughput levels at the top 53 continental U.S. ASPM provided airports.The remaining airports are given generic airport state capacities based on similar runway configurations.Smaller airports usually have less difference between VFR and IFR capacities and it is less likely for a current day demand size to exceed their capacities.Therefore, all airports other than ASPM are considered to be in VFR state all day.Choosing a good weather day like 2/19/2004 increases the likelihood that these airports are actually operating in VFR conditions.A total of 1669 domestic U.S. airports and 160 other airports including international, Alaskan, and Hawaiian airports were simulated as the origin-destination pairs for the 46243 simulated flights. +C. Other ACES Inputs and ParametersThe functionality used for this initial validation includes airport and sector constraints, lateral enroute delay maneuvers, minimum separation at departure meter fix crossings, and winds.Airport and sector capacities constrain the system by capping airport operation throughput rates and sector densities.These constraints drive the delay in the system which may occur at the departure gate, airport taxi surfaces, TRACON, and enroute airspace.Lateral delay maneuvers allow throughput constraints at the airports and sectors to propagate through the enroute airspace.Departure meter fix separation ensures that aircraft adhere to minimum separation requirements as they transition from the TRACON to enroute airspace.Environmental conditions in the NAS are highly dynamic and fluctuate throughout the day.Winds affect to a certain degree the routes proposed for each flight as well as the estimated enroute time.Rapid Update Cycle (RUC) data 11 provides information about en-route winds and is used as a direct input to ACES.The winds are then used in ACES to produce realistic ground speeds given a flights desired cruising airspeed.ACES is capable of addition functionality such as conflict detection and resolution, airline operations control, rerouting, and enhanced terminal area models.These were not utilized in this validation for simplicity. +III. Validation ResultsThis sections discusses the various comparisons made with the available real world data.First airport throughput comparisons were made using ASPM data.Then BTS delay and flight times were compared to their ACES counterparts.Finally, ACES aircraft state data was compared with ASDI individual flight tracks. +A. Airport Throughput Comparisons Using ASPM DataACES collects the time of various stages of flight for each simulated flight which can be used to calculate transit time and delay.Both scheduled and actual times for gate departure, takeoff, departure and arrival meter fix crossing, landing, and gate arrival are captured accurate to the second.Figure 4 shows a diagram of these flight times and how the transit times between them are defined.ASPM publishes flight sums at each of its 53 continental US airports for metrics defined as departure and arrival demands, and actual departures and arrivals per quarter hour.The actual departures and arrivals are equivalent to wheels off and on or actual ACES takeoffs and landings as defined in figure 4. ASPM departure and arrival demands are similar but not quite equivalent to scheduled ACES takeoffs and landings.ASPM counts a scheduled operation any time between it's scheduled and actual wheels off or on time.Therefore, flights with taxi-out delay may be counted for more than one departure demand quarter hour and flights with enroute delay may be counted for more than one arrival demand quarter hour.This is not a true measure of scheduled throughput because the metric is affected by delay which will be compared separately.Figure 5 shows total quarter hourly scheduled and actual takeoffs and landings for the 53 continental US ASPM airports from ACES output data and ASPM.For both takeoffs and landings, ASPM demands are higher than the actual throughput rates illustrating how system delays affect the demand metric.ACES scheduled and actual operations are much closer together.Departure delays cause the actual takeoffs curve to be smoother than the scheduled takeoffs curve.ACES correlation with ASPM varies from 99.1% to 99.6% between scheduled and actual takeoffs.Both scheduled and actual ACES landings correlate 98.3% with ASPM.ACES succeeded in representing 97.9% of ASPMs takeoffs and 92.6% of ASPMs arrivals within the 24 hour period.Individual airport correlations ranged from 73.8% (KBUR) to 97.6% (KDFW) averaging 91.1% for actual takeoffs, and they ranged from 65.4% (KBUR)to 96.6% (KLGA) averaging 84.7% for actual landings.The low correlating airports tended to be lower throughput airports like Burbank, CA (KBUR). +B. Flight Time And Delay Comparisons Using BTS DataBTS provides delay statistics at the individual flight level for certain airlines.It is a much better source for delay and transit time analysis than ASPM summary statistics.Performing summary statistics for all BTS flights and their corresponding subset of ACES flights ensures that the data is comparable and averages will normalize correctly.Due to low fidelity ACES surface modelling and the lack of BTS individual flight unimpeded taxi times, analysis of taxi transit time and delay is moot.More focus is placed on the en-route domain and gate departure delay analysis.Results are skewed by the fact that ACES does not simulate negative delay and over 50% BTS recorded flights for 2/19/2004 departed the gate before the scheduled time.For this reason ACES has 2.4 times as many flights with no delay.It is apparent from comparing numbers of flights with positive delay that ACES overestimates gate departure delay.High delay ACES outliers make a large impact on the mean delay difference of 10.28 minutes per flight.Trimming the sample set by 5% reduceds the mean to 5.44 minutes per flight.The distribution of individual flight delay difference between BTS and ACES around this mean can be seen in figure 6's bar graph.BTS does not include meter fix crossing times, nor may the real meter fixes be the same distance from the airport as generic ACES meter fixes.Therefore en-route analysis was done using in-flight transit times and delays as specified in figure 4. In this case the excessive delay in ACES is caused by a unanticipated relationship between the generic nodal airport model and departure meter fix separation functionality.Figure 9 (a) shows ACES's generic nodal airport model in action with just four departure meter fixes at North, South, East, and West and four arrival meter fixes at the corners.ACES assigns to a flight the closest departure meter fix to the flight's first en-route way point and the closest arrival meter fix to it's last en-route way point.This can be seen in figure 8 where the ACES trajectory intersects the airport TRACON boundaries.When using departure meter fix separation functionality, flights may be delayed at the departure meter fix to meet minimum separation requirements before entering the en-route domain.Airports with high departure throughput can result in flights with unrealistic TRACON delay.Figure 9 (b) shows the percent distance completion for flight AAL1372 with respect to time for ASDI and ACES.Departure meter fix separation induced TRACON delay is the source of ACES's 41 minutes of delay.While a small percent of ACES flights have this unrealistic TRACON delay, most of them are KORD departures followed by KDFW departures; two airports with the highest departure throughput rates.For both airports, one or two meter fixes receive the most departures and produce the majority of the TRACON delay.Both the nodal airport model and departure meter fix queueing work exactly as they were designed to.Increasing airport capacity does not increases meter fix capacity, and to avoid meter fix overload more meter fixes must be added, or meter fixes must be placed more strategically to balance departure distribution between the fixes.Current ACES development includes a terminal area model redesign where these insights are being used to establish requirements. +IV. Summary and ConclusionThis validation proved ACES to be an excellent modeler of the NAS for the level of fidelity utilized.ACES' adequately realistic recreation of current day NAS operations gives greater confidence to future operational concept evaluations using ACES.The validation produced several note worthy insights that will aid in future ACES development.Realistic traffic levels were achieved by generating flight data sets from ASDI NAS messages and using ASPM maximum hourly throughput rates to drive ACES airport capacities.ACES throughputs correlate with ASPM real world data between 98% and 99%.Delay analysis results show how common negative delay is in the real world.According to delay comparisons, ACES somewhat over estimates gate departure delay and actual in-flight transit time with 5% trimmed mean differences of 5.44 and 10.41 minutes respectively.However, ACES is an excellent modeler of scheduled in flight trajectories with a mean difference less than 10 seconds.Further analysis at the trajectory level, identified real world en-route short cuts and ACES departure meter fix bottlenecks at high throughout airports such as KORD and KDFW as probable causes for in flight transit time overestimation.The departure meter fix bottleneck illustrates how two independently designed models can interact in ways that were not intended.As these exposed issues are resolved, further validations will aid the development cycle and increase confidence in ACES.Comparisons by region and validations using different weather days will be especially helpful as new methods of implicitly modeled weather are developed.Figure 1 .1Figure 1.Diagram of the ACES validation process. +Figure 2 .2Figure 2. ASPM published departure statistics for airport KFLL. +Figure 3 .3Figure 3. (a) ASPM published actual departures, arrivals, and total operations per hour for KFLL.Maximums for this day are marked with large dots.(b) The Pareto curve produced when using hourly operation throughput maximums from (a). +Figure 4 .4Figure 4. Flight time and transit time definitions. +Figure 5 .5Figure 5.Total continental US ASPM airport takeoffs (a) and landings (b) per quarter hour smoothed by three quarter hours. +Figure 6 .6Figure 6.Bar graph of numbers of flights with x amount of gate departure delay accompanied by a chart of summary statistics. +Figure 66Figure6illustrates the gate departure delay analysis and displays gate departure delay summary statistics.Results are skewed by the fact that ACES does not simulate negative delay and over 50% BTS recorded flights for 2/19/2004 departed the gate before the scheduled time.For this reason ACES has 2.4 times as many flights with no delay.It is apparent from comparing numbers of flights with positive delay that ACES overestimates gate departure delay.High delay ACES outliers make a large impact on the mean delay difference of 10.28 minutes per flight.Trimming the sample set by 5% reduceds the mean to 5.44 minutes per flight.The distribution of individual flight delay difference between BTS and ACES around this mean can be seen in figure6's bar graph.BTS does not include meter fix crossing times, nor may the real meter fixes be the same distance from the airport as generic ACES meter fixes.Therefore en-route analysis was done using in-flight transit times and delays as specified in figure4.Figure7displays the differences between ACES and BTS in-flight transit time and delay.With a mean difference of 0.16 minutes per flight, ACES is an excellent modeler of mean scheduled in-flight transit time.ACES overestimates actual in-flight transit time and therefore in-flight delay. +Figure6illustrates the gate departure delay analysis and displays gate departure delay summary statistics.Results are skewed by the fact that ACES does not simulate negative delay and over 50% BTS recorded flights for 2/19/2004 departed the gate before the scheduled time.For this reason ACES has 2.4 times as many flights with no delay.It is apparent from comparing numbers of flights with positive delay that ACES overestimates gate departure delay.High delay ACES outliers make a large impact on the mean delay difference of 10.28 minutes per flight.Trimming the sample set by 5% reduceds the mean to 5.44 minutes per flight.The distribution of individual flight delay difference between BTS and ACES around this mean can be seen in figure6's bar graph.BTS does not include meter fix crossing times, nor may the real meter fixes be the same distance from the airport as generic ACES meter fixes.Therefore en-route analysis was done using in-flight transit times and delays as specified in figure4.Figure7displays the differences between ACES and BTS in-flight transit time and delay.With a mean difference of 0.16 minutes per flight, ACES is an excellent modeler of mean scheduled in-flight transit time.ACES overestimates actual in-flight transit time and therefore in-flight delay. +Figure 7 .Figure 8 .78Figure 7. Bar graph of numbers of ACES flights that differ by x amount with BTS in-flight statistics accompanied by a chart of summary statistics. +Figure 9 .9Figure 9. (a) ACES visualization of the nodal airport model implemented at KORD.(b) Percent flight progress for flight AAL1372 with respect to time for ASDI and ACES data. + + + + + + + + + Fast-Time NAS Simulation System for Analysis of Advanced ATM Concepts + + DouglasSweet + + + VikramManikonda + + + JesseAronson + + + KarlinRoth + + + MatthewBlake + + 10.2514/6.2002-4593 + AIAA-2002-4593 + + + AIAA Modeling and Simulation Technologies Conference and Exhibit + Monterey, California + + American Institute of Aeronautics and Astronautics + 2002. August + + + Sweet, D., Manikonda, V., Aronson, J., Roth, K., and Blake, M. 2002. Fast-Time Simulation System for Analysis of Advanced Air Transportation Concepts. In AIAA Modeling and Simulation Technologies Conference and Exhibit,August, Monterey, California, AIAA-2002-4593. + + + + + Airframe Noise Modeling Appropriate for Multidisciplinary Design and Optimization + + SerhatHosder + + + JosephSchetz + + + BernardGrossman + + + WilliamMason + + 10.2514/6.2004-698 + AIAA 2004-698 + + + 42nd AIAA Aerospace Sciences Meeting and Exhibit + Reno, Nevada + + American Institute of Aeronautics and Astronautics + 2004. January + + + Hosder, S., Schets, J., Grossman, B., and Manson, W. 2004. Airframe Noise Modeling Appropriate For Multidisciplinary Design And Optimization. In 42nd AIAA Aerospace Sciences Meetingand Exhibit, January, Reno, Nevada, AIAA 2004-698. + + + + + Airport, airspace, and NAS system capacity studies + + AlbertSchwartz + + + RobertHolladay + + + JohnVanderveer + + + AnthonyBradley + + 10.2514/6.1998-5553 + AIAA 1998-5553 + + + AIAA and SAE, 1998 World Aviation Conference + Anaheim, California + + American Institute of Aeronautics and Astronautics + 1998. September + + + Schwartz, A., Holladay, R., Vanderveer, J., and Brdley, A. 1998. Airport, Airspace, and NAS System Capacity Studies. In IAAA World Aviation Conference, September, Anaheim, California, AIAA 1998-5553. + + + + + The Jet:Wise Model of National Air Space System Evolution + + WilliamPNiedringhaus + + 10.1177/0037549704042029 + + + SIMULATION + SIMULATION + 0037-5497 + 1741-3133 + + 80 + 1 + + 2004. January + SAGE Publications + + + Niedringhaus, W. 2004. The Jet:Wise Model of National Airspace System Evaluation. In SIMULATION, Vol. 80, No. 1, January, pp. 45-58. + + + + + Health hazard evaluation report: HHE-81-042-832, Federal Aviation Administration, New York Air Route Traffic Control Center, Ronkonkoma, New York. + 10.26616/nioshhhe81042832 + no. ASDI-FD-001 + + + + William J. Hughes Technical Center Atlantic City International Airport + + August 2000. 2004. November 1997 + U.S. Department of Health and Human Services, Public Health Service, Centers for Disease Control, National Institute for Occupational Safety and Health + + + Report + NJ. 8 Federal Aviation Administration. National Airspace System Resource Aeronautical Data. ATO-R. Subscription + Aircraft Situation Display To Industry: Functional Description and Interface Control Document. August 2000. Report no. ASDI-FD-001. Volpe Center Automation Applications Division, Cambridge, MA. 6 Federal Aviation Administration Aviation Policy and Plans (APO) Data System. 2004. http://www.apo.data.faa.gov/index.html. 7 National Airspace System En Route Configuration Management Document: Computer Program Functional Specifica- tions Route Conversion and Posting. November 1997. Report no. NAS-MD-312. Federal Aviation Administration, Operational Support National En Route Automation Division, William J. Hughes Technical Center Atlantic City International Airport, NJ. 8 Federal Aviation Administration. National Airspace System Resource Aeronautical Data. ATO-R. Subscription. + + + + + Cuestionamientos al encierro en el devenir sujetos Diálogos del psicoanálisis con la pedagogía social Autores/as Ana Lía López Brizolara DOI: https://doi.org/10.36496/n130.131a5 Palabras clave: Adolescencia, Institución, ciencias sociales, sociedad, subjetivación, desubjetivación, abandono, resiliencia Resumen Este trabajo nos ofrece aportes psicoanalíticos para pensar los procesos de subjetivación en situaciones de encierro en niños y adolescentes. La autora dialoga con las propuestas de Diego Silva Balerio desde la Pedagogía Social para el cuidado de niños/as y adolescentes en situación de máxima vulnerabilidady las alternativas a la internación y la des-internación. Retoma la pregunta: ¿se pueden sustituir funciones llevadas a cabo por los familiares? ¿Qué se juega cuando se sustituyen? Propone pensar cómo en los Centros de internación se conjuga el efecto del dispositivo de poder y el efecto del desamparo estructural. Describe algunos aspectos de las adolescencias y su apremio ante las preguntas acerca de su existencia e identidad y cómo pensarlo en situaciones de encierro. Descargas Biografía del autor/a Ana Lía López Brizolara Miembro Asociado de la Asociación Psicoanalítica del Uruguay Citas Calmaestra Fernández, A. (2017). Giorgio Agamben, ¿Qué es un dispositivo? seguido de El amigo y de La Iglesia y el Reino, Eikasia, 75, 332-337. ISSN-e 1885-5679. Casas de Pereda, M. (1996). Metapsicología del objeto y los fenómenos transicionales. Revista Uruguaya de Psicoanálisis, 83, 50-62. Dobon. (2015). Duelos Congelados. En Consecuencias subjetivas del terrorismo de estado. Buenos Aires: Grama. Freud, S. (1895). Proyecto de Psicología. En Obras Completas. Vol. I. Publicaciones prepsicoanalíticas y manuscritos inéditos en la obra de Freud (pp. 325-393). Buenos Aires: Amorrortu. Freud, S. (1905). Tres ensayos de Teoría Sexual. En S. Freud, Obras Completas Tomo VII. Buenos Aires: Amorrortu. Giorgio Agamben. (2015). ¿Qué es un dispositivo? seguido de El amigo y de La Iglesia y el Reino. Barcelona: Anagrama. Kachinovsky, C. (2005). Multiplicidad de las identidades en un tiempo de exclusiones. Relaciones. Serie: Los diferentes VII. Recuperado de: http://www.chasque.net/frontpage/ relacion/0404/identidades.htm Kachinovsky, C. & Sopeña, A. (2004). Importancia de la música en el proceso adolescente. Música, rock y tribus urbanas. Recuperado de: https://www. apuruguay.org/sites/default/files/Kachinovsky- Sope%c3%b1a.pdf Leiva, J. L. (2004). Autorización y saber. Trabajo presentado en las Jornadas Aniversario años de Escuela (1974-2004). Escuela Freudiana de Buenos Aires, Buenos Aires. Recuperado de http://www.efba.org/efbaonline/ leiva-11.htm López Brizolara, A. L. (2003). Ser adolescente después de la modernidad. Recuperado de: https:// analialopezbrizolara.wordpress.com/2009/11/. López Brizolara, A. L. (2005). Ritualidades Contemporáneas en la adolescencia. A Cien Años de los «Tres Ensayos para una Teoría Sexual» Hospital Pereira Rosell. Montevideo: Revista APPIA. López Brizolara, A. L. (2011). Espacios creados, espacios conquistados. Pertenencias y procesos de subjetivación. Montevideo: Inédita. Maffesoli, M. (1990). El tiempo de las Tribus. Barcelona: Icaria Editorial. Schroeder, D. (2004). La confrontación adolescente hoy: Aspectos imaginarios y simbólicos. Revista APPIA, 15, 170-180. Schvarstein, L. (1991). Psicología social de las organizaciones. Nuevos aportes. Buenos Aires: Paidós. Silva Balerio, D. (2019). Experiencias de institucionalización en la adolescencia: ambivalencias entre una subjetivación cartográfica y la circulación social endogámica. XXXII Seminario de psicopedagogía Social, (págs. 140-144). Barcelona. Silva Balerio, D., & Domínguez Collete, P. (2017). Desinternar, sí. Pero ¿cómo? Montevideo, Uruguay: UNICEF. Wald, A. (2018). Notas sobre vulnerabilidad y desamparo en la infancia. Obtenido de APURUGUAY.ORG: https://www.apuruguay.org/ apurevista/2010/16887247201812708.pdf Winnicott, D. (1971). Realidad y Juego. Buenos Aires: Gedisa. + 10.36496/n130.131a5 + + + + revista uruguaya de psicoanálisis + RUP + 1688-7247 + + 130 + 2004. 2004. 2004 + Asociacion Psicoanalitica Uruguay + Morganville, NJ + + + LLC + 9 AirNav website. 2004. AirNav, LLC, Morganville, NJ http://www.airnav.com. 10 Digital Aeronautical Flight Information File website. 2004. https://164.214.2.62/products/digitalaero/index.cfm 11 Rapid Update Cycle website. National Weather Service Telecommunication Opperations Center. 2004. http://weather.gov/tg/modfiles.html. + + + + + + diff --git a/file834.txt b/file834.txt new file mode 100644 index 0000000000000000000000000000000000000000..8079432611216f777b75c7e3a0a9b7e752f5f307 --- /dev/null +++ b/file834.txt @@ -0,0 +1,509 @@ + + + + +INTRODUCTIONWith today's congested airports, a considerable amount of research has focused on increasing airport throughput.The throughput of a single arrival runway is ultimately constrained by how tightly flights can be spaced as they complete the final approach.Whereas wake separation requirements dictate a minimum allowable spacing for safety, the achieved spacing is nominally a half to one mile greater to account for imprecision associated with aircraft merging and wind uncertainties [1].The current practice of vectoring to properly space merging flows is highly flexible, but demands a high degree of controller attention, and produces relatively varied spacing results.Time-based arrival management concepts such as the 4D Cooperative ARrival MAnager (4D-CARMA) from German Aerospace Center (DLR) [2][3][4] and FAA's Trajectory Based Flow Management (TBFM) [5][6][7] make use of improved estimated-time-of-arrival (ETA) afforded by fixed routing, such as Area Navigation (RNAV) and Required Navigation Performance (RNP), and wind forecasts to develop precise arrival schedules.The question of spacing precision then becomes one of schedule conformance.The more precisely flights can be controlled to meet their scheduled times of arrival, the tighter flights may be scheduled.To this end, these time-based arrival managers employ controller decision support tools to help meet a schedule.Higher schedule conformance can be achieved by controlling aircraft to their scheduled time of arrival with speed control rather than vectoring [8].The increased schedule conformance allows reduced scheduling buffers and increased throughput.However, speed control alone offers a limited range of flexibility to respond to disturbances.For example, maximum speed restrictions make time recovery for late arrivals difficult in terminal airspace.A more feasible path-based method of time recovery could increase schedule conformance, allowing the precision schedulers to reduce their buffers and increase throughput even further.Whereas path-based control in the terminal area is not a new concept, when coupled with scheduling, flights are usually scheduled to the shortest paths, reserving longer paths for exception handling [7,9].However, an investigation offering standard path stretch and shortcut path options to controllers to recover from highly disruptive offnominal events, found that the shortcuts were among the most often used [10].Previous research explored tactical use of shortcut path options to allow time recovery for late arrivals with promising results [11].However, the study focused on a single generic scheduling point and assumed all aircraft had similar arrival precision at the scheduling point.This paper extends previous research by applying the concept of tactical shortcut path options to enhance TBFM arrival schedule conformance and thereby gain system throughput and delay benefits without increasing controller workload.Tactical shortcuts are integrated into a model of TBFM multi-point arrival scheduling at Los Angeles International Airport.TBFM development has placed considerable effort on integrating high precision procedures for equipped aircraft [17] to not only reap additional system benefits, but to incentivize equipage.Mixed aircraft scheduling conformance performance is considered in this study to explore what if any advantage tactical shortcuts can give to high arrival time precision aircraft.The next section summarizes the tactical shortcuts concept explored in previous research and the main findings. +II. BACKGROUNDA flight's schedule consists of scheduled times of arrival (STAs) at coordination points (e.g.meter fixes, merge points, runway thresholds) along a fixed nominal route.The STAs of sequential flights at a given coordination point are spaced to achieve at minimum the required separation between the flights plus an additional scheduling buffer to account for flight arrival time uncertainty.A suitable scheduling buffer can be calculated for a given arrival uncertainty and desired maximum likelihood of a separation violation requiring controller intervention.The tactical shortcuts concept assumes that at some point prior to the coordination point, late flights have the opportunity to take a shortcut path option and recover time.Previous research found that shortcuts designed to recover ~1.5 times the standard deviation of the flight arrival time uncertainty, could allow flights to be scheduled with smaller buffers without increasing the likelihood of separation violation [11].This equated to 11% increase in throughput during high traffic demand periods.During low traffic periods, more benefit was found in scheduling flights directly to the shortcut path with the original larger buffers if there was sufficient slack in the schedule.Essentially, the schedule strategically assigned a nominal or a shortcut path-based on how well the schedule could accommodate the respective buffer size requirements.Because required buffer size also depends on aircraft schedule conformance precision, a similar method may be used to account for mixed aircraft performance.Aircraft equipped to meet a schedule with a higher degree of precision require less buffer and so may be scheduled directly to the shortcut path under the same conditions that would schedule a less equipped aircraft requiring larger buffers to the longer nominal route.One objective of this paper is to apply the tactical shortcuts concept to multi-point scheduling modeled after a real terminal airspace.For this purpose, shortcut path options were designed for arrival routing into Los Angeles International Airport (LAX).The resulting routing provides shortcut opportunities at up to two separate merge points along a single route from meter fix to runway.Another objective of this paper is to explore what if any advantage this method of nominal vs. shortcut path scheduling can give to high arrival time precision aircraft that may incentivize advanced equipage. +III. METHOD +A. Scheduler ModelThe arrival scheduler was modeled after TBFM's terminal metering functionality.This is a multi-point first-come-firstserved constraint-modified scheduler [5].Flights are assigned an arrival runway and STAs at multiple coordination points along a fixed route (e.g. the arrival meter fix, runway threshold, and any merge points in between).Flights are scheduled on a first-come-first-served basis according to their earliest runway ETA with a planning horizon of ~40 minutes (~200 nm) from the runway.The scheduler considers constraints based on minimum required separation at coordination points, and feasible time-to-fly (TTF) ranges between coordination points.The TBFM model in this work uses given TTF ranges between coordination points and time blocks at each point to generate feasible STA time windows at each point [12].The TTF ranges fed to the model are based on predefined nominal and slow speed profiles for the given engine type along the given route.Time blocks are time conversions (based on nominal speed) of the distance-based separation requirements of previously scheduled flights plus a scheduling buffer.In previous research on tactical reconfiguration [13], the required separation distances used by the TBFM scheduling model were 5nm at all coordination points more than 40 nm from the airport, and 3nm at all points within 40 nm of the airport.Some differences are applied to that model for this work.First, more realistic TTFs are calculated based on step-down speed profiles and nominal and slow speed restrictions from route adaptation used in TBFM [14].Second, in-trail wake vortex separation standards are enforced.Third, in this work, the scheduling buffer used to define the blocked times of previously scheduled flights are not time conversions of a constant distance buffer.The time-based buffers are dependent on flights' expected arrival time performance at the coordination point, and the availability of tactical shortcuts for the coordination point.Parametric analyses from the previous tactical shortcuts work [11] were used to generate buffer size functions of expected arrival time uncertainty standard deviation and scheduled path (nominal or shortcut) for each coordination point.The use of speed control within the shortcuts allows the minimum buffer (designed for optimal time recovery) to be used for a desired threshold percentage of flights estimated to require tactical controller intervention to avoid loss of separation.For a threshold of 10% controller intervention, the required scheduling buffers are 1.1σ and 1.8σ when a shortcut is and is not available, respectively, where σ represents the estimated delivery precision based on aircraft equipage [11].A normal distribution with zero mean is assumed.The scheduler considered two control modes within the TRACON.All flights were assumed to be RNAV equipped with a subset also equipped with flight-deck schedule conformance precision capability.Flight-deck Required Time of Arrival (RTA) delivery precision is estimated to be 4 sec [15].Flight-deck interval Management (FIM) delivery precision is estimated to be 4.7 or 2.2 seconds if fast speed profiles are used [16].For the purposes of this research, the RTA or FIM capable aircraft delivery precision is assumed to be σ RTA =4.5 sec.All flights without flight-deck schedule conformance precision capability were controlled from the ground with the aid of ground-based automation tools.Controller Managed Spacing (CMS) ground-based automation tools have been shown to have a delivery precision of 5.2 seconds [8].However, a recent evaluation suggests that CMS delivery accuracy may be as low as 8-10 seconds 1 [17].For the purposes of this research, the ground controlled aircraft delivery precision is assumed to be σ CMS =9 sec.Thus flightdeck control was assumed to have twice the precision (half the standard deviation) as ground-based control.Table 1 shows the scheduling buffers used in this study based on aircraft equipage and tactical shortcut availability. +Table 1. Scheduling Buffer SizeTactical Shortcut Availability Available Not Available Delivery Precision RTA 1.1σ RTA = 4.9 sec 1.8σ RTA = 8.1 sec CMS 1.1σ CMS = 9.9 sec 1.8σ CMS =16.2 secThe analysis will compare metrics for ratios of mixed aircraft performance between RTA and CMS control modes ranging from 0 to 1 in 0.1 increments.1 Estimated 84th percentile (nominal distribution σ) schedule conformance at terminal schedule points. +B. Traffic ScenariosThe tactical shortcuts concept was applied to LAX arrival scheduling due to the availability of LAX adaptation used in recent human-in-the-loop simulations of TBFM terminal metering [17].Figure 1 shows the standard arrival routes from this LAX adaptation.Routes extend from meter fix to runway.The Northwest flows feed only runway 24R, the South flows feed only runway 25L, and the East flows feed both runways.Traffic scenarios were generated in 15-minute increments to meet a desired aircraft per quarter hour set of meter-fix ETAs.Each flight was then randomly assigned a weight class, engine type, and arrival meter fix according to weighted distributions observed from July 2014 historical traffic.Figure 1 shows the baseline LAX arrival routing used along with observed traffic distributions to each meter fix.Observed arrival traffic was comprised of 94% jets and 6% turboprops.Pistons were not considered, as they comprised less than 0.1% of the traffic.Of the three Northwest meter fixes, only jets used FIM and only turboprops used JEFFY, whereas VTU was shared between jets and turboprops.Of the two South meter fixes, only a small percentage of jets used SXC, whereas SHIVE was shared between jets and turboprops.Of the two East meter fixes, only jets used GRAMM, whereas KONZL was shared between jets and turboprops.Dashed lines in Fig. 1 depict segregated routes used by turboprops only until they merge with jets downstream.Although turboprops share the VTU, SHIVE, and KONZL meter fixes with jets, the routes are altitude separated until they reach merge points SADEE, SLI, and PALAC or GAATE respectively.Jet weight classes were comprised of 1% super, 15% heavy, 9% B757, 74% large, and 1% small.Turboprop weight classes were comprised of 13% large and 87% small. +C. Route ScenariosShortcut path options were designed for jet-dominated routes where the geometry allowed.The shortcut path options are shown in orange in Fig. 2 along with the percent reduction in path distance each shortcut provides in the upper right corner.The orange diamonds mark the decision waypoints, before which the decision to use the associated shortcut must be made.The orange dots are new merge points introduced by shortcuts.Red dots are merge points from the nominal route structure that are bypassed when a shortcut is used.These merge points may be scheduled with reduced buffers due to tactical shortcut availability when nominal routing is assigned by the scheduler.Of the six shortcuts shown in Fig. 2, two are final sector shortcuts (JAVSI to SAPPI2 and SLI to HUNDA), and the rest are feeder sector shortcuts.In this study, all of the shortcuts could be used tactically, but only the final shortcuts were allowed to be used strategically by the scheduler.The existence of downstream merge points from the feeder shortcuts would inhibit their ability to offer earlier runway STAs by reducing buffers upstream.Therefore, it would seem advantageous to always schedule to the feeder shortcut routes, as they would provide the same or earlier STAs as the nominal routes.However, given the relatively large error at which flights were delivered to meter fixes, it was expected that the benefit of reserving these feeder shortcuts to tactically aid schedule conformance would be greater.This is why in this study feeder shortcuts were not allowed to be used strategically by the scheduler, only used tactically in simulation.Simulation results will be compared between the baseline routing shown in Fig. 1 and the shortcut routing shown in Fig 2. +D. SimulationTraffic scenarios characterized flights with unique meter fix ETA, meter fix, engine type, weight class, and equipage.If a flight had multiple procedure options available (due to final shortcut and runway options), the preferred procedure was assumed to be the one that resulted in the earliest runway ETA.Flights were ordered by this earliest runway ETA for first-come-first-served scheduling.Each flight was scheduled by trial scheduling each of its procedure options using the buffers in Table 1 according to equipage and shortcut availability, and the procedure producing the earliest runway STA was assigned to the flight and its schedule was frozen.After all flights were scheduled in this way, the simulation processed actual times of arrival (ATAs) at all coordination points in order of the STAs.Flights were not allowed to change sequence.A Gaussian error was added to the closest reachable time to the STA to get a preliminary ATA.For merge points and runways, the Gaussian error was based on the same arrival precision assumed by the scheduler, σ CMS =4.5 sec for equipped and σ RTA =9 sec for unequipped aircraft.Delivery precision to the meter fix was assumed to follow a Gaussian error of σ MF =60 sec without the benefit of precision scheduling and spacing tools in Center airspace.All Gaussian errors were clipped to +/-3σ to avoid extreme outliers.The reachable time assumes speed control authority can only modify the scheduled time-to-fly (TTF) within a certain range.The control authority assumed in the controller speed advisory tool research was +/-15% [18].However results from a recent HITL experiment suggests that the purely speed-based control authority is less than half this [17].For this work speed control authority may increase scheduled TTF by 10% (by flying slower earlier) or decrease scheduled TTF by 5% (by flying faster longer).Note that this model is not centered at zero like the speed advisory tool, but gives more flexibility to increase TTF (slow down) than to decrease TTF (speed up) to be more realistic.These limits on TTF range make it possible for errors to cascade to the adjacent coordination point.If the error at a coordination point is late enough such that decreasing the flight TTF by 5% is unable to catch up to the original STA at the next point, the upstream ATA + (TTF-5%) serves as the closest reachable time to which the next Gaussian error is added.Each sequential point had an independent error applied to it.To ensure reasonable errors were applied, error was clipped to be within the range of a 10 kn head or tail wind error between sequential points.For example, consider the 2-mi segment between LUVYN and MINZA.Assume the final adjusted scheduled TTF for a flight between these points is 30 sec, making the average speed 240 kn.Assuming at most a 10 kn wind error, the 230 kn to 250 kn average speed translates to a 28.8 sec to 31.3 sec range of reasonable actual TTF.Assume the random sampled Gaussian error is 5 sec suggesting that the actual TTF for this segment be 35 sec.This is outside the reasonable error range for such a short distance, so the actual TTF is clipped to 31.3 sec to calculate the preliminary ATA at MINZA.For this segment distance, equipage is unlikely to make much of a difference in precision.However, it would make a difference in the next segment from MINZA to PALAC where the 13.6 mi segment distance at the same average 240 kn speed results in a reasonable actual TTF range of 195.4 sec to 212.4 sec and an error range of -8.8 se to 8.1 sec.The preliminary ATA was then compared to the preceding ATA at the coordination point.If the time spacing is less than the required separation, the preliminary ATA was adjusted back to meet the required separation.Spacing related adjustments that make the actual TTF greater than the scheduled TTF + 10% are assumed to require vectoring.For shortcut options simulations, schedules must be created for the additional merge points (orange dots) and decision points (orange diamonds) seen in Fig. 2. The decision points that are also meter fixes or merge points use ATA-STA error at the decision point to determine shortcut use.The remaining decision points are scheduled without buffers purely to estimate errors (ETA-STA) at the downstream merge point that the shortcut bypasses.The ATA-STA error at a decision point was assumed to be the same as the ETA-STA at its downstream merge point.All flights that were late (ATA>STA) at the decision point used the shortcut.Shortcut use shortened the scheduled TTF of a route segment by the percentages shown in Fig. 2. The adjusted scheduled TTF was then used to calculate the next downstream ATA as described above.Simulations were repeated 100 times for each traffic scenario with unique random errors applied in each simulation. +IV. METRICSThe following metrics were computed for comparative analysis between mixed aircraft performance ratios (ranging from 0 to 1 in 0.1 increments) and routing models (baseline vs. shortcut). +A. ThroughputDemand throughput, scheduled throughput, and actual throughput were hourly rates of ETA, STA, and ATA, respectively.Hourly throughputs were calculated as the numbers of ETAs, STAs, and ATAs at either runway between 60 and 120 minutes into each simulation.This time range represents the most consistent runway loading of each 2-hour traffic scenario.The demand and scheduled throughputs were averaged across the 1000 traffic scenarios generated.Actual throughputs were averaged across 100 simulations for each traffic scenario before averaging again across the 1000 traffic scenarios. +B. Scheduled DelayScheduled delays were segregated between delay absorbed in the TRACON and delay passed back to the Center.TRACON delays were further segregated between delay absorbed by path and by speed.Center delay is the difference between meter fix STA and ETA.TRACON path delay is the difference between the scheduled route runway ETA and the preferred route runway ETA.TRACON speed delay is the difference between runway STA and scheduled route ETA minus Center delay.Total delay is the sum of Center, TRACON path, and TRACON speed delay, which is also the difference between runway STA and preferred runway ETA.Delays were first averaged across all 144 flights in a single scenario and averaged again across the 1000 traffic scenarios. +C. WorkloadThe highest priority controller task is maintaining separation.The percentage of instances when the preliminary ATA must be modified was used to represent speed control workload.Similarly, vectoring workload was measured separately as the percentage of instances a flight's actual TTF on any route segment is greater than the speed control authority range of scheduled TTF +10%.This may happen when an ATA is pushed back significantly. +D. Shortcut UsageShortcut usage was evaluated for the shortcut routing case only as the baseline routing case did not have any shortcuts.Both scheduled and tactical shortcut usage was measured.For all flights with the option of scheduling to a particular shortcut, scheduled shortcut usage is the percentage of flights assigned to the shortcut route by the scheduler.These percentages were calculated across all 1000 traffic scenarios.Only the final shortcuts (JAVSI to SAPPI2 and SLI to HUNDA) were available as schedulable shortcut route options.VTU and FIM flights were used to evaluate JAVSI to SAPPI2 scheduled shortcut usage.SXC and SHIVE flights were used to evaluate SLI to HUNDA scheduled shortcut usage.Tactical shortcut usage was evaluated for all shortcuts (two final and four feeder).For all flights scheduled to the nominal path and able to use a particular shortcut tactically, tactical shortcut usage is the percent of flights that used the shortcut in simulation.These percentages were calculated across all 100 simulations per scenario across all 1000 traffic scenarios. +E. Schedule ConformanceSchedule conformance is the difference between ATA and STA at each coordination point.Standard error (standard deviation from zero) and mean error were calculated at each coordination point across all 100 simulations per scenario across all 1000 traffic scenarios.Standard deviation from the mean was originally considered as a schedule conformance metric.However, the typical shape of the error histograms was asymmetric with peaking close to zero, short tails on the left (early) and long tails on the right (late).Therefore, standard error was chosen rather than standard deviation to represent the error spread. +V. RESULTSThis section presents the comparative analysis of the metrics between baseline and shortcut routing and between CMS and RTA equipage. +A. ThroughputShortcut routing achieved higher scheduled and actual throughput at lower equipage ratios, but this benefit decreased as the RTA ratio increased.Figure 3 shows average demand, scheduled, and actual throughput for baseline and shortcut routing cases as the ratio of RTA equipped aircraft increases from 0 to 1. +Figure 3. Demand, scheduled, and actual throughputThe black demand throughput line is consistently ~72 aircraft per hour.This shows how the traffic scenarios of 18 aircraft per quarter hour at the meter fixes projected the same demand rate to the runways.The dashed and solid colored lines represent baseline and shortcut routing cases respectively.The blue and green lines represent scheduled and actual throughput respectively.Scheduled and actual throughput for both routing cases increases as the ratio of equipped aircraft increases.The actual throughput is consistently ~1 aircraft per hour lower than the scheduled throughput for both routing cases.Both schedule and actual throughput for shortcut routing are ~1.5 flight per hour greater than baseline routing at low RTA ratios.However, this benefit diminished as the RTA ratio increases until no shortcut routing benefit is seen at when all aircraft are RTA equipped.This is because as the average required slot size of aircraft decreases, there are more natural gaps in the schedule and shortcuts make less of an impact. +B. Scheduled DelayA much larger difference in delay was seen between baseline and shortcut routing than between RTA and CMS aircraft.Most of the additional scheduled delay for baseline routing was applied to the Center.Figure 4 shows the average total delay segregated by CMS and RTA equipage for each case and equipage ratio.The dashed and solid colored lines represent baseline and shortcut routing cases respectively.The blue and green lines represent CMS and RTA average total delay respectively.The inverse effect to throughput is seen as delay decreases with increased RTA equipage.Although RTA delay is usually less than CMS delay, the difference is minimal.A much larger difference in delay is seen between baseline and shortcut routing.The shortcut routing delay reduction over the baseline ranges from over 1 minute at low equipage ratios to ~20 sec at high equipage ratios.Center delay is the most responsive to equipage ratio.This is because, even at high equipage ratios, the demand is so high that the TRACON's ability to absorb delay is saturated, pushing all residual delay to the Center.The baseline case path delay is fixed, as scheduling to shortcuts is not an option.The path delay for the shortcut routing case is almost as high as the baseline case suggesting that very few flights were scheduled to shortcuts. +C. WorkloadIn general, both routing cases had much higher (~ 3 times) percentages of controller intervention required to prevent loss of separation than the 20% designed for in the isolated merge point case studied from previous research.This was due to large errors introduced at the meter fixes that cascaded to subsequent flights without any relief in demand to recover.Increasing the spacing buffers is one way to add some relief by pushing more schedule delay to the Center.Another approach that should be explored in future work is to use a more realistic wind error model.Whereas it is realistic to expect sequential flights to the same point to have very different errors, it is less realistic to expect the errors of the same flight at adjacent points along a route to have very different errors due to wind.Figure 6 compares speed control workload between shortcut and baseline routing for varying equipage ratio.The workload percentages reduce very slightly (~1%) as the RTA equipage ratio increases and the average required slot size becomes smaller.The difference between shortcut and baseline routing is far more pronounced.The shortcut routing workload percentages are 8-10% lower than those of the baseline routing. +Figure 6. Speed control workload vs. equipage ratioThere were no instances of vectoring for all simulations.This shows that even though the high meter fix arrival +D. Shortcut UsageOverall, relatively few flights were scheduled to final shortcuts but more were scheduled to shortcuts as the RTA ratio increased.On the other hand, a large percentage of flights used shortcuts tactically, especially when the flight scenario was dominated by one equipage level.Scheduled shortcut results are shown in Fig. 7, and tactical shortcut usage results are shown in Figs.8910.Figure 7 shows the percentage of flights that were scheduled to one of the final shortcuts when available.These results are for the shortcut routing case only as the baseline routing had no shortcuts.Relatively few flights (4-16%) were scheduled directly to the shortcuts.The scheduler assigned most flights to nominal routing to take advantage of reduced scheduling buffers and accommodate the high demand.More flights are scheduled to the shortcuts as the equipage ratio increases.This is because the relatively smaller RTA buffers are sufficient enough in more instances to do without further reducing the buffers with tactical shortcut options.Lower demand traffic scenarios with more natural gaps would be expected to have more flights scheduled to shortcuts. +Figure 7. Percent time shortcut is scheduledIn general, the RTA flights are scheduled to shortcuts only slightly more often due to their smaller buffers.This accounts for the only slightly lower total delay for RTA than CMS flights seen in Fig. 4.Tactical shortcuts are used any time a flight on the nominal route arrives at the tactical shortcut decision point late.In the absence of any systemic factors, the tactical shortcut would be used 50% of the time due to the use of zero mean uncertainty distributions.However, the total percent tactical shortcut usage is consistently high, over 92%.This means that flights arrive at the tactical shortcut decision point late more than 92% of the time.This is due to large meter fix arrival uncertainty and cascading separation delays.Figure 8 shows the percent of tactical shortcut usage segregated by CMS and RTA equipage for each equipage ratio.Shortcut usage is highest when the scenario is dominated by one equipage level.The minority equipage level uses tactical shortcuts less often. +Figure 8. Percent tactical shortcut usage by equipageFigure 9 shows the percent of tactical shortcut usage segregated by shortcut route.A few of the shortcuts are used less often than others.Figure 10 shows the average number of flights with the option of using each tactical shortcut, which is indicative of the traffic load on the route.Tactical shortcut usage is higher on more heavily loaded routes because there are fewer natural gaps and therefore more cascading separation delays causing flights to be late at the shortcut decision points more often.However, even the very light loading of the route offering the RUBEEN to SLI shortcut has 75% shortcut usage.This is due to the asymmetric speed control authority given to recover from large meter fix arrival uncertainty.Flights early to the meter fix may increase their TTF by 10%, whereas late flights may reduce their TTF by only 5%.This shifts the average reachable target ATA later than the STA at the decision point. +E. Schedule ConformanceThe greatest differences in schedule conformance were seen between shortcut and baseline routing rather than equipage or equipage ratio.Therefore, the results presented in this section are average conformance standard error and mean error values across all equipage ratio scenarios.Figure 11 shows the progression of conformance metrics along the highest demand Northwest path, from FIM at the bottom to 24R at the top.Starting at the FIM meter fix, baseline and shortcut routing metrics are very similar as they both reflect the zero mean 60-second standard deviation arrival uncertainty modeled as flights first enter the simulation.As traffic progresses toward 24R, both baseline and shortcut routing cases reduce the standard error, but the shortcut routing case is more effective.The baseline shows a consistent increase in mean error as traffic progresses through increasingly congested merge points.The shortcut routing case shows more erratic mean error behavior along the same path but ultimately the mean error is only a few seconds greater at the runway than the meter fix. +Figure 11. Conformance metrics progression from FIM to 24RConformance metrics along other paths from the Northwest and South had similar yet less pronounced patterns due to lower traffic volume.High traffic volume flows from the East resulted in similar standard error to the FIM flow but much higher mean error.This can be seen in Fig. 12 showing the conformance metrics for feeder merge points organized by flow. +Figure 12. Feeder merge point conformance metricsEven though the East flows serve over half the traffic demand, they not only have fewer shortcut options than other flows, but they do not bypass any merge points and offer very little additional range for scheduled delay.In future work, it would be interesting to see how routing allowing more symmetric shortcut availability affects the results. +VI. CONCLUSIONThis paper presented an evaluation of the tactical shortcuts concept applied to a model of TBFM arrival scheduling at LAX with mixed equipage.The TBFM model reduced the scheduling buffers of flights to merge points with a tactical shortcut due to the increased schedule conformance expected.It also reduced the scheduling buffers of flights equipped to enable higher schedule conformance at any merge point.In the high traffic volume scenarios tested, the increased throughput and reduced scheduled delay benefits seen (for shortcuts vs. baseline routing and as the ratio of equipped flights increased) were directly due to reducing the these buffers and scheduling flights into smaller slots.Due to the increased schedule conformance, the benefits were achievable without increasing (in fact reducing) controller workload.Results showed that the throughput and delay benefits of shortcut routing over baseline routing diminished as the equipage ratio increased but the workload benefits increased.Increasing the equipage ratios and average scheduling slot size was essentially equivalent to reducing the traffic demand with the same scheduling slot size.Therefore, we can assume that when traffic demand is lower and additional throughput is not 0" 10" 20" 30" 40" 50" 60" GRAMM"to"MINZA" KONZL"to"LUVYN" KITEC"to"BAYST" JAVSI"to"SAPPI2" RUBEEN"to"SLI" SLI"to"HUNDA" Average number of flights per simulation that may use each shortcut tactically 0" 10" 20" 30" 40" 50" 60" needed, the workload benefits will still remain.Giving the scheduler the ability to use the shortcuts strategically (as with the final shortcuts in this study) negates this low traffic volume benefit.The tactical shortcuts are no longer being used to reduce buffer and squeeze in extra flights, but to mitigate scheduling disturbances.Reserving tactical shortcut options rather than scheduling to the shortest route makes the schedule more robust.The largest source of scheduling disturbance modeled in this study was large meter fix delivery uncertainty and the purely tactical feeder shortcuts were vital to mitigating meter fix uncertainty disturbances.However, tactical shortcuts could be used to mitigate other types of common disturbances that would cause flights to be late, such as head wind, slowing for turbulence, or deviation around weather.Overall, the tactical shortcuts concept showed potential as a valuable enhancement to TBFM scheduling and spacing.Figure 1 .1Figure 1.Distribution of observed JET and TURBO traffic across standard arrival routes adapted to LAX. +The above distributions were used to randomly generate 1000 2-hour traffic scenarios for an 18 aircraft per quarter hour arrival rate.Previous HITL studies of mixed equipage precision scheduling at LAX observed an average throughput of 71 aircraft per hour with standard deviation less than 3[17].This equates to 18 aircraft per quarter hour for the purposes of generating traffic scenarios in 15-minute increments as described above.Each of the 1000 2-hour traffic scenarios included 8 quarter hours of 18 aircraft each totaling to 144 aircraft per scenario. +Figure 2 .2Figure 2. Shortcut routing and segment distance saved relative to standard arrival routes. +Figure 4 .Figure 5 .45Figure 4. Average total CMS and RTA delay +Figure 55Figure5segregates delay by Center, TRACON speed, and TRACON path delay.Center delay is the most responsive to equipage ratio.This is because, even at high equipage ratios, the demand is so high that the TRACON's ability to absorb delay is saturated, pushing all residual delay to the Center.The baseline case path delay is fixed, as scheduling to shortcuts is not an option.The path delay for the shortcut routing case is almost as high as the baseline case suggesting that very few flights were scheduled to shortcuts. +.1" 0.2" 0.3" 0.4" 0.5" 0.6" 0.7" 0.8" 0.9" 1" uncertainty causes cascading separation delays, they can be managed within the bounds of speed control modeled. +Figure 9 .Figure 10 .910Figure 9. Percent tactical shortcut usage by route + + + +Dr. Jung is an Aerospace Engineer at NASA Ames Research Center.In 2009-2010, he worked on dynamic airspace configuration.From 2011, he has supported ATM Technology Demonstration-1, focusing on terminal area metering and conducting Human-in-the-Loop simulations.Currently he is leading a vehicle and surveillance group in the Unmanned Aerial System Traffic Management (UTM) project, with a goal of safely enabling low altitude operations for manned and unmanned aircraft.Prior to joining NASA, he worked on Global Navigation Satellite System applications at Trimble Navigation Limited. + + + + + + + Benefits of precision scheduling and spacing for arrival operations + + SZelinski + + 10.1109/dasc.2012.6382979 + + + 2012 IEEE/AIAA 31st Digital Avionics Systems Conference (DASC) + Williamsburg, Virginia + + IEEE + October 14-18, 2012 + + + S. Zelinski, "Benefits of precision scheduling and spacing for arrival operations," 31 st Digital Avionics Systems Conference, Williamsburg, Virginia, October 14-18, 2012. + + + + + 4D FMS for Increasing Efficiency of TMA Operations + + BerndKorn + + + AlexanderKuenz + + 10.1109/dasc.2006.313663 + + + 2006 ieee/aiaa 25TH Digital Avionics Systems Conference + Hamburg, Germany + + IEEE + September 3-8, 2006 + + + 25 th Congress of the International Council of the + B. Korn, H. Helmke, and A. Kuenz, "4D trajectory management in the extended TMA: coupling AMAN and 4D FMS for optimized approach trajectories," 25 th Congress of the International Council of the Aeronautical Sciences (ICAS), Hamburg, Germany, September 3-8, 2006. + + + + + Time-based arrival management for dual threshold operation and continuous decent approaches + + HHelmke + + + RHann + + + MUebbing-Rumke + + + DMuller + + + DWittkowski + + + + 8 th USA/Europe Air Traffic Management R&D Seminar + + 2009 + Napa Valley, California + + + H. Helmke, R. Hann, M. Uebbing-Rumke, D. Muller, and D. Wittkowski, "Time-based arrival management for dual threshold operation and continuous decent approaches," 8 th USA/Europe Air Traffic Management R&D Seminar, Napa Valley, California, 2009. + + + + + Controller aids for integrating negotiated continuous decent approaches into conventional landing traffic + + MUebbing-Rumke + + + MTemme + + + + 9 th USA/Europe Air Traffic Management R&D Seminar + + 2011 + Berlin, Germany + + + M. Uebbing-Rumke and M. Temme, "Controller aids for integrating negotiated continuous decent approaches into conventional landing traffic," 9 th USA/Europe Air Traffic Management R&D Seminar, Berlin, Germany, 2011. + + + + + Design and evaluation of the terminal area precision scheduling and spacing system + + HSwenson + + + JThipphavong + + + ASadovsky + + + LChen + + + CSullivan + + + LMartin + + + 2011 + Berlin, Germany + + + USA/Europe Air Traffic Management R&D Seminar + + + H. Swenson, J. Thipphavong, A. Sadovsky, L. Chen, C. Sullivan, and L. Martin, "Design and evaluation of the terminal area precision scheduling and spacing system," 9 th USA/Europe Air Traffic Management R&D Seminar, Berlin, Germany, 2011. + + + + + Evaluation of the terminal area precision scheduling and spacing system for near-term NAS application + + JThipphavong + + + HSwenson + + + LMartin + + + PLin + + + JNguyen + + 10.1201/b12321-12 + + + Advances in Human Aspects of Aviation + San Francisco, CA + + CRC Press + July 21-25, 2012 + + + + J. Thipphavong, H. Swenson, L. Martin, P. Lin, and J. Nguyen, "Evaluation of the terminal precision scheduling and spacing system for near-term NAS application," Applied Human Factors and Ergonomics, San Francisco, CA, July 21-25, 2012. + + + + + Development and evaluation of the terminal precision scheduling and spacing system for off-nominal condition operations + + HNSwenson + + + Jaewoo Jung + + + JThipphavong + + + Liang Chen + + + LMartin + + + JNguyen + + 10.1109/dasc.2012.6382303 + + + 2012 IEEE/AIAA 31st Digital Avionics Systems Conference (DASC) + Williamsburg, Virginia + + IEEE + October 14-18, 2012 + + + H. Swenson, J. Jung, J. Thipphavong, L. Chen, L. Martin, and J. Nguyen, "Development and evaluation of the terminal precision scheduling and spacing system for off-nominal condition operations," 31 st Digital Avionics Systems Conference (DASC), Williamsburg, Virginia, October 14-18, 2012. + + + + + Controller-Managed Spacing - A Human-In-The-Loop Simulation of Terminal-Area Operations + + MichaelKupfer + + + ToddCallantine + + + JoeyMercer + + + LynneMartin + + + EverettPalmer + + 10.2514/6.2010-7545 + + + AIAA Guidance, Navigation, and Control Conference + 9 th USA/Europe Air Traffic Management R&D Seminar + Berlin, Germany + + American Institute of Aeronautics and Astronautics + 2011 + + + M. Kupfer, T. Callantine, L. Martin, J. Mercer, and E. Palmer, "Controller support tools for schedule-based terminal-area operations," 9 th USA/Europe Air Traffic Management R&D Seminar, Berlin, Germany, 2011. + + + + + Arrival Management Architecture and Performance Analysis with Advanced Automation and Avionics Capabilities + + AslaugHaraldsdottir + + + JulienScharl + + + JanetKing + + + MatthewBerge + + 10.2514/6.2009-7006 + + + 9th AIAA Aviation Technology, Integration, and Operations Conference (ATIO) + Hilton Head, South Carolina + + American Institute of Aeronautics and Astronautics + September 21-23, 2009 + + + A. Haraldsdottir, J. Scharl, J. King, and M. E. Berge, "Arrival management architecture and performance analysis with advanced automation and avionics capabilities," AIAA 9 th Aviation Technology, integration and Operations (ATIO) Conference, Hilton Head, South Carolina, September 21-23, 2009. + + + + + Investigating the impact of off-nominal events on high-density ‘green’ arrivals + + ToddJCallantine + + + ChristopherCabrall + + + MichaelKupfer + + + LynneMartin + + + JoeyMercer + + + EverettAPalmer + + 10.1109/dasc.2011.6096195 + + + 2011 IEEE/AIAA 30th Digital Avionics Systems Conference + Seattle, Washington + + IEEE + October 16-20, 2011 + + + Callantine, T. J., C. Cabrall, M. Kupfer, L. Martin, J. Mercer, E. A. Palmer, "Investigating the impact of off-nominal events on high-density 'green' arrivals," 30 th Digital Avionics Systems Conference (DASC), Seattle, Washington, October 16-20, 2011. + + + + + Design considerations for shortcut path-based time recovery + + ShannonZelinski + + 10.1109/dasc.2013.6719601 + + + 2013 IEEE/AIAA 32nd Digital Avionics Systems Conference (DASC) + Syracuse, New York + + IEEE + October 6-10, 2013 + + + S. Zelinski, "Design considerations for shortcut path-based time recovery," 32 nd Digital Avionics Systems Conference (DASC), Syracuse, New York, October 6-10, 2013. + + + + + A Closed-Form Solution to Multi-Point Scheduling Problems + + LarryMeyn + + 10.2514/6.2010-7911 + + + AIAA Modeling and Simulation Technologies Conference + Toronto, Ontario Canada + + American Institute of Aeronautics and Astronautics + 2010 + + + L. Meyn, "A closed-form solution to multi-point scheduling problems," AIAA Modeling and Simulation Technologies (MST) Conference, Toronto, Ontario Canada, 2010. + + + + + Precision Arrival Scheduling for Tactical Reconfiguration + + ShannonZelinski + + 10.2514/6.2014-3152 + + + 14th AIAA Aviation Technology, Integration, and Operations Conference + Atlanta, Georgia + + American Institute of Aeronautics and Astronautics + June 6-10, 2014 + + + S. Zelinski, "Precision arrival scheduling for tactical reconfiguration," AIAA 14 th Aviation Technology, Integration and Operations (ATIO) Conference, Atlanta, Georgia, June 6-10, 2014. + + + + + The dynamic planner: the sequencer, scheduler, and runway allocator for air traffic control automation + + GWong + + NASA TM-2000-209586 + + April 2000 + + + G. Wong, "The dynamic planner: the sequencer, scheduler, and runway allocator for air traffic control automation," NASA TM-2000-209586, April 2000. + + + + + Controlled Time of Arrival Flight Trials -Results and Analysis + + JKlooster + + + ADelamo + + + PManzi + + + + Europe Air Traffic Management R&D Seminar + + 2009 + Napa, CA + + + 8 th USA/ + J. Klooster, A. DelAmo, P. Manzi, "Controlled Time of Arrival Flight Trials -Results and Analysis," 8 th USA/Europe Air Traffic Management R&D Seminar, Napa, CA, 2009. + + + + + An Evaluation of a Flight Deck Interval Management Algorithm including Delayed Target Trajectories + + KurtASwieringa + + + MatthewCUnderwood + + + BryanBarmore + + + RobertLeonard + + 10.2514/6.2014-3148 + + + 14th AIAA Aviation Technology, Integration, and Operations Conference + Atlanta, Georgia + + American Institute of Aeronautics and Astronautics + June 6-10, 2014 + + + K. Swieringa, M. Underwood, and B. Barmore, "An evaluation of a flight deck interval managmeent algorithm including delayed target trajectorires," AIAA 14 th Aviation Technology, Integration and Operations (ATIO) Conference, Atlanta, Georgia, June 6-10, 2014. + + + + + Evaluation of the Terminal Sequencing and Spacing system for Performance-Based Navigation arrivals + + JaneThipphavong + + + JaewooJung + + + HarrySwenson + + + LynneMartin + + + MelodyLin + + + JimmyNguyen + + 10.1109/dasc.2013.6712503 + + + 2013 IEEE/AIAA 32nd Digital Avionics Systems Conference (DASC) + 10 th USA/Europe Air Traffic Management R&D Seminar + Chicago, IL + + IEEE + 2013 + + + J. Thipphavong, J. Jung, H. Swenson, K. Witzberger, M. Lin, J. Nguyen, L. Martin, M. Downs, and T. Smith, "Evaluation of the controller- managed spacing tools, flight-deck interval management and terminal metering capabilities for the ATM technology demonstration #1," 10 th USA/Europe Air Traffic Management R&D Seminar, Chicago, IL, 2013. + + + + + How to compute a slot marker — Calculation of controller-managed spacing tools for efficient descents with precision scheduling + + TPrevot + + 10.1109/dasc.2012.6382318 + + + 2012 IEEE/AIAA 31st Digital Avionics Systems Conference (DASC) + Williamsburg, Virginia + + IEEE + October 14-18, 2012 + + + T. Prevot, "How to compute a slot marker -calculation of controller- managed spacing tool efficient descents with precision scheduling," 31 st Digital Avionics Systems Conference (DASC), Williamsburg, Virginia, October 14-18, 2012. + + + + + + diff --git a/file835.txt b/file835.txt new file mode 100644 index 0000000000000000000000000000000000000000..7f41853761669e717ee8d6cb4fbb8584bb579fa5 --- /dev/null +++ b/file835.txt @@ -0,0 +1,912 @@ + + + + +IntroductionAirspace management concepts attempt to mitigate required traffic flow management and allow more user preference and traffic flexibility.One form of airspace management is Dynamic Airspace Configuration, that is reconfiguring airspace boundaries to correspond with the prevailing demand traffic and allow more throughput.Today's traffic flexibility is limited in part due to largely static airspace design.New algorithmic methods of airspace design are being developed to allow airspace to change more dynamically and conform to more flexible traffic.A prior comparison of algorithm generated airspace configurations identified strengths and weaknesses of three airspace partitioning algorithms [1,2].The different algorithms each had different strengths including reduced flight delay, reduced airspace complexity, and more balanced workload among airspace sectors.However, there were some technical limitations in the previous comparison.Currently, airspace is reconfigured to accommodate traffic demand, and in the future, it will reconfigure more dynamically.However, as a first step, this previous analysis only compared static configurations.Due to airspace design algorithm limitations, the number of sectors was not fixed between compared configurations, making it difficult to assess their relative benefits.These algorithms have since evolved to address identified weaknesses from the comparison as well as address other considerations such as complexity due to reconfiguration and traffic pattern interactions with airspace boundaries.In addition, several other algorithms have since matured to a level sufficient to participate in a comparison study.This paper presents a next round comparison of newly improved airspace design solutions for dynamically reconfiguring airspace in response to nominal daily traffic volume fluctuation.Improved versions of the three algorithmic airspace design methods from the previous study and five additional methods were compared.Each method designed three-configuration airspace solutions with projected simulated flight tracks in Kansas City Center at two times today's traffic levels.In addition, each method used updated traffic projections at each reconfiguration time to update subsequent planned reconfigurations.Data analysis compared not only airspace design benefits between different algorithmic methods, but between design updates from a single method to assess the benefit of dynamically updating planned configurations.Additional compared metrics included traffic pattern complexity with respect to airspace boundaries and reconfiguration complexity.This paper is organized as follows.The Background section provides background on airspace design research and specific design methods compared in this paper.The Method section describes the methods, including experiment setup, scenarios, and metrics.Detailed and summary results are described in the Results section.The paper ends with Conclusions. +BackgroundCurrently, airspace in both the United States and Europe is partitioned into functional blocks that may be combined into fewer large sectors when traffic volume is low or split into more small sectors when traffic volume is high.This is actually a simple and flexible method of adapting to traffic volume fluctuation.However, configuration schedules are generated days in advance based on estimated traffic demand and tactical changes to the configuration schedules are based on managers' personal experience and judgment.Staffing constraints limit the number of sectors that may be open within a particular group of sectors for which each controller is trained.In the United States, these groups of sectors are called areas of specialization (AOS).This also limits feasible combinations of functional airspace blocks within a single area.In addition, overloads may occur in sectors that cannot be further split.One body of research is developing algorithms to build good realistic functional block-based configuration schedules rather than relying on human judgment [3][4][5][6].Other research focuses on redesigning the airspace boundaries themselves [7][8][9][10][11][12][13][14].Many of these methods have performed selfassessments and a few have been compared to each other [1,2], but the assessments focused on the designs themselves and not the cost of reconfiguration.A human-in-the-loop study conducted at NASA Ames tested the feasibility of reconfiguring by moving an airspace boundary on-demand rather than combining and splitting functional airspace blocks [15].The study found that the reconfiguration operation itself was feasible.However, certain characteristics of boundary designs with respect to traffic pattern geometry and of reconfiguration from one design to the next tended to increase controller workload and decrease acceptability.Many airspace design algorithms have since incorporated these traffic pattern and reconfiguration complexity considerations into their methods.This paper compares delay reduction benefits, traffic pattern complexity, and reconfiguration complexity of seven airspace design algorithms that use different approaches.Three of these airspace design methods attempt to address the reconfiguration complexity considerations by using elements of the currently used functional airspace blocks in their design.DAU Slices modifies a given configuration by defining five nmi slices of airspace along the shared edge of a sector pair as Dynamic Airspace Units (DAUs) [16] to effectively move airspace boundaries in five nautical mile increments between sector pairs.CombineSplit uses a set of given functional airspace blocks and desired number of sectors as inputs to group the sectors into configurations [17].FlightLevel [18] starts with the AOS boundaries, and partitions the AOS's vertically by flight level (1,000 ft increments) to achieve the desired number of sectors for a configuration.By contrast, CombineSplit has the option of recombining the base sector units irrespective of AOS boundaries.Inter-AOS reconfiguration is assumed possible in the future with the development of generic airspace tools and procedures that make it easier for individual controllers to work a larger variety of airspace and remain current [19].Four other airspace design methods compared were more freeform and did not use any elements of the current functional airspace block design.A new Graph-based method [14] partitions a graph representation of the filed flight plan structure and assigns airspace to each graph partition trying to keep intersections and major flow paths away from sector boundaries.The final three methods compared are improved versions of those compared in previous work [1,2].These are SectorFlow, CellGeoSect, and Voronoi described below.SectorFlow [10] clusters flight track points attempting to minimize airspace complexity parameters.Airspace is then assigned to each flight track cluster.An improved version of SectorFlow [20] compared in this paper addresses flow pattern complexity by including several parameters that helped the algorithm keep flow intersections away from sector boundaries and using a gradient search to refine the boundaries after the initial partition.CellGeoSect is a hybrid of an airspace cell clustering method compared in previous work [11] and an airspace splitting method called GeoSect [21] used to address traffic pattern complexity.The cell clustering method represents the airspace as a tessellation of hexagonal cells and clusters these cells to maximize flow connectivity within and balance flight count between clusters.GeoSect modifies the resulting design by sequentially removing and redefining the boundary between each pair of sectors to avoid geometric constraints such as the length and boundary crossing angles of major flows.Voronoi [12] represents the airspace using a Voronoi Diagram.A Genetic Algorithm then optimizes the Voronoi Diagram representation to minimize sector overloading.An improved version of this method, used in this comparison, uses a multistage process to incorporate different kinds of constraints into the overall optimization [22].Some constraints added to address traffic pattern complexity include trying to keep intersections and major flow paths away from sector boundaries and minimizing the number of low dwell time flights. +MethodThe following subsections describe the experiment setup and metrics used to compare the seven airspace design methods described above. +Experiment SetupThe experiment focused on airspace above 24,000 ft within Kansas City Center (ZKC) in the United States.ZKC presented a good focus center because its airspace design is moderately complex.The entire center shares a common split between low and high altitude airspace of 24,000 ft.In current operations, the airspace above 24,000 ft is routinely reconfigured between combinations of 27 functional airspace blocks (base sectors) within six areas of specialization (AOS).Figure 1 shows the current ZKC sector design, where color indicates AOS and the altitude split between high and super-high sectors is shown with the super-high altitude sector labels. +Figure 1. Current ZKC Sector DesignThe altitude split between high and super-high altitude sectors ranges between 35,000 and 38,000 ft.Higher altitude over-flights (above 33,000 ft), passing through the center, dominate ZKC's traffic patterns. +Reconfiguration ScenariosTo reduce experiment complexity, all algorithm generated reconfiguration scenarios derived from a three-configuration simplification of historic ZKC airspace operations on 2/8/2007, a nominal day with little weather impact.Lai and Zelinski [23] describes the procedure for processing operational sector combination data and identified three configurations as a reasonable baseline representation for 2/8/2007.All reconfiguration scenarios had the same reconfiguration times and numbers of sectors in each configuration as the Baseline.Given the number of sectors and projected track data for each configuration time period, each algorithm was free to partition the ZKC airspace above 24,000 ft both laterally and vertically.This was a design improvement over the previous study where algorithms partitioned only laterally within two predefined altitude layers. +Figure 2. Reconfiguration Scenario DesignSector capacities for each configuration were assigned using the method presented in Welch et al [24].This capacity estimation method validated well with respect to historical data using a simple quadratic model based on sector volume and the average flight transit time through the sector during the peak traffic period.The Welch capacity estimation method provided an easily implementable improvement over the previously used method based purely on average flight transit time, which tended to underestimate capacity for large sectors and overestimate capacity for small sectors.The Welch model was used to assign Baseline sector capacities as well.Even though Baseline configurations were not modified at each reconfiguration, updated projected flight tracks required that the sector capacities be updated at each reconfiguration. +SimulationSimulations were completed using the Airspace Concept Evaluation System (ACES) [25].ACES has been validated to be a good modeler of en-route trajectories producing similar delay results to realworld operational statistics for good-weather days [26,27].ACES modeled gate-to-gate flight operations on airport surfaces and in terminal and enroute airspace.Air traffic control and traffic flow management models controlled flights during these operations to ensure that airspace capacity constraints were not violated.Lower fidelity models were used for ground and airport modeling and higher fidelity models were used for en-route trajectory modeling, which extended from departure meter fixes to arrival meter fixes.The only constraints imposed in simulation were sector capacities for ZKC airspace above 24,000 ft.Airport and airspace outside of the design scope were unconstrained.It is very difficult to decompose the cause of delays simulated in ACES.Therefore, unconstraining the airports and airspace outside of the design scope ensured that all simulated delay was caused by the ZKC reconfiguration scenario being tested.ACES simulated flight tracks from the 2/8/2007 flight schedule.Without capacity constraints from airports or neighboring centers or weather impacts, the simulated 2007 traffic produced negligible flight delays.To stress the simulation into producing more delay for analysis, a demand generation tool, AvDemand [28], was used to create a two-times traffic schedule by cloning flights from the 2/8/2007 schedule.AvDemand also time-shifted flights within an hour of the original schedule to reduce unnatural demand peaks.At the time of this experiment, airspace design algorithms had not been fully integrated into ACES.Therefore, the iterative simulation process in Figure 3 was used to mimic a closed loop simulation of the Figure 2 +MetricsMetrics were designed to assess the performance of individual configuration designs and reconfigurations between them.Three categories of metrics include, delay, traffic pattern complexity, and reconfiguration complexity.Delay is a user benefit metric.Traffic pattern complexity metrics assess properties of the traffic patterns with respect to airspace boundaries that may affect controller workload.Reconfiguration complexity metrics assess the transition cost from one configuration to the next.Metric details are described below. +DelayDelay is not only costly to airlines and passengers, but it increases uncertainty by altering flight plans.Reduced delay relative to the current-day baseline simulated with 2x traffic quantifies a user benefit for a set of airspace configurations.ACES traffic flow management (TFM) monitors sector capacity and projected sector demand with a 6hour look-ahead time window.TFM issues an entry time restriction to the first flight projected to exceed a particular sector's capacity.Time restrictions may propagate and accumulate as the flight passes through many sectors.Delay may be absorbed en-route with path stretching maneuvers or at the gate as departure delay.At the end of each simulation, the total delay for a flight is the difference between its scheduled and actual gate arrival times.Because airports and airspace outside ZKC and below 24,000 ft were unconstrained, all flight delay was due to high altitude ZKC airspace capacity constraints.However, it is difficult to quantify the individual delay caused by a particular ZKC configuration.Therefore, average total delays were computed for each three-configuration simulation.Let d(s i ) be the average total delay for ith iteration simulation s i .There are three simulation iterations, +Traffic Pattern Complexity MetricsOriginal airspace design algorithms were mostly concerned with minimizing and balancing sector traffic load.However, airspace design must also accommodate traffic pattern geometry to minimize controller cognitive complexity.Most algorithms compared in this study have incorporated some method of aligning the airspace design with traffic patterns to minimize this complexity.The following metrics measure traffic pattern complexities with respect to sector boundaries.Controllers prefer major flows and their intersections to be well within sector boundaries.To guarantee separation, controllers must be aware of flights not only within a sector, but also just outside the sector.Brinton and Cook [29] show how as-flown flight paths have a statistically significant lower percentage of flight time within two miles of current sector boundaries (designed to accommodate these paths) than great-circle or wind-optimal paths.The number of aircraft within a threshold distance of a sector boundary was also included in 17 out of 52 original dynamic density metrics found to be significant for measuring airspace complexity [30].Ideally, flows should stay at least three to five nmi inside the sector boundary to avoid magnifying flight awareness workload of neighboring sectors and to leave room for maneuvering but using a two nmi threshold captures flights that clearly require extra controller attention.Let α(s i c j ) be the percentage of flight tracks within two miles of a sector boundary for the jth configuration in the ith simulation iteration.Let α for a particular method be the average α(s i c j ) of all configurations for all iterations weighted by configuration duration.Controllers require some time to become familiar with a flight entering the sector before it approaches a major intersection.The more time they have, the more efficiently they may control the flow safely through the intersection.Therefore, intersections should be away from sector boundaries.Brinton and Cook [29] show how there are statistically significantly fewer as-flown flight intersections less than ten miles from a sector boundary than great-circle or wind-optimal path intersections.Jung et al [31] found that increased workload during stable configuration periods correlated to a lower average distance between traffic flow intersections and sector boundaries.Let β(s i c j ) be the average distance between traffic flow intersections and sector boundaries for the jth configuration in the ith simulation iteration.Let β for a particular method be the average β(s i c j ) of all configurations for all iterations weighted by configuration duration.Jung et al [31] also found that increased workload during stable configuration periods correlated to the number of flights with short dwell time within a sector.When a flight spends very a small amount time within a sector, controllers often coordinate to directly handoff the flight to the next sector without taking ownership.This causes increased controller workload with no additional service provided to the flight.Let γ(s i c j ) be the average number of short dwell flights (spending less than two minutes within the a sector) per quarter hour per sector for the jth configuration in the ith simulation iteration. +Reconfiguration Complexity MetricsIt is assumed that reconfiguration incurs an operational cost related to transitioning from one configuration to another.Homola et al [32] showed how new on-demand reconfigurations could be implemented to balance sector traffic load and minimize over-capacity time periods without compromising safety, but at the cost of increasing controller task-load and workload ratings.Lee et al [15] and Jung et al [31] identified percent airspace volume and number of aircraft transferred as the primary contributors to reconfiguration workload for the same study.Percent airspace volume transferred impacts controller situational awareness and number of aircraft transferred impacts controller task-load of handing off aircraft to their new sectors.The first step for computing reconfiguration metrics between two configurations is to map their sectors to each other.First, sector pairs are mapped in order of decreasing intersection volume.Then, sectors with no intersecting volume are mapping in order of increasing Housdorff distance.Housdorff distance measures how far one sector is spatially shifted from another [33].Consecutive configurations with different numbers of sectors will have some unmapped sectors assumed to appear or disappear as the sector number increases or decreases, respectively.Let υ + (k 1 ,k 2 ) and υ -(k 1 ,k 2 ) be the volume gained and lost for sector pair (k 1 ,k 2 ) given asυ + (k 1 ,k 2 ) = υ(k 1 ) -υ ∩ (k 1 ,k 2 ) υ -(k 1 ,k 2 ) = υ(k 2 ) -υ ∩ (k 1 ,k 2 )where k 1 is the old sector, k 2 is the new sector, and υ ∩ (k 1 ,k 2 ) is the shared volume between k 1 and k 2 seen in figure 4. +Figure 4. Volume Gained and LostLet v + (k 1 ,k 2 ) and v -(k 1 ,k 2 ) be the percent volume gained and lost with respect to the old sector volume.v + (k 1 ,k 2 ) = υ + (k 1 ,k 2 ) / υ(k 1 ) v -(k 1 ,k 2 ) = υ -(k 1 ,k 2 ) / υ(k 1 )For unmapped appearing sectors, υ + (-,k 2 )=υ(k 2 ) and v + (-,k 2 )=100%.For unmapped disappearing sectors, υ -(k 1 ,-)= υ(k 1 ) and v -(k 1 ,-)=100%.Let V(k 1 ,k 2 ) be a weighted combined volume transfer complexity given byV(k 1 ,k 2 ) = wv + v + (k 1 ,k 2 ) + wv -v -(k 1 ,k 2 )where wv + and wv -are weighting factors.Operational reconfigurations can be completed in roughly five minutes [15].Therefore, the numbers of aircraft gained (n + (k 1 ,k 2 ,t)) and lost (n -(k 1 ,k 2 ,t)) between k 1 and k 2 are the numbers of unique aircraft flying in υ + (k 1 ,k 2 ) and υ -(k 1 ,k 2 ), respectively, during the five minutes preceding reconfiguration time t.Let N(k 1 ,k 2 ,t) be a weighted combined aircraft transfer complexity given byN(k 1 ,k 2 ,t) = wn + n + (k 1 ,k 2 ,t) + wn -n -(k 1 ,k 2 ,t)where wn + and wn -are weighting factors.Lai and Zelinski [23] found that in current reconfiguration operations, there is an average of two aircraft gained and two aircraft lost.Clustering operational reconfigurations into the simplified threeconfiguration set used in this study altered these values because the clustered reconfiguration times were no longer coordinated with the traffic.The aircraft gained and lost metrics are very sensitive to reconfiguration time due to traffic fluctuation.In operation, managers would be free to implement a reconfiguration any time within a range to minimize the aircraft transfer complexity.Therefore, N(k 1 ,k 2 ,t) was calculated for t ranging from thirty minutes before to thirty minutes after the reconfiguration design time in five-minutes increments.It was assumed that the reconfiguration could occur within any of these five-minute intervals, but the entire reconfiguration must be completed within the same interval.This assumption made component-based airspace design methods such as Baseline, CombineSplit, DAU Slices, and FlightLevel, that could reconfigure incrementally, more comparable to the freeform design methods that may require the reconfiguration to occur all at once.It was also assumed that managers would choose the time that minimized the maximum aircraft transition workload. +ResultsThe following subsections present results for the Baseline and designs from seven airspace design methods. +Delay (d)Delay measured the benefits of each airspace design method from the user prospective.Lower delay demonstrated user benefits.Figure 5 shows the average total delay for each of the three-configuration simulation iterations shown in Figure 3. Five of the methods produced lower delay than Baseline with their original designs in s 1 .After the first design update, all but CombineSplit reduced delay below Baseline.The most significant delay reduction is due to the first design update, between s 1 and s 2 .Very little if any delay reduction is achieved with the second design update, between s 2 and s 3 .CellGeoSect showed the most user benefit, reducing the Baseline delay by more than two thirds.In general, freeform methods produced lower delays than methods using Baseline components. +Traffic Pattern Complexity +Flight Track Percentage Close to Boundary (α)The percentage of flights tracks within two nmi of a sector boundary (α) was computed for each configuration and iteration.Figure 6 shows average α results. +Flow Intersection Boundary Proximity (β)Figure 7 shows β for each airspace design method.The β values for Baseline'08 and SectorFlow'08 were calculated from Figure 8 in [28] by multiplying the x and y axis for each column and summing.Baseline and Baseline'08 are very similar and all methods except SectorFlow'08 have higher β than both Baselines.The SectorFlow improvement over SectorFlow'08 moved flow intersections 2.4 miles farther from sector boundaries on average, which is approximately 20 seconds of flight time.Just as with α, FlightLevel has the best β values due to it's use of AOS footprints. +Number of Short Dwell Flights (γ)All of the design methods indirectly try to maximize average flight dwell time through each sector because it is directly related to maximizing sector capacity.However, only Voronoi explicitly tried to minimize the number of short dwell flights.Figure 8 shows γ averages and quartiles for each airspace design method.Voronoi and CombineSplit are the methods with the most similar or lower γ values than Baseline.CombineSplit is very similar to Baseline because it uses the same base airspace volumes.Voronoi is similar or better than Baseline because it is the only method that explicitly tried to minimize the number of short dwell flights.FlightLevel sticks out with γ values that are consistently more than twice that of Baseline.Due to FlightLevel vertical partitioning, as the number of sectors increases, sector vertical range decreases.The difference in γ between FlightLevel and Baseline is entirely due to climbing or descending flights passing through sectors spanning only two or three flight levels.This did not negatively affect FlightLevel's α or β because distances for these metrics are measured relative to lateral boundaries only. +Reconfiguration Complexity +Volume Transition Complexity (V)Figures 9 and 10 show averages and quartiles of all V between pairs of mapped sectors for the first and second reconfiguration, respectively.V was calculated using wv + = wv -= 0.5.As seen in Figure 9, only FlightLevel has lower V than Baseline.All other methods have slightly higher V than Baseline and freeform methods have higher V than those using Baseline components.This trend is exaggerated in the second reconfiguration seen in Figure 10.Most methods produce more varied results than Baseline.High maximums were caused by mapped sector pairs with little or no overlapping volume.V tended to be larger in the second reconfiguration than the first for two reasons.V is a weighted percentage of the sector size prior to reconfiguration and c 2 sector sizes were the smallest.Also, the second configuration reduced the number of sectors causing volume gained to dominate V, whereas volume lost dominated V in the first reconfiguration.There was no limit on how much volume a sector could gain but the most volume a sector could lose was 100%.Figure 12 shows generally higher N values for freeform methods than those using Baseline components.Only CombineSplit has consistently lower second reconfiguration N than Baseline.By contrast, SectorFlow and Voronoi produce two to three times higher N than Baseline. +Results Summary and DiscussionAverage airspace design performance is summarized as a percent increase or decrease from Baseline in Table 1.The ultimate goal of each algorithm was to provide user benefits by reconfiguring airspace.The results show positive delay reduction benefits (d) in all but one algorithm, achieving the algorithms' goal.A few algorithms do better than others at minimizing traffic pattern complexity (α, β, γ), but all do a fairly decent job.In general, algorithms that aggressively change the airspace show more delay reduction benefits but at higher reconfiguration costs (V, N).The reconfiguration costs are expected and are acceptable as long as they are manageable.The three methods using Baseline elements performed very differently.DAU Slices achieved modest benefits with modest negative effects to traffic pattern and reconfiguration cost.This was expected as DAU Slices is the most conservative method, designed to allow small changes to existing airspace design at high reconfiguration frequency.CombineSplit was the only method to worsen d.All other CombineSplit metrics are similar to Baseline.This method was designed for a more tactical application, suggesting configurations every 15 minutes over a two-hour horizon.CombineSplit actually decreased delay in a study comparing DAC benefits when applied to a more tactical two-hour weather rerouting scenario when number of sectors remained the same [34].FlightLevel is the most unique case with widely varying results.It achieved modest d improvement without negatively affecting reconfiguration complexity.However, the traffic pattern results suggest that more research is needed to determine if FlightLevel configurations are feasible.FlightLevel significantly improved α and β metrics because these metrics did not consider vertical boundaries.The significantly worsened γ due to flights climbing or descending through sectors only a few flight levels thick may not be acceptable.Most recent freeform algorithm development has focused on improving traffic pattern complexity.The improvement of SectorFlow from SectorFlow'08 in this area was demonstrated in the Traffic Pattern Complexity subsection.The most aggressive freeform methods (Voronoi, SectorFlow, and CellGeoSect) produced the greatest delay reduction benefit, but they also significantly increased reconfiguration complexity.Voronoi was the only method to reduce delay relative to Baseline without negatively affecting traffic pattern complexity, making this the most attractive method if the reconfiguration complexity increase is manageable.Reconfiguration complexity thresholds when using DataComm-based controller tools such as those used in [15] have yet to be determined.With the right controller tools and further algorithm refinement to reduce reconfiguration complexity, achieving the higher benefits of these more aggressive methods may be feasible. +ConclusionsA fast-time simulation study compared the performance of solutions from new airspace design methods to a representation of current day dynamic airspace operations.Three categories of metrics compared delay reduction benefits, traffic pattern complexity, and reconfiguration complexity.Most methods achieved benefits by decreasing delay, which was augmented by allowing strategic airspace design updates.Most methods also did a reasonable job of keeping traffic pattern complexity low.Methods using design elements from Baseline had more modest benefits and reconfiguration complexity.Freeform airspace design methods achieved the highest benefits and highest increase in reconfiguration complexity.Future research is needed to determine if high reconfiguration complexity is acceptable given the right controller tools.Airspace design methods may also further refine their algorithms to minimize reconfiguration complexity.Figure 22Figure 2 diagrams the reconfiguration scenario design.Each box represents a configuration designed for the corresponding time periods.Orange boxes identify active configurations.Black boxes identify planned configurations that were updated before they could be implemented.The reconfiguration times, shown in Central Standard Time, are identical to the Baseline.First, airspace design algorithms use projected flight tracks for the entire day to design an initial three-configuration scenario, labeled as Config 1, 2, and 3, with 6, 24, and 19 sectors above 24,000 ft, respectively.Flight traffic is simulated through Config 1 from 1:45 AM to 8:15 AM, at which point the airspace design algorithm may use updated projected flight tracks to modify the remaining two configurations.These modified configurations are labeled Config 2ʹ′ and 3ʹ′.The flight traffic simulation continues through Config 2ʹ′ from 8:15AM to 9:00PM, at which point the airspace design algorithm may used updated projected flight tracks to modify the last configuration, Config 3ʹ′ʹ′.The simulation completes through Config 3ʹ′ʹ′ from 9:00PM to 1:45AM. +reconfiguration scenarios. +Figure 3 .3Figure 3. Iterative Simulation Process +s 1 , s 2 , and s 3 , shown as [Config 1, Config 2, Config 3], [Config 1, Config 2ʹ′, Config 3ʹ′], and [Config 1, Config 2ʹ′, Config 3ʹ′ʹ′] in Figure 3. +Figure 5 .5Figure 5. Average Total Delay +Figure 6 .6Figure 6.Average α Results +Figure 7 .7Figure 7. Average β Results +Figure 8 .8Figure 8. Averages and Quartiles for γ Results +Figure 9 .Figure 10 .910Figure 9. V Results for First Reconfiguration +Figure 11 .Figure 12 .1112Figure 11.N Results for First Reconfiguration +5 .5Yellow cells with values close to zero are similar to Baseline.Red cells with negative values are worse and green cells with positive values are better than Baseline.Darker shaded red and green cells have increasingly worse or better results, respectively.Delay performance, d, is based on d(s 3 ) from Figure Traffic pattern complexity performances are based on values from Figure 6 for α, Figure 7 for β, and averages from Figure 8 for γ.Reconfiguration complexity performances are based on averages of the averages from Figures 9 and 10 for V and from Figures 11 and 12 for N. +Table 1 . Airspace Design Performance Summary1DAU SlicesCombineSplitFlightLevelGraph-basedSectorFlowCellGeoSectVoronoid23-57 2518306858α-13 -430-16 -15 -17 -8β063512692γ-22 2-88 -22 -29 -25 16V-18 -11 12-29 -56 -55 -100N-56 11-2-85 -112 -90 -176WorseSimilarBetter + + + + +AcknowledgementsThe authors thank all algorithm developers who submitted reconfiguration scenarios for this study. + + + + + + + + + A Comparison of Algorithm Generated Sectorizations + + ShannonZelinski + + 10.2514/atcq.18.3.279 + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 18 + 3 + + 2009 + American Institute of Aeronautics and Astronautics (AIAA) + Napa Valley, California + + + Zelinski, S., 2009, A Comparison of Algorithm Generated Sectorizations. In 8th USA/Europe ATM R&D Seminar, Napa Valley, California. + + + + + A Comparison of Algorithm Generated Sectorizations + + ShannonZelinski + + 10.2514/atcq.18.3.279 + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 18 + 3 + + 2010 + American Institute of Aeronautics and Astronautics (AIAA) + + + Zelinski, S., 2010, A Comparison of Algorithm Generated Sectorizations. In Air traffic Control Quarterly, Vol. 18, No. 3, pp. 279-301. + + + + + State energy conservation plan for New Mexico + + CVerlhac + + + SManchon + + 10.2172/5979880 + + + 4th USA/Europe ATM R&D Seminar + Santa Fe, New Mexico + + Office of Scientific and Technical Information (OSTI) + 2001 + + + Verlhac, C. and S. Manchon, 2001, Optimization Of Opening Schemes. In 4th USA/Europe ATM R&D Seminar, Santa Fe, New Mexico. + + + + + A Tool to Design Functional Airspace Blocks + + CEBichot + + + NDurand + + + + th USA/Europe ATM R&D Seminar + Barcelona, Spain + + 2007 + + + Bichot, C. E. and N. Durand, 2007, A Tool to Design Functional Airspace Blocks. In 7th USA/Europe ATM R&D Seminar, Barcelona, Spain. + + + + + Ground-water hydrology of northern Napa Valley, California + + DGianazza + + + CAllignol + + + NSaporito + + 10.3133/wri7313 + + + th USA/Europe ATM R&D Seminar + Napa Valley, California + + US Geological Survey + 2009 + + + Gianazza, D., C. Allignol, and N. Saporito, 2009, An Efficient Airspace Configuration Forecast. In 8th USA/Europe ATM R&D Seminar, Napa Valley, California. + + + + + Algorithms for Combining Airspace Sectors + + MichaelBloem + + + ParimalKopardekar + + + PramodGupta + + 10.2514/atcq.17.3.245 + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 17 + 3 + + 2009 + American Institute of Aeronautics and Astronautics (AIAA) + + + Bloem, M., P. Gupta, and P. Kopardekar, 2009, Algorithms for Combining Airspace Sectors. In Air traffic Control Quarterly, Vol. 17, No. 3, pp. 245- 268. + + + + + A constraint-programming formulation for dynamic airspace sectorization + + HTrandac + + + Vu Duong + + 10.1109/dasc.2002.1067897 + + + Proceedings. The 21st Digital Avionics Systems Conference + The 21st Digital Avionics Systems Conference + + IEEE + 2003 + + + Trandac, H., P. Baptiste, V. and Duong, 2003, Airspace Sectorization by Constraint Programming. In Proceedings de la 1re confrence en Recherche Informatique Vietnam & Francophone (RIVF). + + + + + Airspace design process for dynamic sectorisation + + RuedigerEhrmanntraut + + + StuartMcmillan + + 10.1109/dasc.2007.4391888 + + + 2007 IEEE/AIAA 26th Digital Avionics Systems Conference + Dallas, Texas + + IEEE + 2007 + + + Ehrmanntraut, R. and S. McMillan, 2007, Airspace Design Process for DYnamic Sectorisation. In 26th Digital Avionics System Conference, Dallas, Texas. + + + + + + RSConker + + + DAMoch-Mooney + + + WPNiedringhous + + + BTSimmons + + New Process for "Clean Sheet" Airspace Design and Evaluation + Barcelona, Spain + + 2007 + + + th USA/Europe ATM R&D Seminar + Conker, R.S., D.A. Moch-Mooney, W.P. Niedringhous, B.T. and Simmons, 2007, New Process for "Clean Sheet" Airspace Design and Evaluation. In 7th USA/Europe ATM R&D Seminar, Barcelona, Spain. + + + + + Airspace partitioning using flight clustering and computational geometry + + ChrisRBrinton + + + StephenPledgie + + 10.1109/dasc.2008.4702800 + + + 2008 IEEE/AIAA 27th Digital Avionics Systems Conference + St. Paul, Minnesota + + IEEE + 2008. October 26-30 + + + Brinton, C. and S. Pledgie, 2008, Airspace Partitioning Using Flight Clustering and Computational Geometry, In 27th Digital Avionics System Conference, October 26-30, St. Paul, Minnesota. + + + + + Analysis of an optimal sector design method + + MichaelDrew + + 10.1109/dasc.2008.4702801 + + + 2008 IEEE/AIAA 27th Digital Avionics Systems Conference + St. Paul, Minnesota + + IEEE + 2008 + + + Drew, M., 2008, Analysis of an Optimal Sector Design Method. In 27th Digital Avionics System Conference, St. Paul, Minnesota. + + + + + Airspace Sector Redesign Based on Voronoi Diagrams + + MinXue + + 10.2514/6.2008-7223 + AIAA-2008-7223 + + + AIAA Guidance, Navigation and Control Conference and Exhibit + Honolulu, Hawaii + + American Institute of Aeronautics and Astronautics + 2008 + + + Xue, M., 2008, Airspace Sector Redesign Based on Voronoi Diagrams. In AIAA Guidance, Navigation and Control Conference and Exhibit, Au- gust, Honolulu, Hawaii. AIAA-2008-7223. + + + + + Air Traffic Controller Ability Requirements in the U.S. National Airspace System + + STien + + + RHoffman + + 10.4324/9781315242538-12 + + + Staffing the ATM System + Napa Valley, California + + Routledge + 2009 + + + + Tien, S. and R. Hoffman, 2009, Optimizing Airspace Sectors for Varying Demand Patterns Using Multi-Controller Staffing. In 8th USA/Europe ATM R&D Seminar, Napa Valley, California. + + + + + A Spectral Clustering Based Algorithm for Dynamic Airspace Configuration + + JinhuaLi + + + TongWang + + + InseokHwang + + + InseokHwang + + 10.2514/6.2009-7056 + AIAA-2009- 7056 + + + 9th AIAA Aviation Technology, Integration, and Operations Conference (ATIO) + Hilton Head, South Carolina + + American Institute of Aeronautics and Astronautics + 2009. September 21-23 + + + Li, J., T. Wang, and I. Hwang, 2009, A Spectral Clustering Based Algorithm for Dynamic Airspace Configuration, In 9th AIAA Aviation Technology, Integration and Operations Conference, September 21-23, Hilton Head, South Carolina, AIAA-2009- 7056. + + + + + Impact of Airspace Reconfiguration on Controller Workload and Traffic Performance + + PaulLee + + + NancySmith + + + ThomasPrevot + + + JeffreyHomola + + + HwasooLee + + + AngelaKessell + + + ConnieBrasil + + 10.1201/ebk1439834916-c54 + + + Advances in Human Factors and Ergonomics Series + Miami, Florida + + CRC Press + 2010 + + + + Lee, P.U., T. Prevot, J. Homola, H. Lee, Kessell, A., C. Brasil, and N. Smith, 2010, Impact of Airspace Reconfiguration on Controller Workload and Task Performance, In 3rd Intl. Conference on Applied Hu- man Factors and Ergonomics, Miami, Florida. + + + + + Dynamic FPAs: A new method for dynamic airspace configuration + + AlexanderKlein + + + MarkDRodgers + + + Hong Kaing + + 10.1109/icnsurv.2008.4559176 + + + 2008 Integrated Communications, Navigation and Surveillance Conference + Washington, DC + + IEEE + August 2008 + + + Dynamic Airspace Configuration Benefits. Report submitted by M. Rodgers et al, under NASA Contract NNA07BB31C, Washington, DC, August 2008. + + + + + Algorithms for Combining Airspace Sectors + + MichaelBloem + + + ParimalKopardekar + + + PramodGupta + + 10.2514/atcq.17.3.245 + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 17 + 3 + + 2010. September 19-24 + American Institute of Aeronautics and Astronautics (AIAA) + Nice, France + + + In 27th Congress of the International Council of the Aeronautical Sciences + Bloem, M. and P. Gupta, 2010, Configuring Airspace Sectors with Approximate Dynamic Programming, In 27th Congress of the International Council of the Aeronautical Sciences, September 19- 24, Nice, France. + + + + + Flight Level-based Dynamic Airspace Configuration + + KennethLeiden + + + StevenPeters + + + StaceyQuesada + + 10.2514/6.2009-7104 + AIAA-2009-7104 + + + 9th AIAA Aviation Technology, Integration, and Operations Conference (ATIO) + Hilton Head, South Carolina + + American Institute of Aeronautics and Astronautics + 2009. September 21-23 + + + Leiden, K., S. Peters, S. and Quesada, 2009, Flight Level-based Dynamic Airspace Configuration, In 9th AIAA Aviation Technology, Integration and Operations Conference, September 21-23, Hilton Head, South Carolina, AIAA-2009-7104. + + + + + Generic Airspace Concepts and Research + + RichardMogford + + 10.2514/6.2010-9159 + AIAA-2010-9159 + + + 10th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Fort Worth, Texas + + American Institute of Aeronautics and Astronautics + 2010. September 13-15 + + + Mogford, R., 2010, Generic Airspace Concepts and Research, In 10th AIAA Aviation Technology, Integration and Operations Conference, September 13-15, Fort Worth, Texas, AIAA-2010-9159. + + + + + Airspace Sectorization by Dynamic Density + + ChrisBrinton + + + JimHinkey + + + KenLeiden + + 10.2514/6.2009-7102 + AIAA-2009-7102 + + + 9th AIAA Aviation Technology, Integration, and Operations Conference (ATIO) + Hilton Head, South Carolina + + American Institute of Aeronautics and Astronautics + 2009. September 21-23 + + + Brinton, C., K. Leiden, and J. Hinkey, 2009, Airspace Sectorization by Dynamic Density, In 9th AIAA Aviation Technology, Integration and Operations Conference, September 21-23, Hilton Head, South Carolina, AIAA-2009-7102. + + + + + Flow Conforming Operational Airspace Sector Design + + GirishkumarSabhnani + + + ArashYousefi + + + JosephS BMitchell + + 10.2514/6.2010-9377 + AIAA-2010-9377 + + + 10th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Fort Worth, Texas + + American Institute of Aeronautics and Astronautics + 2010. September 13-15 + + + Sabhnani, G., A. Yousefi, and J. Mitchell, 2010, Flow Conforming Operational Airspace Sector Design, In 10th AIAA Aviation Technology, Integration and Operations Conference, September 13-15, Fort Worth, Texas, AIAA-2010-9377. + + + + + Three Dimensional Sector Design with Optimal Number of Sectors + + MinXue + + 10.2514/6.2010-8290 + + + AIAA Guidance, Navigation, and Control Conference + Toronto, Ontario Canada + + American Institute of Aeronautics and Astronautics + 2010. August 2-5 + + + Xue, M., 2010, Three Dimensional Sector Design with Optimal Number of Sectors, In AIAA Guidance, Navigation, and Control Conference and Exhibit, August 2-5, Toronto, Ontario Canada. + + + + + Operational dynamic configuration analysis + + ChokFLai + + + SZelinski + + 10.1109/dasc.2010.5655496 + + + 29th Digital Avionics Systems Conference + Salt Lake City, Utah + + IEEE + 2010. October 3-7 + + + Lai, C.F. and S. Zelinski, 2010, Operational Dynamic Configuration Analysis. In 29th Digital Avionics System Conference, October 3-7, Salt Lake City, Utah. + + + + + Macroscopic Capacity Model with Individual Sector Closing Speed Estimates + + JerryWelch + + + JohnAndrews + + 10.2514/6.2009-7105 + + + 9th AIAA Aviation Technology, Integration, and Operations Conference (ATIO) + Barcelona, Spain + + American Institute of Aeronautics and Astronautics + 2007 + + + Welch, J., 2007, Macroscopic Workload Model for Estimating EnRoute Sector Capacity. In 7th USA/Europe ATM R&D Seminar, Barcelona, Spain. + + + + + Build 4 of the Airspace Concept Evaluation System + + LarryMeyn + + + RobertWindhorst + + + KarlinRoth + + + DonaldVan Drei + + + GregKubat + + + VikramManikonda + + + SharleneRoney + + + GeorgeHunter + + + AlexHuang + + + GeorgeCouluris + + 10.2514/6.2006-6110 + AIAA-2006-6110 + + + AIAA Modeling and Simulation Technologies Conference and Exhibit + Keystone, Colorado + + American Institute of Aeronautics and Astronautics + 2006 + + + Meyn, L., R. Windhorst, K. Roth, D. VanDrei, G. Kubat, V. Manikonda, S. Roney, G. Hunter, A. Huang, and G. Couluris, 2006, Build 4 of the Airspace Concept Evaluation System. In AIAA Modeling and Simulation Technologies Conference an Exhibit, Keystone, Colorado. AIAA-2006-6110. + + + + + Validating The Airspace Concept Evaluation System Using Real World Data + + ShannonZelinski + + 10.2514/6.2005-6491 + + + AIAA Modeling and Simulation Technologies Conference and Exhibit + San Francisco, California + + American Institute of Aeronautics and Astronautics + 2005 + + + Zelinski, S., 2005, Validation of the Airspace Concept Evaluation System Using Real World Data. In AIAA Modeling and Simulation Technologies Conference and Exhibit, San Francisco, California. + + + + + Validating the Airspace Concept Evaluation System for Different Weather Days + + ShannonZelinski + + + LarryMeyn + + 10.2514/6.2006-6115 + + + AIAA Modeling and Simulation Technologies Conference and Exhibit + Keystone, Colorado + + American Institute of Aeronautics and Astronautics + 2006 + + + Zelinski, S. and L. Meyn, 2006, Validation of the Airspace Concept Evaluation System for Multiple Weather Days. In AIAA Modeling and Simulation Technologies Conference an Exhibit, Keystone, Colorado. + + + + + Futuristic US Flight Demand Generation Approach Incorporating Fleet Mix Evolution + + AlexHuang + + + DavidSchleicher + + 10.2514/6.2008-6678 + AIAA 2008-6678 + + + AIAA Modeling and Simulation Technologies Conference and Exhibit + Honolulu, Hawaii + + American Institute of Aeronautics and Astronautics + 2008. August 2008 + + + + Huang, A. and D. Schleicher, 2008, Futuristic US Flight Demand Generation Approach Incorporating Fleet Mix Assumptions. In AIAA Modeling and Simulation Technologies Conference an Exhibit, 18 -21 August 2008, Honolulu, Hawaii. AIAA 2008-6678. + + + + + Analysis of Current Airspace Operations and Implications for Dynamic Airspace Configuration + + ChrisBrinton + + + LaraCook + + 10.2514/6.2008-7224 + AIAA- 2008-7224 + + + AIAA Guidance, Navigation and Control Conference and Exhibit + Honolulu, Hawaii + + American Institute of Aeronautics and Astronautics + 2008. August 18-21 + + + Brinton, C. and L. Cook, 2008, Analysis of Current Airspace Operations and Implications for Dynamic Airspace Configuration. In AIAA Modeling and Simulation Technologies Conference and Exhibit, August 18-21, Honolulu, Hawaii, AIAA- 2008-7224. + + + + + Airspace Complexity Measurement: An Air Traffic Control Simulation Analysis + + PKopardekar + + + + th USA/Europe ATM R&D Seminar + Barcelona, Spain + + 2007 + + + Kopardekar, P., 2007, Airspace Complexity Measurement: An Air Traffic Control Simulation Analysis. In 7th USA/Europe ATM R&D Seminar, Barcelona, Spain. + + + + + Effect of Dynamic Sector Boundary Changes on Air Traffic Controllers + + JaewooJung + + + PaulLee + + + AngelaKessell + + + JeffreyHomola + + + ShannonZelinski + + 10.2514/6.2010-8289 + AIAA-2010-8289 + + + AIAA Guidance, Navigation, and Control Conference + Toronto, Ontario Canada + + American Institute of Aeronautics and Astronautics + 2010. August 2-5 + + + Jung, J., P. Lee, A. Kessell, J. Homola, and S. Zelinski, 2010, Effect of Dynamic Sector Boundary Changes on Air Traffic Controllers. In AIAA Guidance, Navigation, and Control Conference and Exhibit, August 2-5, Toronto, Ontario Canada, AIAA-2010-8289. + + + + + A Human-in-the Loop Exploration of the Dynamic Airspace Configuration Concept + + JeffreyHomola + + + PaulLee + + + ThomasPrevot + + + HwasooLee + + + AngelaKessell + + + ConnieBrasil + + + NancySmith + + 10.2514/6.2010-8293 + + + AIAA Guidance, Navigation, and Control Conference + + American Institute of Aeronautics and Astronautics + 2010 + + + Homola, J., P. Lee, T. Prevot, H. Lee, A. Kessell, C. Brasil, N. and Smith, 2010, A Human-in- the-Loop Exploration of the Dynamic Airspace Configuration Concept. In AIAA Guidance, Navigation and Control Conference and Exhibit, + + + + + Trigger Metrics for Dynamic Airspace Configuration + + ArashYousefi + + + RobertHoffman + + + MarcusLowther + + + BabakKhorrami + + + HerbertHackney + + 10.2514/6.2009-7103 + AIAA-2009-7103 + + + 9th AIAA Aviation Technology, Integration, and Operations Conference (ATIO) + Hilton Head, South Carolina + + American Institute of Aeronautics and Astronautics + 2009. September 21-23 + + + Yousefi, A., R. Hoffman, M. Lowther, B. Khorrami, H. and Hackney, 2009, Trigger Metrics for Dynamic Airspace Configuration, In 9th AIAA Aviation Technology, Integration and Operations Conference, September 21-23, Hilton Head, South Carolina, AIAA-2009-7103. + + + + + Analysis of Regional Airspace Reconfigurations in Presence of Convective Weather + + JJung + + + CFLai + + + SZelinski + + for AS.2.3.05. 30 + + + th Digital Avionics Systems Conference + + 2010. October 16-20, 2011 + + + NASA Milestone report + Jung, J., C.F. Lai, and S. Zelinski, 2010, Analysis of Regional Airspace Reconfigurations in Presence of Convective Weather, NASA Milestone report for AS.2.3.05. 30th Digital Avionics Systems Conference October 16-20, 2011 + + + + + + diff --git a/file836.txt b/file836.txt new file mode 100644 index 0000000000000000000000000000000000000000..b38adac4b2913052f055d96a26e536f4f46c43cd --- /dev/null +++ b/file836.txt @@ -0,0 +1,302 @@ + + + + +This paper describes a method for defining route structure from flight tracks.Dynamically generated route structures could be useful in guiding dynamic airspace configuration and helping controllers retain situational awareness under dynamically changing traffic conditions.Individual merge and diverge intersections between pairs of flights are identified, clustered, and grouped into nodes of a route structure network.Links are placed between nodes to represent major traffic flows.A parametric analysis determined the algorithm input parameters producing route structures of current day flight plans that are closest to todays airway structure.These parameters are then used to define and analyze the dynamic route structure over the course of a day for current day flight paths.Route structures are also compared between current day flight paths and more user preferred paths such as great circle and weather avoidance routing.The current airspace sector design evolved over decades, driven by human-centered separation assurance and an airway system of ground-based navaids.The future system is migrating toward more direct and flexible user-preferred routing.Automation tools and dynamic reconfiguration of airspace boundaries will enable controllers to adapt to changing weather and traffic conditions and increase controller staffing flexibility.Structure-based abstractions of traffic conditions are necessary both to trigger and guide airspace boundary changes and to quickly give controllers situational awareness.Histon et al [1] identified elements of structure-based traffic abstraction that controllers use to maintain airspace situation awareness.Two key elements identified, standard flows and critical points (where major flows cross and merge), have been used as inputs in several automated airspace boundary reconfiguration methods [2,3,4,5].Complex air traffic networks have been developed to study aggregate system behavior and robustness to disturbances [6].Route structure is an abstraction of the air traffic network based on standard flows and critical points.A method of quickly determining route structure is necessary for controllers and airspace configuration to adjust to flexible userpreferred routing.Several methods have been proposed to define the route structure elements of a given set of flight trajectories.Most methods begin with two-dimentional grid structures that capture information about traffic passing through each grid cell.Xue [3] used the heading variance of flights within each grid cell to determine if the cell belonged to a major traffic flow or intersection point and could be used as an airspace boundary constraint.Other methods used flight occupancy counts within each grid cell along with other techniques to bundle flight trajectories into flows [7] or to create an abstract network flow graph of traffic [4].Of the methods that defined route structure connections between flows and critical points, Sabhnani et al [7] proposed two top-down approaches that first identified flows and then defined their intersections to be critical points, whereas work preceding this paper [8] proposed a bottom-up approach that identified critical points first.Li et al [5] took a short cut by refining a complex route structure generated from flight plans.However, this approach depended upon the existing airway structure elements found in today's flight plans.This paper extends preceding work [8] by adding an altitude component to critical point identification and linking critical points together with flows to form a route structure.A parametric analysis is performed to determine the algorithm input parameters that produce route structures that are closest to today's airway structure.These parameters are then used to extract and analyze the dynamic changing route structure over the course of a day for current day flight paths and for more user preferred paths such as great circle and weather avoidance routing.The remainder of this paper is organized as follows.Section 2 describes the method of defining route structure from flight tracks.A parametric analysis identifies optimal method parameters in section 3.Sections 4 and 5 analyze route structures from dynamically shifting time windows of traffic, and from alternate user preferred routing, respectively.Finally, conclusions are presented in section 6. +Defining Route StructureThis section describes how flight track data are processed to define a route structure.First individual intersection points were generated from pairs of flights within lateral and vertical proximity of one another.Intersection points within another set of specified lateral and vertical proximity of one another formed cluster points.Clusters were further grouped into nodes to simplify the structure.Finally, links were placed between nodes with the nominal paths of flights forming each link representing flow paths. +Proximate PointsThe first step in route structure generation identifies points on pairs of flight tracks that come within a specified proximity of one another.Let P = p 1 , ..., p n be an ordered set of latitude/longitude/altitude/time coordinates describing the trajectory of flight P. Similarly, let Q = q 1 , ..., q m describe the trajectory of flight Q.For all points within a specified time range, any point in P that comes within a lateral radius r I and vertical distance z I of any point in Q is a proximate point of P with respect to Q. +IntersectionsMerge and diverge intersections classify the interactions between pairs of flights.When two flights come within a specified proximity of one another, their paths merge.When two flights originally within a specified proximity of one another are no longer within the specified proximity, their paths diverge.For every consecutive sequence of flight track coordinates on P that are proximate to Q, the first point is a merge intersection, and the last point is a diverge intersection.For every intersection on P with respect to Q, an intersection is also identified on Q with respect to P.Boundary intersections occur where flights cross a given airspace boundary.Any two consecutive points on a flight track that are in different specified airspace regions, such as sectors or centers, produce boundary intersections.These are not necessary to generate a route structure.However, if a route structure is generated for only a specified bounded region, boundary intersections make it possible to identify flows that enter or exit the region.Rather than storing single instances of intersection, flight track points are tagged as merge and diverge intersections with respect to multiple other flights and as boundary intersections when straddling a boundary.Each flight track point tagged with merges defines a merge intersection I Mi with position p(I Mi ) identical to the track point position and with weight w(I Mi ) equal to the number of individual merges identified with respect to other flights at this track point.A diverge intersection I Di is defined similarly.A boundary intersection I Bi is different only in that w(I Bi ) ≡ 1 because a single track point cannot intersect the same boundary multiple times.This method allows a single flight track point to define up to three intersections, one of each type.The resulting set of clusters may contain numerous low weight outliers, caused by random traffic, which should be filtered.A different weight threshold should be used for filtering different time ranges of flight track data analyzed.Let τ be the flight track data time range and let ω C be a rate threshold such that cluster C i is filtered if w(C i ) < τ• ω C .This way, the same ω C may be used for various τ with similar filtering effect.Because boundary cluster weights are determined by instances of flights crossing a boundary rather than instances of flight pair intersections, a unique ω C B filters boundary clusters. +Clusters +NodesNodes are at the top level of intersection groupings and serve as the end points for links representing traffic flows.Crossing traffic flows may form merge and diverge clusters very close to one another or merge or diverge clusters may exist very close to the boundary.Therefore, nodes are used to represent groups of clusters of mixed type.Building a network of flows between nodes rather than individual clusters significantly simplifies and strengthens the resulting route structure.Each cluster may belong to only one node.A boundary node refers to any node containing a boundary cluster.Interior nodes refer to nodes without any boundary clusters.The position p(N i ) of node N i is the centroid of all clusters within N i .Node weight w(N i ) is the sum weight of all clusters within N i .Nodes can also be expressed as groups of intersections belonging to its clusters (I → C → N ≡ I → N).The spacial size of node N i is measured by the three dimensional standard deviation, σ(N i ) = (σ x (N i ), σ y (N i ), σ z (N i )), of all I within N i from p(N i ).A different method is used to group clusters into nodes than intersections into clusters.Initially each cluster defines a unique node.Nodes with overlapping σ are too close with respect to their spacial size to serve as traffic flow end points.Therefore, if the distances between nodes d(N i , N j ) < σ(N i ) + σ(N j ) in all three dimensions, then clusters within N i and N j are grouped into a single node.Additionally, if a single track point defines multiple intersections grouped to unique nodes, these nodes are grouped into a single node. +LinksLinks are directed connections between pairs of nodes representing traffic flow.Through I → N mappings, all flights with track points defining intersections, may be expressed as a sequence of nodes along the flight trajectory.Let L i j specify a set of flight segments from N i to N j .Only segments between consecutive nodes along a flight trajectory are considered.Let link weight w(L i j ) be the number of flight segments in L i j .Let P(L i j ) specify the nominal path or flow path of all segments in L i j .Link paths may curve and bend around special use airspace or weather as they travel between nodes.The sample rate used to define link paths is one point every 8 nmi.This sample rate is similar in distance to the 1-minute resolution flight tracks analyzed.As with clusters, the resulting set of links may have many low weight outliers which should be filtered.Let ω L be a rate threshold such that link L i j is filtered if w(L i j ) < τ• ω L .Filtering links also has an additional filtering effect on nodes.Boundary nodes with less than one link and interior nodes with less than two links are filtered.If nodes are filtered, links are recalculated and the process iterates until the number of nodes and links stabilizes. +Parametric AnalysisThere are a number of parameters described in section 2 that affect the fidelity of the resulting route structure.Route structure links and nodes are functionally equivalent to today's airways and their intersections.Results from a range of parameters are analyzed to find values that produce route structure similar to today's airway structure.Flight tracks used to generate the route structures were 1-minute resolution simulated trajectories of flight schedules and filed flight plans from a high traffic volume, good weather Tuesday on 4/21/2009.This analysis focused on high altitude tracks in Kansas City Center (ZKC) during the peak four hours, approximately 15:30-19:30 ZKC local time.Because the ZKC high altitude sector floor was 24,000 ft, only tracks at 24,000 ft and above were considered.There are four types of airways in the US.V-routes (below 18,000 ft) and J-routes (above 18,000 ft) are classic airways using ground-based navaids.T-routes (below 18,000 ft) and Q-routes (above 18,000 ft) are newer area navigation or RNAV-based airways.This analysis focused on the 38 J-routes and 2 Q-routes in high altitude ZKC shown in Fig. 1. +ParametersThere are three types of parameters used to define route structure.The parameters used to define intersection points are r I and z I .filter low weight clusters and links.The range of intersection and cluster parameters explored was designed with aircraft separation tolerances in mind.En-route lateral and vertical aircraft separation tolerances are five nmi and 1,000 ft, respectively.Therefore, the analysis considered values of r I , r C , and r C B between three and 30 nmi and values of z I , z C , and z C B between 500 and 3,000 ft.Recall from section 2 that each flight pair interaction produces two intersection points, one on each flight trajectory, separated by as much as r I laterally and z I vertically.Therefore, r C ≥ r I and z C ≥ z I to guarantee that these points are clustered.Filtering parameters were increased from zero. +Route Structure EvaluationRoute structure results from each set of parameters were evaluated to determine their precision and accuracy.Precision metrics compare route structure to the individual flight tracks they represent and accuracy metrics compare route structure to the current airway structure.Precision metrics include traffic deviations from route structure and route structure intersection representation.Let σ r (L i j ) be the lateral standard deviation, of all flight track segments in L i j from P(L i j ).Let traffic deviation σ r (L) be the average of all σ r (L i j ) weighted by w(L i j ).Route structure intersection representation, ρ, measures the ratio of flight intersections represented by the final route structure.It is the ratio of sum weight of all nodes over the sum weight of all intersections before cluster filtering.If all filtering parameters were zero, ρ would always be one.Route structure accuracy metrics compare route structure with airway structure to find the right balance of abstraction and precision.Let S i j specify the airway segment between airway intersections M i and M j .Just as link nominal path points are positioned every eight nmi along each path, airway path points are placed every eight nmi along each airway segment.Let P(S i j ) be the sequence of (x, y) points every eight nmi along S i j .Unlike links, airway segments do not have a vertical component or a weight.Accuracy metrics include deviation and representation metrics between airway and route structure.The route structure deviation, δ R , is the average lateral distance of each link nominal path point p k (L i j ) from its closest airway segment, weighted by w(L i j ).The airway structure deviation, δ A , is the average lateral distance of each airway path point p k (S i j ) from it's closest link.Airway and route structure representations are the ratios of each structure found to be closest to a piece of the other structure.Let n S i j be the number of path points on P(S i j ).Let n L(S i j ) be the number of points from all P(L) identifying S i j as the closest airway segment.The airway structure representation is defined asλ A = ∑ min[n S i j , n L(S i j ) ]∑ n S i j .Similarly, let n L i j be the number of path points on P(L i j ) and let n S(L i j ) be the number of points from all P(S) identifying L i j as the closest link.The route structure representation is defined asλ R = ∑ [min[n L i j , n S(L i j ) ]• w(L i j )] ∑ [n L i j • w(L i j )]. This is similar to λ A calculation, only λ R is weighted by w(L i j ).Table 1 summarizes the precision and accuracy metrics.The right column indicates the goal of the parametric analysis to either minimize (-) or maximize (+) each metric.Note that the goal is to minimize all deviation metrics and maximize all representation metrics. +ResultsThe goal of the parametric analysis was to find parameters that minimized σ and δ metrics, and maximized ρ and λ metrics.A simple optimization formula was designed to combine the metrics in Table 1 into a single value to minimize by multiplying metrics to minimize and dividing by metrics to maximize.J = σ r (L)• δ R • δ A ρ• λ R • λ AThis function was used to guide the search for route structure parameters that produced the smallest deviation metrics with the largest and most balanced structure representation metrics.Balanced structure representation metrics ensured that the airway and route structure network lengths were similar and thus represented traffic at a similar level of abstraction.Table 2 shows the set of optimal parameter values identified and used in remaining route structure analyses.The optimal intersection parameters were very similar to standard separation criteria with r I equalling the lateral standard and z I only 200 ft greater than the verti-cal standard.As expected, all clustering parameters were greater than intersection parameters.Boundary intersection clustering required a r C B almost three times r C .Merge and diverge clustering produced much larger weight clusters than boundary intersection clustering.Therefore, ω C is eight times larger than ω C B .Links were less seldom formed due to the requirement of full flight trajectory segments between nodes and because they were iteratively filtered.Therefore, ω L was never raised above one.Table 2 Optimal set of route structure generation parameters.r I = 5 nmi z I = 1,200 ft r C = 6 nmi z C = 2,000 ft r C B = 17 nmi z C B = 2,000 ft ω C = 20 w/hr ω C B = 2.5 w/hr ω L = 1 w/hrTable 3 shows the evaluation metrics obtained using the parameters in Table 2.In general, σ r (L) responded mostly to intersection and cluster parameters.As ω C and ω C B were increased, ρ and λ A decreased slightly, and λ R increased.When the filtering parameters were increased too much, ρ and λ A began to fall more rapidly.The higher λ A and λ R , the lower there respective δ A and δ R were.This is because portions of the unrepresented structure began to skew deviation metrics with closest distances to parts of the other structure that clearly did not match.In general, δ A tended to be greater than δ R because many unrepresented links were in similar lateral locations as other links at different altitudes.The airway structure was more prone to skewing δ A due to underutilized airway segments.The optimal parameters were chosen such that λ A and λ R were somewhat balanced and as high as possible.Figure 2 illustrates the route structure generated from the parameters in Table 2. Links are shown as lines with color indicating link weight.Nodes are shown as red dots with size indicating node weight.Because node weight tends to increase very quickly, the square root of node weight is shown to better view relative Table 3 Evaluation metrics obtained from optimal set of route structure generation parameters.σ r (L) = 1.79 nmi ρ = 0.57 δ A = 5.81 nmi λ A = 0.73 δ R = 1.67 nmi λ R = 0.53 weights.Black and gray lines underlying the route structure are the same airway segments and sector boundaries, respectively, shown in Fig. 1.The route structure completely overlaps many of these black lines.Figure 2 visually confirms the accuracy metrics from Table 3.It can be seen that route structure elements that follow the airway structure, follow closely.Some airway segments do not get enough traffic to be represented at all whereas a few links highlight flows where no published airway segment exists.Because route structure is altitude specific, Fig. 2 shows several instances where links appear to overlap or cross without a node.Fig. 2 Route structure generated from optimal parameters overlaying airways. +Robustness VerificationThe parametric analysis calibrated route structure parameters to a single day.To demonstrate that these parameters do not need to be re-calibrated, the same parameters optimized One of the main potential benefits of route structure is that it may be generated dynamically to visualize how the structure changes with time.Figure 6 shows intersection, airway, and route structure representations (ρ, λ A , and λ R ) for each of the route structures from Fig. 3. Intersection and airway representations increase with number of flights, whereas route structure representation decreases.This is because as segments of the airway structure saturate, more and more flows are formed in places where no airway has been published.Average deviation metrics are shown in Fig.!" #!" $!!" $#!" %!!" %#!" &!!" &#!" '!!"($!" ()" (*" ('" (%" !" %" '" *" 7. The average lateral standard deviations of intersection points for their node centroids (σ r (N)) and flight tracks from their link paths (σ r (L)) are stable with respect to number of flights.These metrics depend far more on the route structure generation parameters than the track data processed.Route structure deviation from airways is also very stable.However, airway deviation from route structure correlates heavily to route !" tracks in section 3 were used to generate great circle tracks and weather rerouted tracks for the same peak four-hour time period (15:30-19:30 local) in ZKC.!#$" !#%" !#&" !#'" (" )(!" )'" )&" )%" )$" !" $" %" &" Representations ρ λ A λ R +Great Circle RoutingThe route structure of great circle flight tracks resulted in significantly lower intersection representation than that of flight plan tracks when using parameters from Table 2.When the cluster filtering parameters were reduced to increase intersection representation to a value similar to the flight plan route structure, the number of links and nodes quadrupled with no increase in average link weight and a decease in average node weight.Figure 8 shows great circle route structures with the original parameters on top and reduced cluster filter parameters on the bottom.In addition to increased numbers of links and nodes, Fig. 8 shows how lowering the cluster filters significantly shortened the average link length.This reaffirms the integrity of the originally chosen parameters.The total number of flight pair intersections was actually quite similar between flight plan and great circle routing.Great circle routing simply spread these intersections out more randomly such that only 21% of the traffic intersections were stable enough to form a structure.Figure 8 also shows how great circle route structure rarely conforms to the airway structure.The largest nodes are in different places than the flight plan route structure and are often close to sector boundaries.If the future airspace system is to support user preferred routing, route structures like these will enable designing airspace to accommodate them. +Weather Avoidance ReroutesUser preferred reroutes around weather were simulated [9] for both original flight plans and great circle routing.The simulated weather were stationary contours of percent likelihood that a flight would avoid the area [10].These weather contours blocked several airways in Southeast ZKC.Figure 9 shows the route structures for the flight plan and great circle tracks avoiding weather.The weather contours are shown underlying the route structure with color indicating the percent likelihood of a flight avoiding the area.Both route structures clearly show links of relatively high weight curving to avoid the weather.In the flight plan based route structure, a new node with weight 10,088 ( w(N) ≈ 100) ap- pears just north of the weather.This is much larger than the largest weight node (w(N) = 6, 885, w(N) ≈ 83) from the non weather impacted route structure in Fig. 2.Although the non weather impacted great circle route structure at the top of Fig. 8 does not show any significant structure in the weather location, weather reroutes sufficiently compressed flight tracks to form the structure around the weather seen at the bottom of Fig. 9. +AnalysisRoute structures were compared flight plan and great circle tracks and between original and weather rerouted tracks.Figure 10 shows precision metrics for each of these four route structures.As mentioned in section 5.1, ρ is significantly lower for great circle routing and numbers and weights of nodes and links follow a similar trend.The pre-filtered sum weight of all I was actually only 10% less for great circle than flight plan routing, indicating a similar overall complexity in terms of conflict likelihood.How-ever, great circle routing spread out these intersections such that most were filtered.Weather reroutes actually caused a slight increase in ρ for great circle because of the track compressing effect of the weather obstacle.Route structures had higher σ r (L) for user preferred routing options; σ r (L) was higher for great circle route structures than for flight plan route structures and was higher for route structures with weather reroutes than for routes structure without weather reroutes.In contrast, σ r (N) was far less affected by routing.σ r (N) was slightly greater for weather reroutes than original routing, following a similar although less pronounced trend as σ r (L).However, an opposite trend to σ r (L) appeared in that σ r (N) was slightly less for great circle than flight plan routing.In addition to metrics compared in pervious sections, metrics relating route structure to sector boundaries were compared.Let n L and n N be the average numbers of unique links and interior nodes, respectively, within each sector.Each additional link or node within a sector potentially increases the focus split of a controller.Let n L X be the average number of links entering or exiting a sector.A lower n L X is preferred to minimize sector handoff workload for controllers.Let n L , n N , and n L X be maximum numbers of links, nodes, and boundary crossing links, respectively, in a single sector. Figure 11 shows these sector relative numbers of links and nodes.The total numbers of links and nodes were less for great circle than flight plan route structure but greater for weather reroutes than original routes.The averages in Fig. 11 echo this result.The increased averages can be attributed to weather reroutes that cause flights to fly through a few extra sectors to avoid the weather.Even though the average number of links per sector was less for great circle than flight plan route structure, the maximum number of links was greater due to the shifting location of links with respect to sector boundaries.The same effect is seen for maximum nodes for weather rerouted great circle tracks.Controllers prefer major flows and critical points to be well within sector boundaries.Controllers must be aware of not only flights within a sector, but also just outside the sector boundary to guarantee separation.Ideally flows should stay at least five nmi inside the sector boundary to avoid magnifying flight awareness workload of neighboring sectors.Controllers require some time to become familiar with a flight entering the sector before it approaches a critical point.The more time they have, the more efficiently they may control the flow safely through the critical point.Therefore, the length of flow segments from the point they enter a sector to when they encounter a critical point should be maximized.Let η L n be the percent of points along link nominal paths that are within five nmi of a sector boundary.Let η N n be the percent of nodes that serve as the end point of a link entering a sector within 10 nmi.Let η L w and η N w be η L n and η N n percentages normalized by link and node weights.Figure 12 shows these percentages of link and node sector boundary proximity.η L n and η L w do not change significantly with weather reroutes but they do increase noticeably from flight plan to great circle routing.Great circle routing affects the location of links to be more random with respect to sector boundaries.η N n decreases from flight plan to great circle routing due to the significant reduction total number of nodes.However, the increase in η N w from flight plan to great circle routing shows how the traffic randomizing affect of great circle routing moved some larger nodes too close to sector boundaries.This effect is amplified in the great circle weather reroutes.!"# $!"# %!"# &!"# '!"# (!"# )!"# FP FP weather GC GC weather r σ (L) r σ (N) r (nmi)!" #" $!" $#" %!" %#" n N n N n L n L n L X n L X +ConclusionThis paper presented a method of defining route structure from a given set of flight tracks.A para-metric analysis identified optimal method parameters to define good weather route structures similar to today's published airway structure.Flight tracks simulated from four different days of flight schedules were tested with similar results compared to the airway structure.On average, the route structures deviated only 1.7 nmi from airways.Route structures were generated for four-hour time windows of track data shifted every two hours to demonstrate how route structure can be updated dynamically.Links and nodes of the route structure appeared and increased in weight as flight traffic increased.When airways began to saturate, more and more links began to appear where airways had not been published, but the precision of the route structure in representing the track data remained stable.On average, the standard deviation of flights tracks was 2 nmi from links and 4.7 nmi from nodes.Route structures of great circle routing differed significantly from flight plan routing.Great circle tracks displayed much less structure with lower traffic intersection representation and link and node weights than flight plan tracks.However, great circle route structure had larger average numbers of links and nodes per sector than flight plans, as well as a higher percentage of the route structure close to sector boundaries.Simulated user preferred reroutes around a stationary weather obstacle produced route structures with greater numbers of link sector boundary crossings and average links and nodes per sector.Weather reroutes of great circle tracks resulted in a significant increase in larger weight nodes close to sector boundaries.These results demonstrate how dynamically generated route structures could be useful in guiding dynamic airspace configuration to accommodate traffic volume changes or weather reroutes.They also suggest that while route structure may help controllers retain situational awareness under dynamically changing traffic conditions, it may not help in traffic with little structure such as with great circle routing.NomenclatureA= airway structure B = boundary C = cluster D = diverge I = intersection L = link M = merge N = node n = number P = path or sequence of points p = point position R = route structure r = lateral radial distance S = airway segment w = weight z = vertical distance δ = structure deviation η = boundary proximity percentage λ = structure representation ρ = intersection representation σ = standard deviation τ = flight track data time range ω = filtering threshold 1 Introduction +Clusters characterize groups of intersections within close spacial proximity.Each intersec-tion may belong to only one cluster in an intersection to cluster (I → C) mapping.The position p(C i ) and weight w(C i ) of cluster C i are the centroid and sum weight of the intersections within C i , respectively.Sets I M , I D , and I B are clustered separately to form sets of merge clusters, C M = {C M1 ,C M2 , ...}, diverge clusters, C D = {C D1 ,C D2 , ...}, and boundary clusters, C B = {C B1 ,C B2 , ...}, respectively.An agglomerative clustering technique was used because it handles data outliers well and automatically determines the final number of clusters.Pairs of intersection points are clustered in order of increasing lateral distance if they are within r C and z C of each other.Each grouped point pair is replaced by the group centroid for consideration in the next clustering iteration.The process repeats until all groups with centroids within r C and z C of each other are clustered.Boundary intersections are different enough from merge and diverge intersections, that unique clustering parameters, r C B and z C B , are used to cluster boundary intersections. +Fig. 11Fig. 1 Airways and their intersections within high altitude Kansas City Center (ZKC). +for 4 /421/2009 were used to create route structures for the peak four-hour periods of three other lowweather Tuesdays spread throughout 2009.Ta-ble 4 shows the evaluation metrics as well as a few additional metrics for all four days.Additional metrics include number of flights (n f ), number of links (n L ), number of nodes (n N ), average link weight (w L ), average node weight (w N ), and the average lateral standard deviation of points within each node from the node centroid (σ r (N)).Metrics are grouped into numbers, average weights, representation ratios, and deviations.The values for each metric are very similar when compared between different days, nor are 4/21/2009 metrics always the best.This verifies that the chosen set of parameters are robust to different days of flight schedules.The same parameters are used to analyze dynamic route structure and route structures for user preferred routing in following sections. +Figure 33shows multiple route structures from 4/21/2009 by shifting the four-hour time window in two-hour increments from the peak four-hours shown in Fig. 2. In some instances, links and nodes move slightly, but for this good weather +Fig. 33Fig. 3 Dynamic route structure shifted every two hours. +Fig. 44Fig. 4 Dynamic route structure shifted every two hours. +Fig. 55Fig. 5 Dynamic route structure shifted every two hours. +Fig. 6 Fig. 767Fig. 6 Dynamic route structure shifted every two hours. +Fig. 88Fig. 8 Great circle route structures for original and reduced cluster filters. +Fig. 99Fig. 9 Route structures for flight plan and great circle tracks avoiding weather. +Fig. 1010Fig. 10 Precision metrics comparing flight plan (FP) to great circle (GC) and original routing to weather reroutes. +Fig. 1111Fig. 11 Average and maximum links and nodes per sector. +Fig. 1212Fig. 12 Percentages of links and nodes close to sector boundaries. +Table 11Summary of route structure evaluationmetricsPrecision Metricsσ r (L) = lateral traffic deviation-ρ= intersection representation +Accuracy Metricsδ R = route structure deviation-δ A = airway structure deviation-λ R = route structure representation+λ A = airway structure representation + +Table 44Evaluation metrics from the peak four-hour periods of four low weather Tuesdays in2009.2/104/217/1411/10n f1,098 1,054 1,059 1,064n L285315284268n N129138121125w L9.469.839.8710.77w N510.80 554.88 545.07 604.38ρ0.520.570.510.56λ A0.730.730.750.72λ R0.600.530.580.59σ r (L) 2.261.792.072.05σ r (N) 4.694.695.074.78δ A4.455.815.474.78δ R1.711.671.731.864 Dynamic Route Structure + + + +Copyright StatementThe authors confirm that they, and/or their company or organization, hold copyright on all of the original material included in this paper.The authors also confirm that they have obtained permission, from the copyright holder of any third party material included in this paper, to publish it as part of their paper.The authors confirm that they give permission, or have obtained permission from the copyright holder of this paper, for the publication and distribution of this paper as part of the ICAS2010 proceedings or as individual off-prints from the proceedings. + + + + + + + Introducing Structural Considerations into Complexity Metrics + + JonathanMHiston + + + RJohnHansman + + + GuillaumeAigoin + + + DanielDelahaye + + + StephanePuechmorel + + 10.2514/atcq.10.2.115 + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 10 + 2 + + 2002 + American Institute of Aeronautics and Astronautics (AIAA) + + + Histon, J.M., Hansman, R.J., Aigoin, G., De- lahaye, D. and Puechmorel, S. (2002) "Intro- ducing structural considerations into complexity metrics," Air Traffic Control Quarterly, Vol. 10, No. 2, pp. 115-130. + + + + + Geometric algorithms for dynamic airspace sectorization + + GRSabhnani + + + May 2009 + NY + + + Stony Brook University + + + Ph.D. thesis + Sabhnani, G. R., "Geometric algorithms for dy- namic airspace sectorization," Ph.D. thesis, Stony Brook University, NY, May 2009. + + + + + Three Dimensional Sector Design with Optimal Number of Sectors + + MinXue + + 10.2514/6.2010-8290 + + + AIAA Guidance, Navigation, and Control Conference + Toronto, Ontario, Canada + + American Institute of Aeronautics and Astronautics + August 2010 + + + Xue, M. "Three dimensional sector design with optimal number of sectors," AIAA Guidance, Navigation, and Control Conference and Exhibit, Toronto, Ontario, Canada, August 2010. + + + + + A Weighted-Graph Approach for Dynamic Airspace Configuration + + StephaneMartinez + + + GanoChatterji + + + DengfengSun + + + AlexandreBayen + + 10.2514/6.2007-6448 + AIAA-2007-6448 + + + AIAA Guidance, Navigation and Control Conference and Exhibit + Hilton Head, South Carolina + + American Institute of Aeronautics and Astronautics + August 2007 + + + Martinez, S., Chatterji, G., Sun, D., Bayen, A. "A weighted-graph approach for dynamic airspace configuration," AIAA Guidance, Navigation, and Control Conference and Exhibit, Hilton Head, South Carolina, August 2007, AIAA-2007-6448. + + + + + A Spectral Clustering Based Algorithm for Dynamic Airspace Configuration + + JinhuaLi + + + TongWang + + + InseokHwang + + + InseokHwang + + 10.2514/6.2009-7056 + AIAA-2009- 7056 + + + 9th AIAA Aviation Technology, Integration, and Operations Conference (ATIO) + Hilton Head, South Carolina + + American Institute of Aeronautics and Astronautics + September 2009 + + + Li, J., Wang, T., and Hwang, I. "A spectral clus- tering based algorithm for dynamic airspace con- figuration," 9th AIAA Aviation Technology, Inte- gration and Operations Conference, Hilton Head, South Carolina, September 2009, AIAA-2009- 7056. + + + + + Network Characteristics of Air Traffic in the Continental United States + + BanavarSridhar + + + KapilSheth + + 10.3182/20080706-5-kr-1001.02216 + + + IFAC Proceedings Volumes + IFAC Proceedings Volumes + 1474-6670 + + 41 + 2 + + 2008 + Elsevier BV + Coex, South Korea + + + Sridhar, B. and Sheth, K. "Network Char- acteristics of Air Traffic in the Continental United States," Proceedings of the 17th Inter- national Federation of Automatic Control World Congress, Coex, South Korea, 2008. + + + + + Algorithmic Traffic Abstraction and its Application to NextGen Generic Airspace + + GirishkumarSabhnani + + + ArashYousefi + + + IrinaKostitsyna + + + JosephMitchell + + + ValentinPolishchuk + + + DavidKierstead + + 10.2514/6.2010-9335 + + + 10th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Fort Worth, Texas + + American Institute of Aeronautics and Astronautics + September 2010 + + + Sabhnani, G., A. Yousefi, D.P. Kierstead, I. Kos- titsyna, J.S.B. Mitchell, and V. Polishchuk, "Al- gorithmic traffic abstraction and its application to NextGen generic airspace," 10th AIAA Aviation Technology, Integration, and Operations Confer- ence, Fort Worth, Texas, September 2010. + + + + + Defining critical points for dynamic airspace configuration + + SZelinski + + + + 26th International Congress of the Aeronautical Sciences, Anchorage + Alaska + + 2008 + + + Zelinski, S. "Defining critical points for dy- namic airspace configuration," 26th International Congress of the Aeronautical Sciences, Anchor- age, Alaska, 2008. + + + + + Analysis of Automated Aircraft Conflict Resolution and Weather Avoidance + + JohnLove + + + WilliamChan + + + ChuLee + + 10.2514/6.2009-6995 + AIAA-2009-6995 + + + 9th AIAA Aviation Technology, Integration, and Operations Conference (ATIO) + Hilton Head, South Carolina + + American Institute of Aeronautics and Astronautics + September 2009 + + + Love, J., Chan, W., and Lee, C.H. "Analysis of Automatetd Aircraft Conflict Resolution and Weather Avoidance," 9th AIAA Aviation Tech- nology, Integration and Operations Conference, Hilton Head, South Carolina, September 2009, AIAA-2009-6995. + + + + + An Exploratory Study of Modeling En Route Pilot Convective Storm Flight Deviation Behavior + + RDelaura + + + JEvans + + + + Proceedings of the 12th Conference on Aviation, Range, and Aerospace Meteorology + the 12th Conference on Aviation, Range, and Aerospace MeteorologyAtlanta, Georgia + + 2006 + + + DeLaura, R., and Evans, J., "An Exploratory Study of Modeling En Route Pilot Convective Storm Flight Deviation Behavior," Proceedings of the 12th Conference on Aviation, Range, and Aerospace Meteorology, Atlanta, Georgia, 2006. + + + + + + diff --git a/file837.txt b/file837.txt new file mode 100644 index 0000000000000000000000000000000000000000..e527d4ebc03a29bfaeff214123405315651c9b68 --- /dev/null +++ b/file837.txt @@ -0,0 +1,198 @@ + + + + +I. IntroductionOver the past several decades, the demand for air travel has increased.The National Airspace System (NAS) has evolved with modest advances in weather/wind prediction, faster and quieter airplanes, and by adding new runways and technologies at airports.Each small step plays a part in the attempt to minimize delay as demand continues to grow.However, current technology is fast approaching the point of diminishing returns.New airspace operational concepts will be necessary to reverse this imbalance.These conceptual changes may affect the entire NAS in unpredictable ways.The Airspace Concept Evaluation System (ACES) is a NAS-wide fast-time simulation developed at NASA Ames Research Center.ACES simulates a model of the NAS with interacting agents for center control, terminal flow management, airports, individual flights, and other NAS elements.These agents pass messages between one another similar to real world communications.This distributed agent based system is designed to emulate the highly unpredictable nature of the NAS, making it a suitable tool to evaluate envisioned airspace concepts.Before new concepts can be evaluated, the tool must be used to evaluate the current NAS situation.This will lead to the development of processes that attempt to better understand the interdependencies between elements making up the NAS.Through these processes, the cascading effects that interdependencies produce on the system can be better understood.As a first step, this study was performed to establish an initial characterization of NAS-wide delay.The goal was to use ACES to simulate a variety of demand and capacity scenarios in the NAS to quantitatively establish their effects on system-wide delay.The study included thirty-six simulations, each encompassing a 24-hour demand period.The simulations varied in flight demand and airport operational capacity.Four flight demand cases were studied which included a representation of current-day 2002, an approximate doubling of current-day 2002, and two intermediate flight demand schedules.All cases included domestic commercial passenger flights operating between the top 98 US airports.Capacity was varied with nine cases representing different airport operational capacities.Two cases represented optimum and worst-case airport states.All airports operated under visual flight rules (VFR) during the simulation period for the optimum case.For the worst case, the 30 benchmark airports 8 operated under instrument flight rules (IFR) during the simulation period.The remaining seven intermediate days represented airport operational conditions from historic weather days.Delays for each segment of flight were determined and mapped against demand and capacity.The resulting surface provides a quantification of the effect that demand and capacity have on NAS delay. +II. ACES EnvironmentACES simulates the NAS using individual models for each of the communicating agents within the system.Examples of agents include System Command Center, Air Route Traffic Control Center (ARTCC), and Terminal Radar Approach Control (TRACON).ACES is different from other NAS simulation tools in that it can simulate and record all communications between agents for post analysis.ACES Build 1.2 is the first working version of ACES.This build emphasizes the assembly of an agentbased framework with lower emphasis on developing high-fidelity agent models.Models or functionality which are absent in this current version of ACES include: 1) sector capacity limits, 2) separation constraints, 3) flight plan rerouting, 4) delay in the arrival terminal area and arrival surface, and 5) en-route altitude and cruise speed changes.Additional and enhanced models and improved functionality in these areas are under current development.Two options that are available in ACES Build 1.2 are delay maneuvers and TRACON departure fix separation.Delay maneuvers are lateral en route course alterations used to delay individual flights.The TRACON departure fix separation option provides a simulation of miles-in-trail separation of aircraft at each departure fix. +III. Demand VariationDemand is defined as the number of flights flown in the simulation.It is primarily dependant on the flight data set, which is a list of scheduled flight plans for the simulation.Each flight plan includes departure and arrival airports, cruise altitude and speed, a planned trajectory specified as an array of latitude/longitude coordinates, and scheduled departure time.Each flight is individually flown unconstrained between origin and destination airports during ACES configuration.This establishes the scheduled take-off, landing and gate arrival times.Although no flights are cancelled, not all flight plans are loaded and flown in the simulation.ACES rejects and filters out flights during configuration.Reasons why ACES rejects flights include: 1) when the origin and destination airports are the same, 2) sector boundary definition errors resulting in flights passing through undefined airspace, 3) inadequate number of way points in planned trajectories, and 4) distance between consecutive way points exceeds a pre-defined value.Two flight data sets were readily available for this characterization study.The first was a current-daydemand constructed from recorded traffic for May 17, 2002. 1 The second data approximately doubled the current-day traffic.This data set was developed as a future demand representing a nominal day in the year 2020. 2 The two data sets included domestic commercial passenger flights between pairs of airports within the continental United States' top ninety-eight airports.At run-time, ACES loaded 16,468 flights for the current-day demand set, and 33,186 flights for the future demand set.Intermediate flight data sets were created by filtering flights from the future data set while maintaining similar flight distributions.Let T(A, Q) and L(A, Q) be matrices containing the number of flights scheduled to takeoff and land at airport A at quarter hour Q from the start of the simulation.Let α be the reduction ratio desired.Then T r (A, Q) = αT(A, Q) and L r (A, Q) = αL(A, Q) become the matrices of the number of flights that must be removed per airport per quarter hour to reduce the data set by α.For each flight, let A t and A l be the takeoff and arrival airport.Let Q t and Q l be the scheduled takeoff and landing quarter hours from the start of the simulation.The flight filtering algorithm is as follows.For each loaded flight in the future demand data set:If (T r (A t , Q t ) > 0.5) and (L r (A l , Q l ) > 0.5) :Remove flight and decrement T r (A t , Q t ) and L r (A l , Q l ) by 1 Else:Add flight to new data set.Two intermediate data sets were reduced from the future demand data set with α = 1 6 and α = 1 3 .At run-time ACES loaded 25,597 and 20,461 flights respectively from these intermediate data sets.These data sets combined with the current demand and future demand provided 4 equally distributed variations of demand with similar flight distributions.Figure 1 shows the flight distributions for San Francisco International Airport (SFO) as a running average of scheduled takeoffs and landings per quarter hour starting at midnight Greenwich Mean Time (GMT).It is apparent from the figure that the defined filtering algorithm produced two intermediate demand data sets with similar distributions to that of the future demand set.These three distributions also compare favorably with the current-day demand distribution. +IV. Capacity VariationCapacity has been defined in many ways.En-route capacity has been defined as a function of the total number of aircraft under track control in a sector, average time of a flight in a sector, and ratio of total altitude changes to number of aircraft. 6Airport capacity has been defined as airport operations (arrivals, departures, or total) divided by throughput 7 or most commonly, as airport operations per unit time. 4,5 n all cases, capacity is defined for a single sector or airport.For the purpose of this analysis, a single capacity value encompassing the entire NAS is needed.This single capacity value must be independent of any one airport or region and independent of total demand or delay.The most appropriate NAS-wide capacity calulation that could be made using ACES Build 1.2 is a combined maximum airport capacity defined in terms of operations per quarter hour.ACES models each airport's state per quarter hour of run-time as either Visual Flight Rules (VFR) or Instrument Flight Rules (IFR).This operating state in turn defines the airport's capacity as the maximum number of departures, arrivals and total operations.A simple average of all airport capacities throughout the day would neglect the fact that demand is placed more heavily on some airports than others.Therefore, NAS capacity was calculated as a weighted average of airport capacities throughout the day across the scheduled demand distribution, and normalized by total demand D. The resulting NAS capacity is dependant on the demand distribution and not the total demand itself.NAS capacity units remain the number of operations per quarter hour just as with individual airport capacities.Let the matrices C d (A, Q), C a (A, Q), and C t (A, Q) contain the maximum airport capacities for departures, arrivals, and total operations for airport A at quarter hour Q from the start of the simulation.The respective NAS capacity measurements are given asC N ASa = 1 D L(A, Q) • C a (A, Q) (1) C N ASd = 1 D T(A, Q) • C d (A, Q) (2) C N ASt = 1 2D (L(A, Q) + T(A, Q)) • C t (A, Q) (3)where T(A, Q) and L(A, Q) are the takeoff and landing distribution matrices defined in section III.The variation of each airport's state throughout the day can be used to emulate the effects of terminal area weather.These variations affect airport capacity and therefore NAS capacity.Higher NAS capacity values result when the majority of airports are operating in VFR conditions during most of the day.Lower NAS capacity values result when IFR conditions occur more often.The greatest NAS capacity for a given demand results when all airports are in VFR conditions all day.A worst case low capacity day was defined as a day when the top 30 continental airports are in IFR conditions all day.Intermediate NAS capacities were calculated using airport state conditions taken from Aircraft System Performance Metrics (ASPM) analysis weather reports for various historical days.A report by Metron Aviation provided convenient access to this data for 7 different days. 3Three of these days were reported as low weather days.The remaining four were reported as days where weather affected delay.Since the simulations are based on the top 98 continental airports, and ASPM reports provide airport state data for only the top 49, the remaining airports were always assumed to be in a VFR state. +V. Delay CalculationsACES can collect all messages between its various agents, but only aircraft state messages were necessary for this analysis.The aircraft state message for each flight includes the origin and destination airports, and the scheduled and actual simulation times for gate departure, takeoff, landing, and gate arrival.For each flight, let t d , t t , t l , and t a define the actual gate departure, takeoff, landing, and gate arrival times.Similarly, let t ds , t ts , t ls , and t as define the scheduled times.Delay can be attributed to departure delay d d , takeoff surface delay d t , enroute delay d l , or landing surface delay d a .The sum of these is total delay d total .The following are the calculations for each delay element.Figure 2 shows a graphical representation of how delay is attributed to the various stages of flight.As stated earlier, arrival surface delay d a is not currently modelled and will be zero for every flight.d d = t d -t ds (4) d t = (t t -t ts ) -d d (5) d l = (t l -t ls ) -d d -d t (6) d a = (t a -t as ) -d d -d t -d l (7) d total = d d + d t + d l + d a (8) +VI. ResultsA total of 36 simulations were run across the 4 demand sets and 9 weather days (see Figure 3(a)).The incurred departure delay, as a percentage of total delay, increases significantly from approximately 0% to 30% as demand increases and capacity decreases.As a result of the departure delay increase, takeoff surface delay decreases from 70% to 45% and enroute delay decreases from 36% to 21%.As stated earlier, ACES Build 1.2 does not currently cancel flights.Every flight must complete its flight plan regardless of the delay incurred due to congestion.When demand exceeds capacity, flights are held at their gates, and departure delay increases more rapidly than either surface or enroute delay.Figure 4(a) shows a graph of demand and total NAS capacity for each weather day.The lines represent the average capacity for each weather day.The data points show the deviation of each capacity from the average.The averaged standard deviation of the NAS capacity values for all demand sets is 0.34 total operations per quarter hour.The averaged standard deviation of the NAS capacities for the three larger demand sets with similar flight distribution is 0.18 total operations per quarter hour.The averaged standard deviation is higher when the current demand data set is included because the NAS capacity calculation is dependent on the flight demand distribution.However, a clear distinction in NAS capacity can still be seen between the weather days.It is only for the 3 low weather days that current demand NAS capacity deviates enough to cross neighboring capacity days.When demand is compared to delay, quadratic curves produce the lowest and most random residuals; the values having a maximum of 0.32 minutes and an average of 0.05 minutes.Figure 4(b) shows what the delay surface looks like when demand data are interpolated using quadratic curves.Figure 5 shows comparisons of delay to capacity for each demand set.The data show a general linear However, as demand increases, nonlinearities in the data grow larger.This suggests that other factors are influencing the results.In general the days with higher delay than the linear trend have a majority of east coast airports in IFR state.The days with lower delay have very few east coast airports in IFR state.Therefore, regional concentrations of weather located in high traffic areas may be one source of these nonlinearities. +VII. ConclusionThis delay characterization study provides an initial quantification of the significant rise in delay experienced in the current NAS due to increasing demand.The results make a strong case for developing new airspace concepts that provide increased system capacity, and for continuing development of ACES, a NAS-wide simulation system, to assess their performance.Immediate future steps include a comparison of simulated ACES results to real world data, and an exploration of the effects of regional weather on the performance of the entire NAS.As ACES functionality increases, there we be limitless possibilities for exploring the interdependent and cascading effects that each element of the NAS has on the system as a whole.Figure 1 .1Figure 1.Takeoff and landing flight distributions at SFO for each demand data set. +Figure 2 .2Figure 2. Graphical representation of delay per stage of flight. +Delay results for 36 runs across 4 demand sets and 9 weather days. +Delay results are mapped to their respective weather days. +Figure 3 .3Figure 3. Delay surface results. +Figure 3 (3Figure 3(b) shows how the delay results correspond to their respective weather days as described in section IV.The all-VFR weather day provided the greatest NAS capacity and had the least delay.The top-30-IFR weather day resulted in the greatest delay and least capacity.The intermediate weather days, February 11, 2001, June 20, 2000, and July 12, 2001 were described as being low weather days.3They contributed little to delay in simulation.The data shows high NAS capacities and corresponding low NAS delays for these days.The remaining 4 intermediate weather days were characterized as days where weather contributed to delay.The data shows increasing delay with decreasing capacity, and more variability in the results for these days. +Demand vs. NAS Capacity for each weather day. +NA S ca pa cit y average delay (min) (b) Delay surface as a quadratic extrapolation with respect to demand. +Figure 4 .4Figure 4. Demand Analysis Results +Figure 5 .5Figure 5. Delay vs. Capacity + + + + + + + + + KTG: A Fast-Time Kinematic Trajectory Generator for Modeling and Simulation of ATM Automation Concepts and NAS-wide System Level Analysis + + YingchuanZhang + + + GoutamSatapathy + + + VikramManikonda + + + NikhilNigam + + 10.2514/6.2010-8365 + NASA contract NAS2-00015 + + + AIAA Modeling and Simulation Technologies Conference + + American Institute of Aeronautics and Astronautics + February 2003 + + + Creation of a Modeling and Simulation Capability to Support NAS-Wide Analysis of Advances ATM tools and Concepts. CTOD 7.19 Software Demonstration and Validation Report. NASA contract NAS2-00015. February 2003. + + + + + Define Scenarios for Transportation Demand and Economic Analysis. NASA contract + + Lmi + + + December 2003 + + + LMI, Define Scenarios for Transportation Demand and Economic Analysis. NASA contract. December 2003. + + + + + Aggregate Statistics of the National Airspace System + + JimmyKrozel + + + BobHoffman + + + StevePenny + + + TarynButler + + 10.2514/6.2003-5710 + + + AIAA Guidance, Navigation, and Control Conference and Exhibit + + American Institute of Aeronautics and Astronautics + October 2002 + + + J. Krozel, B. Hoffman, S. Penny, and T. Butler. Selection of Datasets for NAS-Wide Simulation Validations. NASA contract RTO73. October 2002. + + + + + National Airspace System demand and capacity modeling + + IFrolow + + + JHSinnott + + 10.1109/5.47724 + + + Proceedings of the IEEE + Proc. IEEE + 0018-9219 + + 77 + 11 + + November 1989 + Institute of Electrical and Electronics Engineers (IEEE) + + + I. Frolow and J. Sinnott, National Airspace System Demand and Capacity Modeling. In Proceedings of the IEEE, Vol. 77, No. 11, November 1989. + + + + + The Aviation System Analysis Capability Airport Capacity and Delay Models + + DLee + + + CNelson + + + GShapiro + + + + Logistics Management Institute + + April 1998 + McLean, Virginia + + + D. Lee, C. Nelson, and G. Shapiro. The Aviation System Analysis Capability Airport Capacity and Delay Models. Logistics Management Institute, McLean, Virginia. April 1998. + + + + + ARTS IIIA Terminal Baseline Research + + RMogford + + + KAllendorfer + + + JGalushka + + Report. DOT/FAA/CT-TN99/7 + + April 1999 + + + R. Mogford, K. Allendorfer, and J. Galushka. ARTS IIIA Terminal Baseline Research Report. DOT/FAA/CT-TN99/7. April 1999. + + + + + FAA Report on Airport Capacity; Volume I — National Summary. Office of Systems Engineering Management, Federal Aviation Administration, U.S. Department of Transportation, Washington, D.C. 20590. January 1974. Various paging + + CConner + + + DRutishauser + + 10.1177/004728757401300244 + + + + Journal of Travel Research + Journal of Travel Research + 0047-2875 + 1552-6763 + + 13 + 2 + + August 2001. 2001. July 2004 + SAGE Publications + Hampton, Virginia + + + Airport Capacity Benchmark Report + Department Of Transportation and Federal Aviation Administration + C. O'Conner and D. Rutishauser. Enhanced Airport Capacity Through Safe, Dynamic Reductions in Aircraft Separation: NASA's Airctaft VOrtex Spacing System (AVOSS). Langely Research Center, Hampton, Virginia. August 2001. 8 Department Of Transportation and Federal Aviation Administration. Airport Capacity Benchmark Report 2001. http://www.faa.gov/events/benchmarks/. July 2004. + + + + + + diff --git a/file838.txt b/file838.txt new file mode 100644 index 0000000000000000000000000000000000000000..ba489e604a5f28324196a6153a36b680c67b1c22 --- /dev/null +++ b/file838.txt @@ -0,0 +1,458 @@ + + + + +INTRODUCTIONNASA is committed to demonstrating a concept of integrated arrival, departure, and surface operations by 2020 under the Airspace Technology Demonstration 2 (ATD-2) sub-project [1].This will be accomplished in three phases, starting with a demonstration of flight specific time-based surface metering at Charlotte Douglass International Airport (CLT) [2].ATD-2 tactical metering capability is based on NASA's Spot And Runway Departure Advisor (SARDA) which has been tested successfully in human-in-the-loop simulations of CLT [3].SARDA makes use of surface surveillance data and surface modeling to estimate the earliest takeoff time for each flight active on the airport surface or ready for pushback from the gate.The system then schedules each flight to its assigned runway and assigns a target pushback time displayed to ramp controllers as an advisory gate hold time.The objective of this method of surface metering is to move as much delay as possible to the gate to minimize surface congestion and engine on-time while keeping sufficient pressure on the runway to maintain throughput.This flight specific approach enables greater flight efficiency and predictability, facilitating trajectory-based operations, which ATD-2 aims to achieve.Throughout ATD-2 project formulation and system development, researchers have continuously engaged with stakeholders and future users, uncovering key system requirements for tactical metering that SARDA did not address.These include:• Pushback time advisories do not change after the pilot calls in ready for pushback so that controllers may communicate a single gate hold to the pilot at this time.• Make use of and incentivize improved accuracy of air carrier-provided estimates of pushback ready time.• Facilitate integration with strategic scheduling consistent with the Surface Collaborative Decision Making Concept of Operations [4].This paper presents a benchmark fast-time simulation evaluation of the tactical scheduler designed to meet these new metering requirements.The scheduler is used to calculate gate holds for flights.Flight operations with the gate holds are simulated using a fast-time medium-fidelity simulation of CLT.A key parameter used in the generation of gate holds, the taxi time delay buffer, is analyzed to determine the value that moves the most delay from the runway queue to the gate without drying out runway operations.Fast-time simulation is used to rapidly develop and test new scheduler design features and future fasttime simulation studies will build upon this benchmark evaluation to explore other design considerations.The rest of this paper is organized as follows.Section II presents a background of surface metering research.Then the ATD-2 tactical scheduler used for surface metering is presented in section III.Simulation details and results from a benchmark evaluation of the ATD-2 tactical scheduler are presented in sections IV and V, respectively.Finally, conclusions and future work are presented in section VI. +II. BACKGROUNDPrior demonstrations of automation aided surface metering in the U.S. have focused on strategic approaches to enable Collaborative Decision Making (CDM) between flight operators, Air Navigation Service Providers, and other stakeholders in surface and departure operations as defined in the Surface-CDM Concept of Operations [4].Both the Collaborative Departure Queue Management (CDQM) approach evaluated at Memphis International Airport [5] and the Ground Management Program (GMP) at John F. Kennedy International Airport [6,7] used ration by schedule algorithms to allocate Active Movement Area (AMA) entry slots to different carriers to manage the length of departure runway queues below a threshold value.The methods differed in their mechanism to enable carrier flexibility.CDQM abstracts the individual flight based slots by allocating to carriers a specific number of aircraft for each that may enter the AMA within 10-minute intervals.The carrier then has the flexibility to decide which of its flights to fill each allocated slot.On the other hand, the GMP assigns each slot to a specific aircraft.Air carriers are still given the flexibility to swap flights according to their priorities and manage their own ramp areas, but this flight specific method also allows the system to monitor compliance with AMA entry times to ensure that everyone is treated fairly.The most tactical form of surface metering field demonstrated in the U.S. is a test of pushback rate control at Boston Logan International Airport [8,9].This method is not flight specific, suggesting the number of aircraft that should be allowed to push back in the next 15 minutes to maintain a count of active aircraft predetermined to place sufficient pressure on the runways without over-congesting the surface.Flights are released on a first-comefirst-served basis as they call in ready for pushback.Departure MANager (DMAN) [10] is a time-based tactical surface metering system operational in Europe.A demonstration of DMAN implemented at Athens Airport provided published results [11].DMAN calculates a runway time and corresponding start-up (gate pushback in U.S.) time for each flight to maintain departure queues of two or three aircraft at the runways.Conceptually, DMAN is intended to collaborate with surface and arrival management counterparts (AMAN and SMAN) to establish arrival and departure sequences and optimize surface movement plans [12].The trials at Athens Airport implemented DMAN alone, using flight clearances as events to update calculations.Thus, as flights were issued start-up clearances, the start-up advisories for all flights still at the gate were updated.SARDA also employs time-based tactical surface metering, but has thus far been demonstrated in human-in-the-loop simulation [3], not in the field.Whereas DMAN scheduler updates in the Athens demonstration were event based, the SARDA scheduler is updated every 10 seconds using real-time surface surveillance data.These updates ensure the most accurate information is used to predict runway usage.However, rapid updates also open the potential for fluctuating advisories, which Ramp controllers at a busy airport like CLT find challenging.Therefore, a new requirement for ATD-2 tactical metering is that all advisories must freeze as soon as the pilot calls in ready for pushback so that Ramp controllers may communicate a single hold time when responding to pilot ready calls.ATD-2 aims to bridge the gap between strategic CDM and tactical metering by using air carrier provided estimates of pushback or start-up time called Earliest Off Block Time (EOBT).SARDA developers found that historically available EOBT updates based on air carrier "L-times" were too inaccurate for tactical metering.Therefore, SARDA used EOBTs only to define the group of flights within the tactical planning horizon (within 10 to 15 minutes of EOBT).Flights within the tactical time horizon that had not yet called in ready were scheduled opportunistically so that hold advisories would be available for each in case it was the next to become ready.In anticipation of improved EOBT accuracy from American Airlines, the hub operator at CLT, the ATD-2 tactical scheduler will schedule flights within the tactical time horizon that have not yet called in ready for pushback more realistically, and incorporate S-CDM principles into the scheduling prioritization scheme to prevent EOBT gaming and stabilize tactical scheduler advisories in the presence of updating EOBTs. +III. TACTICAL SCHEDULERThe tactical scheduler is supported by three basic functions: trajectory prediction, runway scheduling, and advisory generation.These functions may be performed differently according to the runway usage prediction accuracy of flights in different states of surface operation.The most basic set of flight groups in order of descending runway usage predictability are• Landing Arrival -arrivals predicted to land on the scheduled runway or any other runway that imposes spacing constraints on the scheduled runway.• Taxi Arrival -arrivals predicted to taxi across the scheduled runway.• Taxi Departure -departures actively taxiing on the airport surface (already pushed back from the gate) predicted to take off from the scheduled runway.• Gate Departure Ready -departures for which the pilot has called in ready for pushback but the flight is being held at the gate.• Gate Departure Planned -departures at the gate with EOBT within the tactical scheduling horizon but for which the pilot has not yet called in.• Gate Departure Uncertain -departures at the gate with EOBT outside the tactical scheduling horizon and for which the pilot has not yet called in ready.Other groups may be included to distinguish flights with emergency or exempt status or departures affected by traffic management initiatives, but these cases are not analyzed in this paper.Flights may jump from one group to another between scheduler calls as their state of surface operation changes. +A. Trajectory PredictionTrajectory prediction uses flight state and intent information to generate an Earliest Runway Usage Time (ERUT) for all flights predicted to use each scheduled runway.Four types of runway use are predicted: arrivals landing on the scheduled runway, arrivals landing on other runways which impose spacing constraints on the scheduled runway (e.g., intersecting or converging runways), arrivals taxiing across the scheduled runway on the way to their arrival gates, and departures assigned to takeoff from the scheduled runway.Trajectory prediction calculates a flight's ERUT differently depending on its flight group.Table I summarizes how ERUTs are calculated for each flight group.These calculations are explained in more detail below.In the field, Time-Based Flow Management (TBFM) [13], the currently operational tool responsible for arrival metering, will calculate Scheduled Times of Arrival (STAs) for arrivals and send them to the tactical scheduler.Therefore, the Landing Arrival group, consisting of the first two runway usage types, uses landing STAs as ERUTs.For all other groups, nominal taxi paths between gates and runways and nominal speeds for various aircraft types in different areas of the airport surface are used to calculate an Unimpeded Taxi Time (UTT) between current position and the scheduled runway for each flight.For the Taxi Arrival and Taxi Departure groups, trajectory prediction adds UTT to current time to get each flight's ERUT.If a Taxi Arrival has not yet landed, UTT is added to its landing STA for its landing runway to get the ERUT for its taxi crossing runway.The advisory time at which to release a flight for pushback, known as the Target Off Block Time (TOBT), is frozen once a flight calls in ready.Therefore, trajectory prediction for Gate Departure Ready flights adds UTT to the flight's frozen TOBT to get ERUT.For Gate Departure Planned group flights, trajectory prediction adds UTT to the EOBT to get ERUT.Because Gate Departure Uncertain group EOBTs are not expected to be very accurate, trajectory prediction adds UTT to current time to get ERUT just in case one of these flights calls in ready and jumps straight to the Gate Departure Ready group. +B. Runway SchedulingAll flights predicted to use the runway are scheduled one at a time in an order that depends on their flight group.Flight groups are scheduled in priority order of decreasing runway usage predictability listed in the previous section.All flights within a single flight group are scheduled before moving on to the next flight group.Within each flight group, flights are scheduled in the order of their ERUTs.The only exception is the Gate Departure Planned flight group, which is ordered by the airline posted Scheduled Off Block Time (SOBT) in compliance with S-CDM ration-by-schedule convention.Using SOBT rather than EOBT to order the flights for scheduling prevents EOBT gaming and stabilizes tactical scheduler advisories in the presence of updating EOBTs.Time-based spacing requirements between each leaderfollower pair of runway use types and aircraft weight categories are adapted to a particular airport and used as constraints for runway scheduling.Only wake separation requirements between departures on the same runway depend on aircraft weight categories.Distance-based rules were converted to time-based separations between departures and arrivals on the same and interdependent runways, and flights taxiing across runways.ATD-2 also considers other constraints from traffic management initiatives such as Miles-in-Trail (MIT), Expect Departure Clearance Time (EDCT), and Approval Request/Call for Release (APREQ/CFR).However, these are not included in this study.Within a single scheduling call, each scheduled flight creates spacing constraints for all flights scheduled after it.If a flight's ERUT does not meet all constraints, the flight is delayed and rechecked for constraint violations.The first time meeting all constraints is assigned to the flight as its Target Runway Usage Time (TRUT) and is used to define spacing constraints for subsequently scheduled flights.The Gate Departure Uncertain group is the only group that is scheduled opportunistically, meaning that these flights are inserted into runway slots when available and do not create constraints for subsequent flights when scheduled. +C. Advisory GenerationWhen all flights have been scheduled and assigned a TRUT, TOBTs are calculated for all departures still at the gate by subtracting the gate-to-runway UTT times a constant, A, and delay buffer, B, from the TRUT.TOBT is calculated asTOBT = TRUT -A(UTT) -B,(1)where A ≥ 1 and B ≥ 0. Factor A accounts for congestion delay the flight may encounter along it's taxi route and assumes this delay is greater the more unimpeded time the flight has left to travel.Factor B adds queue delay where flights line up from the end of the runway.The resulting TOBTs are presented to ramp controllers as advisory hold times for metering. +D. Scheduler DynamicsThe tactical scheduler is called every 10 seconds updating TRUTs and TOBTs.Flights move from the Gate Departure Uncertain group to the Gate Departure Planned group when their EOBTs move within the tactical planning horizon.Flights that become ready between scheduler calls are moved from the Gate Departure Uncertain or Gate Departure Planned group to the Gate Departure Ready group and their TOBTs are frozen at the last TOBT update given when they were in their previous group.This ensures that a static hold advisory is available for a flight as soon as the pilot calls in ready.Because the Gate Departure Planned group flights create constraints for subsequently schedule flights, they can't jump in front of each other by calling in earlier than their EOBTs.Because the Gate Departure Uncertain group flight do not create constraints, these flights can jump in front of each other by calling in early.But they can't jump in front of Gate Departure Planned flights (because those flights are scheduled earlier and create constraints) unless there is a sufficient gap in the schedule due to low demand. +IV. SIMULATIONFast-time simulations were conducted to evaluate tactical metering performance for varying scheduler parameters. +A. Simulation EnvironmentThe Surface Operations Simulator and Scheduler (SOSS) [14] is used to simulate surface operations at CLT. SOSS simulates both departure and arrival aircraft movements within a network of nodes and links representing the airport surface.Fig. 1 shows a map view of the SOSS CLT surface model identifying the different types of nodes and links.In South flow, runways 18R and 23 are used only for arrivals.Runways 18C and 18L are used predominantly for departures with only occasional use for arrivals to offload 18R.Departure routes begin at a gate node, transition through ramp, spot, taxiway, and departure queue nodes, and takeoff from departure nodes.Arrival routes begin at an arrival node, transition through taxiways, a spot, and ramp nodes, ending at a gate.Some arrivals may need to taxi across a runway via crossing nodes, such as 18R arrivals crossing 18C.If a gate conflict between and arrival and departure is predicted, the arrival will be diverted and held at a hardstand node until the departure vacates the gate [15].For the purposes of this study, only arrivals are sent to hardstands to avoid gate conflict induced grid-lock without impacting departure scheduling.Arrival landing STAs and subsequent entry into the surface model are predetermined by the traffic scenario.Whereas departure SOBTs are predetermined by the traffic scenario, actual gate pushback may be controlled by scheduler TOBT assignments.Departures with assigned TOBTs are held at the gate until current time is greater than or equal to TOBT.Otherwise, departures are released as soon as they are ready.For this evaluation, EOBTs remain static and no ready time uncertainty was modelled.Therefore, all flight ready times are equal to their EOBTs which are equal to their SOBTs as specified in the traffic scenario.Aircraft are allowed to progress along predefined routes through the node-link network as long as they do not violate separation constraints specific to each part of the network.To avoid violating separation constraints, flights may stop at taxiway intersections and form queue lines at departure runways, resulting in taxi-time uncertainties due to surface congestion.Although SOSS has the capability of modeling uncertainty in various areas of the airport surface, the only uncertainties simulated were surface congestion related taxi-time uncertainties.This minimum level of uncertainty modeling was desired to perform a benchmark fast-time simulation evaluation of the ATD-2 tactical scheduler.The simulation time step was set to 0.5 sec and the scheduler was called every 10 seconds. +Figure 1. CLT Surface Model +B. Traffic ScenarioAs a hub airport for American Airlines, CLT traffic is characterized by tightly spaced departure and arrival banks.A four-hour traffic scenario was generated using CLT surface surveillance data from March 11, 2016, covering two such banks when CLT was operating in South flow.Fig. 2 shows the arrival and departure demand on each runway in 15-minute bins. +C. Scheduler ParametersScheduler parameters include the planning horizon and taxi time delay buffer parameters.The start of the planning horizon was set to 10 minutes prior to EOBT. +Delay buffer parameters include values A and B from (1).Value A was set to 1.05 to account for a minor amount (5%) of congestion related delay.Value B is the key parameter which determines how much delay may be absorbed on the surface, passing any excess delay to the gate.To analyze how the amount of delay pushed back to the gate affects scheduler performance, multiple simulations were completed, each with a different B value.Value B was increased from 0 in 1-minute intervals until the simulation resulted in no delay passed to the gate as would be the case with no surface metering. +D. Evaluation MetricsThe goal of surface metering is to minimize surface delay and move as much delay as possible from active taxi in the ramp of AMA to the gate with engines off, without negatively impacting runway throughput.To that end, surface metering tries to regulate the flow of departures to the runway to meet the predicted runway departure rate and maintain a queue of flights at the end of the runway just long enough to account for uncertainty.The following delay, runway usage time, throughput, and departure queue metrics are designed to evaluate the performance of timebased surface metering utilizing tactical scheduling in the presence of taxi time uncertainties. +1) DelayTotal departure delay d dep is measured as the difference between a departure flight's ERUT and Actual Runway Usage Time (ARUT).Surface metering attempts to move a portion of the delay from taxiing (in the ramp or AMA) to the gate.Thus, total delay is segregated into gate delay d gate , ramp delay d ramp , and AMA delay d AMA .Gate delay is measured as the difference between a flight's EOBT and Actual Off Block Time (AOBT).The transition from ramp to AMA occurs at the Movement Area entry Time (MAT).Ramp delay is obtained by subtracting gate delay from the difference between a flight's earliest and actual MAT (AMAT -EMAT).AMA delay is then obtained by subtracting gate and ramp delay from total departure delay.d dep = ARUT -ERUT (2)d gate = AOBT -EOBT (3)d ramp = AMAT -EMAT -d gate (4)d AMA = d dep -d gate -d ramp (5)Total arrival delay d arr is the difference between an arrival flight's Earliest Gate Arrival Time (EGAT) and Actual Gate Arrival Time (AGAT), where EGAT is calculated by adding UTT between runway and gate to arrival STA.In this paper, because no arrival time uncertainty was modeled, arrival STA is the same as Actual Time of Arrival (ATA).Total arrival delay is segregated into AMA delay d AMA and ramp delay d ramp .The transition from AMA to ramp occurs at the Ramp Entry Time (RET).AMA delay for arrivals is the difference between a flight's earliest and actual RET (ARET -ERET).Ramp delay for arrivals is then obtained by subtracting AMA delay from total arrival delay.EGAT = STA + UTT ( 6)d arr = AGAT -EGAT(7)d AMA = ARET -ERET(8)d ramp = d arr -d AMA (9) +2) Ruway Usage Time PredictionBecause TOBT is frozen at flight ready time, surface metering is dependent on runway scheduler predictions of TRUT made at flight ready time.Let a flight's predicted runway usage time error be e = ARUT -TRUT(t ready ), (10) where t ready is the flight ready time. +3) Throughput PredictionDue to taxi time uncertainties and lack of control beyond gate push back, the tactical scheduler is unlikely to predict the exact sequence of flight operations on the runway, which will negatively impact e. Tactical scheduler performance is evaluated in a less sequence sensitive way by calculating runway departure throughput prediction errors.Let t be current time and t be a future time for which departure throughput for a given runway is predicted.Let runway departure throughput rate R(t,t) be the number of departures scheduled to use the runway between time t-15 minutes and t, calculated at time t.R(t,t) is the actual departure rate calculated at time t, when all flights included in the rate have already departed.R(t + 15,t) is the earliest rate that is based entirely on predictions or flights that have not yet departed.Let prediction error beE(t,t) = R(t,t) -R(t,t). (11)At CLT, nominal taxi times can take up to 25 minutes.Allowing for some gate hold time, prediction errors for t up to ~35 minutes greater than t may affect surface metering performance.The most relevant throughput predictions to surface metering are analyzed by computing E(ARUT, t ready ) for each flight. +4) Departure QueueAnother metric for tactical scheduler evaluation is departure queue size.Queues are desired to be shorter to minimize congestion related delay.However, they should be long enough to account for taxi time uncertainty and keep the runways from going dry.Four queue values per runway are used: ramp count q ramp , AMA count q AMA , taxi count q taxi , and the queue line q line .The first three queues are simple counts of departures taxiing in the ramp, AMA, or either (ramp or AMA) bound for each runway.The queue line is the number of aircraft lined up from the end of the runway waiting to depart.This is calculated in post processing by identifying the uninterrupted cascading chain of departures scheduled to use the runway that are within close proximity (200 meters) of one another extending out from the runway entrance node.The amount of delay buffer used to calculate TOBTs from ERUTs directly influences the average queue size.Good departure throughput predictability should enable surface metering to provide more queue stability, which should enable smaller queues to maintain maximum runway utilization. +V. RESULTSSimulations of increasing B from (1) were performed until d gate was zero for all departures at a B value of 14 minutes.At a B value of 10, d gate was zero for all departures using runway 18C.Therefore, results focus on B values between 0 and 10.Although extra arrival delay is expected due to additional gate conflicts as arrivals wait longer for departures held at their gates, much of this delay was absorbed as extra taxi path distance for arrivals using hardstands, which is not captured in the presented arrival delay calculations.Whereas departures absorbed most of their active taxi delay in the AMA, arrivals absorbed most of their delay in the ramp.In general, the arrival taxi-in routes are much longer than departure taxi-our routes.Also, the areas where arrivals tended to stack up waiting for use of common taxi route segments happened to be in the ramp area. +A. Delay +B. Runway Usage Time Prediction ErrorFig. 5 shows the average and standard deviation of runway usage time prediction error, e, across all departures at their ready times for each simulation of varying B value.On average, departures use the runway later than predicted as indicated by the positive average e for each simulation.This bias may be because ERUTs used to derive TRUTs are based on unimpeded transit time.This bias could be reduced by subtracting some expected congestion delay from the ERUT calculations shown in Table I.The average e increases with B and begins to level off as B exceeds 6 minutes.Despite releasing departures earlier from the gates (higher B values), these departures encounter so much additional congestion delay as to cause their actual use of the runway to be even later.Standard deviation of e also increases with B. This indicates that it is more difficult to predict the departure sequence in the presence of additional congestion caused by releasing departures from their gates earlier. +C. Throughput Prediction ErrorFig. 6 shows the average and standard deviation error in throughput prediction made at each departure's ready time E(ARUT, t ready ), referred to as simply E below.For B < 2, runway throughput is over predicted (E is negative) on average.For these simulations, departures were not released from the gate early enough to provide sufficient pressure on the runways allowing instances of excess separation that decreased actual throughput.Like e, the average E increases with B and begins to level off as B exceeds 6 minutes.Unlike e, the standard deviation of E is similar for all values of B as this metric is less sensitive to departure sequence. +D. Departure QueueDeparture queues were analyzed by calculating the maximum and average values separately for each runway and two hours of simulation time capturing a single demand cycle.Results are presented for only the first two hours of simulation time encompassing the larger of the two demand cycles as can be seen in Fig. 2. Figs.7 and8 show maximum and average queue sizes for 18L and 18C, respectively.For both runways, the q AMA results are similar to the q line results.The average q AMA is always slightly higher than the average q line because most of the line is contained within the AMA.At B = 10, maximum q line exceeds maximum q AMA for 18L when the line extends out into the ramp.Fig. 9 shows how the last departure counted as part of q line is in the ramp (blue nodes and links).For 18L, the average and maximum q AMA and q line increase with B, whereas the q ramp metrics do not.This is probably because there is very little room in CLT's ramp area near 18L, and so the q ramp saturates quickly.The fluctuations in maximum q ramp between simulations are due to subtle differences in the traffic jams encountered in each simulation, but the ramp area near 18L appears to saturate at q ramp = 6 departures.max q taxi shows signs of saturation for B > 5 when it levels off to 13 departures.It is interesting to note that the maximum q ramp jumps higher and the q AMA dips lower for B = 7, suggesting that a ramp traffic jam gave some relief to the AMA in this run.For 18C, the average queue results show minimal increases between simulations.There is more room in both the ramp area and AMA near 18C than near 18L.Whereas areas near 18C did not saturate, traffic jams near 18L may have restricted the flow of departures to 18C.The maximum queues for 18C actually occur during relatively light demand periods when departures to 18C are allowed to flow more freely.In actual operations, interviews with CLT ramp managers reveal that they try to maintain a q taxi of ~10 departures.The maximum q taxi for 18L is within one departure of this desired queue value for B values between 2 and 5.However, for B > 5, the maximum q taxi values for 18L are assumed to be undesirable. +VI. SUMMARY AND CONCLUSIONSA model of the ATD-2 tactical scheduler was implemented in fast time simulation and evaluated for varying values of taxi time delay buffer, B from (1).The two most interesting shifts in scheduler evaluation metrics occurred when B was increased from 1 to 2 minutes.Average total delay decreases began to level off and the average throughput prediction at flight ready time shifted from over prediction (negative error) to under prediction (positive error).The goal of surface metering is to move as much delay as possible from taxiing to the gate without negatively impacting total delay.Because throughout is lost when the departures are over metered, it is better to under predict throughput and allow the queues to temporarily increase than to over predict and allow the queues to reduce to 0 and let the runway go dry.In actual operations, CLT ramp managers prefer maintain a count of departure taxiing to a particular runway ~10, which runway 18L exceeds for B > 5.For these reasons, B values in the range of 2 to 5 minutes are recommended for future SOSS simulation studies for the ATD-2 tactical scheduler implemented at CLT.Other factors, such as additional uncertainties and human factors, may influence the desired delay buffer, requiring it to be calibrated separately in the field.This study provides a benchmark for fast-time exploration of design options for other ATD-2 tactical scheduler features.Future research building off of this benchmark will include additional uncertainties, external traffic management initiatives, and airline priority flights.Figure 2 .2Figure 2. Runway Demand +Fig. 33shows a stacked plot of average departure d gate , d ramp , and d AMA for each simulation as B was increased.Total departure delay d dep can be viewed as the sum of the stacked bars.As expected, d gate decreases and active taxi delay (d ramp + d AMA ) increases as B is increased.This verifies that B is an effective parameter to control the amount of delay moved from active taxi to the gate.Compared to d AMA , d ramp changes very little.This means that most if not all of the delay the lower B values transfer to the gate is queue delay (time waiting in line in the taxiways to use the runway).Note that d dep is noticeably greater when B < 2. When B is too low, the d gate is increased more than d AMA can be decreased due to starving the runway. +Fig. 44Fig.4shows a stacked plot of average arrival d AMA and d ramp for each simulation as B was increased.Total arrival delay d arr can be viewed as the sum of the stacked bars.Although average d arr is consistently much larger than d dep , it did not appear to be sensitive to departure taxi time delay buffers.Although extra arrival delay is expected due to additional gate conflicts as arrivals wait longer for departures held at their gates, much of this delay was absorbed as extra taxi path distance for arrivals using hardstands, which is not captured in the presented arrival delay calculations.Whereas departures absorbed most of their active taxi delay in the AMA, arrivals absorbed most of their delay in the ramp.In general, the arrival taxi-in routes are much longer than departure taxi-our routes.Also, the areas where arrivals tended to stack up waiting for use of common taxi route segments happened to be in the ramp area. +Figure 3 .3Figure 3. Gate, Ramp, and AMA Departure Delay +Figure 4 .4Figure 4. AMA and Ramp Arrival Delay +Figure 5 .Figure 6 .Figure 7 .567Figure 5. Runway Usage Time Prediction Error +Figure 9 .9Figure 9. 18L Maximum Queue Line +TABLE II.FLIGHT GROUP TRAJECTORY PREDICTIONFlight GroupPositionERUT CalculationLanding ArrivalairborneLanding STAairborneLanding STA + UTTTaxi ArrivalsurfaceCurrent time + UTTTaxi DeparturesurfaceCurrent time + UTTGate Departure ReadygateTOBT + UTTGate Departure PlannedgateEOBT + UTTGate Departure UncertaingateCurrent time + UTT + + + + +ACKNOWLEDGMENT This work was sponsored by NASA Airspace Operations and Safety Program's Aviation Technology Demonstration 2 (ATD-2) subproject. + + + + + + + + + A Concept for Integrated Arrival/Departure/Surface (IADS) Traffic Management for the Metroplex + + SZelinski + + + RCoppenbarger + + + + Airspace Technology Demonstration + + 2 + 2 + August 2015 + ConOps Synopsis + + + NASA Ames Research Center + + + unpublished + S. Zelinski and R. Coppenbarger, "A Concept for Integrated Arrival/Departure/Surface (IADS) Traffic Management for the Metroplex," Airspace Technology Demonstration 2 (ATD-2) ConOps Synopsis, NASA Ames Research Center, August 2015, unpublished. + + + + + ATD-2 Concept of Use (ConUse) for Charlotte-Douglas International Airport for the Phase 1 Baseline Integrated Arrival, Departure, Surface (IADS) Demonstration + + Nasa Atd-2 Conuse + + + Team + + + + Working Paper Series: Informal Publications of the ATD-2 Sub Project + + June 2016 + + + unpublished + NASA ATD-2 ConUse Team, "ATD-2 Concept of Use (ConUse) for Charlotte-Douglas International Airport for the Phase 1 Baseline Integrated Arrival, Departure, Surface (IADS) Demonstration," Working Paper Series: Informal Publications of the ATD-2 Sub Project, June 2016, unpublished. + + + + + Call for Papers + + MHayashi + + + THoang + + + YCJung + + + WMalik + + + HLee + + + VLDulchinos + + 10.1027/2192-0923/a000067 + + + Aviation Psychology and Applied Human Factors + Aviation Psychology and Applied Human Factors + 2192-0923 + 2192-0931 + + 4 + 2 + + June 2015 + Hogrefe Publishing Group + Lisbon, Portugal + + + M. Hayashi, T. Hoang, Y. C. Jung, W. Malik, H. Lee, and V. L. Dulchinos, "Evaluation of Pushback Decision-Support Tool Concept for Charlotte Douglas International Airport Ramp Operations," Eleventh USA/Europe Air Traffic Management Research and Development Seminar (ATM2015), Lisbon, Portugal, June 2015. + + + + + An Integrated Collaborative Decision Making and Tactical Advisory Concept for Airport Surface Operations Management + + GautamGupta + + + WaqarMalik + + + YoonJung + + 10.2514/6.2012-5651 + + + 12th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference and 14th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference + + American Institute of Aeronautics and Astronautics + June 2014 + + + FAA Air Traffic Organization Surface Operations Office Directorate "U.S. Airport Surface Collaborative Decision Making (CDM) Concept of Operations (ConOps) in the Near-Term: Application of the Surface Concept at United States Airport," June 2014. + + + + + Field test results of Collaborative Departure Queue Management + + ChrisBrinton + + + SteveLent + + + ChrisProvan + + 10.1109/dasc.2010.5655527 + + + 29th Digital Avionics Systems Conference + Berlin, Germany + + IEEE + June 2011 + + + C. Brinton, S. Lent, C. Provan, T. Prevost, and S. Passmore, "Collaborative Departure Queue Management: An Example of Airport Collaborative Decision Making in the United States," Ninth USA/Europe Air Traffic Management Research and Development Seminar (ATM2011), Berlin, Germany, June 2011. + + + + + Analysis of a Surface Congestion Management Technique at New York JFK Airport + + AlexNakahara + + + TomReynolds + + + ThomasWhite + + + ChrisMaccarone + + + RonDunsky + + 10.2514/6.2011-6987 + + + 11th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference + Virginia Beach, Virginia + + American Institute of Aeronautics and Astronautics + September 2011 + + + A. Nakahara, T. G. Reynolds, T. White, C. Maccarone, and R. Dunsky, "Analysis of a Surface Management Technique at New York JFK Airport," 11 th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, Virginia Beach, Virginia, September 2011. + + + + + Assessing the impacts of the JFK Ground Management Program + + StevenStroiney + + + BenjaminLevy + + + HarshadKhadilkar + + + HamsaBalakrishnan + + 10.1109/dasc.2013.6712508 + + + 2013 IEEE/AIAA 32nd Digital Avionics Systems Conference (DASC) + Syracuse, New York + + IEEE + October 2013 + + + S. Stroiney, B. Levy, H. Khadilkar, and H. Balakrishnan, "Assessing the Impacts of the JFK Ground Management Program," 32 nd Digital Avionics Systems Conference (DASC), Syracuse, New York, October 2013. + + + + + Demonstration of reduced airport congestion through pushback rate control + + IoannisSimaiakis + + + HarshadKhadilkar + + + HamsaBalakrishnan + + + TomGReynolds + + + RJohnHansman + + 10.1016/j.tra.2014.05.014 + + + Transportation Research Part A: Policy and Practice + Transportation Research Part A: Policy and Practice + 0965-8564 + + 66 + + June 2011 + Elsevier BV + Berlin, Germany + + + I. Simaiakis, H. Khadilkar, H. Balakrishnan, T. G. Reynolds, R. J. Hansman, B. Reilly and S. Urlass, "Demonstration of Reduced Airport Congestion Through Pushback Rate Control," Ninth USA/Europe Air Traffic Management Research and Development Seminar (ATM2011), Berlin, Germany, June 2011. + + + + + Dynamic control of airport departures: Algorithm development and field evaluation + + ISimaiakis + + + HBalakrishnan + + 10.1109/acc.2012.6315592 + + + 2012 American Control Conference (ACC) + Montreal, Canada + + IEEE + July 2012 + + + I. Simaiakis, and H. Balakrishnan, "Dynamic Control of Airport Departures: Algorithm Development and Field Evaluation," American Control Conference, Montreal, Canada, July 2012. + + + + + Tactical Departure Management with the Eurocontrol / DLR DMAN + + DBöhme + + + + Sixth USA/Europe Air Traffic Management Research and Development Seminar (ATM2005) + Baltimore, Maryland + + June 2005 + + + D. Böhme, "Tactical Departure Management with the Eurocontrol / DLR DMAN," Sixth USA/Europe Air Traffic Management Research and Development Seminar (ATM2005), Baltimore, Maryland, June 2005. + + + + + Departure flow control through takeoff sequence optimization: Setup and results of trials at Athens airport + + MeilinSchaper + + + GinaTsoukala + + + RaniaStavrati + + + NikosPapadopoulos + + 10.1109/dasc.2011.6096182 + + + 2011 IEEE/AIAA 30th Digital Avionics Systems Conference + Seattle, Washington + + IEEE + October 2011 + + + M. Schaper, G. Tsoukala, R. Stravrati, and N. Papadopoulos, "Departure Flow Control Through Takeoff Sequence Optimization: Setup and Results of Trials at Athens Airport," 30 th Digital Avionics Systems Conference (DASC), Seattle, Washington, October 2011. + + + + + Trajectory based ground movements and their coordination with departure management + + MeilinSchaper + + + IngridGerdes + + 10.1109/dasc.2013.6712662 + + + 2013 IEEE/AIAA 32nd Digital Avionics Systems Conference (DASC) + Syracuse, New York + + IEEE + October 2013 + + + M. Schaper and I. Gerdes, "Trajectory Based Ground Movements and Their Coordination with Departure Management," 32 nd Digital Avionics Systems Conference (DASC), Syracuse, New York, October 2013. + + + + + Email: Its decision support systems inroads—An update + + EdwardJLusk + + 10.1016/j.dss.2005.01.001 + + + + Decision Support Systems + Decision Support Systems + 0167-9236 + + 42 + 1 + + July 10, 2017 + Elsevier BV + + + NextGen Air Traffic Decision Support Systems. (n.d.) http://www.faa.gov/nextgen/update/progress_and_plans/decision_support_ systems/ [cited on July 10, 2017]. + + + + + Validation of Simulations of Airport Surface Traffic with the Surface Operations Simulator and Scheduler + + RobertDWindhorst + + + JustinVMontoya + + + ZhifanZhu + + + SergeiGridnev + + + KatyGriffin + + + AdityaSaraf + + + SteveStroiney + + 10.2514/6.2013-4207 + + + 2013 Aviation Technology, Integration, and Operations Conference + Los Angeles, California + + American Institute of Aeronautics and Astronautics + August 2013 + + + R. D. Windhorst, J. V. Montoya, Z. Zhu, S. Gridnev, K. Griffin, A. Saraf, and S. Stroiney, "Validation of Simulations of Airport Surface Traffic with the Surface Operations Simulator and Scheduler," 13 th Aviation Technology, Integration, and Operations (ATIO) Conference, Los Angeles, California, August 2013. + + + + + Modelling and Simulating Airport Surface Operations with Gate Conflicts + + SZelinski + 0000-0002-0008-0149 + + + RWindhorst + + 10.1017/aer.2018.111 + + + The Aeronautical Journal + Aeronaut. j. + 0001-9240 + 2059-6464 + + 123 + 1259 + + November 2017 + Cambridge University Press (CUP) + London, United Kingdom + + + S. Zelinski and R. Windhorst, "Modeling and Simulating Airport Surface Operations with Gate Conflicts," Royal Aeronautical Society Flight Simulation Conference, London, United Kingdom, November 2017. + + + + + + diff --git a/file839.txt b/file839.txt new file mode 100644 index 0000000000000000000000000000000000000000..5c08f102c9a067842313be0f037f4513e70e9950 --- /dev/null +++ b/file839.txt @@ -0,0 +1,438 @@ + + + + +I. IntroductionHE development of concepts and algorithms for efficiently managing airport surface operations is a critical area of Air Traffic Management research supporting the Next Generation Air Transportation System.Time-based metering or surface scheduling concepts currently in development are expected to improve not only airport efficiency, but also predictability, enabling integration with time-based concepts managing other phases of flight.Surface schedulers are used to predict when flights will use capacity constrained surface resources such as runways and gates, and generate advisories informing controllers when to clear or hold flights at key surface locations to minimize congestion and maximize throughput.NASA develops and tests surface scheduling concepts with both fast-and real-time (human-in-the-loop) simulations.Real-time simulations provide critical data about interactions between the automation and human operator components of a concept.However, real-time simulations have relatively high software development and staffing costs compared to fast-time simulations, limiting the number of scenarios and variables that may be studied this way.Although fast-time simulation allows a much larger problem trade space to be studied compared to real-time, humancentric operations must be modeled and are often difficult to validate.Together, real-time and fast-time simulation improve the understanding of surface scheduling and operations by using fast-time simulation to expand the scope of understanding and real-time simulation to refine understanding in areas of human interaction.NASA has developed a fast-time simulation of airport surface operations called the Surface Operations Simulator and Scheduler (SOSS) 1 to rapidly develop and test surface scheduling concepts.SOSS simulates aircraft moving through a network of surface taxiways between gates and runways.Aircraft must conform to operating rules such as separation constraints.SOSS also has the ability to connect to a scheduler, to which it passes aircraft state information and from which it may receive commands such as release times for specific flights at specific nodes along their surface route.This ability enables researchers to use SOSS to develop and test new surface scheduling concepts and algorithms.SOSS airport surface models have been used to study future operations at airports in the U.S.A. (Dallas Fort Worth International Airport (DFW) [2][3][4] and Charlotte Douglass International Airport (CLT) 5- 9 ) and outside (Hamburg Airport in Germany 10 and Incheon Airport in South Korea 11 ).Currently, researchers are using SOSS to develop surface metering schedulers to support the NASA Air Traffic Management Technology Demonstration 2 (ATD-2) project's field demonstration of time-based surface metering at CLT. 8,9,12 Much of SOSS development for this work has been driven by the challenges in modelling and simulating CLT operations.First, CLT's surface configuration, including dual use runways, intersecting runways, converging runways, and taxiways crossing runways, requires a complex set of runway operating constraints for safe operations.CLT also has limited taxiways including several single lane taxiways, that make managing operations to prevent surface gridlock a challenge.Finally, CLT is a busy hub airport characterized by tightly spaced arrival and departure banks (temporary rise in demand), which result in heavy use of limited gates.The departure banks precede the arrival banks just in time for departures to vacate gates for arrivals.If a departure pushback from the gate is late or an arrival using the same gate is early, this causes a gate conflict, which, if left unmanaged, can lead to surface gridlock.In real operations, hardstand areas are used to temporality park aircraft out of the way to avoid gate conflicts.Surface metering concepts like ATD-2 have the potential to increase gate conflicts as more flights are held at the gates to keep taxiways free of congestion, therefore, this is an important phenomenon to model and include in simulations used to develop metering algorithms.This paper describes the SOSS simulation platform and the modeling approach taken for each of the aforementioned challenges associated with CLT surface operations, particularly gate conflicts.Models of CLT hardstands and new SOSS functionality were developed to address gate conflicts.Several approaches to managing gate conflicts with and without the use of hardstands were simulated and their effects on surface operations compared. +II. Surface Operations Simulator and SchedulerThis section provides an overview of the SOSS simulation platform.The overview describes the airport surface model and how aircraft move through this model and adhere to various separation constraints.A description of how SOSS interacts with an externally modeled surface scheduler is also provided. +A. Airport ModelA SOSS airport model specifies a network of nodes and links as an undirected graph, where nodes represent points on the airport surface and links represent straight paths between nodes.Figure 1 shows a diagram of the SOSS airport model for CLT.Long rectangles outline CLT's four runways labeled for South configuration, where runways 18R and 23 are used only for arrival operations, and runways 18C and 18L are used predominantly for departure operations.Nodes and links are color-coded by type, which reflects their operational function.A flight taxi route is specified by a sequence of nodes, connected by links, traversing the airport model.Because links define straight paths between nodes, curved route segments are approximated by a series of shorter links between closely spaced nodes.A departure route always begins at a gate node (gray) and moves through a sequence of ramp nodes and links (blue) to a spot node (yellow), which marks the transition point between the ramp area (controlled by air carriers) and the active movement area (AMA) (controlled by the air navigation service provider).From the spot node, the departure then moves through a sequence of taxiway nodes and links (green) and departure queue nodes and links (cyan) until it reaches a departure node (cyan) from which it will take off.An arrival route always begins at an arrival node (cyan), then moves through a sequence of taxiway nodes and links to a spot node, then through a sequence of ramp nodes and links until it reaches a gate node.Some arrival routes may need to taxi across a runway, in which case they will pass through crossing entry and exit nodes (red) on either side of the runway.For this work, another node type was introduced to represent hardstands (purple), where aircraft may park temporarily within the ramp area out of the way of other ramp traffic.A total of 11 hardstand nodes were included in the airport model shown in figure 1, eight in the lower left of the ramp area near runway 18C, and three in the uppermost ramp area near runway 18L.A flight scenario contains information about all the flights in the simulation, including simulation entry time, gate and runway assignment, aircraft type, and flight plan.A list of all routes that flights are allowed to traverse within a simulation are specified in a separate route set.For each departure specified in the flight scenario, the route set must contain at least one route from the assigned gate to a departure node for the assigned runway.Similarly, for each arrival specified in the flight scenario, the route set must contain at least one route from an arrival node for the assigned runway to the assigned gate.The first route specified in the route set for each unique gate-to-runway or runway-togate assignment pair is the default route for that assignment pair.If the route set specifies more than one route for an assignment pair, any flight with that assignment pair may be rerouted to one of the specified alternate routes during the simulation.Alternate routes were used to model hardstand operations.For each default route, 11 alternate hardstand routes were defined, each passing through one of the 11 hardstand nodes with start and end nodes identical to those in the default route.Routes were designed to keep traffic along each link flowing in one direction as much as possible.Figure 2 shows examples of default and hardstand routes.Two examples each are given for departures and arrivals.Each example shows only one of several hardstand routes defined for the default route.The other hardstand routes are similar but pass through other hardstand nodes.The example shows how some hardstand routes may be comparable in length to their default routes, whereas some are much longer than their default routes -most are longer. +B. Aircraft MovementSOSS's aircraft movement model uses kinematic equations of motion to move flights along their assigned routes.Aircraft parameters specify acceleration and deceleration values and a set of target speeds for different areas of the surface for each aircraft type (e.g.B737, A380, etc.).As the aircraft transitions from one area of the surface to another and the target speed changes, the model will use the specified acceleration or deceleration values for that aircraft type to speed up or slow to the new target speed.Ref. 4 describes the aircraft movement model in more detail and Ref. 6 provides an example speed profile using aircraft parameters. +C. Conflict Detection and ResolutionAircraft movement may be interrupted by other traffic when taxi separation constraints are imposed.SOSS predicts when two aircraft will come into conflict and resolves the conflict by slowing or stopping one of the flights based on right-of-way rules.Taxi separation constraints in CLT's complex and often congested ramp area surrounding the gates were not included in early CLT surface scheduling studies due to gridlocking issues. 5,6Later, SOSS's conflict detection and resolution functionality was enhanced as described below to impede gridlock and allow these constraints to be included.In-trail conflicts, in which an aircraft is predicted to overtake another traveling in the same direction, are handled by slowing the trailing aircraft to avoid getting too close to the one in front.Merging conflicts, in which two aircraft paths intersect, are handled by slowing or stopping the aircraft predicted to arrive at the intersection last, to let the other one pass.Head-on conflicts, in which two aircraft are predicted to lose separation travelling in opposite directions on the same link, are handled as follows: first, the set of links that make up the common path between the two aircraft is determined; then, the aircraft predicted to reach its nearest link in the common set last is slowed or stopped outside the common path to let the other aircraft pass.Conflict detection and resolution is performed for only two aircraft at a time.There is a potential for three or more aircraft entering into a gridlock situation, especially where there is a tight network with short links or if one of the aircraft involved is large and requires a large taxi separation.High demand flight scenarios often gridlock in simulation.In some cases, the flight scenario may be modified to allow a simulation to complete without gridlock by removing flights or changing their simulation entry time or aircraft type.The gate conflict management approaches described later reduce the occurrence of simulation gridlock without having to modify the flight scenario. +D. Runway Separation ModelIn previous work, SOSS runway separation operations were implemented by purely time-based constraints (specified by tables of minimum time required for one runway operation type and aircraft weight class to follow another) and the scheduler used the same time-based constraints to calculate metering control times. 5Many surface separation rules are distance-based in actual operations.Therefore, uncertainty introduced by the discrepancy between distance-based separation rules for operations and time-based separation rules for scheduling was lost.The SOSS runway separation model has since been enhanced to more accurately reflect the tactical runway separation rules used in actual operations with the addition of distance-based rules.Three types of operations use runways: arrival, departure, and taxi crossing.Arrivals enter the simulation on final approach several miles from the runway threshold with no opportunity to hold before landing at the arrival node.Consecutive arrivals are assumed to be sufficiently separated by their simulation entry times in the flight scenario.SOSS implements all other runway separation constraints by holding or releasing departures at the departure node and by holding or releasing crossers at the crossing entry node.Departure and crossing operations are handled tactically in first-come-first-served order.All time-based separation rules are used by the scheduler as well.However, the scheduler uses time-based approximations of the more recently implemented distance-based separation rules included below.The scheduler time-based approximations are tuned based on observations of the simulated operations following the distance-based rules.Aircraft are held or released for takeoff based on the following rules:• Consecutive departures on the same runway must be time separated based on the weight class.All departures following Small or Large weight class departures must be separated by at least 60 s.Small or Large departures following Heavy or B757 departures must be separated by at least 90 s.Heavy or B757 departures following Heavy or B757 departures must be separated by at least 120 s. • A departure following an arrival or a crosser on the same runway may be released for takeoff no earlier than one second after the other aircraft reaches its runway or crossing exit node, respectively.• A departure preceding an arrival on the same runway may be released for takeoff only if the arrival is at least 1500 m from reaching the runway entry node.• A departure on 18L preceding an arrival on the intersecting runway 23 may be released for takeoff only if the arrival is at least 1500 m from reaching its runway entry node.• A departure on 18L following an arrival on the intersecting runway 23 may be released for takeoff only if the arrival is at least 600 m past its runway entry node.This puts the arrival past where the runways intersect.• A departure on 18C preceding an arrival on the converging runway 23 may be released for takeoff only if the arrival is at least 3334 m (1.8 nmi) from reaching its runway entry node.• A departure on 18C following an arrival on the converging runway 23 may be released for takeoff no earlier than 1 second after the arrival reaches its runway entry node.Aircraft are held/released for crossing a runway based on the following rules:• Consecutive releases from the same crossing entry node must be separated by at least five seconds.• A crossing aircraft preceding an arrival may be released from its crossing entry node only if the arrival is at least 1000 m from reaching its runway entry node.• A crossing aircraft following an arrival or departure operation may be released from its crossing entry node only if the other aircraft has passed where the taxiway crosses the runway.This condition is implemented as a unique minimum distance past the runway threshold based on the crossing location. +E. Scheduler InterfaceSOSS may connect to schedulers via a socket using a protocol called the Common Algorithm Interface (CAI).The user sets the frequency of scheduler calls.For each call, SOSS sends the scheduler information for all flights currently active in the simulation as well as flights that will enter the simulation within a user specified planning horizon.Flight information includes information about the aircraft (e.g.call sign, weight class, type), state (e.g.location, heading, speed), route, and any constraints imposed by external traffic management initiatives.A key piece of departure information included is the time the departure expects to push back from the gate, also known as its Earliest Off Block Time (EOBT).The scheduler uses the flight information it is given to calculate and send back to SOSS times of release at specific nodes along each flight's route.If a flight arrives at a node before its scheduled release time, SOSS holds the flight at the node until the release time.If a flight arrives at a node without an assigned release time or the release time has already passed, SOSS allows the flight to continue along its route.It is up to the scheduler to set or update release times for specific flights at nodes along their routes.The tactical scheduler for the ATD-2 departure metering concept nominally sets release times for departures at gate nodes.This release time is known as the departure's Target Off Block Time (TOBT).The scheduler may also use the CAI to change a flight's taxi route as long as the flight has not been released from its first node yet (gate node for departures or arrival node for arrivals).All flight reroutes to hardstands are implemented before the flight is released from its first node.Any departure may be rerouted to a hardstand prior to pushing back from the gate by changing its route from the default to one of the associated alternate hardstand routes.Similarly, any arrival may be rerouted to a hardstand by changing its route before reaching the arrival node.Once the route is changed, the scheduler may set release times at nodes along the new route.For the purposes of modeling hardstand operations integrated with departure metering, the scheduler will set release times not only for departure gate nodes, but also for departure and arrival hardstand nodes.At every scheduler call, for each flight f and each node p along the flight's assigned route r, there are three types of times stamps.Earliest time E f,p,r is the earliest time f can arrive at, or be released from, p calculated by projecting unimpeded transit time from current position along r.The target time T f,p,r is the time produced by the tactical scheduler for f to be released from p.The T f,p,r is always greater than or equal to E f,p,r .Only some T f,p,r are sent to SOSS as controlled release times (e.g. for departure gate nodes and hardstand nodes).Others are merely predictions of when f will be released from p based on scheduling constraints.Actual time A f,p,r is the actual time f was released from p in simulation.The time duration U f,p1,p2,r is the unimpeded or minimum time it would take for f to travel from p 1 to p 2 along r in the absence of other flights, which is equal to the time duration between E f,p1,r and E f,p2,r .All flight routes contain at most one of each of the node types, gate, spot, hardstand, departure, and arrival.All route options for the same flight use the exact same gate, spot, and departure or arrival nodes.Only the hardstand nodes and intermediate ramp nodes between the spot and gate via the hardstand differ between route options for the same flight.Let a p denotation of G, S, and H represent the gate, spot and hardstand nodes along r, respectively.Let a p denotation of R represent either the departure or arrival node on the runway.Using this notation, the abbreviations EOBT and TOBT, commonly used in previous work, are represented by E f,G,r and T f,G,r , respectively.This deviation from previously used nomenclature is adopted to add clarity in describing gate conflict management logic considering more than one flight or route.For each gate conflict, two flights are considered, one arrival and one departure, for which f is denoted as A and D, respectively.For each flight, at most two routes are considered, the original route assigned to the flight and next available hardstand route, for which r is denoted as O and H respectively.Table 1 summarizes denotations described above.E = Earliest A = Arrival G = Gate O = Original T = Target D = Departure S = Spot H = Hardstand A = Actual H = Hardstand U = Unimpeded R = Runway +III. Modeling Hardstand OperationsHardstands are temporary parking areas for flights that do not have access to a gate.Hardstands are often used as remote gates, and passengers are bussed between terminals and hardstands for boarding and deplaning.Hardstands are also used to temporarily park a flight out of the way when it's gate is not available, and is how they are used in this study to avoid predicted gate conflicts. +A. Predicting Gate ConflictsIn actual operations, a gate conflict is typically discovered and resolved as follows.Ramp Control is contacted by an arrival flight at the spot asking for clearance to enter the ramp.At this point, Ramp Control would either see that the arrival gate is empty and give the arrival clearance to proceed through the ramp to its gate, or discover that the gate is still occupied, which constitutes a gate conflict.Because in simulation the decision to change an arrival's route must be made prior to entering the arrival node, the scheduler must predict gate conflicts rather than react to them as they occur.When the scheduler is called, it will first update the runway schedule for all flights (T f,R,O ), then update T f,G,O for all departures based on the runway schedule.The scheduler will then search for all pairs of arrivals/departures assigned to the same gate and check for gate conflicts.A gate conflict is predicted for a given arrival/departure pair (A, D) when E A,G,O is within a given gate separation buffer b of T D,G,O for the same gate as follows. +Gate Conflict: EA,G,O < T D,G,O + b.(1)In previous work 8 before SOSS hardstand models were developed, gate conflicts were detected using a b equal to one minute and resolved by updating the time T D,G,O to equal max(E A,G,O -1 minute, E D,G,O ).For this work, gate conflicts were predicted using a more conservative b of five minutes.Interviews with CLT ramp controllers suggested five minutes was a reasonable buffer for defining gate conflicts in actual operations. +B. Gate Conflict ManagementFour gate conflict management approaches were developed and modeled for this work: No Hardstand, Departure Hardstand, Arrival Hardstand, and Dual Use Hardstand.These approaches allow different combinations of resolutions involving the departure and/or the arrival in a given gate conflict.Two types of departure resolutions are used in this study: early release from the gate without changing the taxi route, and early release from the gate while rerouting to the hardstand.Departure resolutions are implemented when the departure is ready for pushback, referred to as ready time.In operations, ready time is indicated by a voice call from the pilot to Ramp Control.In SOSS, departures are initialized and occupy the gate for a user specified time duration prior to ready time.The departure's "arrival" at the gate node is only recognized and recorded at ready time.SOSS has the capability to model ready time uncertainty relative to E f,G,r , however, ready time uncertainty was not modelled in this study and, therefore, ready time equaled E f,G,r for all departures.The only type of arrival resolution used in this study is rerouting to the hardstand.Arrival resolutions are implemented when the arrival is predicted to land within 100 seconds to ensure there is time to change the arrival's taxi route before it lands. +No HardstandThis approach resolves predicted gate conflicts by releasing the departures from the gate once they are ready for pushback by setting T D,G,O equal to current time.Because only departure early release resolutions are allowed, under this approach gate conflicts are identified and resolutions implemented at departure ready time only. +Departure HardstandThe problem with the No Hardstand approach is that it gives an unfair advantage to departures with gate conflicts and it can potentially make the runway schedule less predictable.The Departure Hardstand approach addresses this issue by allowing the departure to push back from the gate at ready time and to be metered from a hardstand rather than at the gate, thus freeing up the gate for use by the arrival flight.The departure is assigned the next available hardstand route r=H and T D,H,H is calculated to allow the flight to reach the runway at its original target time T D,R,O .The T D,H,H calculation is similar to the T D,G,O calculation for gate hold advisories described in previous work 9 by back calculating release time from T D,R,O and inserting delay buffers A and B as follows.T D,G,O = T D,R,O -A(U D,G,R,O ) -B (2) T D,H,H = T D,R,O -A(U D,H,R,H ) -B,(3)where A ≥ 1 and B ≥ 0. Factor A accounts for congestion delay the flight may encounter along its taxi route and assumes this delay is directly proportional to the flight's remaining unimpeded travel time.Factor B captures queue delay incurred when flights line up from the end of the runway.Because most hardstand routes are longer than default routes, there is a possibility that by the time the flight is ready to pushback, the hardstand route does not enable the flight to meet its original target runway time.This is checked by comparing the difference in unimpeded transit time (buffered A) and the gate release times between the original route and hardstand route.At ready time, ifA(U D,G,R,H -U D,G,R,O ) £ T D,G,O -currenttime, the departure is released right away and sent to the hardstand.Otherwise, the resolution is identical to No Hardstand, where the departure retains the original route and is released from the gate right away.Because only departure early release and departure-to-hardstand resolutions are allowed, under this approach gate conflicts are identified and resolutions implemented at departure ready time only. +Arrival HardstandWhereas the Departure Hardstand approach may resolve a gate conflict caused by holding a ready departure, it will not resolve a gate conflict that occurs before the departure is ready.The Arrival Hardstand approach addresses this issue by sending conflicting arrivals to the hardstand instead of departures.In this approach, every time a gate conflict is predicted, the arrival is assigned a hardstand route.An Arrival Hardstand approach was used in previous work 9 to avoid the impact that departure resolutions might have on departure metering results being studied.Whereas in previous work, the arrival was released from the hardstand when actual departure pushback was detected, in this work the T A,H,H is designed to get the arrival to the gate b after departure pushback.The T A,H,H is calculated asT A,H,H = T D,G,O -A(U A,H,G,H ) + b.(4)Because only arrival-to-hardstand resolutions are allowed, gate conflicts are identified and resolutions are implemented only when the arrival is predicted to land within 100 seconds. +Dual HardstandAlthough the Arrival Hardstand approach appears to be a simple way to resolve gate conflicts without disrupting departure metering, getting arrivals to the gate on time is a higher priority for airlines than minimizing active taxi time for departures or maintaining a predictable runway schedule.CLT is a hub airport where most of the passengers arriving have ~45 minutes to deplane and transit the busy terminals to make a connecting departure.Delaying arrival gate time increases the probability of missed connections.The Dual Use Hardstand approach is designed to strike a balance between Departure Hardstand and Arrival Hardstand by sending departures or arrivals to the hardstand when most appropriate.In this approach, if a gate conflict is predicted when the arrival is predicted to land within 100 seconds, the arrival is sent to the hardstand only if the E A,G,O is within b of the departure E D,G,O .If a gate conflict is predicted at departure ready time, the logic first checks to see if the arrival is already being held at or is on its way to a hardstand.If not, the arrival either has not yet been rerouted to a hardstand, or has already been released from the hardstand, in which case the departure is sent to the hardstand in accordance with the Departure Hardstand logic.Because the resolutions may involve the arrival or departure, gate conflicts are identified and given resolution opportunity both when the arrival is predicted to land within 100 seconds, and at departure ready time.Therefore, it is possible to use both an arrival and departure resolution action for the same conflict.This may be necessary if conditions change after implementing the first action or the first action could not fully resolve the conflict.A specific example is discussed in results section V.A.Table 2 summarizes the scheduler logic used in each of the above gate conflict management approaches. +IV. Experiment SetupA. Simulation Parameters Four SOSS simulations were completed and compared, one for each gate conflict management approach in Table 2, each using a b of five minutes.Each simulation was run using a time step of 0.5 seconds and the tactical scheduler was called every 10 seconds.In all simulations, the tactical scheduler used values of 1.05 for A and 2.0 minutes for B. A detailed description of the tactical scheduler can be found in previous work. 9ne simulation was attempted without any gate conflict management (i.e., no early gate releases or hardstand rerouting actions), but a group of flights gridlocked in the ramp area near 18L, and the simulation could not be completed.Therefore, no results are presented for this incomplete simulation. +B. Traffic ScenarioAs a hub airport for American Airlines, CLT traffic is characterized by tightly spaced departure and arrival banks.A four-hour traffic scenario was generated using CLT surface surveillance data from March 11, 2016, covering two such banks when CLT was operating in South flow.Because this scenario was generated from surveillance data, arrival and departure demand reflects actual operations in which arrivals landed early or departures pushed back late relative to their airline schedules, which is why predicted gate conflicts were expected to occur naturally in the scenario.The scenario contains a total of 175 arrivals and 199 departures.Fig. 3 shows the arrival and departure demand on each runway in 15-minute bins. +C. Evaluation MetricsFour categories of metrics were used to evaluate and compare each gate conflict management approach: gate time separation, scheduler predictability, surface transit times, and surface counts.These are defined as follows. +Gate Time SeparationActual gate time separation is the difference in actual gate times between the arrival and departure involved in an identified gate conflict.The actual separation delta d is calculated by subtracting b from the actual gate separation.d = A A,G,r -A D,G,r -b.(5) +Scheduler PredictabilityScheduler predictability measures how well the departure runway usage time can be predicted at ready time.Scheduler predictability is measured at ready time because this is the last opportunity the scheduler has to update the T D,G,O , after which they are frozen to provide stable gate hold advisories for ramp controllers.Let a departure's runway usage time prediction error be the difference between its actual runway time and target runway time when it became ready for pushback calculated ase = A D,R,r -T D,R,O (t ready ),(6)where t ready indicates that the T D,R,O was calculated at the flight's ready time. +Surface Transit TimesSurface transit time t s (f) measures the length of time that flight f spent in surface area s.Surface areas times includes active taxi times in the ramp and in the AMA, and hold times spent at the gate and hardstand.Gate hold t G (f) and hardstand hold t H (f) times for flight f are calculated as the difference between when the flight arrived at and was released from the gate or hardstand respectively.Calculations for arrivals and departures are indicated with f denotations of A and D, respectively.tG (D) = A D,G,r -t ready ,(7) t H(A) = A A,H,r -t H ,(8) t H(D) = A D,H,r -t H ,(9)where t H is the time the flight arrived at the hardstand node.Note that gate hold is calculated for departures only and the departure's ready time t ready is the same as when it "arrives" at the gate node in simulation.Hardstand hold is calculated similarly for arrivals and departures.Ramp taxi t Ramp (f) and AMA taxi t AMA (f) times are calculated for arrivals (f=A) and departures (f=D) as follows.t Ramp (D) = A D,S,r -A D,G,r -t H (D),(10)t Ramp (A) = A A,G,r -A A,S,r -t H (A),(11)t AMA (D) = A D,R,r -A D,S,r ,(12)t AMA (A) = A A,S,r -A A,R,r .(13)Note that both arrival and departure ramp taxi calculations exclude hardstand hold, so that only time in active taxi is captured.Arrival and departure calculations differ only in their direction of travel between surface nodes. +Surface CountsSurface count " (t) is the number of flights within each area of the airport surface s at time t.Areas of the airport surface used for counts are the same as those used for surface transit times.They include gate hold count # $ (t), hardstand hold counts % $ (t) and % & (t), ramp taxi counts '()* $ (t) and '()* & (t), and active movement area taxi counts $+$ $ (t) and $+$ & (t), with each segregated between arrivals and departures, indicated with superscripts A and D, respectively.Whereas surface transit times measure elapsed time for each flight, surface counts measure numbers of flights for each simulation time step. +V. ResultsThis section discusses the results for each of the four evaluation metrics described in section IV.C.Because these results were produced using a model of CLT, they may not hold for another airport with other runway and traffic conditions. +A. Gate Conflict Resolutions and Gate Time SeparationEach simulation produced gate conflicts between the same 13 flight pairs.Figure 4 shows the number of flight pairs in each simulation belonging to each of five types of resolutions.Values are stacked such that each row sums to 13 total number of flight pairs.Figure 5 The resolution type "None" means that the gate conflict was predicted but no resolution could be implemented.Four such conflicts arose while using management approaches in which only departure resolutions were allowed (No Hardstand and Departure Hardstand).The departures from these four flight pairs were not assigned any gate delay by the scheduler, and were unable to push back from the gate any earlier to resolve the gate conflict.Figure 5 shows how none of these flight pairs (red Xs) achieved the desired separation as seen from their negative d values.Other No Hardstand and Departure Hardstand resolutions failed to achieve desired separation because, although the departures were released as early as possible rather than waiting for the scheduler assigned T D,G,O , this was not early enough to avoid the gate conflict.Although the Departure Hardstand approach succeeded in moving many of the No Hardstand approach early released departures to the hardstand, the departure gate release times are the same.Differences in d The Arrival Hardstand approach produced only one gate conflict violation (green X close to -1).This was because in the time between when the arrival was released from the hardstand and the flight became ready for pushback, the scheduler updated the departure T D,G,r to a later time, causing the departure to hold at the gate longer than earlier expected.This gate conflict was resolved using the Dual Hardstand approach (pink X) by sending the departure to the hardstand as well.Overall, these results show that arrival resolutions are necessary to fully resolve gate conflicts due to the large proportion of cases where departures are not ready to pushback early enough to resolve the conflict. +B. Scheduler PredictabilityFigure 6 shows departure runway usage time prediction error (e) results for each simulation.Average and standard deviation are calculated for three sets of departures: those that were involved in one of the 13 identified gate conflicts (Conflict), those that were not involved in gate conflicts (Other), and all departures regardless of their involvement in gate conflicts (All).Negative and positive average e values indicate that flight used the runway earlier and later than predicted, respectively.The management approaches appear to affect the standard deviation of e for all departures, and affect the average e of flights involved in gate conflicts more than other flights.The No Hardstand approach releases departures involved in gate conflicts early causing their average runway usage time to be earlier than predicted, which decreases predictability for all flights (indicated by high standard deviation values).By moving some of these departures to the hardstand, the Departure Hardstand approach decreases the average e but the standard deviation of e for all flights does not decrease much.The Arrival Hardstand approach produces an average e for departures involved in gate conflicts that is similar to that of other flights.Additionally, the standard deviation of e for all flights is lower under the Arrival Hardstand approach.This is due to the fact that departures are not involved in any of the gate conflict resolutions.The Dual Hardstand approach produces average and standard deviation of e for all flights similar to that of the Arrival Hardstand approach.The impact on the departure scheduler of sending a few departures to the hardstand in addition to the arrivals is only apparent in the average and standard deviation of e for departures involved in a gate conflict.Overall, these results show that departure resolutions have the greatest impact on scheduler predictability, and their use should be limited if maintaining scheduler predictability is a priority. +C. Surface Transit TimeFigure 7 shows average surface transit time (t) results for each simulation.The t results in the bar chart are stacked such that the total height of each bar represents the average total transit time between landing and gate for arrivals and between pushback ready and takeoff for departures.Three sets of stacked averages are shown for each management approach and flight type (arrivals and departures): flights involved in gate conflicts (Conflict), flights not involved in gate conflicts (Other), and all flights regardless of gate conflict involvement (All).The t results for the No Hardstand and Departure Hardstand approaches are similar, with the most notable difference in results for departures involved in gate conflicts.Whereas both approaches remove all gate holding for departures involved in gate conflicts, the Departure Hardstand approach transfers this gate holding to hardstand holding.In both approaches, the ramp taxi times are much higher for arrivals involved in gate conflicts than other arrivals due to the gate conflict violations with negative d values in Fig. 5.The arrivals involved in these violations are forced to stop in the ramp and wait for the departures to vacated their gates.The Arrival Hardstand and Dual Hardstand approaches add a large amount of hardstand hold time to the total transit time for arrivals involved in gate conflicts.The active taxi time (Ramp and AMA) for other flights is slightly higher for arrivals and lower for departures than in the management approaches that do not include arrival resolutions.Sending these arrival to the hardstand just delays their surface congestion impact.Because the arrival banks are slightly later than departure banks as seen in Fig. 3, this shifts the congestion impact slightly from departures to arrivals.Much of the gate holding for departures involved in gate conflicts in the Arrival Hardstand approach is converted to hardstand holding in the Dual Hardstand approach without impacting other departures.Overall, these results show the cost of using arrival resolutions.Sending arrivals to the hardstand increases total transit time for the arrivals involved in gate conflicts ~6 minutes and other arrivals ~1 minute on average. +D. Surface CountsFigure 8 shows the maximum surface count (n) results for each simulation.As expected, hardstand hold counts % $ and % & , appear only for management approaches allowing arrival-to-hardstand and departure-to-hardstand resolutions, respectively.No more than three aircraft occupy hardstands at one time, suggesting that the 11 hardstand nodes modeled are more than enough to handle gate conflicts arising from high volume CLT traffic in simulation.For all management approaches, arrival ramp taxi count '()* $ is the largest surface count, more than double any other count, making arrivals in the ramp the largest source of congestion.The management approaches with arrival resolutions (Arrival Hardstand and Dual Hardstand) reduce the maximum '()* $ and increase the maximum $+$ $ by sending arrivals to the hardstand.The capacity of the ramp is temporarily reduced by the complex routing of arrivals exiting hardstands.Holding arrivals at the hardstand, released them into the ramp during a higher arrival volume period and pushed the line of arrivals waiting to enter the ramp farther into the AMA.The maximum departure taxi counts '()* differ by no more than one or two aircraft between simulations suggesting the management approach does not impact departure congestion as much as arrival congestion.This may be because departure congestion is managed by metering at the gates, whereas arrivals must be allowed to enter the taxiways as soon as they land.The Departure Hardstand approach has higher maximum # & than all other approaches.Whereas the No Hardstand approach is expected to have lower # & due to departure early release resolutions, it is unclear why the Arrival Hardstand and Dual Hardstand approaches have lower # & as well.Overall, these results highlight arrivals as the greatest source of surface congestion.Whereas departure surface congestion is managed by metering at the gates, arrivals must enter the active taxiways as soon as they land.Sending arrival to hardstands manages gate conflicts, not arrival surface congestion.Developing models for managing arrival surface congestion is left for future research. +VI. Summary and ConclusionsThis paper described the SOSS fast-time simulation platform used to rapidly develop and test surface scheduling concepts.New SOSS models and functionality for hardstand operations were developed to simulate gate conflict management approaches using hardstands at CLT.Four gate conflict management approaches were simulated and compared.The No Hardstand approach resolved predicted gate conflicts by releasing metered departures from the gate early.The Departure Hardstand approach introduced the option of sending an early released departure to a hardstand to be metered.The Arrival Hardstand approach resolved gate conflicts by sending arrivals to the hardstand instead of departures.The Dual Hardstand approach allowed resolutions releasing the departure from the gate early and sending either or both arrival and departure to the hardstand.The gate conflict management approaches allowing arrivals to go the hardstand (Arrival Hardstand and Dual Hardstand) were most successful in resolving gate conflicts.They also produced more consistent scheduler predictability between departures involved in gate conflicts and other departures.However, these approaches increased the average total surface transit time ~6 minutes for arrivals sent to the hardstand and ~1 minute for other arrivals.Due to its relative simplicity over the Dual Hardstand approach, the Arrival Hardstand is the best approach to use for tactical scheduler development at when departures are subject to short tactical delays as seen in this study.The Dual Hardstand approach may show more advantage for its additional complexity when departures are subject to longer strategic delays from external traffic management initiatives, which is a subject for future study.Figure 2 .2Figure 2. Example default and hardstand routes for departures and arrivals. +Figure 1 .1Figure 1.Airport Model of CLT. +Figure 3 .3Figure 3. Runway demand. +shows the d value for each flight pair.Positive values of d indicate that the gate conflict was successfully resolved meeting the desired separation buffer b of five minutes.Negative values of d indicate that the gate time separation was less than b.Note that d may not be less than -5 minutes as this would mean that the arrival and departure flights occupied the gate at the same time and violated taxi separation constraints described in section II.C. +Figure 4 .Figure 5 .45Figure 4. Numbers of resolution types per management approach +Figure 6 .6Figure 6.Average and standard deviation of runway usage time error (e) for each management approach. +Figure 7 .7Figure 7. Average surface transit times (t) for each management approach. +Table 1 . Denotations for Surface Time Stamps and Time Duration1Time Stamp or DurationFlight (f)Node (p)Route (r) +Table 2 . Scheduler Logic for Managing Predicted Gate Conflicts2Management ApproachScheduler LogicNo Hardstandif (Gate Conflict predicted at departure ready time)TD,G,O = current timeDeparture Hardstandif (Gate Conflict predicted at departure ready time)if (A(UD,G,R,H -UD,G,R,O) £ TD,G,O -current time)assign flight D to route HTD,H,H = TD,R,O A(UD,H,R,H) -BTD,G,H = current timeelseuse No HardstandArrival Hardstandif (Gate Conflict predicted 100 sec prior to predicted arrival landing time)assign flight A to route HTA,H,H = TD,G,O -A(UA,H,G,H) + bDual Use Hardstandif (Gate Conflict predicted 100 sec prior to predicted arrival landing time)if (EA,G,O < ED,G,O + b)use Arrival Hardstandif (Gate Conflict predicted at departure ready time)if (flight A is not at or on-route to a hardstand)use Departure Hardstand + & and $+$ & + + + + +AcknowledgmentsThis work was sponsored by NASA Airspace Operations and Safety Program's Aviation Technology Demonstration 2 (ATD-2) subproject.Figure 8. Maximum surface counts (n) for each management approach. + + + + + + + + + Validation of Simulations of Airport Surface Traffic with the Surface Operations Simulator and Scheduler + + RobertDWindhorst + + + JustinVMontoya + + + ZhifanZhu + + + SergeiGridnev + + + KatyGriffin + + + AdityaSaraf + + + SteveStroiney + + 10.2514/6.2013-4207 + + + 2013 Aviation Technology, Integration, and Operations Conference + Los Angeles, California + + American Institute of Aeronautics and Astronautics + August 2013 + + + + Windhorst, R. D., J. V. Montoya, Z. Zhu, S. Gridnev, K. J. Griffin, A. Saraf, and S. Stroiney, "Validation of Simulations of Airport Surface Traffic with the Surface Operations Simulator and Scheduler," AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, 12-14 August 2013, Los Angeles, California. + + + + + A Simulator for Modeling Aircraft Surface Operations at Airports + + ZacharyWood + + + SivakumarRathinam + + + YoonJung + + + MatthewKistler + + 10.2514/6.2009-5912 + + + AIAA Modeling and Simulation Technologies Conference + Chicago, Illinois + + American Institute of Aeronautics and Astronautics + 10-13 August 2009 + + + Wood, Z., M. Kistler, S. Rathinam, and Y. Jung, "A Simulator for Modeling Aircraft Surface Operations at Airports," AIAA Modeling and Simulation Technologies Conference, 10-13 August 2009, Chicago, Illinois. + + + + + An Integrated Gate Turnaround Management Concept Leveraging Big Data/Analytics for NAS Performance Improvements + + WilliamWChung + + + GChadchad + + + RHochstetler + + 10.2514/6.2016-3909 + + + 16th AIAA Aviation Technology, Integration, and Operations Conference + Washington, D.C + + American Institute of Aeronautics and Astronautics + June 2016 + + + + Chung, W., G. Chadchad, and R. Hochstetler, "An Integrated Gate Turnaround Management Concept Leveraging Bif Data/Analytics for NAS Performance Improvements," AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, 13-17 June 2016, Washington, D.C. + + + + + Towards a Fast-time Simulation Analysis of Benefits of the Spot and Runway Departure Advisor + + RobertWindhorst + + 10.2514/6.2012-4975 + + + AIAA Guidance, Navigation, and Control Conference + Minneapolis, Minnesota + + American Institute of Aeronautics and Astronautics + August 2012 + + + + Windhorst, R., "Towards a Fast-time Simulation Analysis of Benefits of the Spot and Runway Departure Advisor," AIAA Guidance, Navigation, and Control Conference, 13-16 August 2012, Minneapolis, Minnesota. + + + + + Benefits Assessment of a Surface Traffic Management Concept at a Capacity-Constrained Airport + + KatyGriffin + + + AdityaSaraf + + + PeterYu + + + StevenStroiney + + + BenjaminLevy + + + GustafSolveling + + + John-PaulClarke + + + RobertWindhorst + + 10.2514/6.2012-5533 + + + 12th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference and 14th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference + Indianapolis, Indiana + + American Institute of Aeronautics and Astronautics + September 2012 + + + + Griffin, K. J., A. Saraf, P. Yu, S. R. Stoiney, B. S. Levy, G. Solveling, J. Clarke, and R. D. Windhorst, "Benefits Assessment of a Surface Traffic Management Concept at a Capacity-Constrained Airport," AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, 17-19 September 2012, Indianapolis, Indiana. + + + + + Analysis of Airport Surface Schedulers Using Fast-time Simulation + + JustinVMontoya + + + RobertDWindhorst + + + SteveStroiney + + + KatyGriffin + + + AdityaSaraf + + + ZhifanZhu + + + SergeiGridnev + + 10.2514/6.2013-4275 + + + 2013 Aviation Technology, Integration, and Operations Conference + Los Angeles, California + + American Institute of Aeronautics and Astronautics + August 2013 + + + + Montoya, J., R. Windhorst, Z. Zhu, S. Gridnev, K. J. Griffin, A. Saraf, and S. Stroiney, "Analysis of Airport Surface Schedulers Using Fast-time Simulation," AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, 12-14 August 2013, Los Angeles, California. + + + + + Recommendations for NextGen Airport Surface Traffic Scheduling Algorithms: A Fast-time Simulation-based Perspective + + AdityaSaraf + + + KatyGriffin + + + SteveStroiney + + + RobertDWindhorst + + + ValentinoFelipe + + 10.2514/6.2013-4276 + + + 2013 Aviation Technology, Integration, and Operations Conference + Los Angeles, California + + American Institute of Aeronautics and Astronautics + August 2013 + + + + Saraf, A., K. Griffin, S. Stroiney, V. Felipe, and R. Windhorst, "Recommendations for NextGen Airport Surface Traffic Scheduling Algorithms: A Fast-time Simulation-based Perspective," AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, 12-14 August 2013, Los Angeles, California. + + + + + Departure queue prediction for strategic and tactical surface scheduler integration + + ShannonZelinski + + + RobertWindhorst + + 10.1109/dasc.2016.7778082 + + + 2016 IEEE/AIAA 35th Digital Avionics Systems Conference (DASC) + Sacramento, California + + IEEE + September 2016 + + + + Zelinski, S. and R. Windhorst, "Departure Queue Prediction for Strategic and Tactical Surface Scheduler Integration," 35 th Digital Aviation Systems Conference (DASC), 25-29 September 2016, Sacramento, California. + + + + + Assessing tactical scheduling options for time-based surface metering + + ShannonZelinski + + + RobertWindhorst + + 10.1109/dasc.2017.8101901 + + + 2017 IEEE/AIAA 36th Digital Avionics Systems Conference (DASC) + St. Petersburg, Florida + + IEEE + September 2017 + + + + Zelinski, S. and R. Windhorst, "Assessing Tactical Scheduling Options for Time-Based Surface Metering," 36 th Digital Aviation Systems Conference (DASC), 17-21 September 2017, St. Petersburg, Florida. + + + + + Performance evaluation of the approaches and algorithms using Hamburg Airport operations + + ZhifanZhu + + + NikolaiOkuniek + + + IngridGerdes + + + SebastianSchier + + + HanbongLee + + + YoonJung + + 10.1109/dasc.2016.7778081 + + + 2016 IEEE/AIAA 35th Digital Avionics Systems Conference (DASC) + Sacramento, California + + IEEE + September 2016 + + + + Zhu, Z., N. Okuniek, I. Gerdes, S. Schier, H. Lee, and Y. Jung, "Performance Evaluation of the Approaches and Algorithms Using Hamburg Airport Operations," 35 th Digital Aviation Systems Conference (DASC), 25-29 September 2016, Sacramento, California. + + + + + Operational Characteristics Identification and Simulation Model Verification for Incheon International Airport + + YeonjuEun + + + DaekeunJeon + + + HanbongLee + + + ZhifanZhu + + + YoonCJung + + + MyeongsookJeong + + + HyounkyongKim + + + EunmiOh + + + SungkwonHong + + + JunwonLee + + 10.2514/6.2016-3161 + + + 16th AIAA Aviation Technology, Integration, and Operations Conference + Washington, D.C + + American Institute of Aeronautics and Astronautics + June 2016 + + + + Eun, Y., D. Jeon, H. Lee, Z. Zhu, Y. C. Jung, M. Jeong, H. Kim, E. Oh, S. Hong, and J. Lee, "Operational Characteristics Identification and Simulation Model Validation for Incheon International Airport," AIAA Aviation Technology, Integration, and Operations (ATIO) Conference, 13-17 June 2016, Washington, D.C. + + + + + A Concept for Integrated Arrival/Departure/Surface (IADS) Traffic Management for the + + SZelinski + + + RCoppenbarger + + + + Zelinski, S. and R. Coppenbarger, "A Concept for Integrated Arrival/Departure/Surface (IADS) Traffic Management for the + + + + + Airspace Technology Demonstration 2 (ATD-2) ConOps Synopsis + + Metroplex + + + + NASA Ames Research Center + + August 2015 + + + unpublished + Metroplex," Airspace Technology Demonstration 2 (ATD-2) ConOps Synopsis, NASA Ames Research Center, August 2015 (unpublished) + + + + + + diff --git a/file840.txt b/file840.txt new file mode 100644 index 0000000000000000000000000000000000000000..784896abd67658ccd38b1416e11b64dcd974e5f6 --- /dev/null +++ b/file840.txt @@ -0,0 +1,202 @@ + + + + +I. INTRODUCTIONNASA is committed to demonstrating an Integrated Arrival, Departure, Surface (IADS) concept at Charlotte Douglas International Airport (CLT) within the next few years [1].The surface scheduling capability of the proposed IADS system includes a strategic component to enable planning and collaborative decision making among airlines and air traffic managers, and a tactical component to enable flexibility and maximize efficiency and throughput.The strategic scheduler derives from the Surface-Collaborative Decision Making (S-CDM) concept's Departure Reservoir Management (DRM) capability [2].The tactical scheduler derives from a first-comefirst-served (FCFS) with heuristics version of NASA's Spot And Runway Departure Advisor (SARDA) [3,4].Thus far the strategic and tactical scheduling components have been developed separately and research is required to study how they will interact.The end goal of these schedulers is the same, to maximize both the efficiency and predictability of surface operations under various operational constraints.However, they use different metrics to drive their scheduling solutions to this shared goal.Whereas the tactical scheduler uses flight delay to drive its schedule, the strategic scheduler uses queue length.Here queue length can be defined as the number of aircraft using or waiting to use surface resources.As a first step toward integrating the tactical and strategic surface schedulers, this paper explores the effect of tactical scheduling on strategic scheduling.Both schedulers attempt to meter surface traffic by assigning target release times.A Target Off Block Time (TOBT) is the suggested time to release a departure for gate pushback.A Target Movement Area entry Time (TMAT) is the suggested time to release a departure from the spot and transition from the airline controlled ramp area of the surface to air traffic controlled taxiways.The tactical scheduler assigns TOBTs no more than 15 minutes in advance.TOBTs are updated every 10 seconds until they are within 2 minutes of current time, when they are frozen.When the strategic scheduler predicts runway queues will exceed an upper threshold, it assigns TMATs as much as 90 minutes in advance as part of a Departure Metering Program (DMP).An implemented DMP may be extended, or adjusted if predictions change.However, unlike tactical TOBTs, strategic individual TMATs are not as flexible to change in response to minor uncertainty.The envisioned integration of these schedulers expects that the strategic TMATs will be given to the tactical scheduler, which then assigns TOBTs designed to meet the strategic TMATs within a defined compliance time range.If successful, the strategic adds an element of stability and predictability to surface metering, enabling a collaborative decision making mechanism by which air carriers can negotiate on scheduling decisions (e.g.slot swapping) for equitability.Meanwhile, the tactical scheduler is still allowed to make small adjustments to maximize efficiency in response to uncertainty (e.g.pushback readiness, taxi-time, arrival operations).An important research question is how large the compliance time range surrounding strategic TMATs should be to allow tactical improvements.The answer to this question will be influenced not only by uncertainties in surface trajectory predictions, but by the inherent differences in how the strategic and tactical surface schedulers generate their respective metering times.This paper analyzes the effect of tactical scheduling on strategic scheduler predictability.Section II describes the simulation environment and strategic and tactical schedulers in detail.In section III, strategic queue predictions and target times are analyzed using three different queue definitions to drive the strategic scheduler.Conclusions are presented in section IV. +II. SIMULATION +A. Surface OperationsThe Surface Operations Simulator and Scheduler (SOSS) [5] is used to simulate surface operations at CLT both with and without tactical surface metering.In the tactical surface metering simulations, SOSS holds aircraft at the gate until their target gate departure times.Without surface metering, SOSS releases aircraft from the gate as soon as they are ready.SOSS simulates both departure and arrival aircraft movements within a network of nodes and links representing gates, ramp, taxiways, and runways.Aircraft are allowed to progress along predefined routes through the node-link network as long as they do not violate separations constraints specific to each part of the network.Although SOSS has the capability of modeling uncertainty in various areas of the airport surface, uncertainty was not modeled for the purposes of this analysis. +B. Traffic ScenarioA SOSS traffic scenario is a list of departure and arrival flight information including aircraft type, flight plan, runway and gate assignment, and time the flight enters the simulation.Arrival and departure flight plans include the Standard Terminal Arrival Route (STAR) or Standard Instrument Departure (SID) respectively.Arrivals enter the simulation at the final approach fix and departures enter the simulation at the gate at their published pushback time or p-time.A four-hour scenario was developed based on CLT surface surveillance data from March 11, 2016.This was a relatively high traffic volume day with low weather impact when CLT was operating in South flow most of the day.Surface surveillance data was used assign the actual gate and runway used by each flight.Nominal routings between gates and runways within the SOSS node-link network were generated using rules learned from CLT ramp controllers.Fig. 2 shows the demand for each runway as numbers of arrivals and departures per 15-minutes.As a central hub for American Airlines, CLT traffic is characterized by cycles of a departure push (departure demand peak) vacating the gates, followed closely by an arrival rush (arrival demand peak) occupying the gates, followed by a lull in operations when passengers transit between connecting flights.The four-hour scenario in Fig. 2 covers two such cycles.Analysis of strategic scheduler calculations with up to 90-minutes lookahead cannot begin until at least 90 minutes into the simulation.This fourhour scenario allows analysis of a complete departure-arrival cycle beginning at 120 minutes. +C. Strategic Surface SchedulerTable I summarizes the abbreviations used for times at surface points throughout this section.The strategic surface scheduler is modeled after the DRM, which uses a manually input target queue length per runway to calculate TMATs.The DRM periodically calculates and compares queue predictions to the target queue length, and recommends a DMP including TMATs for all flights within the DMP active time period, only when the queue is predicted to exceed the target queue length by a user defined amount.The strategic scheduler for this analysis differs from the DRM in that TOBTs are calculated rather than TMATs so that they may be compared with tactical scheduler generated TOBTs.The strategic TOBT esimate is simply TMAT minus unimpeded ramp transit time.Both strategic and tactical TOBT calculations begin by generating Predicted Take Off Times (PTOTs) and back calculating the target times to achieve the desired queue length.However, the DRM employs a more complex method of assigning PTOT slots FCFS by Initial Off Block Time to inhibit gaming as airlines update their Earliest Off Block Time (EOBT) estimates.For the purposes of this analysis, the calculation of PTOTs is made simpler by assuming EOBTs are not updated by the airlines.In this analysis, DMPs are not implemented.Therefore, the strategic scheduler continually updates both queue predictions and TOBTs every minute. +1) Queue PredictionThe strategic scheduler calculates a range of queue predictions from current time to 90-minute lookahead in oneminute intervals.These queue predictions are updated every minute using new current state information.Three different definitions of queue length for a given runway are considered.Queue q1 is the number of active aircraft on the ramp and taxiways.Queue q2 is the number of aircraft on just the taxiways.Queue q3 is the number of aircraft on the taxiways waiting in line to use the runway.Note that the flights in q3 are a subset of q2, which are a subset of q1.For all queue definitions, an aircraft exits the queue at its Actual Take Off Time (ATOT).Actual Queue Entry Times (AQETs) for each queue definition are as follows.A flight's AQET1 is when it is released from departure gate, also know as the Actual Off Block Time (AOBT).A flight's AQET2 is when it is released from the spot and transitions from the ramp to the active movement area (taxiways), also known as the Actual Movement Area entry Time (AMAT).A flight's AQET3 is when it has exceeded its unimpeded taxi out time since entering the active movement area.This is equal to the flight's last updated Earliest Take Off Time (ETOT) before AMAT.The queue prediction method begins by calculating Earliest Queue Entry Times (EQETs) and Predicted Take Off Times (PTOTs) for each flight.For all flights currently at the gate: EQET1 equals the EOBT, which is the maximum of the flight's p-time and the current time; EQET2 equals the Earliest Movement Area entry Time (EMAT) projected from EOBT; and EQET3 equals the ETOT projected from EMAT.For all flights currently on the ramp, AQET1 is used instead of EQET1 and EQET2 is the updated EMAT projected from current position.For all flights currently on the taxiways, AQET2 is used instead of EQET2.EQET1 and EQET2 are updated every minute as EOBT and EMAT are updated.Whereas EQET3 is updated when the flight is still at the gate or in the ramp, it is not updated once the flight enters the active movement area, and this last EQET3 will in fact become AQET3 when current time catches up to it.Fig. 3 shows how EQETs update until they become AQETs as the flight transitions from gate to ramp to taxi.Notice how when the flight is in taxi, ETOT continues to update, whereas AQET2 remains frozen from the time the flight entered the active movement area (AMAT).Flight PTOTs are calculated by scheduling flights to the runway in FCFS fashion, taking into account departure runway spacing constraints.The FCFS sequence is determined by first sorting all flights by ETOT and modifying to ensure that precedence constraints are met.For example, a fast aircraft lined up behind a slow aircraft on the same taxiway may have an earlier ETOT, but the slow aircraft has precedence and must be scheduled first because it is physically impossible for the fast aircraft behind it to pass on the taxiway.Departure runway spacing constraints include wake vortex separation, departure fix miles-in-trail, and runway departure rate.The strategic scheduler does not consider arrival specific spacing constraints (imposed by intersecting or converging arrival runways, or arrivals landing on or crossing the scheduled runway) because accurate arrival runway assignments and landing times are not expected to be available to the strategic surface scheduler two hours in advance.Arrivals are instead accounted for with runway departure rate spacing constraints.Departure rate is envisioned to be a user defined input to the strategic scheduler manually updated by the Departure Reservoir Coordinator as airport configuration and other surface conditions change.When arrivals are expected to impact departure operations on a particular runway, a lower departure rate would be entered, which would increase the minimum inter-departure spacing constraint.The current queue lengths per runway, are the numbers of flights that have an AQET but no ATOT (entered but not yet exited the queue).Predicted queue lengths per runway are calculated by sorting all EQETs and PTOTs by increasing time.Then for each EQET and PTOT, the current queue length is incremented or decremented respectively, and set for the +2) Target Off Block Time CalculationStrategic TOBTs are back calculated from Target Queue Entry Times (TQETs) designed to maintain the target queue length.For a target queue length Q, the TQET of the ith flight in runway takeoff sequence is equal to the PTOT of the (i-Q)th flight in sequence.If the (i-Q)th flight has already exited the queue (taken off), then the TQET of the ith flight equals its EQET.Unimpeded transit time is then subtracted from TQET to get TOBT, depending on the queue definition.TOBT1 is equal to TQET1.Unimpeded ramp transit time is subtracted from TQET2 to get TOBT2.Both unimpeded ramp transit and taxi times are subtracted from TQET3 to get TOBT3. +D. Tactical Surface SchedulerThe tactical surface scheduler is modeled after a FCFS version of SARDA.As with the strategic scheduler, TOBT calculation begins with assigning PTOTs to flights.Like the strategic scheduler, the tactical scheduler sequences flights for scheduling by sorting all flights by ETOT and modifying it to ensure that precedence constraints are met.Whereas the strategic scheduler considers only departures, the tactical scheduler also considers arrivals landing on intersecting and converging runways, as well as arrivals landing on or taxing across the scheduled runway.Arrival landing times are treated as hard constraints, but arrivals crossing the scheduled runway are sequenced along with the departures and scheduled in turn.In addition, the tactical scheduler attempts to resolve gate conflicts by prioritizing flights that would otherwise block the gate for incoming arrivals.As the tactical scheduler explicitly considers arrival specific runway spacing constraints, it does not use departure rate as a constraint.After the tactical scheduler calculates PTOTs, it calculates a TOBT for each flight by subtracting unimpeded transit time (through ramp and taxi) and a delay buffer from its PTOT.The delay buffer varies linearly with the unimpeded transit time such that the buffer equals a constant c multiplied by the unimpeded transit time.The constant c is tuned to a particular airport configuration to provide just enough delay buffer to keep the runways from going dry.A delay buffer of 3% was used for this study because this is was the buffer used in recent human-in-the-loop SARDA simulations of CLT [3].The tactical scheduler calculates TOBTs for all flights with EOBT within 15 minutes of current time and updates every 10 seconds.When a flight's TOBT is within 2 minutes of current time, it is frozen and no longer updated. +E. Strategic-Tactical Scheduler InteractionThis study models the strategic scheduler in predictive mode, so there is no information passed from the strategic to the tactical scheduler.However, when the tactical scheduler assigns a TOBT to a flight, this tactical TOBT updates the EOBT the strategic scheduler uses to calculate queue predictions and strategic TOBTs.To facilitate analysis of open-loop strategic TOBTs with as little tactical information as possible, the strategic scheduler still calculates TOBTs for flights with tactical TOBTs.Final implementation of the strategic scheduler will likely directly use tactical TOBTs and other tactical scheduling information rather than recalculating them, but this is left for future study. +III. RESULTSSeveral analyses were completed using the output of two SOSS simulations, one with and one without tactical surface metering, hereafter referred to as the "Tactical" and "Baseline" simulations, respectively.First the tactical scheduler performance results are presented, followed by strategic scheduler performance analyses of queue and TOBT prediction. +A. Tactical Scheduler PerformanceFig. 5 compares taxi times and delays between the baseline and tactical scheduling simulations.The boxes and whiskers show quartiles and means for all individual flight taxi times (transit time from gate to runway) and delays within each simulation.The tactical scheduler substantially reduces departure taxi time without affecting arrival taxi time.It does this by moving most of the taxi delay to the gate where flights may wait with engines off, burning less fuel and producing less emissions.Not only is taxi delay reduced, but it is far more consistent, improving predictability and enabling the capability to coordinate in an IADS context.Fig. 6 compares the queue lengths at runway 18C between simulations.Queue q1 is largest for both Baseline and Tactical simulations because this counts all flights in the ramp and taxiways, whereas q2 and q3 count only flights in the taxiways.Similarly, q2 is always greater than or equal to q3 as q3 is the subset of q2 flights that have exceed their unimpeded taxi time.The similarity between q2 and q3 indicates that most flights in the taxiways had exceeded their unimpeded taxi time.Compared to Baseline, the Tactical simulation substantially reduces all queue lengths by holding flights at the gate and delaying their queue entry times.However, Tactical simulation queue lengths are still quite erratic because the tactical scheduler does not target a desired queue length with it's method of departure metering.Both simulations' departure rates are substantially lower than demand during the demand peaks, pushing more operations to later time bins as flights are delayed.The dips in departure rate at 60 and 180 minutes are due to arrival demand peaks interfering with departure operations.Most of the time, the tactical scheduler is able to produce a similar departure rate to that of the baseline except during the last hour when tactical scheduler departure rates are lower. +B. Queue Prediction AnalysisDeparture rate is envisioned to be a user defined input to the strategic scheduler manually updated by the Departure Reservoir Coordinator as airport configuration and other surface conditions change.The departure rate is used as a flexible calibration factor to compensate for the lack of arrival runway usage rather than an explicit value.This analysis compares queue prediction accuracy between calibrated static and variable departure rate inputs.Root mean square error was chosen to measure prediction accuracy.Let e(i,j) be the queue prediction error for time i at a lookahead time of j minutes.e(i,j) = q(i,j) -q(i,0)(where q(i,0) is the actual queue length at time i (i.e. the lookahead time is 0) and q(i,j) is the queue prediction for time i at a lookahead time of j minutes.Let σ(i,*) be the root mean square for all lookahead times between 0 and 90 minutes for queue prediction errors of time i.𝜎(𝑖, * ) = 𝑒(𝑖, 𝑗) ! /90 !" !!!(2)Let σ(*, j) be the root mean square for all queue prediction errors between time 120 and 240 at a lookahead time of j minutes. +𝜎( * , 𝑗) =(, ) !/120!"# !!!"#(3)Let σ(*,*) be the overall root mean square for all queue prediction errors between time 120 and 240 for all lookahead times between 0 and 90 minutes. +𝜎( * , * ) = , !/(120 * 90)!!!"#,!!!" !!!"#,!!!(4)Static departure rates assign a single departure rate for each runway that is used to set minimum separations for the strategic scheduler PTOTs used to calculate queue predictions.A range of static departure rates was tested in increments of 0.1 departures per quarter hour.For each runway, the departure rate that minimized σ(*,*) was selected.The calibrated static departure rates were found to be 7.3 and 7.6 departures per quarter hour for runways 18C and 18L respectively.Variable departure rates use a unique departure rate per runway per quarter hour.Variable departure rates were modeled after the simulated number of departures per quarter hour (as seen in Fig. 7 for 18C and similarly for 18L) plus a constant offset value.A range of offset values was tested.For simulation time (minutes) operations per 15-minutes each runway, the offset value that minimized σ(*,*) was selected.The calibrated offset values applied to the variable departure rates were found to be 1.7 and 1.3 departures per quarter hour for runways 18C and 18L respectively.Because these variable rates were derived from simulated departure rates after the fact, they represent a zero uncertainty prediction, setting an upper limit of expectations for departure queue predictability using variable departure rates.Fig. 8 shows σ(*,*) for all three queue definitions for each runway using static and variable departure rate modes.The improvement in queue prediction of variable over static departure rate mode is substantial.Not only does the use of variable departure rates reduce the overall root mean square error for both runways, but the errors are more similar between runways.Note that there is very little difference in predictability between queue definitions.As Fig. 6 shows, q1 is always larger than the other queue definitions.Therefore, one could argue that the predictability normalized by average queue length is actually best for q1.Either way, the remaining comparisons focus on q1 as the results are very similar between queue definitions.Figs. 9 and 10 show σ(*, j) and mean error, respectively, as a function of lookahead time for each runway and departure rate mode.It is interesting to note that in all cases, the prediction root mean square error peaks around 30 minutes of lookahead time and then reduces, rather than continuing to increase with lookahead.The mean error also undulates with lookahead time and the undulations differ between runways.These root mean square and mean error effects are likely due to the dramatic peaks and valleys of the traffic demand.The use of variable departure rate mitigates these effects but does not eliminate them completely.Fig. 11 and 12 show static and variable departure rate σ(i,*), mean error and actual queue length as a function of prediction time for runways 18C and 18L respectively.For each runway, the σ(i,*) trends for static and variable departure rates are similar, although the variable departure rate errors are lower.In both cases, the σ(i,*) values appear to be high and mean error low (under predicted queue length) when the actual queue length is high.0" 0.18C"Sta-c" 18C"Varied" 18L"Sta-c" 18L"Varied"Queue"1" Queue"2" Queue"3" runway and departure rate mode overall root mean square error (number of flights)0" 15" 30" 45" 60" 75" 90"18C"Sta0c" 18C"Varied" 18L"Sta0c" 18L"Varied" lookahead time (minutes) root mean square error (number of flights) It is clear that variable departure rate provides substantial improvement to queue prediction over static departure rates.However, the variable departures rates in this analysis were modeled after actual departure rates with no uncertainty.It is unclear if manually updated departure rates will enable the strategic scheduler to achieve this level of queue prediction accuracy and stability.Future research should explore automated methods of departure rate prediction to feed queue prediction.Future research should also explore incorporating any arrival data available in the strategic time frame directly into the strategic scheduler rather than relying on solely departure rate to approximate arrival impact.!1# !0.5# 0# 0.5# 1# 1.5# 2# 0# 15# 30# 45# 60# 75# 90# 18C#Sta2c# 18C#Varied# 18L#Sta2c# 18L#Varied#When a similar analysis of queue predictability is performed on the simulation with tactical scheduling, queue prediction overall root men square error is nominally doubled.Fig. 13 shows how when lookahead time exceeds 15 the mean error rises rapidly and continues to over predict queue length.This is expected because the tactical scheduler actively meters the departures to lower queue lengths as seen in Fig. 6.Beyond 15 minutes lookahead, the strategic scheduler does not have the benefit of knowing what if any gate holds (TOBT) the tactical scheduler will assign to flights.This is why the next section analyzes TOBT predictability as a measure of strategic scheduler performance in the presence of tactical scheduling. +C. TOBT Prediction AnalysisIn the simulation with tactical scheduling, flights push back from the gate according to the last TOBT assigned by the tactical scheduler such that AOBT is equal to this final tactical TOBT.Strategic TOBT error for a given flight can then be measured relative to AOBT.A flight's TOBT error can also change with time as strategic queue predictions and TOBT calculations are updated prior to AOBT.To facilitate an analysis of time varying TOBT error that is comparable between simulations with and without tactical scheduling, the flight's EOBT was chosen as a reference time for TOBT lookahead.Let ε(f,j) be the TOBT error for flight f calculated at a lookahead time of j minutes prior to its EOBT.ε(f,j) = TOBT(f,j) -AOBT(f)(5)As with the queue prediction analysis, root mean square error was chosen to measure TOBT predictability and calibrate target queue lengths that best align the strategic and tactical schedulers.Let * , be the root mean square error for all flights with EOBT between 120 and 240 minutes calculated with a lookahead time of j minutes prior to EOBT.𝜌( * , 𝑗) = 𝜀(𝑓, 𝑗) ! /𝑁 !"#$ ! !!"# !"#$ ! !!!" (6)where N is the number of flights for which EOBT is between 120 and 240 minutes into the simulation.Whereas overall root mean square error was used to calibrate static departure rates and variable departure rate offsets for queue prediction, the effects of tactical scheduling on TOBT prediction error with respect to lookahead time make overall root mean square error less suitable for calibrating target queue lengths for strategic TOBT calculation.Therefore, * ,30 was used to calibrate target queue lengths, which will now be shown.To illustrate the target queue length calibration process, Figs. 14 and 15 show * , and mean TOBT prediction error, respectively, as a function of lookahead time for a sample range of target queue lengths (0-6) for runway 18C using static departure rate and queue 1.Notice how both TOBT prediction error metrics show a distinct change in behavior around 15 minutes lookahead time, which is also the tactical scheduler planning horizon with respect to EOBT.At 15 minutes lookahead, the tactical scheduler assigns tactical TOBTs to flights, updating their EOBTs and trajectory predictions used for strategic queue prediction and TOBT calculation.Both strategic TOBT prediction error metrics stabilize around 30 minutes lookahead, which is why 30 minutes lookahead was chosen to calibrate target queue lengths.In Figs. 14 and 15 the target queue length with minimum root mean square error at 30 minutes lookahead is highlighted in blue.As expected, target queue lengths are greatest for q1 which counts the total number of flights active in ramp or taxi, and become smaller for q2 as the number of flights in ramp are removed, and smallest for q3 which focuses on the number of flights waiting in line for the runway.It is interesting to note that the calibrated target queue lengths are slightly smaller than average actual queue lengths produced by the tactical scheduler seen in Fig. 6.This may be because congestion related surface delays in the ramp and taxiways inflate the actual queues higher than the strategic scheduler predictions using unconstrained ramp and taxi transit times.In general the target queue lengths are small for the purposes of initiating and adjusting DMPs, which is the strategic schedulers intended function.For example, the DRM looks for queue predictions dipping below a user defined lower threshold to indicate the need for DMP compression.This would be impossible for the q3 target queue lengths of zero.In actual implementation, the strategic scheduler may be used more conservatively with higher target queue lengths to manage the level of residual delay the tactical scheduler will need to apply.Whereas mean TOBT prediction errors for most queue definitions are less than 1 minute, all root mean square errors are quite large (between 12 and 15 minutes).This is due to the tendency of the tactical scheduler to resequence flights at the runways in response to interactions with arrivals.This is especially true where gate conflicts between departure and arrivals are concerned.As the departure demand increases requiring more average gate delay, the likelihood of departures predicted to have a gate conflict with an arrival increases.This in turn increases the likelihood of early pushback to resolve gate conflicts changing the scheduled runway departure sequence.This tactical scheduler response to gate conflicts not only compounds tactical/strategic scheduler deviation, but leaves the system at risk to gaming if an airline should intentionally assign arrival gates to promote priority treatment of its departures.The tactical scheduler may be modified to plan for some delay to be absorbed in the ramp either by queuing at the spot or waiting in hard stands in response to gate conflicts rather than shifting to the delay to another flight and resequencing flights for runway usage. +IV. CONCLUSIONThis study analyzed strategic surface scheduler predictability to facilitate future integration with tactical scheduling.Queue prediction accuracy was used to measure strategic predictability in a simulation of CLT surface operations without tactical scheduling.The use of variable departure rates as a strategic scheduler input was shown to substantially improve queue predictions over static departure rates.This illustrates the importance of incorporating the anticipated effects of arrival operations on departures when scheduling without explicit arrival information.A comparison of strategic assigned target gate push back times with actual gate pushback times was used to measure strategic predictability in a simulation with tactical scheduling.Whereas the strategic scheduler can be tuned to predict average delays similar to the tactical scheduler by calibrating target queue lengths, it is very difficult for the strategic scheduler to accurately predict individual flight delays assigned by tactical scheduler.is due to the difference in flight sequence at the runway between the tactical and strategic schedulers.Reducing the effect of tactical departure resequencing in response to gate conflicts with arrivals is expected to reduce tactical/strategic scheduler deviation in future studies.Fig. 1 shows a map view of the SOSS CLT surface model, highlighting characteristics of South flow operations.Runways are labeled in light blue and yellow to indicate their predominant use for arrivals and departures respectively.When departure demand exceeds runway capacity, departure queues form along the taxiways to the right of 18C and to the left of 18L, indicated by green arrows.Crossing light blue and yellow arrows highlight runway constraints where arrivals impact departure operations and consequently departure queue length.Intersecting runways 18L and 23 cause arrivals on 23 to impact departures on 18L.Converging runways 18C and 23 cause arrivals on 23 to impact departures on 18C.Arrivals on 18R must cross 18C as they taxi to the gates, impacting departures on 18C. +Fig. 1 .1Fig. 1.CLT surface model. +Fig. 2 .2Fig. 2. Runway demand for the four-hour scenario. +Fig. 3 .3Fig. 3. Queue entry time updates. +or PTOT.The predicted queue length is then retrieved from the list of EQETs and PTOTs and recorded every 60 seconds out to 90 minutes from current time. +Fig. 44Fig. 4 diagrams how information is passed among SOSS and the tactical and strategic schedulers.Notice how information and parameters flow only into the strategic scheduler.Its outputs are metrics for analysis. +Fig. 4 .4Fig. 4. Simulation information flow. +Fig. 5 .5Fig. 5. Simulation taxi times and delays. +Fig. 6 .6Fig. 6.Simulation queue lengths at 18C. +Fig. 77Fig. 7 compares the departures per quarter hour at runway 18C between simulations. +Fig. 7 .7Fig. 7. Simulation demand and departure rates at 18C. +Fig. 8 .8Fig. 8. Queue prediction overall root mean square error for the Baseline simulation. +Fig. 9 .9Fig. 9. Queue prediction root mean square error vs. lookahead time for the baseline simulation. +Fig. 10 .10Fig. 10.Queue prediction mean error vs. lookahead time for the baseline simulation. +Fig. 11 .11Fig. 11.Queue prediction root mean square and mean error vs. prediction time for runway 18C for the baseline simulation. +Fig. 12 .12Fig. 12. Queue prediction root mean square and mean error vs. prediction time for runway 18L for the baseline simulation. +Fig. 13 .13Fig. 13.Queue prediction mean error vs. lookahead time for the tactical scheduling simulation. +Fig. 14 .14Fig. 14.Sample TOBT prediction root mean square errors vs. lookahead time for the tactical scheduling simulation. +Fig. 15 .15Fig. 15.Sample TOBT prediction mean errors vs. lookahead time for the tactical scheduling simulation. +Figs. 1616Figs. 16, 17, and 18 show the calibrated target queue lengths and associated root mean square and mean error, respectively, for each queue definition and runway/departure rate mode. +Fig. 16 .16Fig. 16.Calibrated target queue lengths for the tactical scheduling simulation. +Fig. 17 .17Fig. 17.TOBT prediction root mean square error at 30 minutes lookahead for calibrated target queue lengths for the tactical scheduling simulation. +Fig. 18 .18Fig. 18.TOBT prediction mean error at 30 minutes lookahead for calibrated target queue lengths for the tactical scheduling simulation. +TABLE I .IABBREVIATIONS FOR TIMES AT SURFACE POINTS +Abbreviation Prefix = Time Type Abbreviation Suffix = Surface Point18R18C taxiway runway crossings taxiwa ysgates departure ramp queue18L intersecting runways23operations per 15-minutes0" 5" 10" 15" 20"0"15"runway usage time bin (15-minutes) 30" 45" 60" 75" 90" 105" 120" 135" 150" 165" 180" 195" 18C"Arr" 18C"Dep" 210" 225" 240" 18L"Arr" 18L"Dep" 18R"Arr" 23"Arr"convergingrunwaysAActualOBTOff Block TimeEEarliestMATMovement Area entry TimePPredictedQETQueue Entry TimeTTargetTOTTake Off Time + + + + +ACKNOWLEDGMENTThis research is funded by NASA's Airspace Operations and Safety Program in support of the Airspace Technology Demonstration 2 (ATD-2) subproject. + + + + + + + + + A Concept for Integrated Arrival/Departure/Surface (IADS) Traffic Management for the Metroplex + + SZelinski + + + RCoppenbarger + + + + Airspace Technology Demonstration + + 2 + 2 + August 2015 + ConOps Synopsis + + + NASA Ames Research Center + + + unpublished + S. Zelinski and R. Coppenbarger, "A Concept for Integrated Arrival/Departure/Surface (IADS) Traffic Management for the Metroplex," Airspace Technology Demonstration 2 (ATD-2) ConOps Synopsis, NASA Ames Research Center, August 2015, unpublished. + + + + + An Integrated Collaborative Decision Making and Tactical Advisory Concept for Airport Surface Operations Management + + GautamGupta + + + WaqarMalik + + + YoonJung + + 10.2514/6.2012-5651 + + + 12th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference and 14th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference + + American Institute of Aeronautics and Astronautics + June 2014 + + + unpublished + FAA Air Traffic Organization Surface Operations Office Directorate "U.S. Airport Surface Collaborative Decision Making (CDM) Concept of Operations (ConOps) in the Near-Term: Application of the Surface Concept at United States Airport," June 2014, unpublished. + + + + + Evaluation of Pushback Decision-Support Tool Concept for Charlotte Douglas International Airport Ramp Operations + + MHayashi + + + THoang + + + YCJung + + + WMalik + + + HLee + + + VLDulchinos + + + + Europe Air Traffic Management R&D Seminar + + June 2015 + Lisbon, Portugal + + + Tenth USA + M. Hayashi, T. Hoang, Y. C. Jung, W. Malik, H. Lee, and V. L. Dulchinos, "Evaluation of Pushback Decision-Support Tool Concept for Charlotte Douglas International Airport Ramp Operations," Tenth USA/Europe Air Traffic Management R&D Seminar, Lisbon, Portugal, June 2015. + + + + + Exact and Heuristic Algorithms for Runway Scheduling + + WaqarMalik + + + YoonCJung + + 10.2514/6.2016-4072 + + + 16th AIAA Aviation Technology, Integration, and Operations Conference + Washington, D. C. + + American Institute of Aeronautics and Astronautics + June 2016 + + + W. A. Malik and Y. C. Jung, "Exact and Heuristic Algorithms for Runway Scheduling," AIAA Aviation Confrence, Washington, D. C., June 2016. + + + + + Validation of Simulations of Airport Surface Traffic with the Surface Operations Simulator and Scheduler + + RobertDWindhorst + + + JustinVMontoya + + + ZhifanZhu + + + SergeiGridnev + + + KatyGriffin + + + AdityaSaraf + + + SteveStroiney + + 10.2514/6.2013-4207 + + + 2013 Aviation Technology, Integration, and Operations Conference + Los Angeles, California + + American Institute of Aeronautics and Astronautics + August 2013 + + + R. D. Windhorst, J. V. Montoya, Z. Zhu, S. Gridnev, K. Griffin, A. Saraf, and S. Stroiney, "Validation of Simulations of Airport Surface Traffic with the Surface Operations Simulator and Scheduler," 13 th Aviation Technology, Integration, and Operations (ATIO) Conference, Los Angeles, California, August 2013. + + + + + + diff --git a/file841.txt b/file841.txt new file mode 100644 index 0000000000000000000000000000000000000000..2d52c96e7f9666f7488500680f0cf7566f3b507b --- /dev/null +++ b/file841.txt @@ -0,0 +1,203 @@ + + + + +I. IntroductionTraffic Management Units (TMUs) found in US Air Route Traffic Control Centers (ARTCCs) monitor flight traffic, weather patterns, and airspace and route status, identify potential problems to traffic flow, and respond with strategic and tactical adjustments to the flow.TMU responsibilities include very complex and time-intensive interactions with multiple facilities, area supervisors, and the control room floor.Due to constantly changing weather and traffic demand, there is often a small window of opportunity in which adjustments can be implemented without causing unforeseen complications.Adjustments made farther upstream are often preferable, resulting in the least impact on flight time and distance.However, the current practice tends toward postponing the adjustment decision and letting the scenario play out until the potential problem is more certain due to time-consuming verbal coordination required across facilities.The FAA's Optimized Route Capability (ORC) concept is envisioned to assist TMUs by conducting much of the laborious data gathering and interpretation necessary to identify, evaluate, and respond to changing conditions. 1,2 e ORC concept is envisioned to interpret and integrate disparate data sources and consider multiple objectives (e.g.weather avoidance, competing flow demands, workload considerations) when generating optimal airspace and routing configurations that allow for timely and seamless coordination beyond the immediate and adjoining control facilities.This concept requires ORC to provide reroute recommendations for individual flights when projected demand exceeds capacity in defined areas of the National Airspace System (NAS).ORC is also required to take into account the effect of its recommendations on metrics such as extra track distance, delay, and resource loading that are currently either unavailable or too difficult for a traffic management coordinator to discern.This data-driven approach to formulating routing solutions is envisioned to facilitate cross facility coordination, allowing for timely and seamless implementation of changes.The original research assessment identified Metroplex environments, where there are many airport interactions, high traffic volume, and complex resource-utilization challenges, as particularly well-suited for the ORC concept. 1 The concept not only provides automation support to identify alternate arrival and departure routes to underutilized resources, but to dynamically utilize airspace nominally reserved for segregating arrivals and departures or flows to or from other airports.NASA has recently teamed with the FAA to facilitate initial algorithm development.This initial phase of ORC focuses on intelligent offloading of congested arrival routes into high density airports. 3High demand at a few arrival meter fixes can cause unacceptable arrival scheduling delay to be passed from terminal to en-route airspace, while other arrival meter fixes to the same Terminal Radar Approach Control (TRACON) have available capacity.ORC will monitor projected arrival scheduling delay at each arrival meter fix.If excessive delay is predicted, ORC evaluates the effects of available reroute combinations from the over-utilized arrival meter fix to those that are less utilized, and selects a relatively low-cost set of individual flight reroutes to recommend to the TMU for implementation.This paper describes the ORC concept for intelligent offloading of congested arrival routes (hereafter referred to simply as ORC).The initial ORC algorithm and it's implementation in a fast-time simulation model of Houston International Airport (IAH) arrival operations is presented.A fast-time simulation of an aggressive arrival rush into IAH's Northeast corner post meter fix is conducted to exercise and stress the algorithm and provide proof-of-concept.An analysis of ORC algorithm behavior is presented along with a discussion of possible directions for algorithm improvement. +II. ORC Intelligent Offloading ConceptOffloading congested arrival routes is a function currently performed for only airborne flights, by necessity, and in a reactive manner.When it is apparent to a traffic management coordinator (TMC) in the ARTCC TMU that a group of airborne flights will overload an arrival fix to an extent that en-route vectoring to deliver a manageable feed to terminal arrival controllers will become unmanageable, the TMC selects flights to reroute to other under-utilized arrival fixes.However, the problem is usually recognized so late that only the last few flights in the group can be rerouted en-route, leaving the remainder of the problem to be handled tactically as transition airspace reroutes or TRACON vectors.This not only utilizes less efficient trajectory options, but puts additional strain on the already overloaded en-route arrival sectors and TRACON.Predeparture aircraft are not generally considered for reroutes to alternate meter fixes.Rather, they are delayed on the ground until a slot opens up in the congested arrival stream, which may take far longer than rerouting to an alternate meter fix with acceptable additional fuel burn.Figure 1 depicts a notional representation of offloading congested arrival routes.Four orange dots depict arrival meter fixes.Blue lines feeding into the meter fixes are Standard Terminal Arrival Routes (STARs).Triangles represent arrival flights.Due to the abundance of flights in the Northwest, the orange highlighted flight is projected to have excessive meter fix delay indicating meter fix overload.Flights originally bound for the overloaded meter fix are highlighted in green.Several have the option to reroute to an alternate meter fix, shown as dashed orange lines to STAR transitions to neighboring meter fixes.After analyzing route efficiency and scheduling costs associated with these available reroute options, ORC selects a subset of these reroutes to display in the TMU as suggested reroutes to alleviate congestion.One of the challenges associated with the earlier ORC concept was a lack of reliable capacity estimation capabilities to identify problems. 4Flow capacity has typically been managed by placing a miles-in-trail (MIT) restriction at a metering point along the flow.But this method of defining flow capacity requires an operations expert to periodically evaluate the situation and adjust the capacity.This often results in inaccurate or excessive Traffic Management Initiatives.ORC's new focus on arrival route congestion enables it to adopt a TBFM approach to capacity driven by arrival scheduling.a ORC monitors TBFM arrival schedules 90 minutes in advance, leaving ample opportunity for strategic reroutes with less impact to flight time and distance.Because of the uncertainty associated with estimating arrival schedules 90 minutes in advance, a user-defined threshold for projected meter fix delay is used to predict arrival route congestion.Even when arrival scheduling is not being used as intended for terminal metering, TMCs use large meter fix delay estimates as indicators of arrival route congestion requiring their attention.IAH TMCs identified seven minutes as the amount of delay seen on the TBFM time-line they associate with the need for rerouting or passing flow restrictions upstream via Traffic Management Initiatives.When ORC predicts an arrival fix overload (flight meter fix delay exceeding seven minutes for IAH), ORC reroute options for all flights bound for the overloaded fix (including pre-departure aircraft within the 90-minute planning horizon) are filtered to determine reroute availability for each individual flight.All possible combinations of available individual flight reroutes are then prioritized and evaluated to select a solution that provides acceptable arrival fix delay with minimal impact to flight operators.ORC then displays the suggested set of individual flight reroutes to TMCs along with the cost metrics used to select them to help TMCs determine whether to implement the reroutes as suggested, modify them and recalculate their estimated effects, or ignore them. +III. ORC Automation FunctionsThe ORC concept relies on the integration of its own decision support automation with other automation such as the FAA's Traffic Flow Management System (TFMS) for trajectory and meter fix arrival time prediction and Time Based Flow Management (TBFM) for estimates of arrival scheduling delay pushed into en-route airspace.The concept is also dependent on the human TMC's ability to quickly evaluate and implement recommended individual flight reroutes.Figure 2 shows the control loop where ORC decision support automation functions (blue) interact with external automation functions (orange) and the TMC.The main focus of this paper is on the ORC decision support automation.Whereas envisioned inputs and outputs to external human and automation systems are discussed, the manner in which they interface is not discussed in detail.As shown in Fig. 2, ORC automation functions are broken down into four basic functions discussed in detail in the following subsections.These functions are repeated periodically as flight states and constraints are updated.Predict Arrival Route Congestion monitors estimated flight delay at the meter fix from the TBFM arrival scheduler.ORC continues to monitor delay through each update cycle until arrival route congestion is predicted.Determine Reroute Availability filters all flights to just those that currently have a feasible reroute to another meter fix available.Evaluate Reroute Combinations prioritizes the list of available reroutes and probes the Trajectory Predictor and Arrival Scheduler functions to generate evaluation metrics for combinations of available reroutes.Finally, when a set of reroutes is found for which the evaluation metrics satisfy solution criteria, Recommend Reroute Solution displays the reroute solution to the TMC.The TMC may modify the solution (or come up with a completely different solution) and instruct the decision support automation to re-evaluate to see the effect of the new solution on evaluation metrics.It is envisioned that the chosen ORC solution will be displayed on the TFMS system for expedited coordination and delivery of the information.From the ORC algorithm perspective, reroutes take effect when inputs to Predict Arrival Route Congestion from the external Trajectory Predictor and Arrival Scheduler reflect the change in flight plans.a TBFM arrival scheduling is also known as terminal metering among air traffic operations experts. +III.B. Determine Reroute AvailabilityAll ORC reroutes are direct route segments from a flight's current position to transition fixes for existing STARs, streamlining the process of rerouteing to the alternate meter fix.However, not all direct routes from current position to STARs may be feasible.Therefore filters are used to generate a subset of available individual reroute options.At a minimum, meter fix and geographical filters are used to limit the available reroute options.The meter fix filter ensures that all flights eligible for ORC reroutes are currently assigned to a meter fix for which arrival route congestion has been predicted.This saves ORC the trouble of evaluating reroutes that do not offload the source of the predicted problem.The geographical filters attempt to address common operational constraints for reroutes with generic rules related to published adaptation data.Two different geographical filters were considered.First, flights were required to be within a +/-90 degree capture angle from the arrival transition leading to the alternate meter fix.This ensures that if a flight flew direct to the transition fix, it would need to turn no more than 90 degrees to join the STAR transition path once reaching the transition fix.Second, flights may not be within 5 minutes of crossing a Center boundary when rerouting.Operations experts insisted that they would not reroute a flight so close to hand-off between Centers due to the additional coordination required.This was enforced by filtering flights from reroute consideration if their projected dead reckoning trajectory (based on current position, heading, and speed) crossed a Center boundary within 5 minutes.Figure 3 illustrates the geographical filters applied to IAH.The orange dots are meter fixes and the blue lines are STAR transitions to their respective meter fixes.A sample transition fix used to enter the STAR is labeled for each meter fix.All flights shown in this example are approaching the Northeast meter fix assumed to be overloaded and in need of offloading.Flight 3 is projected to cross a Center boundary within 5 minutes and so it is filtered from reroute consideration.Flights 1 and 2 are not projected to cross a Center boundary within 5 minutes.The arrival procedure capture angle filter determines which reroute options are available to them.Blue shaded regions highlight +-90 degree capture angle ranges around candidate arrival transitions beginning at SAT, CVE, SWB, and JEPEG.Even though the flights are within capture angle range of SWB, it leads to the same overloaded meter fix to which the flights are already headed, so SWB is not a reroute option.Flight 1 is within capture angle range of CVE and so this reroute option is highlighted in yellow.Flight 2 is within capture angle range of both CVE and JEPEG and so both these reroute options are highlighted in yellow.Based on discussions with operations experts, TMCs are hesitant to reroute international flights due to crew and fuel constraints.Therefore, an international flight filter removed from reroute consideration all flights with a 4-letter origin airport code not beginning with "K".Additionally, flights that had already been rerouted previously by ORC, were filtered from reroute consideration. +III.C. Evaluate Reroute CombinationsGiven a set of available reroute options, Evaluate Reroute Combinations calculates cost metrics by probing the external trajectory prediction and arrival scheduling functions with various reroute options.Reroute cost metrics only require a single probe of the trajectory predictor per reroute, whereas scheduling cost metrics require probing the arrival scheduler for every possible combination of reroutes, which is much more computationally intensive.Therefore, reroute cost metrics are used to prioritize the set of combinations to hasten finding a suitable solution.ORC considers a single reroute cost metric, flight time difference, measured as the change in runway ETA between the current route and the reroute, based on nominal speed profile.ORC considered a single scheduling cost metric which was excess meter fix delay (d i -D, where D = 7 minutes).Any reroute combination for which all flights had a scheduling cost less than or equal to zero was considered to be a complete solution.However, if no complete solutions existed, a scheduling cost function was used in the following Recommend Reroute Solution process to identify the best partial solution.Figure 4 shows the Evaluate Reroute Combinations process.The process begins by calculating reroute cost metrics for the set of available reroutes from the Determine Reroute Availability function.The reroutes are sent to the trajectory predictor to calculate meter fix ETAs, which are extended to runway ETAs using nominal TRACON transit time based on the same adaptation used by the arrival scheduler.The reroute runway ETA is compared with the original runway ETA to calculate the flight time difference.The cost of a combination of reroutes is simply the sum of the respective individual reroute costs.Next, all possible combinations of reroutes are sorted by reroute cost into a prioritized list.Scheduling costs are computed for each combination of reroutes in order of priority by probing the arrival scheduler with meter fix and ETA updates to the rerouted flights and comparing d i with D. Evaluate Reroute Combinations ends when either a combination is found to be a complete solution (d i -D <= 0 for all flights), or the entire prioritized list has been evaluated.Cost metrics for all evaluated combinations are stored for consideration in the Recommend Reroute Solution function. +III.D. Recommend Reroute SolutionThe final ORC automation function selects a reroute combination to recommend to the TMC.A scheduling cost function was designed to select a complete solution should one exist, but otherwise select a partial solution that reduced the scheduling cost as uniformly as possible among flights.The objective of the scheduling cost function was to minimize the maximum excess meter fix delay for an individual flight.For all complete solutions, the maximum excess meter fix delay is less than or equal to zero.All other reroute combinations constituting a partial solution will have a value greater than zero; however the minimum partial solution ensures than no single flight is left with larger excess delay than necessary.The outcome of this reroute solution selection process is that reroute cost is only considered (via candidate solution prioritization) in so far as a complete solution can be found and all combinations of lower priority (higher reroute cost) are not evaluated.If no complete solution is found and the entire prioritized list is evaluated, partial solution selection is driven by scheduling cost alone.The assumption is that the systemic goal of reducing maximum excess delay is more important than individual flight efficiency. +IV. Simulation ImplementationFor the purposes of demonstrating a proof-of-concept, ORC automation was implemented in a fasttime simulation of IAH arrivals.Models of TFMS trajectory prediction and TBFM arrival scheduling were integrated with ORC automation coded in the Python scripting language to complete the simulation. +IV.A. Trajectory PredictionThe Future ATM Concepts Evaluation Tool (FACET) 5 was used to model the Trajectory Prediction function.FACET is a simulation software tool developed by NASA.FACET can quickly generate and simulate thousands of aircraft trajectories using aircraft performance profiles, airspace models, weather data, and flight schedules and flight plans.It models trajectories for the climb, cruise, and descent phases of flight for each type of aircraft using the Base of Aircraft Data (BADA) database provided by Eurocontrol. 6FACET provides researchers with a simulation environment for preliminary testing of advanced ATM concepts through its Application Programming Interface (API).Given a traffic scenario of flight plans, FACET was used to simulate all flights along unimpeded trajectories using nominal-speed profiles and historic wind data.In order to make use of FACET's capability of trajectory prediction, alternate flight plans or routes via different IAH arrival meter fixes were generated according to the published STARs, and these flight plans were fed into FACET for generating unimpeded trajectories without conflict detection and resolution.The ETAs at alternative meter fixes were then retrieved and used as inputs to the arrival scheduler described in following section.To include the wind effect, the Rapid Refresh (RAP) data provided by National Oceanic & Atmospheric Administration (NOAA) were used as wind data resource, which is reloaded hourly in FACET.As FACET uses the nominal airspeed for a given aircraft type when generating trajectories, speed restrictions in STARs were imposed additionally in FACET via API script.Runway ETAs were estimated by adding to the meter fix ETAs the nominal zero-wind transit time from meter fixes to their primary runways according to TBFM adaptation. +IV.B. Arrival SchedulingResearchers consulted with IAH Center and TRACON controllers and TBFM experts to understand desired characteristics of arrival flows transitioning from Center to Terminal airspace.As both the triggering and evaluating mechanism for ORC, it is extremely important for the Arrival Scheduling function to provide a scheduled arrival flow that is desirable for both Center and TRACON controllers.A simplified model of the TBFM arrival scheduling currently fielded at IAH was developed to reduce computation time for iterative use.Consultation with Houston subject matter experts ensured the simplified algorithm retained key operational considerations.Arrivals enter Houston TRACON (I90) via four arrival gates positioned in the northwest, northeast, southwest, and southeast corners.IAH arrivals enter I90 via six main meter fixes, one for each Southern gate and two for each Northern gate.Figure 5 shows the STAR transition routes to the Northern meter fixes feeding IAH.The transitions to the more Northern meter fixes MPORT and WHACK are the preferred arrival routes.The dual meter fixes at each of the Northern gates are often treated as a single stream due to the complexity of blending two streams when there is a merge shortly after entering the TRACON or runway dependencies.This is known as mirrored metering.In the future, ORC may be used to balance flows to the same arrival gate by modeling TRACON merge point scheduling constraints, but this proof-of-concept focuses on offloading to different arrival gates.Therefore, traffic scenarios were generated such that all flights filed to the preferred MPORT and WHACK meter fixes.As seen in Fig. 5, only one transition fix BRKAT does not have a published STAR transition to the preferred meter fix.BRKAT is rarely if ever used in practice and so traffic scenarios did not include flights using this transition.On a first-come-first-served basis the scheduler computes the earliest feasible schedule times for a flight at the meter fix and runway subject to their respective simplified separation constraints and allowable transit time ranges between them. 7 +IV.B.1. Separation ConstraintsI90 TBFM nominal separation constraints at the meter fixes are set to 6 NM with the exception of the northwest meter fixes, which are set to 7 NM.This provides a 1-2 NM buffer over the 5 NM minimum to account for compression.The winds aloft in Houston airspace are predominantly from the West, often subjecting flows from the West to strong tailwinds, and flows from the East to strong headwinds.Experienced Center controllers feeding the northeast corner can deliver to the northeast meter fixes with as little as 5.5 NM separation to accommodate high demand.Whereas Center controllers feeding the northwest corner insist that higher ground speeds due to westerly winds require them to deliver to the northwest meter fixes with 8-10 NM separation.I90 terminal airspace controllers expressed concern at how the increased use of optimized profile descents (OPD) has not only limited their use of lateral delay maneuvers on these flights but has reduced the nominal transit time in terminal airspace giving them less opportunity to absorb delay with speed adjustments.This makes the Center's ability to deliver flows properly spaced for compression all the more important.When distance separation requirements are given, the ORC arrival scheduler model converts the distances into time slots using the airspeeds published in STARs.The published airspeeds are 280 kn or 250 kn at the meter fix for all long and short side STARs respectively b .These speeds equate to 77.1-100.8sec time slots between arrivals at the meter fix to accommodate 6-7 NM separation with no wind.The same time slots would provide 8.1-9.8NM separation in a 100 kn tail wind, which is consistent with the 8-10 NM separation preferred by northwest corner controllers.Therefore, the model's use of published airspeeds to calculate static time separation requirements accommodates the controller's desire to modify distance separation to match wind conditions.The arrival scheduler accommodates Miles-In-Trail (MIT) restrictions placed on a meter fix by replacing the nominal 6 or 7 NM separation with MIT distance.The MIT distance is converted to time in the same manner as the nominal separation, using published airspeeds.As with the nominal separations, for any given MIT, the time separation requirement is held constant such that the actual distance separation fluctuates with wind.Runway distance separations are specified in TBFM adaptation by weight class pair of ahead/behind aircraft.The scheduler distance to time slot conversion assumes a landing speed of 145 kn.The time slots corresponding to the most common ahead/behind aircraft distance separations of 2.5 nm in 2-runway configurations and 3.2 nm in 3-runway configurations are 62.1 and 79.5 sec respectively.Table 1 summarizes the nominal distance separation to time slot conversions for the ORC arrival scheduler model.Note that although only the most common runway separation (Large Jet behind a Large Jet) is shown in the table, the scheduler uses the entire runway distance separation matrix containing all weight class combinations. +IV.B.2. Transit Time RangesThe arrival scheduler computes the transit time from meter fix to runway along predefined nominal terminal routes published in TBFM adaptation.The transit time range constraint used by the arrival scheduler is the nominal transit time (calculated using nominal airspeed in no wind) plus the maximum amount of delay that can be absorbed with speed along the route, called a delay buffer.All residual delay is passed back to Center airspace as meter fix delay.Currently fielded TBFM at IAH uses a single configurable delay buffer parameter (set at 60 sec) for all terminal routes.Even though there is some variation in the amount of delay different terminal routes can absorb with speed, it was determined that the uniform 60 sec delay buffer was sufficient for the purposes of estimating scheduled meter fix delay 90 minutes in advance.b Long side routes must complete a base leg turn onto final approach, which lengthens the route allowing meter fix airspeeds feeding long side routes to be greater than for short side routes. +V. Proof-of-concept ExperimentIn fast-time simulation, the TMC shown in Fig. 2 is circumvented by immediately implementing all ORC suggested reroutes.When ORC predicts arrival route congestion, the reroutes in the resulting ORC solution are implemented by updating FACET flight plans.The next simulation time cycle begins with the FACET trajectory prediction using the updated routes one minute of simulation time later. +V.A. Experiment SetupThe proof-of-concept focuses on clear-weather arrival operations into IAH during a rush from the Northeast challenging enough to exercise and stress the ORC algorithm.The day Oct 22, 2014 was chosen to seed a traffic scenario.On that day, IAH had high traffic volume, low weather impacts and was dominated by the W3 runway configuration (West flow, 3 arrival runways) which has the greatest runway capacity.To eliminate the effects of mirrored metering as discussed in section IV.B, flight plans were altered as necessary such that all flights filed RNAV STARs utilizing one of the four major meter fixes, WHACK (Northeast), MPORT (Northwest), GMANN (Southwest) and LINKK (Southeast).Whereas this day may have resulted in a few ORC suggested reroutes, a more aggressive arrival demand was desired to stress the algorithm so that it's behavior could be analyzed.FAA operations experts created a more challenging rush from the Northeast by cloning 12 of the 151 original flights and placing them at peak times to increase the load at WHACK.Figure 6 shows the scenario demand on the four IAH meter fixes in 15-minute increments.Between 11:30 and 12:00, the northeast rush hits the northeast corner-post meter fix, WHACK.During this time, not only is the demand on WHACK high, but it is more than twice the demand on any other single meter fix.This scenario was simulated with and without automatic implementation of ORC reroutes to demonstrate the ORC reroute selection process and its effects on arrival operations.The ORC simulations used a 90minute planning horizon, meaning at any time during the simulation, ORC had knowledge of only flights with runway ETAs within 90 minutes. +V.B. ResultsThis section analyzes the ORC reroute solutions and compares arrival operations between the baseline and ORC simulations.Another reason for ORC triggering shortly after implementing a solution is that the previous solution may have only been a partial solution that did not completely remove the excess meter fix delay.For each ORC action, Figure 9 shows the scheduling cost (maximum excess meter fix delay) of the original routing and the ORC selected reroutes.At time 10:55 the excess delay just exceeds the 7-minute threshold and ORC finds a complete solution by moving a single flight South from WHACK to LINKK (see Figure 8 at 10:55).However, all successive ORC solutions are only able to partially solve the problem by reducing the scheduling cost, but not to zero.One reason ORC may not find a complete solution may be the geographical filtering constraint excluding flights within 5 minutes from crossing a Center boundary for reroute consideration.Figure 7 shows how many flights entering the 90-minute planning horizon around the 700-mile range ring that could have been useful reroute candidates (flights highlighted orange and their followers) are just outside the Memphis Center boundary.Another reason for partial solutions may be that the airport runways were saturated by the demand resulting in excess meter fix delays at LINKK and MPORT when flights were offloaded to these meter fixes from WHACK.Figure 10 shows the cumulative difference in runway STAs between the original routing and ORC solution.These differences correspond to the flights rerouted to alternate meter fixes and any succeeding flights within the planning horizon who's STA to their original meter fix changed as a result of moving the rerouted flights.Notice how in most cases the ORC solution saved time.However, at 11:02 the ORC solution to move a single flight from WHACK to MPORT (see Figure 8 at 11:02) resulted in more time lost for the rerouted flight than saved by advancing the runway schedule of successive flights.ORC chose this solution because the maximum excess meter fix delay was reduced (see Figure 9 at 11:02).A traffic manager might have ignored this recommendation. +V.B.2. ORC vs. BaselineThe analyses presented in this section compare arrival operations between simulations with and without the automatic implementation of ORC reroutes.Whereas the previous analyses of ORC reroute solutions focused on the subset of flights within the ORC planning horizon at the time ORC selected each solution, the results in this section present a more global picture of the effect of ORC reroutes on arrival operations.Figure 11 shows the airport demand and schedule load time histories for the simulations with original (baseline) routing and ORC reroutes.Each column represents the total number of ETAs or STAs for all runways within 15-minute bins The plot focuses on the time range between 11:00 and 14:00 where the effects of the Northeast arrival rush and differences between original and ORC routing can be seen.Through 12:30, original and ORC airport loading is identical although the effects of scheduling can be seen in the differences between ETA and STA loading for both.For example, 5 flights with ETAs in the 11:15 bin are shifted to STAs in the 11:30 bin in both the original and ORC simulations.The differences between original and ORC airport loading begin in the 12:45 bin where the same number of flights are scheduled to land but ORC reroutes shift several ETAs to the next bin.Consequently, these ORC reroutes allowed more flights to be scheduled to land in the 13:00 bin, increasing airport capacity utilization.ORC rerouting allowed the airport to largely recover from the airport demand/capacity imbalance in the 13:15 bin, when the original routing STA count is double the ETA count and the airport is still recovering from delays applied to ETAs from previous bins.Figure 12 shows cumulative delays for each simulation (with and without ORC reroutes) segregated by the source of the delay.The Enroute and TRACON path delays are differences in nominal transit time through each domain between original and final routing.Enroute path delay is the difference in meter fix ETA between each flight's original route and reroute to an alternate meter fix.TRACON path delay is a similar difference in runway ETA minus the enroute path delay.Schedule delay is the difference between the reroute ETA (original route if not rerouted) and final STA at the runway.The original routing simulation without ORC did not have any path delay because there were no reroutes.ORC reroutes increased nominal flight time enroute by 41 minutes and in the TRACON by 11 minutes.However, this caused much less scheduled delay to be applied.The ORC reroutes reduced total delay by 113 minutes or 26%.Note that this reduction in total delay is far greater than what can be seen within the ORC planning horizon in Figure 10.All the total delay savings (i.e.landing time difference) from Figure 10 sum to only 15 minutes.However, these planning horizon limited calculations do not account for the fact that each instance of delay or flight time savings may propagate back to all flights beyond the planning horizon.Figure 1 .1Figure 1.Offloading congested arrival routes to alternate meter fixes +Figure 2 .2Figure 2. ORC automation functions +Figure 3 .3Figure 3. Geographical reroute filters +Figure 4 .4Figure 4. Evaluate Reroute Combinations process +Figure 5 .5Figure 5. Published STAR transitions to North meter fixes feeding IAH. +Figure 6 .6Figure 6.IAH arrival traffic scenario demand meter fix load per quarter hour. +Figure 77Figure 7 shows a map view of IAH arrival flight positions (green triangles) at 11:00, approximately 1 hour before the full force of the Northeast rush will hit WHACK.The dashed circles are range rings emanating from IAH in increments of 100 miles.Orange triangles highlight flights that are estimated to have meter fix delay exceeding 7 minutes which ORC would detect as predicted arrival route congestion.This scenario was simulated with and without automatic implementation of ORC reroutes to demonstrate the ORC reroute selection process and its effects on arrival operations.The ORC simulations used a 90minute planning horizon, meaning at any time during the simulation, ORC had knowledge of only flights with runway ETAs within 90 minutes. +Figure 7 .Figure 8 .78Figure 7. Map view of traffic scenario arrival rush from the Northeast. +Figure 9 .9Figure 9.Comparison of scheduling cost before and after each ORC action. +Figure 10 .10Figure 10.Difference in cumulative runway STA after each ORC action. +Figure 11 .11Figure 11.Airport demand and schedule loading per 15 minutes. +Figure 12 .12Figure 12.Cumulative delay segregated by source. +Table 1 .1Distance separation to time slot conversion.meter fix separation (long side)distance (NM) speed (kn) time slot (sec)NW meter fix7.028090.9other meter fixes6.028077.1meter fix separation (short side) distance (NM) speed (kn) time slot (sec)NW meter fix7.0250100.8other meter fixes6.025086.4common runway separationdistance (NM) speed (kn) time slot (sec)2-runway configurations2.514562.13-runway configurations3.214579.5 + Downloaded by NASA AMES RESEARCH CENTER on August 17, 2016 | http://arc.aiaa.org| DOI: 10.2514/6.2016-4357 + of 13 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on August 17, 2016 | http://arc.aiaa.org| DOI: 10.2514/6.2016-4357 + of 13 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on August 17, 2016 | http://arc.aiaa.org| DOI: 10.2514/6.2016-4357 + of 13 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on August 17, 2016 | http://arc.aiaa.org| DOI: 10.2514/6.2016-4357 + American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on August 17, 2016 | http://arc.aiaa.org| DOI: 10.2514/6.2016-4357 + of 13 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on August 17, 2016 | http://arc.aiaa.org| DOI: 10.2514/6.2016-4357 + of 13 American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on August 17, 2016 | http://arc.aiaa.org| DOI: + 10.2514/6.2016-4357 + of + American Institute of Aeronautics and Astronautics Downloaded by NASA AMES RESEARCH CENTER on August 17, 2016 | http://arc.aiaa.org| DOI: 10.2514/6.2016-4357 + + + + +AcknowledgmentsThe authors would like to recognize members of the original ORC concept development team, Philip Basset and Lisa Smith of FAA, Kenneth Hailston, Robert Giacomazzo, Carl Bernten, and Joel Hiks of Booz Allen Hamilton, and Mike Babbidge of Delta Airlines. + + + +VI. ConclusionThis proof-of-concept analysis of ORC applied to a northeast arrival rush into IAH demonstrates how strategically offloading flights from congested arrival routes can potentially provide substantial benefits to arrival operations.The ORC algorithm actively seeks to minimize excess meter fix delay which has the potential to reduce traffic manager workload and the use of other traffic management initiatives upstream such as miles-in-trail and ground delays.It also has the additional potential benefit of maximizing airport capacity utilization and reducing arrival scheduling delays.Future research is needed to explore whether the use of other scheduling cost functions driving the ORC solutions, or incorporating reroute cost into the selection of partial solutions, can further increase these additional potential benefits.This phase of ORC assumed low weather impact conditions.Following phases will incorporate strategic weather avoidance routing to make this a valuable concept in all weather conditions. + + + + + + + Practice for Application of Federal Aviation Administration (FAA) Federal Aviation Regulations Part 21 Requirements to Unmanned Aircraft Systems (UAS) + 10.1520/f2505 + + + Federal Aviation Administration. Optimized Route Capability -Preliminary operational requirements + Washington, D.C.; Washington, D.C. + + ASTM International + July 2013. October 2013 + + + Federal Aviation Administration + Federal Aviation Administration. Optimized Route Capability -Preliminary functional analysis, Washington, D.C., July 2013. 2 Federal Aviation Administration. Optimized Route Capability -Preliminary operational requirements, Washington, D.C., October 2013. + + + + + Optimized Route Capability (ORC) Intelligent Offloading of Congested Arrival Routes + + ShannonZelinski + + + MinXue + + + PhilipBassett + + 10.2514/6.2016-4357 + + + 16th AIAA Aviation Technology, Integration, and Operations Conference + Washington, D.C. + + American Institute of Aeronautics and Astronautics + September 2015 + + + Federal Aviation Administration. Optimized Route Capability -A Revised Concept Of Operations, Washington, D.C., September 2015. + + + + + MITRE progress report No. 2 + + HBateman + + + CGuensch + + + SHeitin + + + SKamine + + 10.2172/5963005 + MTR130600 + + November 2013 + Office of Scientific and Technical Information (OSTI) + McLean, VA + + + Optimized Route Capability (ORC) Research Assessment + + + MITRE Technical Report + Bateman, H., C. Guensch, S. Heitin, and S. Kamine, Optimized Route Capability (ORC) Research Assessment, MITRE Technical Report MTR130600, McLean, VA, November 2013. + + + + + FACET: Future ATM Concepts Evaluation Tool + + KarlDBilimoria + + + BanavarSridhar + + + ShonRGrabbe + + + GanoBChatterji + + + KapilSSheth + + 10.2514/atcq.9.1.1 + + + Air Traffic Control Quarterly + Air Traffic Control Quarterly + 1064-3818 + 2472-5757 + + 9 + 1 + + June 2000 + American Institute of Aeronautics and Astronautics (AIAA) + Napoli, Italy + + + Bilimoria, K., B. Sridhar, G. Chatterji, K. Sheth, and S. Grabbe, FACET: Future ATM Concepts Evaluation Tool, 3rd USA/Europe Air Traffic Management R&D Seminar, Napoli, Italy, June 2000. + + + + + Sensitivity Analysis of Aviation Environmental Impacts for the Base of Aircraft Data (BADA) Family 4 + 10.2514/6.2021-0457.vid + No. 2009-009 + + + Base of Aircraft DATA (BADA) Aircraft Performance Modelling Report + + American Institute of Aeronautics and Astronautics (AIAA) + 2009 + + + EEC Technical/Scientific Report + Eurocontrol Experimental Centre, Base of Aircraft DATA (BADA) Aircraft Performance Modelling Report, EEC Tech- nical/Scientific Report No. 2009-009, 2009. + + + + + A Closed-Form Solution to Multi-Point Scheduling Problems + + LarryMeyn + + 10.2514/6.2010-7911 + + + AIAA Modeling and Simulation Technologies Conference + Toronto, Canada + + American Institute of Aeronautics and Astronautics + August 2010 + + + Meyn, L., A Closed-Form Solution to Multi-Point Scheduling Problems, AIAA Modeling and Simulation Technologies Conference, Toronto, Canada, August 2010. + + + + + + diff --git a/file842.txt b/file842.txt new file mode 100644 index 0000000000000000000000000000000000000000..209ef3d89c279d701d1bfd473ba61c9462665e6e --- /dev/null +++ b/file842.txt @@ -0,0 +1,292 @@ + + + + +Surface metering is used when the predicted demand exceeds the airport capacity during departure banks.Holding flights at the gate transfers excess taxi time from the departure runway queue back to the gate, prior to engine start.As of March 31, 2020, it is estimated a total of 3,832 hours of engine run time reduction that resulted in 5,075,981 pounds of fuel savings and 15,634,022 pounds of CO2 emission reduction [3].Surface scheduling and metering is enabled by the availability and quality of flight readiness time.Major airlines participating in the ATD-2 field evaluation provide an Earliest Off-Block Time (EOBT) for each departure flight.The EOBT represents the best estimate of a flight's earliest pushback time.A good quality EOBT helps the scheduler achieve accurate demand prediction at each runway and hence, achieve reliable scheduling to keep the surface less congested while maintaining overall airport throughput.Along with other data inputs, the surface scheduler uses the EOBTs to predict takeoff times of departure flights and calculate recommended gate holds.Gate hold advisories are communicated to Ramp Control through the ATD-2 Ramp Traffic Console (RTC) interface.Airlines are motivated to improve the accuracy of EOBT data for the benefit of reduced excess taxi time and fuel savings achieved through surface metering.A preliminary study of EOBT quality impact on surface metering [4] [5] shows that the EOBT accuracy has a direct impact on the length of the departure gate hold time assigned by ATD-2's surface scheduler.To date, GA operations have not been included in surface metering at CLT.One reason is that, unlike commercial flights at CLT's main ramp, pilots departing from the GA ramp area do not contact the Ramp Tower prior to taxi.At the main ramp, the Ramp Controller communicates gate hold advisories for surface metering.In contrast, GA flights' first communication is with the Ground Controller at the FAA's Airport Traffic Control Tower (ATCT).Other challenges in GA operations include less-predictable departure times and the ability to communicate up-to-date readiness and intent information to the ATD-2 scheduler.Excluding GA flights before they enter the Airport Movement Area (AMA) from surface metering and scheduling may have a potential impact on the efficiency of airport operations that has not been investigated. +B. Mobile App Technology Enables Two-Way Communication with GA OperationsA mechanism is needed to enable two-way communication between GA operations and the ATD-2 scheduler prior to their entry into the AMA.In response to that need, The MITRE Corporation has developed a prototype Mobile Application (App) to enable two-way information sharing between GA pilots and NASA's IADS traffic management system [6] [7].MITRE has conducted a field test of their prototype Mobile App technology at CLT with a small group of Corporate Flight Operators.Using a mobile device, pilots submit a ready time to the ATD-2 scheduler.This ready time, referred to as a "Ready-to-Taxi Time", or RTT, represents the pilot's best prediction of when they will arrive at the edge of the GA ramp area and be ready to contact Ground Control for their taxi clearance to enter the AMA.The RTT submitted by GA pilots is analogous to the EOBT submitted by airlines for departure flights at the main ramp.It provides ATC and the ATD-2 scheduler with more accurate and up-to-date departure readiness information.In return, the GA pilot receives flight-specific schedule information, generated by the ATD-2 scheduler, such as assigned runway and expected takeoff time, and information about traffic management restrictions in place.Fig. 1 shows an example timeline of how a GA pilot may incorporate using the Mobile App to submit readiness information into their pre-departure procedures. +C. Research ObjectiveBecause of the GA flights' unique operational challenges, their impact on surface operations at CLT, particularly during surface metering, has not been fully understood.Research questions include: 1) whether GA flights have an unintended advantage over commercial airline flights which adhere to metering advisories at the gates?and 2) within a traffic bank, what GA traffic concentration affects the airline flights the most?Since the grand scheme of surface metering depends on quality of predictability for performance, questions exist around whether including GA flights with RTT data in surface metering might improve overall runway throughput, taxi times, delays, and fuel consumption.Note that the concept of surface metering "gate holds", as applied to commercial flights at a parking gate, is that the aircraft is held at the gate, prior to push back.How it would be applied in GA operations has yet to be explored.It may mean that a GA flight holds somewhere in the GA ramp area, and possibly, delays engine start.However, for the purpose of this simulation, the term "gate hold" is applied generally to GA operations to mean they were held prior to entering the AMA.Specific procedures for applying "gate holds" in GA operations are beyond the scope of this simulation.This paper investigates the impact of GA flights on the efficiency of airport operations using IADS's surface metering programs at CLT.The study uses fast-time simulations and explores two sets of variables.The first is the GA traffic demand that is quantified by the number of GA flights in the overall flight traffic at the airport and their flight times during traffic peak hours.The other variable is the GA flight readiness information, RTT, including provision and accuracy.The paper is organized as follows: Section II describes the simulation environment including system setup, traffic scenarios, metrics and measurements, and the RTT model development.Section III presents the simulation results using the given scenarios and RTT model.The paper concludes with a summary in Section IV. +II. Simulation environment description A. AirportThe simulations used the CLT airport in a south simultaneous traffic flow configuration.Fig. 2 shows the airport layout at CLT.While Runway 18R is for arrival only, Runways 18C and 18L are dual-use runways for both departures and arrivals in this runway configuration.A Fixed Base Operator (FBO), where some GA/BA passengers meet their pilot, is located at the General Aviation Ramp on the east side of the airport, shown in the green box in Fig. 2. The simulations in this paper use the same configuration that was built and validated in the previous study about the impact of EOBT uncertainty [4] with the addition of a newly developed RTT model for GA flights, as shown in Fig. 3. NASA's Surface Operations Simulator and Scheduler (SOSS) [8] was used for these simulations.SOSS connects to the ATD-2 Tactical Surface Scheduler through the Surface Modeler.The traffic scenarios are the inputs to the simulator and define the GA traffic level and demand density, as well as scheduled airline flights.The EOBT Model provides the EOBT updates for the airline aircraft in the main terminals.The RTT Model provides the estimated readiness times and accuracy for GA flights.The Surface Modeler used for fast-time simulations is adapted from ATD-2's Surface Model used in real-time field demonstration, which is used for data exchanges and integration.One of the core concepts is to group aircraft by their priorities and predictability of flight readiness in scheduling [9].When the predicted traffic demand exceeds the runway capacity and the predicted excess taxi time is above a preset threshold, surface metering is triggered.The Tactical Surface Scheduler schedules flights from high-priority group to low-priority group.In the simulations, both airline departures at gates that provide EOBT updates and GA aircraft providing an RTT estimate at stand are placed in the same highpriority group.The GA aircraft with no RTT estimate are placed in the low-priority group for scheduling priority and are exempt from surface metering, and, therefore, start taxiing from the parking stand whenever they are ready, emulating the current GA ramp operation without Mobile App communication. +C. Simulation configurationsIn this study, four simulation configuration groups are considered.They are designed to support two sets of simulation variables as described in the research objectives.The baseline traffic scenario for all groups is derived from actual operation data from Bank 2 traffic at CLT on February 14, 2018.There is a total of nine traffic banks daily at CLT.Bank 2 is one of the busiest banks in the morning traffic.The typical traffic pattern in Bank 2 is that the departure demand peaks before the arrival demand starts to build up.In this baseline scenario, there are 85 airline departures and 83 airline arrivals, respectively.Of the 85 departures, 43 take off from Runway 18C and 42 from 18L, respectively.Of the 83 arrivals, 5, 30, and 48 land on Runways 18C, 18L and 18R, respectively. +Simulation Group 1: GA flight request distributionThe purpose of the first simulation group was to investigate the impact of GA flights' requested departure time distribution.Four different departure time distribution patterns were examined, with each scenario having ten GA flights: 1) uniformly distributed over the bank, 2) concentrated in the beginning of the bank, 3) concentrated in the middle of the bank, and 4) concentrated in the end of the bank.GS flights in this simulation group did not provide RTTs and so were exempt for surface metering.It was expected that GA traffic demand at the peak time, i.e., case 3, would have the largest impact on the airport performance in a surface metering situation. +Simulation Group 2: Number of GA flightsThe second simulation group increases the number of GA flights uniformly distributed in the bank to examine any adverse impact to airline flights, such as increase of taxi times or gate hold times (when metering is on).As with Group 1, GA flights did not provide RTTs. +Simulation Group 3: RTT percentageIn the third simulation group, a certain percentage of GA flights were assumed to provide RTTs with similar accuracy as observed in MITRE data described in section E. The percentage of GA flights providing RTTs was varied for ten uniformly distributed GA flights.Those GA flights submitting RTTs are assigned the same scheduling priority as the airline flights that have EOBTs at main terminals and are expected to comply with any assigned metering hold at the parking stands.The performance (e.g., taxi time) of the GA flights will be examined together with airline flights. +Simulation Group 4: RTT accuracyIn the last simulation group, all ten uniformly distributed GA flights were assumed to provide RTTs with varying accuracy.The RTT model generates the various accuracy levels used in the simulations.The expectation was that as the RTT accuracy improves, taxi time reduction will occur for both airline and GA flights during surface metering. +D. Performance metricsTo analyze the impact on the efficiency of airport operations, a set of performance metrics were considered.They included gate hold time of metered GA and airline flights, time series of gate hold time, taxi out time of GA and airline flights, time series of taxi out time, and runway throughput. +E. RTT modelIn this section, we describe how the RTT accuracy model is developed using a data driven approach.For this work, we have used the data collected by MITRE for GA flights that departed from CLT between October 20th, 2017 and September 17th, 2019.The proposed RTT model is a combination of two probabilistic quantities: 1) the timing of the updates up to the Actual RTT (ARTT), and 2) the accuracy of the updated RTTs as a function of the timing.The proposed statistical model that generates RTT accuracy is used in the fast-time simulation for evaluating the impact of RTT accuracy on airport surface operations.1. RTT Data Analysis GA/BA flight data collected by MITRE was used for the RTT data analysis.Out of 1,656 GA flights in this data set, 301 flights submitted at least one RTT and had AMA entry times from Airport Surface Detection Equipment, Model X (ASDE-X).The AMAentry time represents the GA flight's Actual RTT, that is, the time they arrived at the edge of the GA ramp area.In this analysis, two main variables were analyzed: 1) RTT accuracy, and 2) RTT update interval.The RTT accuracy is defined as the difference between the ARTT and pilot's last RTT estimate.The RTT update interval represents the time difference between two RTT updates of the same flight, or between the last RTT update time and ARTT. +Fig. 4 RTT accuracy changes over timeFig. 4 shows the RTT accuracy changes in the lookahead time window [-30 min, 0 min] as time approaches the ARTT.The plot represents the RTT estimate values submitted by the pilots within the lookahead time window.This plot shows that in general, the RTT errors tend to decrease to negative values as the time progresses, which implies that the RTT prediction becomes more conservative, as it approaches ARTT.The majority of the GA flights (85%) submitted their RTTs once (dark red markers), approximately 10% GA flights submitted two RTT updates (red color), and about 5% provided more than two RTT updates. +RTT Model DevelopmentFor the RTT model development, a two-step approach was used, which was introduced in the EOBT modeling in [4].First, the RTT update times of each flight were modeled within a [-30 min, 0 min] lookahead time window as the time approached ARTT.Then, the RTT accuracy was modelled at each update time calculated in the first step.The model produced RTT accuracy distribution similar to the actual RTT error distribution using the probability distributions provided by commercial software libraries like Apache Commons Math package [10] and MATLAB [11].2), to fit the average accuracy trend along the lookahead time.In (2), x and y represent the lookahead time with respect to the ARTT time and the mean RTT error (ARTT-RTT) at x, respectively, with the regression coefficients, c0 and c1.y = c0 + c1  x (2)Next, a sequence of probability distributions was fitted to the actual RTT error data (referred to as PD3).The logistic probability distribution model was the closest to the actual RTT error distribution provided by the actual data.Fig. 7 shows the histogram of actual RTT accuracy values, and the red line in the chart shows the selected Logistic probability distribution model.The Kullback-Leiblar divergence (also called relative entropy) [12] was used to measure how much one probability distribution was different from the reference probability distribution.The Kullback-Leibler divergence of the selected model with respect to the actual distribution was 0.4064, the lowest, or the best, fit of all candidate models.The RTT accuracy, Y, can be modelled using Eq. ( 3):Y = c0 + c1  Xk + random(PD3) ()where Xk is the RTT update time from the previous model in Eq. (1).From the data analysis using the actual RTT data at CLT, the following parameters were calculated in +RTT Model ValidationWith the given parameters in Table I, the RTT values were generated from the proposed RTT model for model validation.The two scatter plots in Figs. 8 and9 show the actual RTT error and the generated RTT errors using the model proposed in Eq. ( 3), respectively, along the lookahead time window [-30 min, 0 min] as the time approaches ARTT.The RTT values from the RTT accuracy model cannot be exactly the same as the actual RTTs for an individual GA flight, but it is shown that the distributions are visually quite similar and the Kullback-Leibler divergence is small (0.0389). +III. Simulation ResultsThis section presents the simulation results and analyses of the four simulation groups.During Bank 2 at CLT, almost all GA flights use the east runway (18L/36R) which is closest to the GA ramp areas.In the simulations, all GA flights were assigned to Runway 18L.The results shown in the analysis are for Runway 18L only. +A. Simulation Group 1: GA flight request distributionIn this simulation group, ten GA flights were added to the baseline scenario.This number reflects the Bank 2 non-passenger flight statistical analysis.Their flight ready times from the GA ramp were distributed in the bank in four ways:• Case 1: uniformly distributed over the whole bank,• Case 2: in the beginning 30 minutes of the bank,• Case 3: in the middle 30 minutes of the bank, and • Case 4: in the last 30 minutes of the bank Twenty Monte Carlo simulation runs were conducted for each case.Figs 10 and 11 show the average total gate hold times and the average total taxi-out times, respectively, over the twenty simulation runs.Error bars indicate 95% confidence level.Since the GA flights provided no RTTs in this group, they were not subject to metering (and zero gate hold time).In Case 2, where the GA traffic concentration was in the beginning of the bank, airline flights had longer gate hold and taxi-out times, rather than Case 3 as originally expected.It reveals that the GA traffic was competing with the airline traffic for the runway at the beginning of the bank.Additionally, because the GA aircraft immediately started taxiing into the AMA, after being designated as "ready", and joined a departure queue earlier, they became higher priority in the scheduler than the airline aircraft at the gate.This resulted in longer gate holds and taxi-out times for the airline flights.Case 3, where the GA traffic concentration was in the middle of the bank, had the second largest impact to airline flights.This is because although the airline departure traffic began to decrease, there was sufficient demand on this dual-use runway due to the inbound arrivals.Case 4, where the GA traffic was in the end of the bank, had the least impact on the airline flights.Also, in Case 4, the total GA taxi-out time increased compared to Cases 2 and 3.The probable reason for the increase in taxi out time is that the arrival traffic on the dual-use runway at the end of the bank caused the GA aircraft to wait longer for the runway use.The results in this simulation group show that GA traffic concentrated in the beginning 30 minutes of the bank had the largest impact on airline flights, in metering hold time, taxi out time, and throughput.One reason is that airline departure demand starts early in the bank, which is then overlapped by GA traffic.Another reason lies in the unpredictability of GA operation, which tends to cause longer gate holds from the surface scheduling algorithm and longer taxi-out times for airline aircraft. +B. Simulation Group 2: Number of GA flightsIn this group, six different numbers of GA flights were added to the baseline scenario.Their readiness time perturbation was evenly distributed over the entire traffic bank.Table II shows the numbers of GA flights in each of the six cases.Twenty simulation runs were conducted for each case.Statistically, Case 6 of this group is the same as Case 1 of Simulation Group 1.Table II Figs 16 and 17 show the average total gate hold times and the average total taxi-out times, respectively, over the Monte Carlo simulation runs.When GA traffic increased, the airline flights experienced longer metering hold times, as expected.At the same time, the total taxi-out time of the airline flights also showed an increasing trend with Case 2 as an outlier, which probably was due to the demand timing of the two GA flights with respect to the airline flights in the baseline scenario.The increase of GA traffic and its less predictability had a negative impact on the airline flights in both gate hold and taxi-out times, or on the metering effectiveness for taxi out time reduction.The increased taxi time of GA flights shown in Fig. 17 was due to the increased number of GA flights.Overall in this simulation group, the simulation results show that the increase of GA traffic affected the airline flights with earlier metering start times, higher hold times, taxi out times and takeoff delays.In particular, because of the less predictability of GA traffic demand, the effectiveness of surface metering to mitigate taxi out times was affected negatively. +C. Simulation Group 3: Percentage of GA flights having RTTsThe RTT model was introduced in Simulation Group 3 and provided the RTT values for GA flights prior to departing the parking stands in the GA hangar.The RTT model was configured to match the accuracy of the actual data from MITRE.Ten GA flights had their ready times perturbed uniformly over the entire bank in the baseline scenario.The percentage of the GA flights that provided their RTTs were varied in six cases: 0%, 20%, 40%, 60%, 80% and 100%.The 0% case represents the current operations without Mobile Applications, whereas the 100% case assumes that all the GA flights provide controllers with the estimated AMA entry times in advance.The GA flights which provided RTTs were considered in the high priority group in runway scheduling and had to comply with the metering hold at parking stand.Forty simulations were run for each case.This doubled number of simulations runs helped expand the statistical sample size to achieve a good accuracy of each target percentage of GA flights that provided RTTs.Figs 22 and 23 show the average total gate hold times and the average total taxi out times, respectively.Fig. 22, shows that the total hold times of the GA flights increase as more GA flights provided RTTs, but the hold times of the airline flights show no statistically significant difference.This implies that the impact on the gate hold time of airline flights by GA flights with or without RTTs is small.GA flights that submit RTTs compete with airline flights for slots from the scheduler and hold at park stand.In contrast, the GA flights that do not provide RTTs directly affect the airline flights by leaving the parking stand freely when ready.Nonetheless, the overall gate hold times of both airline and GA flights increased when more GA flights had RTTs.In correlation with the increase in metering hold times, the taxi out time chart in Fig 23 shows a decreasing trend in both airline and GA flights.It indicates that when a greater percentage of GA flights submitted RTTs and complied with the metering times, the interruption to the metering ecosystem dropped and benefitted the taxi out time reduction.The runway throughputs are depicted in Figs 26 and 27.No obvious differences are found between cases in either plot.That indicates that the RTT provision of GA flights can increase metering hold time but decrease taxi out times of both airline and GA aircraft, helping maintain departure throughput.In summary, the results show that overall taxi out times can be reduced when more GA flights provide RTTs and are in compliance with metering hold times.In other words, GA flights submitting RTTs and complying to metering times can help improve the airport performance. +D. Simulation Group 4: GA flight RTT accuracyIn Simulation Group 4, four different RTT accuracy levels of GA flights were tested, as shown in Fig. 28.Case 1 is the ideal case where the pilots' estimated RTTs are exactly the same as the actual RTTs.Case 3, the baseline case, is the current level of RTT accuracy from the data collected by MITRE.Cases 2 and 4 have the accuracy levels better and worse than the baseline case, respectively, for comparison.The horizontal axis is the lookahead time towards the actual RTT.The vertical axis is the RTT accuracy measured as the difference of the actual RTT and the estimated RTT by pilots.The data of the plots were collected from actual simulations. +Fig. 28 RTT accuracy casesIn each case, ten GA flights were uniformly added across the baseline traffic scenario with the varied RTT accuracy level.Forty simulations were conducted for each case as well.Figs 29 and 30 show the comparison among the four RTT accuracy levels for metering hold and taxi out times, respectively.The metering hold times of GA flights went up steadily when RTT accuracy became worse.However, the additional hold times did not translate to taxi out time reduction.The reason likely lies in the fact that the added hold times were induced primarily by the inaccuracy of RTT prediction rather than scheduling.In particular, the conservative prediction errors (i.e., estimated RTT is later than ARTT) resulted in unnecessary hold times and consequently takeoff delays as seen in the runway throughput plot, Fig. 31.For airline flights, no statistically significant differences are observed in the first three cases.In Case 4, the taxi out times of both airline and GA flights had a small but noticeable amount of increase, compared to the averages of the other three cases, possibly due to the increased uncertainty from the RTT prediction errors.The time series of metering hold and taxi-out times, and the runway throughputs, not shown in the paper, had little difference among the four RTT accuracy levels.The simulation results in this group show that the RTT accuracy affected mainly the GA flights' throughput.The RTT accuracy at or better than the baseline from the actual data had no significant impact on the airline flights.On the other hand, RTT error worse than the baseline level may have a negative impact on both airline and GA flights due to the increased prediction errors. +IV. ConclusionsIn this study, fast-time simulations were used to investigate the impact of GA flight operations on airport performance.The ATD-2 surface metering scheduling algorithm was used for Bank 2 traffic hours at CLT.The objectives of the study are to investigate how the current GA operations may affect surface management efficiency benefits, such as taxi out time reduction from surface metering, and potential merits of including GA flights in the surface scheduling program.The simulations were conducted in four simulation configuration groups.The first two groups focused on the GA departure demand concentration and the number of GA flights in the bank.They were designed to study the impact of current GA operation.In the next two simulation groups, an RTT model was developed based on the actual operational data from MITRE.The model sets percentages of GA flights providing RTT predictions and varies the RTT accuracies in the simulations.These two simulation groups studied how airport performance under surface metering were affected by GA flights submitting RTTs to the scheduler system, through their mobile devices for example.The simulations in the first two groups show that:• The concentration of GA departure traffic at the beginning of the bank had the most significant impact on airline departures.It resulted in longer metering hold and taxi-out times for airline flights.It also triggered the scheduler to start surface metering earlier.GA traffic concentration in the middle of the bank had the second largest impact on airline flights' performance.GA traffic concentration at the end of the bank and evenly spread out across the bank had the least impact on airline flights.However, GA departure concentration at the end of the bank had the longest taxi-out times because of the arrival traffic demand for the same runway the second half of the bank.• The increase of GA traffic affected the airline flights by initiating an earlier surface metering time, longer hold times, longer taxi-out times, and longer takeoff delays.In particular, less of GA flights appeared to affect the metering effectiveness in reducing taxi-out time for airline flights.The results from the next two simulation groups show that:• When more GA flights provided RTTs and followed metering hold advisory, the overall taxi-out times were steadily reduced by a relatively small amount (probably due to the small overall GA traffic level), and at the same time the runway throughputs and takeoff delays remained the same.This finding indicates readiness information from GA flights via mobile devices, such as MITRE's prototype Mobile App, is beneficial to airport performance with surface metering.• When the RTT accuracy is the same as or better than the baseline level, no statistically significant impact on airline flights was seen in all performance metrics.When RTT errors were larger than the baseline level, both airline and GA flights showed longer taxi-out times due to the increased uncertainty introduced by the prediction errors.For GA flights, larger RTT prediction errors caused longer metering hold times that subsequently led to takeoff delay.Despite the relatively small percentage of GA traffic at CLT, the results showed the clear impact of the current GA-operation levels on airport performance in the surface metering environment.Obtaining readiness information from more GA flights may also impact surface operations.Further research may consider other airports such as DAL where GA operations make up a greater percentage of the traffic demand.Fig. 11Fig. 1 Example timeline showing how a Corporate GA pilot may incorporate the Mobile App into their pre-departure procedures.MITRE's field demonstration has shown that Mobile App technology can enable two-way communication flow between GA operators and surface schedulers.However, the communication procedures and requirements related to surface metering for GA flights have not yet been explored.Although not investigated in this field test, Mobile App technology has the potential to support the inclusion of GA flights in surface metering.For example, Mobile App technology might be used to provide metering hold advisories to GA pilots.Although this research has included primarily business aviation (BA) flights, other GA aircraft, cargo, and military flights have the same communication limitation and similar impacts on airport performance.They are included in the GA flights in this study.C. Research Objective Because of the GA flights' unique operational challenges, their impact on surface operations at CLT, particularly during surface metering, has not been fully understood.Research questions include: 1) whether GA flights have an unintended advantage over commercial airline flights which adhere to metering advisories at the gates?and 2) within a traffic bank, what GA traffic concentration +Fig. 22Fig. 2 CLT airport diagram Fig. 3 Fast-time simulation setup B. Simulation setupThe simulations in this paper use the same configuration that was built and validated in the previous study about the impact of EOBT uncertainty[4] with the addition of a newly developed RTT model for GA flights, as shown in Fig.3.NASA's Surface Operations Simulator and Scheduler (SOSS)[8] was used for these simulations.SOSS connects to the ATD-2 Tactical Surface Scheduler through the Surface Modeler.The traffic scenarios are the inputs to the simulator and define the GA traffic level and demand density, as well as scheduled airline flights.The EOBT Model provides the EOBT updates for the airline aircraft in the main terminals.The RTT Model provides the estimated readiness times and accuracy for GA flights.The Surface Modeler used for fast-time simulations is adapted from ATD-2's Surface Model used in real-time field demonstration, which is used for data exchanges and integration.One of the core concepts is to group aircraft by their priorities and predictability of flight readiness in scheduling[9].When the predicted traffic demand exceeds the runway capacity and the predicted excess taxi time is above a preset threshold, surface metering is triggered.The Tactical Surface Scheduler schedules flights from high-priority group to low-priority group.In the simulations, both airline departures at gates that provide EOBT updates and GA aircraft providing an RTT estimate at stand are placed in the same highpriority group.The GA aircraft with no RTT estimate are placed in the low-priority group for scheduling priority and are exempt from surface metering, and, therefore, start taxiing from the parking stand whenever they are ready, emulating the current GA ramp operation without Mobile App communication.C.Simulation configurationsIn this study, four simulation configuration groups are considered.They are designed to support two sets of simulation variables as described in the research objectives.The baseline traffic scenario for all groups is derived from actual operation data from Bank 2 traffic at CLT on February 14, 2018.There is a total of nine traffic banks daily at CLT.Bank 2 is one of the busiest banks in the morning traffic.The typical traffic pattern in Bank 2 is that the departure demand peaks before the arrival demand starts to build up. +Fig. 55Fig. 5 Number of RTT updates Fig. 6 RTT update time Fig. 7 RTT accuracy distribution +Fig. 88Fig. 8 Actual RTT accuracy Fig. 9 Modeled RTT accuracy +Fig. 10 1 Figs 1210112Fig. 10 Total gate hold time of group 1 Fig. 11 Total taxi out time of group 1 +Fig. 12 Gate hold time series of group 1 Fig. 13 Taxi out time series of group 11211Fig. 12 Gate hold time series of group 1 Fig. 13 Taxi out time series of group 1 In the time series for Case 2, surface metering of the airline flights started early around 20 minutes into the bank, and so the GA flights dominated the early part of the bank as shown in the taxi out time series and 'pushed' the airline flights to metering hold.Case 3 shows the similar metering start time to Case 1, but higher metering time bars because of the GA traffic concentration in the middle of the bank.Case 4 shows the least impact on airline aircraft metering which is consistent with the total numbers in Fig. 11. +Fig. 14 Runway throughput chart of group 1 Fig. 15 Cumulative runway throughput of group 1 Figs11Fig. 14 Runway throughput chart of group 1 Fig. 15 Cumulative runway throughput of group 1 Figs 14 and 15 show the runway throughput metric.A departure runway operation is registered at takeoff and an arrival runway operation occurs at landing.The stacked bar charts in Fig. 14 measure the average numbers of runway operations in five-minute bins.Fig. 15 plots depict the cumulative operations for departures only to focus on airline and GA departure throughputs.In Case 2, the bar chart shows the GA aircraft took a significant portion of runway takeoff slots in the early part of the bank and pushed airline departure aircraft to later takeoffs, resulting in takeoff delay of airline aircraft.This is also observed in the cumulative throughput plots in Fig. 15 where the airline departure (in the middle plot) of Case 2 exhibited lower throughput up to about 70 minutes into the simulation time compared to the other cases.A similar situation can be seen in Case 3, in which the GA aircraft concentration in the middle of the bank caused lower airline flight throughput from 60 to 105 minutes.The highest runway throughput in five-minute bins +Fig. 16 Total gate hold time of group 2 Fig. 17 Total taxi out time of group 2 Fig. 182218Fig. 16 Total gate hold time of group 2 Fig. 17 Total taxi out time of group 2 +Fig. 20 Runway throughput chart of group 2 Fig. 21221Fig. 20 Runway throughput chart of group 2 Fig. 21 Cumulative runway throughput of group 2 +Fig. 2222Fig. 22 Total gate hold time of group 3 Fig.23 Total taxi out time of group 3 +Fig. 24 Gate hold time series of group 3 Fig. 25 Taxi out time series of group 333Fig. 24 Gate hold time series of group 3 Fig. 25 Taxi out time series of group 3 The time series bar charts of metering hold and taxi out times are displayed in Figs 24 and 25, respectively.The metering hold time charts show that the percentage of GA flights submitting RTTs did not accelerate nor delay metering start times, because of the same reasons as described in the total number analysis, i.e., the gate hold time of airline flights by GA flights with or without RTTs is small.The metering hold time distributions of the airline flights are very similar to each other.The increase of GA flights' metering hold count and time is evident as more RTTs had been submitted.At the same time, the taxi-out times of both airline and GA flights decreased across the bank, consistent with the results in the total numbers. +Fig. 2626Fig. 26 Runway throughput chart of group 3 Fig.27 Cumulative runway throughput of group 3 +Fig. 29 Total gate hold time of group 4 Fig. 30 Total taxi out time of group 4 Fig. 314431Fig. 29 Total gate hold time of group 4 Fig. 30 Total taxi out time of group 4 Fig. 31 Cumulative runway throughput of GA flights of group 4 +TABLE I .ITable I Parameters used in the RTT ModelRTT ModelProbability Distribution or Regression ModelParametersRTT update timePD1: Log-normal PD2: Weibull = 0.114848,  = 0.303685 A = 11.3599, B = 1.54735RTT accuracyPD3: Logistic Linear Regression = -0,306186,  = 2.27737 c0 = -2.538, c1 = -0.1908 +Number of GA flightsCase No123456Number of GA Flights0246810 + + + + +AcknowledgementThis study is supported by the NASA ATD-2 project, funded by Airspace Operations and Safety Program (AOSP).The authors are grateful to The MITRE Corporation for their collaboration and research support. + + + + + + + + + Practice for Application of Federal Aviation Administration (FAA) Federal Aviation Regulations Part 21 Requirements to Unmanned Aircraft Systems (UAS) + 10.1520/f2505 + + + + General Aviation and Part 135 Activity Surveys -CY 2017 + + ASTM International + null + + + Federal Aviation Administration, "General Aviation and Part 135 Activity Surveys -CY 2017," https://www.faa.gov/data_research/aviation_data_statistics/general_aviation/CY2017 + + + + + Airspace Technology Demonstration 2 (ATD-2) Phase 1 Concept of Use (ConUse) + + YJung + + NASA/TM-2018-219770 + + February 28, 2018 + + + Y. Jung, et al., "Airspace Technology Demonstration 2 (ATD-2) Phase 1 Concept of Use (ConUse)," NASA/TM-2018-219770, February 28, 2018. + + + + + ATD Quart-Monde + 10.1007/springerreference_75803 + + + Publications of the ATD-2 Subproject + + Springer-Verlag + 2020 + + + NASA ATD-2 team + NASA ATD-2 team, "ATD-2 Benefits Mechanism," Publications of the ATD-2 Subproject, 2020 + + + + + Fast-Time Simulation for Evaluating the Impact of Estimated Flight Ready Time Uncertainty on Surface Metering + + HanbongLee + + + YoonCJung + + + ShannonJZelinski + + + ZhifanZhu + + + VaishaliHosagrahara + + 10.1109/dasc43569.2019.9081711 + + + 2019 IEEE/AIAA 38th Digital Avionics Systems Conference (DASC) + San Diego, California + + IEEE + Sep 2019 + + + H. Lee et al., "Fast-Time Simulation for Evaluating the Impact of Estimated Flight Ready Time Uncertainty on Surface Metering," 38th Digital Avionics System Conference, San Diego, California, Sep 2019. + + + + + Evaluating the Impact of Uncertainty on Airport Surface Operations + + SandeepBadrinath + + + HamsaBalakrishnan + + + EmilyClemons + + + TomReynolds + + 10.2514/6.2018-4242 + + + 2018 Aviation Technology, Integration, and Operations Conference + Atlanta, Georgia + + American Institute of Aeronautics and Astronautics + 2018. Jun 2018 + + + S. Badrinath et al., "Evaluating the impact of uncertainty on airport surface operations," 2018 AIAA Aviation Forum, Atlanta, Georgia, Jun 2018. + + + + + Concepts for delivering IFR clearances and exchanging pre-departure data using mobile devices + + PDiffenderfer + + 10.1109/icnsurv.2018.8384968 + + + 2018 Integrated Communications, Navigation, Surveillance Conference (ICNS) + + IEEE + Apr 2018 + + + P. Diffenderfer et al., "Concepts for Delivering IFR Clearances & Exchanging Pre-Departure Data Using Mobile Devices," 2018 Integrated Communications Navigation and Surveillance (ICNS) Conference, Apr 2018. + + + + + Using Mobile Devices for IFR Clearance Delivery and Release and Data Exchange + + PaulADiffenderfer + + + SaraAWilkins + + + KevinMLong + + 10.2514/6.2018-3350 + + + 2018 Aviation Technology, Integration, and Operations Conference + Georgia + + American Institute of Aeronautics and Astronautics + June 2018 + + + P. Diffenderfer et al., "Using Mobile Devices for IFR Clearance Delivery and Release and Data Exchange," AIAA Aviation Forum Atlanta, Georgia, June 2018. + + + + + Validation of Simulations of Airport Surface Traffic with the Surface Operations Simulator and Scheduler + + RobertDWindhorst + + + JustinVMontoya + + + ZhifanZhu + + + SergeiGridnev + + + KatyGriffin + + + AdityaSaraf + + + SteveStroiney + + 10.2514/6.2013-4207 + + + 2013 Aviation Technology, Integration, and Operations Conference + Los Angeles, California + + American Institute of Aeronautics and Astronautics + Aug 2013 + + + R. D. Windhorst et al., "Validation of simulations of airport surface traffic with the Surface Operations Simulator and Scheduler," AIAA Aviation Technology, Integration and Operations Conference, Los Angeles, California, Aug 2013. + + + + + Scheduling Lessons Learned During Phase 1 Field Evaluation of the ATD-2 Integrated Arrival, Departure, Surface Concept + + WCoupe + + + + 13th USA/Europe Air Traffic Management Research and Development Seminar (ATM2019) + Vienna, Austria + + June 2019 + + + W. Coupe et al., "Scheduling Lessons Learned During Phase 1 Field Evaluation of the ATD-2 Integrated Arrival, Departure, Surface Concept," 13th USA/Europe Air Traffic Management Research and Development Seminar (ATM2019), Vienna, Austria, June 2019. + + + + + Appendix A: History of MATLAB and The MathWorks, Inc. + + Matlab + + 10.2514/5.9781600861628.0425.0428 + + + + Basic MATLAB®, Simulink®, and Stateflow® + + American Institute of Aeronautics and Astronautics + + + + + MATLAB, https://mathworks.com/products/matlab.html + + + + + Information, Kullback + + SKullback + + 10.1002/9781118445112.stat01635 + + 1959 + Wiley + + + reprinted in 1978 + S. Kullback, "Information Theory and Statistics," John Wiley & Son, 1959, reprinted in 1978: ISBN 0-8446-5624-9. + + + + + + diff --git a/file843.txt b/file843.txt new file mode 100644 index 0000000000000000000000000000000000000000..0ff9715e23bd7392b80ee02d2d24fbe8d572239e --- /dev/null +++ b/file843.txt @@ -0,0 +1,398 @@ + + + + +I. INTRODUCTIONAt airports under winter snow or freezing temperature with precipitation conditions, aircraft must be inspected for deicing service, and, if requested as the result of inspection, be deiced before takeoff.Deicing service is a procedure to remove frost, ice, slush, or snow from aircraft for safe operation [1] [2].An anti-icing procedure often follows to provide protection against the formation of snow, frost, or ice accumulation on the aircraft surface.A winter season at airports in the northern hemisphere can last a few months when deicing services are needed.It is estimated by the US Environmental Protection Agency (EPA) [3], based on the snowfall and aircraft operations criteria, that in the US there are over 200 primary commercial airports that have potentially significant deicing operations.Deicing aircraft at airports has a big impact on the surface traffic and ultimately, air transportation business.It requires careful planning, implementation and management to reduce extra delays and flight interruptions.For example, the aircraft deicing operation plans at Dallas Fort Worth International Airport (DFW) and Charlotte Douglas International Airport (CLT) consist of detailed rules, procedures, deicing locations and clearly defined roles and responsibilities [4] [5].Depending on the airport infrastructure and space availability, deicing services can be conducted at centralized locations in the ramp or the airport movement area, or at the areas near terminal gates.In the latter case, deicing may occur before pushback or before taxiing after pushback.Centralized deicing, which typically takes place in a remote area away from terminals, has the environmental benefit of being able to collect a high percentage of the sprayed deicing fluid for recycling, and reduce glycol contamination risk.Operationally, it helps reduce the aircraft and service vehicle traffic congestion in the gate area, resulting in better gate access for arrivals.It also allows aircraft to be deiced closer to the departure runway before takeoff, which helps cut down the odds of the anti-ice agent exceeding its Hold Over Time (HOT) [6].One primary concern for centralized deicing, however, is that it may create a bottleneck at deicing facilities, resulting in delays.Careful planning and construction of the deicing zones are, therefore, a common practice to reduce traffic hindrance.In addition, well scheduled deicing operations in connection to departure takeoffs can improve the overall airport performance under snow conditions.The SESAR's De-icing Management Tool [7] under the SESAR Technological Solution portfolio defines itself as a system capable of improving the predictability of aircraft deicing operations at European airports with A-CDM deployment.The solution envisions deicing operations as a part of normal operations rather than adverse conditions in the winter period.The conceptual tool has two key functions: an accurate estimate of the deicing operations duration and a calculation of deicing sequences to optimize the deicing resources.The recent development and field test of various strategies and algorithms in airport traffic management decision support tools conducted by NASA have demonstrated advanced capabilities for airport surface operations [8] [9].Although the new capabilities bring various operational benefits in airport operations, they are primarily designed for the operations under normal weather conditions.Winter operations with deicing services remain a challenge.Controllers require extra effort in assigning deicing services and resources to aircraft that request deicing, and mentally account for deicing operations in departure schedules.Deicing operations add extra uncertainty to the aircraft and directly impact the performance of the decision-making process.The work described in [10] studied the deicing operations at four northern US airports (ORD, MSP, DTW and BOS).It suggested three approaches to estimating departure taxi out times in the event of deicing operations: 1) use a constant deicing time plus a buffer for deicing time uncertainty, 2) attempt to model the deicing time to improve the estimates, and 3) only estimate the taxi time from deicing pad to runway.A recent work [11] considered the deicing operations in a Mixed Integer Linear Programming (MILP) based departure runway scheduler, where the deicing zone assignment was modeled as a decision variable in the optimization formulation.The deicing duration inside the zone was estimated using the average deicing time for each aircraft (size) category.The results showed that an optimal solution could be obtained for both takeoff and deicing queue sequences maximizing the runway throughput.In this paper, we present the work jointly performed by NASA and KARI to model the deicing operations at Incheon International Airport (ICN) and integrate the model into an airport departure scheduler using fast time simulation.The purpose of the research and the contribution of the paper is to help understand deicing operations in airport traffic movement, evaluate deicing service resource management strategy, and eventually incorporate deicing operation support in airport decision support tools.The rest of the paper is organized as follows.Section II introduces the deicing operations at ICN. Section III describes the development of the deicing model and its integration with a departure runway scheduler.Section IV shows the simulation setup and validation.Section V presents the results and analysis based on the Monte Carlo simulation runs using the deicing model with three different zone assignment settings.The comparison with the actual deicing operations is also provided.The paper concludes in Section VI with the summary and future work. +II. DEICING OPEATIONS AT ICN +A. Deicing FacilitiesFig. 1 shows the ICN airport layout [12][13].The airport has three physical parallel runways.In the north flow operation, which is the prevailing flow configuration, departure aircraft take off from 33L and 34 and arrivals land on 33L, 33R and 34.In the south flow situation, departure aircraft use 15R and 16, and arrivals land on 15L, 15R and 16.The Passenger Terminal and Concourse are located in the south.As of 2018, a new passenger terminal, Terminal 2 was opened for operation in the Expanded area.At ICN, aircraft are deiced inside the designated deicing zones in the ramp area.There are total of seven deicing zones as of 2019.The latest addition, T Center, was constructed recently near Terminal 2 and was put in operation in January 2018.Each deicing zone has several deicing pads where aircraft park for deicing.Each pad can accommodate a specific aircraft (wingspan) category according to ICAO Aerodrome Reference Code [14].TABLE I lists the number of pads for each aircraft category at each zone.For example, A South zone has two pads that can support deicing for Category B aircraft, three for E and two for F, respectively.A pad that can service a larger aircraft category can also service smaller ones, but the reverse is not true. +B. Deicing OperatorsAircraft are deiced by the certified deicing service providers, or deicing operators.Some of them are airline affiliates, and some independent.Before each winter season starts, the deicing operators sign service contracts with airlines to provide deicing services.A deicing operator can only perform deicing services for the airlines with whom it has a service contract.In addition, Incheon International Airport Corporation (IIAC), who manages the airport operations, assigns specific deicing zone(s) to the deicing operators to use for deicing services.These service and facility contract restrictions often constrain controllers from using all available deicing resources.This is regarded as a major source of inefficiency in deicing resource allocation. +C. Deicing Operations SequenceOn a deicing day at ICN, departure aircraft will be inspected for deicing.The inspection is done at the gate/stand area before an aircraft is ready for pushback.For an aircraft that needs deicing, the operation sequence follows the steps listed below:• Pilot contacts Deicing Position (DP) at ramp control for deicing request • DP assigns a deicing zone to the aircraft • Pilot calls when ready and obtains pushback clearance • Aircraft leaves gate and taxies to the assigned zone • Aircraft arrives at the zone and gets the pad assignment from DP • Aircraft taxies to the pad, and deicing commences • Pilot contacts Incheon Delivery of ATC for predeparture clearance during deicing • Aircraft leaves deicing zone after completion of deicing and taxies to departure runway to take off The main decision for the DP is the deicing zone assignment.It involves mental calculation to balance the zone workload, taking into consideration various operational constraints such as airline contract, aircraft type, departure runway assignment, terminal and gate location.During the deicing operation, the pilot contacts Incheon Delivery, a position usually at the ATC Traffic Management Unit (TMU), to get an ATC pre-departure clearance.This is to ensure that the aircraft will depart, after deicing, in time to meet the anti-ice fluid's HOT requirement. +D. Operation Data Analysis D.1 Deicing DaysThe deicing operation data in the 2015-2016 winter season was analyzed for deicing model development.The season was three months long from Dec 2015 to Feb 2016.The average number of daily departures of the three months was about 450 flights.Fig. 2 shows the numbers of aircraft deiced for the top ten deicing days of the season.The busiest day occurred on Dec 3, 2015 when a total of 190 deicing operations were recorded.They accounted for about 40% of total departures on that day. +D.2 Deicing Zone AssignmentsThe deicing operations were found in all six zones.TABLE II shows the deicing zone usage percentages.The A South and M South zones were the two most used ones.They were the natural choices for departure aircraft from the passenger terminal gates because their locations are close to the departure runways 33L and 34.A total of 66% of the deicing operations were assigned to them.The A North and M North zones had a total of 27% combined usage.They accounted mostly for the departures to the runways 15R and 16 for the same reason of being close to the departure runways.Despite the obvious advantage of using south zones for north flow departures and north zones for south flow, there were occasions where other zones were chosen due to other reasons such as a service contract constraint and traffic conditions.TABLE III summarizes the percentages of zone usages and departure runways combinations.70% of deicing operations (56% plus 14%) took place at zones that are close to departure runways.The other 30% (18% plus 12%) used the zones that were further away from the departure runways. +Top Ten Daily Deicing Operations +D.3 Deicing Zone TimeThe deicing zone time was measured as the total time that an aircraft stays in a deicing zone.It includes the taxi time inside the zone, deicing time on the pad, engine shut down and spool up time, and waiting time.Fig. 3 depicts the histogram of the zone time from the data records.Some outliers due to the data inaccuracy, such as the spike around the zero, are evident.The purple line represents the normal distribution fit after filtering out the data points equal to zero or greater than 60 minutes.The overall mean and standard deviation of deicing zone time for all aircraft after the filtering was 23.7 and 11.0 minutes, respectively.But larger aircraft in categories E and F took more time inside the zone, as shown in the table.At ICN, the majority of aircraft are of category C, such as A320 and B737, and category E like B787.• the length of time that the aircraft will stay inside the zone for deicing The model is integrated with a departure scheduler.The scheduler calculates the runway takeoff sequence for the aircraft ready for or taxiing to departure.Using the target takeoff times, it then computes the target taxi clearance times for aircraft within a deicing zone to exiting the deicing zone after completion of deicing.The scheduler also issues target pushback times from the gate for both aircraft requesting deicing bound for the deicing zone, and aircraft not requesting deicing bound for the runway.Fig. 4 shows the data flow between the deicing model, the departure scheduler, and the simulator.The flight data from the simulator consist of call sign, aircraft type, assigned gate and runway, aircraft position, aircraft status (ready at gate, taxi, inside zone, etc.).The deicing data are the deicing model output, including zone assignment and deicing time.The scheduler uses the flight data and deicing data to calculate the schedules.The simulator uses the scheduler output data to taxi aircraft from gates to deicing zones and runways.When a departure aircraft is ready for pushback at the gate, the deicing model first decides if the pilot requests deicing or not, and if so, assigns a deicing zone to the aircraft.The scheduler schedules the pushback time for the aircraft to taxi to the assigned zone.Once the aircraft is inside the zone, the deicing model produces the amount of time that the aircraft will stay in the zone for deicing.At the completion of the deicing, the simulator notifies the scheduler to schedule the aircraft for takeoff.When a departure aircraft is ready for pushback at the gate, if the deicing model decides that the pilot does not request deicing, the scheduler schedules the aircraft for takeoff and provides the target pushback time for the aircraft to taxi from gate to runway. +A. Deicing Model Details A.1 Deicing RequestThe deicing request is modeled using a normalized uniform distribution.The model samples the distribution and compares the returned value to a deicing request rate parameter to set the deicing request.The deicing request rate is defined as the percentage of departure aircraft that need deicing.For example, a request rate of 40% corresponds to 40% of total departure aircraft requesting deicing.If the sampling result is smaller than 0.4, the aircraft requests deicing, otherwise no deicing is needed.So, the chance of an aircraft requesting deicing operation is 40% in this example. +A.2 Zone AssignmentThe zone assignment is heuristic in nature.The algorithm considers the gate/stand location, runway, aircraft category, and zone traffic load conditions.It attempts to minimize the taxi distance from zone to runway and the waiting time in a deicing zone queue.It uses a preconfigured priority zone list for each gate area and runway combination.For instance, an aircraft from a Terminal 1 gate to Runway 33L has the zone list in the order of A South followed by A North.The deicing model will assign the aircraft to the zone in the front of the list if the zone's traffic load condition is under a preset threshold.A zone's traffic load condition is aircraft category dependent.This is because the deicing pads in a zone have different aircraft category support capabilities.For example, the A South zone has only two pads that can accommodate category F aircraft.For an aircraft of a given category, the zone's load condition is defined as the ratio of the number of aircraft in the same category currently assigned to the zone to the number of pads which can support the category or larger.For example, if the A South zone has three category F aircraft assigned to its two category F pads, the zone load condition is 1.5 for a category F aircraft.If the preset threshold is larger than the zone load condition, the algorithm allows additional aircraft to be assigned to this zone.Otherwise, the next zone in the list will be considered.If all the zones in the priority list can no longer take a new assignment, the aircraft will be held at gate until the zone traffic is reduced.With this heuristic, the larger the threshold, the larger the zone queue size.On the other hand, the smaller the threshold, the smaller the zone queue size, and more likely an aircraft will be held at the gate in heavy deicing demand.The deicing operator contract constraints are not modeled in this study. +A.3 Deicing TimeWhen an aircraft enters its assigned deicing zone, the model produces a time duration for the aircraft to stay in the zone.This is achieved by sampling the normal time distribution based on the actual deicing data analysis described in the previous section.The model takes aircraft category as input to account for the deicing time impact by aircraft size.This is done by using the unique mean deicing time by category.After the given time duration, the aircraft will be scheduled for takeoff by the scheduler. +B. Departure SchedulerThe departure scheduler uses a first come/ready, first served algorithm with a set of priority groups.It produces three schedule times to the simulator in this study, as follows:• Target off block time from gate to deicing zone for a departure requesting deicing • Target off zone time from deicing zone to runway for a departure after deicing• Target off block time from gate to runway for a departure not requesting deicing The scheduler calculates the target takeoff times for the departures along with the arrival landing times subject to the runway capacity constraints.The algorithm groups the aircraft in the following priorities in descending order to schedule the runway use times:1. Arrivals 2. Departures taxiing to runway 3. Departures ready in deicing zone after deicing 4. Departures ready at gate to runway Departures not ready at gate or deicing zone are not considered in runway scheduling.Within each of the groups, the first come/ready, first served rule applies.For two aircraft taxiing to the same runway, the earliest times of arrival at runway based on unimpeded taxi times are used to decide which one gets scheduled first.If two aircraft at gates are ready for the same runway, their ready times will determine the order of consideration.The gate pushback time to runway and the taxi clearance time from deicing zone to runway are calculated backwards from their target takeoff times, respectively.In this study, the target gate pushback time of a departure to the deicing zone is the earliest time that a zone is assigned, i.e., the scheduler clears a deicing aircraft at the gate once the aircraft has a zone assignment. +IV. SIMULATION SETUP AND VALIDATION +A. Simulation SetupNASA's Surface Operation Simulator and Scheduler (SOSS) [15] fast time simulator was used in the simulations.The traffic scenario was created based on the actual flight record on Dec 3, 2015 between 0800 to 1300 local hours.It was in the north flow configuration.Runway 33L was a departure only runway, 34 was an arrival-departure mix-use runway, and 33R an arrival only runway.The flight ready times of departures in the scenario were set as the actual gate pushback times of the operations.The arrival landing times matched the actual operations, too. +B. System ValidationBefore the simulations, the system was validated using the full day flight data on Dec 3, 2015.In this validation traffic data, Runways 33L and 34 were used for both arrivals and departures, and runway 33R was for arrivals only.In the validation run, aircraft followed the operation data in the records, including gate, runway, departure off block time, deicing zone assignment, and time spent inside the zone.Then, the results of the validation run were compared against the actual data.Adjustments to the simulator configuration parameters, such as engine spool-up time and taxi speed, were made until the validation was subjectively satisfied.The major validation metrics are shown in the following. +B.1 Taxi Out TimeFig. 5 is the comparison of mean and median values of the taxi out times.The taxi out time was measured from actual gate out time to wheels-off time.The deicing zone time was included in the taxi out time for deicing aircraft.The comparison indicates an overall good match of the simulation to the operation, except for the simulated non-deice aircraft to runway 34 taxiing a bit faster than operation (top middle plot).B.2 Runway Throughput Fig. 6 and Fig. 7 plot the accumulated runway throughput comparisons for 33L and 34, respectively.They are the accumulated counts of wheels-off (departures) and wheelson (arrivals) in local time.The closeness in the comparison gives reasonable confidence that the system is able to simulate the runway capacity appropriately. +V. RESULTS AND ANALYSISThe results described in this section were obtained from the three primary sets of 20 Monte Carlo simulation runs that varied by deicing zone load threshold.The perturbation variables of the deicing model included the departure deicing request rate and deicing time.The deicing request rate was set at 40%, matching the actual operation data for the busiest deicing day observed on Dec 3, 2015.The means and standard deviation of the deicing time were also set according to the operation data analysis (Fig. 3).Fig. 10 shows the simulated deicing time distributions in two aircraft wingspan category groups, C or D and E or F. The dominant aircraft categories are C and E shown in percentage in TABLE IV, where the mean deicing times from the simulations are compared to the values in the actual operation.They matched well in all categories.Three deicing zone assignment priority lists were configured according to the gate and runway assignments:• Terminal gates to Runway 33L: [A South, A North],• Cargo gates to Runway 33L: [D South, D North, A South], and • All gates to Runway 34: [M South, M North] Three deicing zone load thresholds, 100%, 150%, and 200%, were used, one for each of the three sets of Monte Carlo runs. +A. Zone AssignmentsFig. 11 shows the percentages of zone assignments from the three simulations of different deicing zone load thresholds.For comparison, the zone usage of the actual operation is also included in the last column.The model used the first-choice zones (A South, D South and M South) for the majority of the deicing requests.In the actual operation, however, the zone assignments appear more spread out.For example, 9% of departures to 33L used M North, which has a long taxi distance to the departure runway.One possible reason was the deicing operator and airline contract constraint, as mentioned earlier.Among the three simulations, the bigger the zone load threshold, the more aircraft tend to be assigned to the firstchoice zones.At the 200% threshold setting, almost all (98% and 100% for 33L and 34, respectively) deicing operations were assigned to the first-choice zones (A South, D South, and M South).At 100% threshold, which represents no overload of the zone capacity, on the other hand, 20% and 15% aircraft were assigned to the second-choice zones for 33L (19% A North and 1% D North) and 34, respectively.It indicates that at 40% deicing request rate and the overall departure traffic demand, the first-choice deicing zone capacity would be able to meet the deicing demand 80% to 85% of times.This assumes that each zone operates at its full capacity in this study. +B. Zone Queue SizeAlthough the first-choice zone in the heuristic model has the shortest taxi distance to the departure runway, overloading it may cause aircraft to incur additional waiting time in the deicing zone when a queue builds up.Fig. 12 and Fig. 13 show the zone queue sizes for 33L and 34, respectively.The zone queue size is defined as the number of aircraft either inside a zone or taxiing toward the zone.It is plotted as the average aircraft count in 10-minute bins along time.The mean and maximum values of the firstchoice zone queues for aircraft from the Terminal gates, A South and M South, are also displayed along the bar charts.The simulations show that overloading the first-choice zones increases their queue sizes, as expected.It will be interesting to see the potential aircraft taxi time change as the queue size grows, which will be analyzed later in this section. +C. Deicing Aircraft Gate HoldThe deicing model holds aircraft at the gate when all the deicing zones in the priority list have reached their loading threshold.To see how it works, two additional simulations were run with the deicing request rates at 30% and 50% without zone overloading, i.e., zone overload threshold of 100%.The data from these two runs were used only for the gate hold analysis.TABLE V shows the results together with the default 40% request rate.The percentage of deicing gate hold is the number of aircraft experiencing gate hold divided by the total number of deicing aircraft.The mean hold time is the total gate hold time divided by the number of holds.At 30% request rate, there was no gate hold, because of the low demand with respect to the zone capacity.From 40% to 50%, the simulations showed increased gate hold occurrence and time.Not only did the gate hold percentage increase, but the mean hold time also increased.Further research is needed to study the heavy deicing situation in the overall traffic management strategy. +D. Deicing Aircraft Taxi Out Time and PredictabilityThe taxi out time is measured from gate out to wheelsoff.It consists of the taxi time from gate to deicing zone, deicing time, and taxi time from deicing zone to wheels-off. +D.1 Mean Taxi Out TimesFig. 14 shows the mean taxi out times of the deicing aircraft.The bar charts visualize the total taxi out time, gate to zone, and zone to wheels-off times.For comparison, the mean taxi out times of actual operations are also plotted.The three simulations showed significantly shorter taxi times than the actual operations in all three measurements, likely because of the zone assignment differences due to deicing operator and airline contract constraints discussed earlier.Among the simulations, there were no noticeable differences in the gate to zone times.This is probably because of the tradeoff between the longer taxi distance when aircraft were assigned to the second-choice zones, and the extra waiting time in the first-choice deicing zone queue as discussed in the zone queue size analysis (Section V.B).On the other hand, the zone to wheels-off times show visible decrease as the threshold increases from 100% to 200%, corresponding to the increased use of first-choice zones.Note that less zone to wheels-off time is desirable for anti-ice fluid HOT compliance.The overall taxi out times appear to be influenced by the zone to wheels-off times, but the overall taxi out time differences between 150% and 200% thresholds are less noticeable. +D.2 Taxi Out Time PredictabilityFig. 15 shows the taxi out time variances in the same three taxi time measurements.The variance is represented by the standard deviation of the taxi time.Less variance leads to better predictability, which in turn, helps the scheduler build robust schedules.The overall taxi out variance is a combination of the variances of the three times, i.e., gate to zone time, deicing time in the zone, and zone to wheels-off time.In the three simulations, the deicing time was modeled with the same statistic.Therefore, the overall taxi out time predictability was a function of the gate to zone and the zone to wheels-off time variances.The results reveal that with increased zone overloading thresholds, the gate to zone time predictability decreased, but the zone to wheelsoff predictability improved.The likely reason for the degradation of the gate to zone time predictability is that more aircraft had to wait in the zone queues, which added more uncertainty to the taxi time.On the other hand, more first-choice zone assignments due to the increased zone overload helped the aircraft taxi shorter distances to the departure runways and therefore improved the taxi time predictability.The opposing interests in the deicing zone assignments for taxi time predictability for scheduling aircraft from gate to deicing zone and from zone to runway suggest a tradeoff solution may be considered in further investigation. +E. Zone and Runway ThroughputsTo analyze the relationship between the deicing zone and runway throughputs, Fig. 16 plots the accumulated zone and departure runway throughputs together.For each departure runway, the zone throughput is the total zone throughputs for the aircraft to the same runway.The results show that the runways are able to catch up with the deicing zone throughputs most of time.In other words, it was the zone operations that dictated the airport departure rate.One notable observation is at 100% threshold on 33L (top left plot) where the runway throughput underran the zone throughput for a time, perhaps due to the longer taxi times after deicing from A North zone.It should be noted, however, that the nominal runway departure rates were used in the simulations.Further study would look into whether a reduced runway operation rate should be imposed during snow weather conditions.The simulations showed the better performance, with less taxi out times and smaller taxi out time variance, than the actual operation.Among the three simulations, where different zone load thresholds were tested, the results showed no visible differences in average gate to zone times due to the tradeoff between the longer taxi distance when aircraft were assigned to the second-choice zone, and the extra waiting time in the first-choice zone queue.The average zone to wheels-off times showed a visible decrease as the zone load threshold increased, due to the increased assignments of first-choice zones, which is good for anti-ice fluid HOT compliance.The overall taxi out times were influenced mainly by the zone to wheels-off times.But between 150% and 200% thresholds, the differences of the average zone to wheels-off times were less noticeable.In addition to the average taxi out times, we also analyzed the taxi out time predictability over three segments: overall taxi out, gate to zone, and zone to wheels-off.The results indicated that in all three simulations the zone to wheels-off time had the best predictability, whereas the overall taxi out time showed the worst predictability.From the scheduling point of view, it makes sense to schedule deicing aircraft to depart from the zone rather than from the gate.Among the three simulations, when more first-choice zones were assigned, the predictability from gate to zone decreased, which would have a negative impact on scheduling aircraft from gate to zone.On the other hand, the predictability from zone to wheels-off improved, which would be beneficial to scheduling aircraft from zone to departure runway.These observations suggest a possible tradeoff in zone assignment strategy.Lastly, the deicing zone and runway throughputs were analyzed.The simulations showed that the runway throughputs closely followed zone throughputs.In other words, for deicing aircraft the zone throughput could be the main bottleneck to determine the overall departure throughput.For future work, the following improvements will be considered: 1) refine the deicing model to model dynamic zone capacity (e.g., deicing truck count), and the fatigue factor of the deicing service operators over time, 2) add zoneairline contract restrictions, and 3) include possible runway operation rate degradation under snow day conditions.Fig. 11Fig. 1 ICN Airport Layout +Fig. 22Fig. 2 Top Ten Daily Deicing Operations at ICN +Fig. 33Fig. 3 Zone Time Distribution III.DEICING MODEL AND DEPARTURE SCHEDULER The deicing model for ICN has been developed according to the current day deicing procedure and operation data analysis.It models three deicing related decision processes: • Pilot's request for deicing operation (for simulation only) • deicing zone assignment, and +Fig. 44Fig. 4 Data Flow among Deicing Model, Departure Scheduler and Simulator +Fig. 55Fig. 5 Taxi Out Times +Fig. 66Fig. 6 Accumulated Runway Throughput on 33L +Fig. 88Fig. 8 Departure Queue Count +Fig. 1010Fig. 10 Simulated Deicing Time Distributions +Fig. 1111Fig. 11 Zone Assignments +Fig. 1212Fig. 12 Zone Queue Size for 33L +Fig. 1414Fig. 14 Mean Taxi Out Times +Fig. 1515Fig. 15 Taxi Out Time Variances +Fig. 1616Fig. 16 Zone and Runway Throughputs VI.SUMMARY AND FUTURE WORK This paper described a deicing model developed for winter snow day operations at Incheon International Airport.Integrated with a departure scheduler, the model provided deicing zone assignment to the scheduler and modeled deicing times to the simulator in three Monte Carlo simulations by varying zone load threshold.The model's zone assignment logic uses a heuristic algorithm that considers the zone locations with respect to the gate and runway assignment as well as the zone workload conditions.The model does not reflect the constraints of the airline operator deicing contracts.The simulation traffic scenario was derived from the snow day operations at ICN on Dec 3, 2015.The simulations showed the better performance, with less taxi out times and smaller taxi out time variance, than the actual operation.Among the three simulations, where different zone load thresholds were tested, the results showed no visible differences in average gate to zone times due to the tradeoff between the longer taxi distance when aircraft were assigned to the second-choice zone, and the extra waiting time in the first-choice zone queue.The average zone to wheels-off times showed a visible decrease as the zone load threshold increased, due to the increased assignments of first-choice zones, which is good for anti-ice fluid HOT compliance.The overall taxi out times were influenced mainly by the zone to wheels-off times.But between 150% and 200% thresholds, the differences of the average zone to wheels-off times were less noticeable.In addition to the average taxi out times, we also analyzed the taxi out time predictability over three segments: overall taxi out, gate to zone, and zone to wheels-off.The results indicated that in all three simulations the zone to wheels-off +TABLE I .IDeicing Zones and PadsCategoryBCDEFA South232A North1041M South231M North21D South411D North411T Center44 +TABLE II .IIZone Usage in 2015-2016 Winter SeasonZoneA SouthM SouthA NorthM NorthD SouthD NorthUsage38%28%22%5%6%1%TABLE III. Zone Usage and Departure RunwaysRunway 33L, 34Runway 15R,16South Zones56%18%North Zones12%14% +TABLE IV .IVSimulated Deicing Time StatisticCategoryCDEFSimulations21.720.725.724.0(51%)(4%)(42%)(3%)Operation21.020.625.224.9 +TABLE V .VDeice Aircraft Gate HoldDeicingPercentage ofMean DeicingRequest RateDeicing Gate HoldGate Hold Time(%)(%)(min)3000402.244.89505.786.08 + + + + +ACKNOWLEDGMENTThis work is a collaborative effort by the KARI and NASA Ames Research Center.It was supported by KARI's MIDAS (Management on Integrated Operations of Departure, Arrival, and Surface) project and NASA's ATD-2 (Airspace Technology Demonstration 2) project.The authors are grateful to the IIAC operation management team for sharing its expert knowledge of ICN deicing operations. + + + + + + + + + Aircraft ground support equipment. Specific requirements + + Icao + + 10.3403/30299188 + + + 2016 + BSI British Standards + + + ICAO, "Aircraft Ground De-Icing/Anti-Icing Operations -ICAO Requirements and Guidance, 2016," http://www.icao.int/safety/airnavigation/OPS/Pages/Aircraft-Ground- De-IcingAnti-Icing-Operations.aspx + + + + + Fixed-wing advisory board seeks FAA circular + + Us Faa + + 10.1016/s0894-8321(88)80042-1 + + + + AeroMedical Journal + AeroMedical Journal + 0894-8321 + + 3 + 4 + 6 + 2004 + Elsevier BV + + + US FAA, "Advisory Circular: Ground Deicing and Anti-Icing Program, 2004, http://www.faa.gov/documentLibrary/media/Advisory_Circular/AC1 20-60B.pdf + + + + + Applying Whole Effluent Toxicity Testing to Aircraft Deicing Runoff + 10.17226/22125 + + + 2015 + Transportation Research Board + + + US EPA, "Preliminary Data Summary Airport Deicing Operations (Revised), 2015, https://www.epa.gov/sites/production/files/2015- 06/documents/airport-deicing-pds-2000.pdf + + + + + Departure Efficiency Benefits of Terminal RNAV Operations at Dallas-Fort Worth International Airport + + RalfMayer + + 10.2514/6.2006-7774 + + + + 6th AIAA Aviation Technology, Integration and Operations Conference (ATIO) + + American Institute of Aeronautics and Astronautics + + + + Dallas Fort Worth International Airport, "Aircraft Deice Operations Plan 2018-2019 Winter Weather Season," https://www.dfwairport.com/cs/groups/webcontent/documents/webas set/p2_626762.pdf + + + + + Airport De-icing Plan + + Charlotte-Douglas InternationalAirport + + + 2015-2016 + + + Charlotte-Douglas International Airport, "Airport De-icing Plan," 2015-2016, + + + + + Index to FAA Office of Aviation Medicine reports: 1961 through 1978. + + "faaUs Faa + + + TimeHoldover + + + Guidelines + + 10.1037/e440062004-001 + + + + American Psychological Association (APA) + + + US FAA, "FAA Holdover Time Guidelines," https://www.faa.gov/other_visit/aviation_industry/airline_operators/a irline_safety/media/FAA_2018-19_HoldoverTables.pdf + + + + + SESAR solution ID #116 De-icing Management Tool + + Sesar + + + + + SESAR, "SESAR solution ID #116 De-icing Management Tool,"https://www.sesarju.eu/sites/default/files/solutions/01_CN_Sol ution_116_De-icing_Management_Tool.pdf + + + + + Call for Papers + + MHayashi + + 10.1027/2192-0923/a000067 + + + Aviation Psychology and Applied Human Factors + Aviation Psychology and Applied Human Factors + 2192-0923 + 2192-0931 + + 4 + 2 + + June 2015 + Hogrefe Publishing Group + Lisbon, Portugal + + + M. Hayashi et al., "Evaluation of Pushback Decision-Support Tool Concept for Charlotte International Airport Ramp Operations," 11 th USA/Europe Air Traffic Management Research and Development Seminar (ATM2015), Lisbon, Portugal, June 2015. + + + + + Field Evaluation of the Baseline Integrated Arrival, Departure, Surface Capabilities at Charlotte Douglas International Airport + + YJung + + + + 13th USA/Europe Air Traffic Management Research and Development Seminar (ATM2019) + Vienna, Austria + + June 2019 + + + unpublished + Y. Jung et al., "Field Evaluation of the Baseline Integrated Arrival, Departure, Surface Capabilities at Charlotte Douglas International Airport," 13th USA/Europe Air Traffic Management Research and Development Seminar (ATM2019), Vienna, Austria, June 2019 (unpublished). + + + + + Improving runway queue management: Modifying SDSS to accommodate deicing + + BruceWilson + + + PatrickHurley + + + PaulDiffenderfer + + 10.1109/icnsurv.2011.5935292 + + + 2011 Integrated Communications, Navigation, and Surveillance Conference Proceedings + Herndon, VA, USA + + IEEE + 2011 + + + + Improving Runway Queue Management: Modifying SDSS To Accommodate Deicing + Bruce Wilson et al., "Improving Runway Queue Management: Modifying SDSS To Accommodate Deicing," IEEE Integrated communications navigation and surveillance conference proceedings, Herndon, VA, USA, pp 16-1 -16-12, 2011. + + + + + Airport Surface Traffic Scheduling with Consideration of De-Icing Operations + + YEun + + + September 9-14, 2018 + Belo Horizonte, Brazil + + + 31st Congress of the International Council of the Aeronautical Sciences + Y. Eun et al., "Airport Surface Traffic Scheduling with Consideration of De-Icing Operations," 31st Congress of the International Council of the Aeronautical Sciences, Belo Horizonte, Brazil, September 9-14, 2018. + + + + + The CDM Co-Ordinator + 10.1002/9781444302554.ch7 + + + + CDM Regulations 2007 Procedures Manual + + Blackwell Publishing Ltd. + + + + + Incheon International Airport Corporation, "A-CDM Operational Manual," https://www.airport.kr/co_file/ko/file06/A- CDM_Operation_Manual_ENG.pdf. + + + + + Operational Characteristics Identification and Simulation Model Verification for Incheon International Airport + + YeonjuEun + + + DaekeunJeon + + + HanbongLee + + + ZhifanZhu + + + YoonCJung + + + MyeongsookJeong + + + HyounkyongKim + + + EunmiOh + + + SungkwonHong + + + JunwonLee + + 10.2514/6.2016-3161 + + + 16th AIAA Aviation Technology, Integration, and Operations Conference + Washington, DC + + American Institute of Aeronautics and Astronautics + June 2016 + + + Y. Eun et al., "Operational Characteristics Identification and Simulation Model Validation for Incheon International Airport," 16 th AIAA Aviation Technology, Integration and Operations Conference, Washington, DC, June 2016. + + + + + Perhitungan Aerodrome Reference Code (ARC) pada Rumpin Airfield, Berdasarkan Regulasi International Civil Aviation Organization (ICAO) + + HartonoHartono + + + DanartomoKusumoaji + + + AbdulAziz + + + MuhammadRusmin + + 10.33506/rb.v6i2.1142 + + + + Jurnal Teknik Sipil : Rancang Bangun + JTSRB + 2476-8928 + 2614-4344 + + 6 + 2 + 52 + + Universitas Muhammadiyah Sorong + + + ICAO Aerodrome Reference Code, https://www.skybrary.aero/index.php/ICAO_Aerodrome_Reference_ Code + + + + + Validation of Simulations of Airport Surface Traffic with the Surface Operations Simulator and Scheduler + + RobertDWindhorst + + + JustinVMontoya + + + ZhifanZhu + + + SergeiGridnev + + + KatyGriffin + + + AdityaSaraf + + + SteveStroiney + + 10.2514/6.2013-4207 + + + 2013 Aviation Technology, Integration, and Operations Conference + Los Angeles, CA + + American Institute of Aeronautics and Astronautics + August 12-14, 2013 + + + R. D. Windhorst et al., "Validation of Simulations of Airport Surface Traffic with the Surface Operations Simulator and Scheduler," AIAA Aviation Technology, Integration and Operations Conference, Los Angeles, CA, August 12-14, 2013. + + + + + + diff --git a/file844.txt b/file844.txt new file mode 100644 index 0000000000000000000000000000000000000000..f69b2c0d7fe1af7d52afe6f3692f86bf7d70660b --- /dev/null +++ b/file844.txt @@ -0,0 +1,458 @@ + + + + +I. INTRODUCTIONDeparting flights at major U.S. airports are often subject to Traffic Management Initiatives (TMIs) in order to alleviate congestion and delay induced by the imbalance between air traffic demand and capacity across the National Airspace System.Strategic TMIs such as the Ground Delay Program and Airspace Flow Program produce an Expect Departure Clearance Time (EDCT) to flights at their departure airports to control the air traffic flow into constrained resources.One commonly used tactical TMI is the Approval Request (APREQ), which is typically issued by the Air Route Traffic Control Center (ARTCC or Center) to assign runway departure times (called release times) to affected flights at airports within the Center, merging into a congested overhead stream.Among all departures at Charlotte Douglas International Airport (CLT) from January 2018 through February 2019, 32,337 flights (10.6% of all departures) were subject to APREQ, EDCT, or both restrictions.Out of 26,733 APREQ flights (8.8% of all departures), 33.2% of these APREQ flights did not meet the given compliance window (within two minutes before and one minute after the assigned release time) [1].TMI non-compliance often result in lower predictability of overhead stream and thus airspace inefficiency, as well as underutilization of the airspace capacity.In addition, the efforts to meet the runway release times of controlled flights at the airport can lead to higher controller's workload, and possible surface delay and inefficient runway utilization.NASA has been developing and testing a suite of decision support capabilities for the Integrated Arrival, Departure, and Surface (IADS) operations.Under NASA's Airspace Technology Demonstration 2 (ATD-2) sub-project, through a close partnership with the Federal Aviation Administration, air carriers, airports, and general aviation community [2], the IADS system is being evaluated in a field demonstration conducted in three phases.The Phase 1 and 2 IADS capabilities provide enhanced operational efficiency and predictability of flight operations through data exchange and integration, surface metering, and automated coordination of release time of controlled flights for overhead stream insertion.The users of the IADS system include personnel at CLT Air Traffic Control Tower (ATCT), American Airlines ramp tower, CLT Terminal Radar Approach CONtrol (TRACON), and Washington and Atlanta ARTCCs.According to the operations data analysis, it is estimated that about 5.1 million pounds of fuel savings and 15.7 million CO2 emission reduction, equivalent to planting over 100,000 urban trees, were achieved during the period of field evaluation between October 2017 and April 2020 [3][4].Some of the benefits came from gate holds of departures, including both controlled and non-controlled flights subject to surface metering.The ATD-2 IADS system provides ramp controllers with pushback hold advisories for controlled flights at gates to meet the given release times, even when surface metering is not active.In addition to gate hold advisories and surface metering, another main capability in the ATD-2 IADS system provides electronic negotiation procedures for APREQ flights [5][6].Without the ATD-2 IADS system, the Call-For-Release (CFR) procedures that had been used at CLT to negotiate the release times of APREQ flights were followed through land-line voice communications.When a pilot of a flight under an APREQ restriction calls for pushback, the ATCT Traffic Management Coordinator (TMC) calls the Center TMC to request a release time and provides the best estimate of when the flight will be ready to depart from the runway.The Center TMC enters the projected runway departure time in the Time Based Flow Management (TBFM) system, assesses the availability of a slot at the meter point, and responds to the ATCT TMC with a release time that is predicted to enable the flight to fit into the overhead stream.Using the ATD-2 IADS system enables nonverbal, electronic coordination of release times at CLT. Prior to pushback, the ATD-2 surface scheduler estimates the Earliest Feasible Takeoff Times (EFTTs) of APREQ flights by which the aircraft will reach the runway with a high level of confidence.These times are shown on the timeline display for the ATCT TMC.When the ATCT TMC selects an APREQ aircraft on the timeline and requests its release time, the Integrated Departure Arrival Capability (IDAC) implemented in the TBFM system at the Center searches for potential windows of release times that would allow the aircraft to be inserted in the available slots in the overhead stream over the constrained meter point.IDAC calculates a runway release time based on the flight's EFTT and slot availability and sends the time electronically to the ATCT.The improved prediction accuracy of takeoff times by the ATD-2 surface scheduler enables the ATCT TMC to coordinate release times with the Center while the aircraft are still at the gates.The surface scheduler calculates Target Off-Block Time (TOBT) based on the negotiated release time.This would allow the controlled flight to be held at the gate until its TOBT and reach the runway to take off within the given compliance window.Also, the electronic coordination procedure makes the renegotiation process easier and faster when the aircraft is predicted to arrive at the runway earlier or later than its release time.The renegotiation of the APREQ time, even while taxiing after pushback, would allow the aircraft to take an earlier slot in the overhead stream, thus resulting in an earlier runway release time and taxi time reduction.The field data at CLT showed that the electronic release time negotiation provided by the ATD-2 IADS system improved APREQ compliance and reduced response time in approving release times [1].However, the impact of controlled flights on overall airport surface performance has not been evaluated.It is essential to investigate the TMI compliance together with other performance metrics such as taxi time and throughput in different operation situations.Fast-time simulation can be used to investigate the impact of controlled flights on airport performance that may not be clearly shown by actual data observations.Through fast-time simulation, for instance, the same traffic condition can be iterated with and without pushback hold advisories provided by ATD-2 surface scheduler.It can also aid in revealing the relationship between the APREQ compliance and other performance indicators such as taxi time reduction brought by ATD-2 gate hold advisories and surface metering.NASA has developed and improved a fast-time simulation tool for evaluating new concepts in airport surface operations, called Surface Operations Simulator and Scheduler (SOSS).From previous research [7][8], a SOSS simulation model for CLT was created and validated against actual operations data, and incorporated with the ATD-2 surface scheduler, providing pushback hold advisories for departures when excess taxi out time exceeds a target value.This model was used to evaluate the impacts of estimated flight ready times on surface metering [7] and of general aviation flights on airport performance [8].In this paper, fast-time simulations using SOSS are performed to evaluate the benefits of the ATD-2 IADS system capabilities with respect to controlled flights.SOSS calculates reasonable release times from available overhead slots, provides pushback advisories for the controlled flights even when surface metering is inactive, and evaluates the airport performance regarding taxi time, runway throughput, and departure queue length.For selected traffic scenarios during busy time periods at CLT having a typical number of controlled flights (i.e., about 10% of total departures), three cases are simulated depending on the control level of surface metering.In the first case, surface metering is off, in which all the departures push back from gates once they are ready, representing the normal operations prior to using the ATD-2 IADS system.In this case, a few controlled flights are expected to leave their gates too early and have long taxi times to meet the assigned release times.In the second case, surface metering is off, but pushback hold advisories are applied to controlled flights only, so that they are held at gates for a certain time to meet the assigned release times for takeoffs, instead of waiting in the queue.In the last case, surface metering is turned on, where a subset of departures, including both controlled and non-controlled flights, are advised to be held at gates to mitigate surface congestion and reduce excess taxi times.Comparison between these cases will show the benefits/costs of managing controlled flights aided by the ATD-2 surface scheduler and the effectiveness of surface metering when controlled flights exist.It is challenging to mimic various tactical operations of controllers in the fast-time simulation model.Controllers may initiate several different actions in order to meet the assigned release time.The controlled flight can be directed to either a hardstand in the ramp area or a designated holding area on the movement area; use by-pass taxiways for intersection takeoffs (takeoffs that start at some point other than the end of the runway, usually at an intersection of the runway with a taxiway); or assign a flight to wait on the opposite side of the runway, all while maintaining smooth takeoffs and landings with the maximum runway utilization.In this study, it is assumed that all the controlled flights use the inner queue taxiway for takeoffs, whereas non-controlled flights enter the runway through the outer queue taxiway.In this way, a controlled flight can wait for takeoff until the given release time, if it arrives at the runway too early.This assumption also allows a late controlled flight to cut in line and take off right away.Renegotiation of the release time, which is available in the ATD-2 IADS system at the field, is not considered as part of this study.With these modeling approaches and assumptions, Section II describes the fast-time simulation environment, including the CLT airport configuration, traffic scenarios, simulation setup, and performance metrics.It also describes the TBFM assigned delay model to generate reasonable release times for APREQ flights based on historical data.Section III compares the simulation results between the three cases described above, in terms of compliance rate, surface efficiency, and controller's potential workload.This paper concludes with a results summary and future work in Section IV. +II. SIMULATION ENVIRONMENT +A. CLTThe simulations used a model of the CLT north flow configuration.According to an earlier study [1], the north flow configuration has a higher percentage of APREQ flights compared to the south flow configuration.Fig. 1 shows the airport layout of CLT with three parallel runways (36R/18L, 36C/18C, 36L/18R) and one diagonal runway (23/5).In the north flow configuration, all three parallel runways, 36R, 36C and 36L, are used for arrivals, and two runways near the main terminal, 36R and 36C, are used for departures.Runway 5 is not used for takeoff or landing.By-pass taxiways are used for intersection departures to make last minute fine-tuned adjustments, usually to have a controlled flight depart ahead of some other flights already in the departure queue or to slightly delay a controlled flight when it would arrive at the runway early, normally more than 4-5 minutes before its release time.As shown in Fig. 1, there are two departure queues, which are inner and outer queues, for departures in each runway, 36R and 36C. +B. Simulation SetupThe fast-time simulations in this study used NASA's Surface Operations Simulator and Scheduler (SOSS) [7][8][9][10] connected to the ATD-2 Tactical Surface Scheduler [11] through the Surface Modeler, as shown in Fig. 2. Apart from facilitating a smooth exchange of input and output between SOSS and the scheduler, the Surface Modeler also contains an Earliest Off-Block Time (EOBT) model [7] and TBFM assigned delay model.The EOBT model provides the estimated flight ready time updates for commercial airline aircraft in the main terminals.The release times for APREQ flights are computed using the TBFM assigned delay model, which will be explained in Section II.C.The traffic sscenarios used in the simulation were created based on actual flight data at CLT for four days during Bank 2, historically one of the most congested time periods (9-11am) at CLT. Table I shows the actual number of departures, arrivals, and APREQ flights on each of the four selected scenario dates.To assess the impact of controlled flights on airport performance, three configuration cases for each scenario were designed.In Case 1, surface metering is off, and all departures are cleared to push back from the gates when ready.In Case 2, surface metering is off, but the controlled flights receive pushback time advisories from the Tactical Surface Scheduler.In Case 3, surface metering is on, and non-controlled flights are subject to metering hold at the gates when the airport surface is congested.The non-controlled flights receive the same pushback time advisories as in Case 2. In all three cases, the simulation tries to meet the release times of the controlled flights calculated by the TBFM assigned delay model.In the given scenarios, several APREQ and EDCT flights were observed.To simplify the simulation, this study focuses on the APREQ flights only, which means that controlled flights hereinafter are equivalent to APREQ flights. +C. TBFM Assigned Delay ModelFor an APREQ flight, its release time is computed as the sum of the EFTT estimated by the scheduler and additional delay assigned by TBFM, considering the slot availability in the overhead stream, called TBFM assigned delay.This is depicted in Fig. 3.In this study, a statistical model to generate the TBFM assigned delay was developed based on actual operations data at CLT, since the proposed fast-time simulation framework was not connected to the TBFM system nor did it have controller inputs.The model was based on actual TBFM assigned delay values from four months of historical data from January and February of 2018 and 2019.About 60% of the TBFM assigned delay values for the APREQ compliant flights in actual data were zero seconds, whereas the rest of the 40% had positive TBFM assigned delay values.Several probability distribution models were fitted to the data to find the best fit.Generalized Extreme Value (GEV) probability distribution model [12] fitted the positive TBFM assigned delay data the best.Fig. 4 shows the actual data for the month of January 2018 and the two closest fitting probability distributions, GEV and Loglogistic.Apart from visual comparison, both the Anderson-Darling test [13] and One-sample Kolmogorov-Smirnov test [14] assessed the GEV distribution as the better fit to the actual data.The GEV distribution parameters based on the actual data are shown in Table II. +TABLE II TBFM ASSIGNED DELAY MODEL PARAMETERSThe proposed TBFM assigned delay model sets 60% of the controlled flights to have zero for their TBFM assigned delay values, and the remaining 40% of the APREQ flights follow the GEV probability distribution using the parameters listed in Table II.For model validation, the TBFM assigned delay values generated from the proposed model were compared with actual data, as shown in Fig. 5.The two-sample Kolmogorov-Smirnov test indicated that the two TBFM assigned delay value distributions from the actual data and the proposed model were similar. +D. Simulation LimitationsHuman interventions by the ATCT TMC and controllers play a critical role in controlled flight operations on the airport surface.For instance, when the TMC notices that a controlled flight already taxiing in the movement area has a release time that is either 20 minutes in the future or predicted to miss its release time, the TMC will try to renegotiate a new release time at an earlier or later time, respectively.In addition, ground controllers use various tactics to insert controlled flights in the right position of a departure sequence to meet the release times while not blocking taxiway traffic, such as holding them at the hardstand or taxiway and using the by-pass taxiway for intersection takeoffs via the inner queue.However, it is challenging to model the controller's tactical maneuvers for the controlled flights in fast-time simulations.Therefore, some assumptions were made in the simulations for this study, considering the following limitations:• The release time renegotiation after pushback was not considered.In other words, the release times were assumed final at pushback time.According to operational data at CLT in January 2018, about 22.5% of APREQ flights had release times updated when the aircraft were taxiing in the ramp and airport movement areas.In other words, about 77.5% of APREQ flights had their final release times at or before pushback.• In this study, it is assumed that a controlled flight absorbs its excess taxi out time in the by-pass departure queue only.In the simulations, the scheduler can hold a controlled aircraft only at the gate.Once it started taxiing, there was no hold short maneuver at the taxiway intersections instructed by the scheduler.In addition, to prevent a controlled flight waiting for its release time in the departure queue from blocking the traffic behind, all the controlled aircraft were assigned to the by-pass departure queue (inner queue shown in Fig. 1), and the non-controlled aircraft to the normal, outer queue.• The use of the inner queue for the intersection takeoff approach worked well in the simulations except when two controlled flights were taxiing to the same runway within a short time window and the release time of the leading aircraft was far in the future.In real operations, the controller could hold the leading aircraft in a holding area early or divide the two aircraft between different departure queues.In the simulation, however, this situation could result in the second controlled aircraft extruding from the inner queue and blocking traffic.Fig. 6 illustrates an example of such a situation at Runway 36C, where two controlled flights (highlighted in yellow) block the non-controlled flights (highlighted in cyan) from entering the outer queue.In this study, a few runs had to be eliminated in order to avoid this gridlock situation. +III. SIMULATION RESULTS AND ANALYSISIn this study, forty simulation runs were implemented with perturbated variables for each scenario and each case for data collection.The three cases tested include: Case 1) Metering off, all departure flights push back when ready; Case 2) Metering off, controlled flights held at gates as advised; and Case 3) Metering on, all departure flights follow gate hold advisories.The perturbation variables were the TBFM assigned delay, EOBT, and pushback ready time.The TBFM assigned delay came from the model described in the previous section.The EOBT values were produced by the EOBT model developed in [7].The pushback ready times were generated by SOSS.This section presents the simulation results and analysis from the first scenario on 1/24/2018 which had the most APREQ flights, but the simulation results from other scenarios showed similar trends.Table III lists the number of flights and runway assignments in the scenario.The number of APREQ flights in parentheses are included in the total counts of departure flights. +A. APREQ compliance under three conditionsThe APREQ compliance was measured as the difference between the release time and the actual wheels-off time, in minutes.The compliance window for APREQ flights is within two minutes before to one minute after the release time.The compliance rates of Case 2 and Case 3 were very close to each other at 65-66% which were also close to the average number found in actual operations [1].Case 1 showed higher performance by about eight percent.When a controlled flight arrived at the inner queue before its compliance window, it waited in the queue.Thus, those controlled flights that taxied to the runway in time would most likely meet the compliance window.On the other hand, if a controlled flight was not able to taxi to the runway before the release time window, i.e., arriving in the inner queue either already inside the compliance window or later, it would not wait.Those aircraft that arrived at the runway after the compliance window would be marked as non-compliant.For those aircraft that arrived at the runway inside the -2/+1-minute compliance window, some of them were able to take off to meet the release times.The others, however, could still miss the compliance window, if there was an arrival landing or if another non-controlled departure was already in the runway position for takeoff.In real operations, if a controlled flight is predicted in advance to arrive in the queue later (or earlier) than the release time, the ATCT traffic manager may renegotiate with the Center TMC for a later (or earlier) release time to avoid the non-compliance situation.This renegotiation may cause extra workload to both the traffic manager and the controller.However, the SOSS simulation lacks this capability of human intervention as described earlier.Although the controlled flights in Case 1 had a better compliance rate, they spent more time on the surface.Fig. 8 shows the relation between the amount of waiting time in the inner queue and the compliance.In the scatter plots, the inner queue waiting time, on the horizontal axis, was measured as the time a controlled aircraft waited in the queue before takeoff.The corresponding compliance value is plotted along the vertical axis.Each marker represents a controlled flight's compliance versus its inner queue waiting time.The area under the dashed line represents the compliance window of -2/+1 minute from the release time.The plot shows that in the Case 1 condition, where the aircraft pushed back from the gate when ready, more controlled flights waited in the queue and spent longer time there until the given release time as well, compared to Case 2 and Case 3 where the controlled flights were subject to gate hold advisories.The controlled flights in Case 1 achieved better overall compliance at a higher cost through increased taxi time and more fuel burn.Again, human interventions such as trying to renegotiate for an earlier release time, if available, would help reduce the excess taxi time in this case.In all three cases, more aircraft going to Runway 36R missed the compliance window in the simulations shown by the orange markers at the zero inner queue time and above the horizontal dash line.This is also evident in the runway breakdown compliance rates shown in Fig. 9.The reason appeared to be that the tactical scheduler used in the simulation underestimated taxi times when assigning release times to 36R, which may be due to the heavily mixed departure and arrival traffic at this runway during the bank.In summary, among the three simulation conditions, Case 1, where the controlled flights were not subject to gate hold, showed better compliance rate, but the aircraft had to absorb higher taxi times on the airport surface than the other two cases.The compliance performance of the controlled flights taking off from Runway 36R was evidently lower than Runway 36C in all three cases, which was likely caused by the underestimated taxi out times due to the heavier traffic with more arrivals at the runway.Both situations, i.e., arriving at the runway earlier and later, could be mitigated by human interventions in real operations, but they were not available in the simulations. +B. Surface Efficiency and Runway ThroughputThis section examines the airport performance metrics related to efficiency and throughput, including gate hold, taxi out time, and departure throughput by runway.In the Case 2 and Case 3 conditions, the Tactical Surface Scheduler calculates the TOBTs for controlled aircraft at gates.In Case 2, where surface metering is off, only the APREQ flights were given TOBTs as pushback advisories for controllers to meet the release time, whereas in Case 3 (surface metering on) both controlled and non-controlled flights were given TOBTs based on the surface metering algorithm.The objective of surface metering is to reduce excess taxi out times by holding departing aircraft at the gates, thus effectively shifting some surface delay from the queue to the gate.Fig. 11 shows the total gate hold time for each case.As designed in the simulation setup, Case 1 had no gate hold.Case 2 and Case 3 showed a similar amount of total gate hold time (22.3 and 21.0 minutes) for the APREQ flights.The gate hold for these controlled flights helped reduce the inner queue waiting time, as shown in the previous inner queue time analysis.In Case 3, the non-controlled flights were also subject to gate hold due to surface metering, which can help reduce excess taxi out time and surface congestion.Table IV shows the total and average (in parentheses) hold times in minutes grouped by the two runways for the Case 2 and Case 3 conditions.The data show that in both cases, Runway 36R had more total gate hold time for the APREQ flights than 36C because of five APREQ flights were assigned to 36R and only three to 36C.For the non-controlled flights, Case 3 had significantly more gate hold time at 36R than 36C.This was probably due to more arrivals landing on Runway 36R, and the scheduler had to hold departures at the gates longer to manage the taxi out times.Table V shows the gate hold percentages for Case 3. Between the two departure runways, the total percentage of aircraft that were assigned to Runway 36R but held at the gates (25.4%) was more than twice of those departing from Runway 36C (10.5%) because of the heavier traffic at 36R.Overall, between the controlled and non-controlled flights, more than half of the controlled flights (54.1%) were held at the gates, and less than twenty percent (16.5%) of the non-controlled flights were held at the gates.This reflects the difference in scheduling objectives.The scheduling for controlled flights aims to deliver the aircraft to runways to meet the release times; where more flights have future release times, more gate hold can be advised.On the other hand, the scheduling strategy for non-controlled flights is to reduce their excess taxi out time.Fig. 12 illustrates the total taxi out time for the three cases.The taxi out time was measured as the duration from pushback start to wheels-off.For the controlled flights, it includes the waiting time in the departure queue for their release times.In Fig. 12, the top bar chart shows the total taxi out times for the APREQ flights, and the bottom bar chart shows the total taxi out times for the non-controlled flights.For the controlled flights to either runway, the taxi out times show a decreasing trend from Case 1 to Case 3. In Case 1, the controlled flights had more taxi out time.The extra taxi out time included the waiting time in the inner queue, as shown earlier in Fig. 10.In Case 2 and Case 3, the controlled flights spent less time on the surface than in Case 1.However, their compliance rates were also lower as indicated in Fig. 7. Despite the small amount of difference among the three cases, they suggest a possible tradeoff between the two performance metrics.For the noncontrolled flights, a slight decreasing trend from Case 2 to Case 3 is also visible.In Case 1 and Case 2, the non-controlled flights were not metered, so they showed similar performance.Compared to Case 2, Case 3 showed better taxi performance because of the gate hold applied to both the controlled and noncontrolled flights.The amount of improvement at 36C was less than that at 36R since less aircraft going to 36C were held, as shown at Table V.The average taxi out time per flight is shown in Fig. 13.The controlled flights produced greater taxi out times than the noncontrolled flights due to the controlled release time conformance.In fact, the average difference in Case 1 was about four minutes, compared to two minutes for Case 2 and Case 3. The average numbers exhibited the similar decrease trend from Case 1 to Case 2 and Case 3, which is consistent with the total taxi out time numbers.The runway throughput comparison is shown in Fig. 14 (a) and (b).The throughput includes both controlled flights and non-controlled flights, as well as arrivals.Each bar represents the runway usage count in a 5-minute time bin.At Runway 36C, the peak departures happened between 40 and 60 minutes into the simulations in all three cases, with five departures in five minutes that corresponded to about 60-second runway separation.No loss of departure throughput was found.At Runway 36R, many arrival landings (green bars) can be found.The highest runway usage occurred at 90 minutes into the simulations, where six combined departure and arrival runway operations were observed.In general, the runway throughput charts for the three cases were very similar to each other because the simulations filled the runway slots with either controlled or non-controlled flights, whenever they were ready to take off, without leaving any gaps. +C. Controller's potential workloadHuman controllers are an integral part of airport surface operations and managing their workload is important to system performance.One of the expected benefits of providing decision support tools is to enhance system efficiency without increasing controller workload [15][16].Although it is beyond the capability of fast-time simulations to directly measure controller workload, there are two metrics in this study that can be explored to compare the notional workload differences between the three cases.The assumption is that workload is associated with the number and duration of aircraft operating on the airport surface.The first metric is the departure aircraft count in the movement area.A congestion factor was derived from this metric.It was computed as the fraction of time that the departure aircraft count in the movement area exceeds a threshold value over the whole simulation time.The larger the fraction value is, the longer the surface is congested.For example, if the number of departures on the surface has been greater than a threshold for half the simulation time, the congestion factor would be 0.5.The second metric is the direct measurement of inner queue waiting time.In actual operations, if a controlled flight experiences a lengthy delay, a controller may hold it short on taxiways and use the by-pass intersection takeoff through an inner queue until the flight's assigned release time, which would require the controller to monitor the situation.In this study, the inner queue was used to absorb the surface delay of controlled flights.So, the amount of inner queue waiting time was used as an indicator of the overall delay consumed on the surface.Therefore, the longer a controlled aircraft waited in the queue, the more workload was deemed to be required by the controller.Table VI displays the two measurements along with the compliance rate for the APREQ flights assigned to Runway 36R, the busier of the two runways.The threshold value for the aircraft count on the movement area was set to ten for the congestion factor.This value was arbitrarily chosen for comparison purposes only.It does not imply a subjective mental threshold of the surface congestion condition to controllers.The compliance rate and the congestion factor are shown in a ratio, and the inner queue time is in minutes.The results indicate that in the Case 1 condition, where the controlled flights pushed back when ready, had the highest compliance rate, but at the same time, may require more controller attention because of the long inner queue waiting time.Case 2 and Case 3 experienced relatively lower compliance rates, but workload demand was lower as expressed by the inner queue time.Case 1 and Case 2 showed the same congestion factor, because surface metering was off in both cases.In contrast, Case 3, where surface metering was on, showed noticeable drop in the congestion level.It is evident that controlled flight operations may have directly impacted controller's workload.A human factors investigation of the relationship between controlled flight compliance and controller's workload is warranted and can be a potential topic for future work. +IV. CONCLUSIONSIn this work, a fast-time simulation-based study was conducted to investigate the effects of controlled flight operations on airport performance during Bank 2 at Charlotte Douglas International Airport (CLT).First, a TBFM assigned delay model was created using actual operational data at CLT.In the simulations, the delay value was added to the Earliest Feasible Takeoff Times (EFTTs) of controlled flights, estimated by the ATD-2 Tactical Surface Scheduler, in order to obtain realistic release times for the controlled flights.Depending on the level of gate holding applied to controlled and noncontrolled flights, three simulation cases were configured with NASA's fast-time simulation engine (SOSS), ATD-2 Tactical Surface Scheduler, and the release time model.In the first case, surface metering was off, and neither controlled flights nor noncontrolled flights were held at the gates.All departures pushed back at their ready times.In the second case, surface metering was inactive, but the controlled flights were subject to gate hold by the scheduler to meet their release times.In the third case, surface metering was on, and so all the controlled and noncontrolled departure aircraft were subject to gate hold.Because there were no human interventions in the fast-time simulation environment, release time renegotiation while the aircraft was taxiing was not considered, and all the controlled flights were assumed to use the by-pass inner queues for intersection takeoffs.In summary, simulation results showed that:• Without surface metering or gate holds (Case 1), better release time compliance rates were observed because the controlled aircraft left the gate earlier once they were ready to push back.However, they spent more taxi out time on the airport surface.The taxi out time included the time waiting in the departure runway queue before their release times.The waiting location could be at the hardstand in the ramp or at a designated area on the movement area in real operations.• The controlled flights in the other cases, Cases 2 and 3, with gate holds produced shorter taxi out times and benefited from pushback hold advisories, but had lower release time compliance rates, compared to Case 1.• Surface metering (along with gate holds) in Case 3 showed additional taxi out time reduction over gate holds alone for controlled departures in Case 2, because some of the non-controlled departures were also held at the gates, which helped reduce surface congestion.• Between the two mixed-use runways of 36C and 36R, the controlled flights to 36R showed a worse compliance rate compared to 36C due to more arrival traffic which led to underestimated taxi out time to the runway.The compliance performance difference was consistent with actual operational data analysis [1].• Runway throughput of all departures among the three cases showed similar performance.One of the key performance metrics for controlled flights is the compliance of their release times.It would be desirable for a controlled flight to taxi to the runway earlier rather than later for compliance.However, arriving too early would result in extra taxi out time and increase controller workload.Using the amount of excess taxi out time of the controlled flights and the departure count on the surface, a notional workload indicator was calculated for the three cases.It showed that in without surface metering or gate holds, where the controlled departures push back at their ready times, the notional controller workload was higher than the other two cases.The TMC and controllers play significant roles in controlled flight operations.The lack of human intervention logic in the fast-time simulations, such as release time renegotiation during taxi, imposed limitations to this study.Future work may consider adding a controller heuristic logic to model controller inputs and support release time renegotiation during taxiing.The scenarios in this study were simplified to have APREQ flights only as controlled flights, but the future study can be extended to investigate more complicated scenarios having both APREQ and EDCT flights that have different compliance windows.Fig. 1 .1Fig. 1.CLT airport layout (as of 2018) +Fig. 22Fig. 2 Data flow between SOSS, Surface Modeler, TBFM Assigned Delay Model, and Tactical Surface Scheduler +Fig. 3 .3Fig. 3. Definition of APREQ times +Fig. 4 .4Fig. 4. Positive TBFM assigned delay distribution +Fig. 5 .5Fig. 5. TBFM assigned delay model validation +Fig. 6 .6Fig. 6.Controlled flight blocks departure traffic behind +Fig. 77Fig. 7 shows the histograms of the overall APREQ compliance for the three configuration cases.The bin size of the histograms is one minute.The two vertical dashed lines represent the -2/+1 minutes of the compliance window.The compliance rate, annotated at each histogram, is the fraction of the APREQ flights whose compliance values fall inside the compliance window. +Fig. 7 .7Fig. 7. Histogram of APREQ compliance rates for three cases +Fig. 8 .8Fig. 8. Compliance vs. inner queue waiting time (in minutes) +Fig. 9 .9Fig. 9. Compliance rate by runway +Fig. 1010Fig.10displays the average controlled flight inner queue waiting times per aircraft in minutes for the three cases.Case 1 had longer queue times than Case 2 and Case 3 because the aircraft had pushed back at their ready times and tended to arrive at the runway before the release time window.The inner queue times of Case 2 and Case 3 were comparable since in both cases, the controlled aircraft were subject to gate hold for their release times.Between the two runways, the average queue times for 36C were longer than 36R, suggesting more controlled aircraft taxied to 36C with more breathing room for the compliance window. +Fig. 10 .10Fig. 10.Mean inner queue time per flight (in minutes) +Fig. 11 .11Fig. 11.Total gate hold time (in minutes) +Fig. 12 .12Fig. 12.Total taxi out times (in minutes) +Fig. 13 .13Fig. 13.Average taxi out time per flight (in minutes) +Fig. 1414Fig.14 (a).Runway 36C departure throughput +TABLE I FLIGHTIArrival Count36C36RAll36C36R36LAll1/24/201846 | 343 | 597 (89 | 8)103647931/25/201850 | 334 | 390 (84 | 6)1038561042/13/201856 | 332 | 394 (88 | 6)13753912/18/201851 | 233 | 389 (84 | 5)63857101COUNTS FOR TRAFFIC SCENARIOSScenario DatesDeparture Count (Non-controlled | APREQ) +TABLE IV .IVGATE HOLD TIME BREAKDOWNCase 2Case 3RunwayControlledNon-controlledControlledNon-controlled36C9.2 (5.6)07.4 (4.8)11.9 (3.4)36R13.2 (5.2)013.6 (5.1)41.7 (4.4) +TABLE V .VPERCENTAGE OF FLIGHTS HELD AT GATES IN CASE 3Runway36C36ROverallControlled55.0%53.5%54.1%Non-controlled7.6%25.9%16.5%Total10.5%25.4%17.9% +TABLE VI .VICONGESTION FACTOR VS.NOTIONAL WORKLOADCase 1Case 2Case 3Compliance rate0.610.480.52Congestion factor0.070.070.01Inner queue time (min)3.20.70.9 + + + + +ACKNOWLEDGMENTThe authors would like to thank Mr. Pete Slattery, ATC/Traffic Management Specialist, for sharing his knowledge and experience about air traffic control at CLT, especially in handling controlled flights. + + + + + + + + + Electronic Departure Approval Requests in ATD-2 Daily Operations + + ToddJCallantine + + + RobertStaudenmeier + + + LindsayStevens + + + JeremyCoupe + + + AndrewChurchill + + 10.2514/6.2019-2934 + + + AIAA Aviation 2019 Forum + Dallas, TX + + American Institute of Aeronautics and Astronautics + 2019. June 2019 + + + T. J. Callantine, R. Staudenmeier, L. Stevens, W. J. Coupe, and A. Churchill, "Electronic Departure Approval Requests in ATD-2 Daily Operations," 2019 AIAA Aviation Forum, Dallas, TX, June 2019. + + + + + Airspace Technology Demonstration 2 (ATD-2) phase 1 Concept of Use (ConUse) + + YJung + + + February 2018 + + + Y. Jung, et al., "Airspace Technology Demonstration 2 (ATD-2) phase 1 Concept of Use (ConUse)," February 2018. + + + + + Field evaluation of the baseline integrated arrival, departure, and surface capabilities at Charlotte Douglas International Airport + + YCJung + + + WJCoupe + + + ACapps + + + SEngelland + + + SSharma + + + + 13th USA/Europe Air Traffic Management Research and Development Seminar + + June 2019 + Vienna, Austria + + + Y. C. Jung, W. J. Coupe, A. Capps, S. Engelland, and S. Sharma, "Field evaluation of the baseline integrated arrival, departure, and surface capabilities at Charlotte Douglas International Airport," 13th USA/Europe Air Traffic Management Research and Development Seminar, Vienna, Austria, June 2019. + + + + + ATD Quart-Monde + 10.1007/springerreference_75803 + + May 2020 + Springer-Verlag + + + ATD-2 team. revision v2 + ATD-2 team, "ATD-2 Benefits Mechanism," revision v2, May 2020. + + + + + Evaluation of approval request/call for release coordination procedures for Charlotte Douglas international airport + + LindsayStevens + + + LynneMartin + + + EricChevalley + + + HanbongLee + + + KimberlyJobe + + + SavitaVerma + + + BonnyParke + + + VictoriaDulchinos + + 10.1109/dasc.2017.8101983 + + + 2017 IEEE/AIAA 36th Digital Avionics Systems Conference (DASC) + St. Petersburg, FL + + IEEE + September 2017 + + + L. Stevens, L. Martin, K. Jobe, B. Parke, E. Chevalley, H. Lee, S. Verma, and V. Dulchinos, "Evaluation of Approval Request/Call for Release Coordination Procedures for Charlotte Douglas International Airport," 36th Digital Avionics Systems Conference (DASC), St. Petersburg, FL, September 2017. + + + + + Evolution of Electronic Approval Request Procedures at Charlotte Douglas International Airport + + LindsayStevens + + + ToddJCallantine + + + RobertStaudenmeier + + 10.1109/dasc.2018.8569318 + + + 2018 IEEE/AIAA 37th Digital Avionics Systems Conference (DASC) + London, UK + + IEEE + September 2018 + + + L. Stevens, T.J. Callantine, and R. Staudenmeier, "Evolution of Electronic Approval Request Procedures at Charlotte Douglas International Airport," 37th Digital Avionics Systems Conference (DASC), London, UK, September 2018. + + + + + Fast-Time Simulation for Evaluating the Impact of Estimated Flight Ready Time Uncertainty on Surface Metering + + HanbongLee + + + YoonCJung + + + ShannonJZelinski + + + ZhifanZhu + + + VaishaliHosagrahara + + 10.1109/dasc43569.2019.9081711 + + + 2019 IEEE/AIAA 38th Digital Avionics Systems Conference (DASC) + San Diego, CA + + IEEE + 2019 + + + H. Lee, Y. C. Jung, S. J. Zelinski, Z. Zhu, and V. Hosagrahara, "Fast- Time Simulation for Evaluating the Impact of Estimated Flight Ready Time Uncertainty on Surface Metering," 38th Digital Avionics Systems Conference (DASC), San Diego, CA, 2019. + + + + + Impact of General Aviation Operations on Airport Performance Through Fast-Time Simulations at Charlotte-Douglas International Airport + + ZhifanZhu + + + VaishaliAHosagrahara + + + HanbongLee + + + YoonCJung + + + DeborahLBakowski + + 10.2514/6.2020-2916 + + + AIAA AVIATION 2020 FORUM + + American Institute of Aeronautics and Astronautics + 2020. June 2020 + + + Z. Zhu, V. Hosagrahara, H. Lee, Y. C. Jung, and D. L. Bakowski, "Impact of General Aviation Operations on Airport Performance Through Fast- Time Simulations at Charlotte Douglas International Airport," 2020 AIAA Aviation Forum, June 2020. + + + + + Validation of Simulations of Airport Surface Traffic with the Surface Operations Simulator and Scheduler + + RobertDWindhorst + + + JustinVMontoya + + + ZhifanZhu + + + SergeiGridnev + + + KatyGriffin + + + AdityaSaraf + + + SteveStroiney + + 10.2514/6.2013-4207 + + + 2013 Aviation Technology, Integration, and Operations Conference + Los Angeles, CA + + American Institute of Aeronautics and Astronautics + August 2013 + + + R. D. Windhorst et al., "Validation of simulations of airport surface traffic with the Surface Operations Simulator and Scheduler," AIAA Aviation Technology, Integration and Operations Conference, Los Angeles, CA, August 2013. + + + + + Assessing tactical scheduling options for time-based surface metering + + ShannonZelinski + + + RobertWindhorst + + 10.1109/dasc.2017.8101901 + + + 2017 IEEE/AIAA 36th Digital Avionics Systems Conference (DASC) + St. Petersburg, FL + + IEEE + September 2017 + + + S. Zelinski and R. Windhorst, "Assessing tactical scheduling options for time-based surface metering," 36th Digital Avionics Systems Conference (DASC), St. Petersburg, FL, September 2017. + + + + + ATD-2 Phase 3 Scheduling in a Metroplex Environment Incorporating Trajectory Option Sets + + WilliamJCoupe + + + YoonJung + + + LiangChen + + + IsaacRobeson + + 10.1109/dasc50938.2020.9256509 + + + 2020 AIAA/IEEE 39th Digital Avionics Systems Conference (DASC) + 13th USA/Europe Air Traffic Management Research and Development Seminar + Vienna, Austria + + IEEE + June 2019 + + + W. J. Coupe, H. Lee, Y. Jung, L. Chen, and I. Robeson, "Scheduling improvements following the phase 1 field evaluation of the ATD-2 integrated arrival, departure, and surface concept," 13th USA/Europe Air Traffic Management Research and Development Seminar, Vienna, Austria, June 2019. + + + + + The Appropriate Extreme Value Distribution for Extreme Returns: A Look at GEV& GL + + YuanAn + + 10.14738/assrj.311.2371 + + + + Advances in Social Sciences Research Journal + ASSRJ + 2055-0286 + + 3 + 11 + + Scholar Publishing + + + Generalized Extreme Value distribution, https://gmao.gsfc.nasa.gov/research/subseasonal/atlas/GEV-RV- html/GEV-RV-description.html + + + + + The Kolmogorov-Smirnov Goodness-of-Fit Test for a Single Sampl + 10.1201/9781420036268.ch7 + + + + Handbook of Parametric and Nonparametric Statistical Procedures + + Chapman and Hall/CRC + + 14 + + + Anderson-Darling Test, https://www.itl.nist.gov/div898/handbook/eda/section3/eda35e.htm [14] Kolmogorov-Smirnov Goodness-of-Fit Test, https://www.itl.nist.gov/div898/handbook/eda/section3/eda35g.htm + + + + + Human-in-the-Loop (HITL) Simulation -Final Results + + LK SStevens + + + YCJung + + + WJCoupe + + + HLee + + + BKParke + + 09%20ATD2_ART2_HITL_Results_2 019-03-06.pdf + + + + Assessment of Ramp Times + + March 2019 + + + Stevens, L. K. S., Jung, Y. C., Coupe, W. J., Lee, H., Parke, B. K., "Assessment of Ramp Times (ART 2) Human-in-the-Loop (HITL) Simulation - Final Results," March 2019. https://aviationsystemsdivision.arc.nasa.gov/publications/atd2/tech- transfers/4_Evaluation_Results/4.09%20ATD2_ART2_HITL_Results_2 019-03-06.pdf + + + + + Alternatives for Scheduling Departures for Efficient Surface Metering in ATD-2: Exploration in a Human-in-the-Loop Simulation + + BonnyKParke + + + LindsayK SStevens + + + WilliamJCoupe + + + HanbongLee + + + YoonCJung + + + DeborahLBakowski + + + KimberlyJobe + + 10.1007/978-3-030-20037-4_28 + + + Advances in Human Error, Reliability, Resilience, and Performance + Washington, D. C. + + Springer International Publishing + July 2019 + + + + Parke, B., Stevens, L. K. S., Coupe, W. J., Lee, H., Jung, Y. C., Bakowski, D. L., and Jobe, K., "Alternatives for Scheduling Departures for Efficient Surface Metering in ATD-2: Exploration in a Human-in-the-Loop Simulation," 10th International Conference on Applied Human Factors and Ergonomics, Washington, D. C., July 2019. + + + + + + diff --git a/file845.txt b/file845.txt new file mode 100644 index 0000000000000000000000000000000000000000..063db893fd9a88a3b5b63780796097bb2b51fc76 --- /dev/null +++ b/file845.txt @@ -0,0 +1,732 @@ + + + + +INTRODUCTIONIn today's commercial aviation system, airport operation is a critical component that has great impact on system performance.According to a study of Air Traffic Management (ATM) operational performance conducted in 2013 by EUROCONTROL and Federal Aviation Administration (FAA) [1], airports contribute to 56% and 86% of total Air Traffic Flow Management (ATFM) delay in Europe and US, respectively.Insufficient capacity, operational uncertainties, and lack of coordination and information sharing among stakeholders and service providers lead to excessive taxi delay and missed opportunities to optimize resource utilization.According to [2], a total of 32 million minutes of taxi-out delay and 13 million minutes of taxi-in delay were estimated at major US airports in 2009.Increased taxi delay translates to extra fuel burn and emissions.Domestic flights in the United States emit about 6 million metric tons of CO 2 , 45,000 tons of CO, 8,000 tons of NO x , and 4,000 tons of Hydrocarbons (HC) in total during taxi-out to runway takeoff [3].In the European Aviation Environmental Report [4], the excess of CO 2 emissions generated by the inefficiency of the taxi-in and taxiout phases in European airports is estimated at 229 kg per flight, or 2 million metric tons in total in 2014.Airport operations impose many challenges due to complex processes and uncertainties.The need to improve airport operations in terms of operational efficiency, predictability, and throughput has drawn much attention within Air Navigation Service Providers (ANSPs) and the ATM research community across the Atlantic [5] [6].The lack of information sharing and coordination in airport operations is a driving motivation for the Collaborative Decision Making (CDM) concept that has recently been introduced into both European and US systems [7] [8].Both Airport-Collaborative Decision Making (A-CDM) in Europe and Surface-Collaborative Decision Making (S-CDM) in the US allow for planning of airport surface operations at a strategic level in order to meet the goal of demand/capacity balancing.In the A-CDM concept of operations, for instance, estimated landing times of arrival flights are continually sent to the airport by the Central Flow Management Unit (CFMU) of the Network Management Operations Center (NMOC) and used to update the subsequent (turnaround) outbound flight plan.The takeoff times (also known as target takeoff times) of the outbound flights, calculated by the A-CDM function, are sent to CFMU to compare with their departure (runway) slots and adjusted as necessary.On the other hand, tactical scheduling tools, which are the focus of this paper, provide advisories to Air Traffic Control Tower (ATCT) controllers and ramp/apron operators to support strategic decision-making and planning, and at the same time provide tactical decision support as flights prepare to depart.These advisories implement tactical schedules, e.g.gate pushback times, taxi operations, and runway departure sequences that aim to reduce taxi delay while maximizing airport throughput under various operational constraints.The German Aerospace Center (DLR) and the National Aeronautics and Space Administration (NASA) are independently developing and testing new concepts, algorithms and tools to address these challenges in airport operations.Both teams have succeeded in innovating and testing new concepts and algorithms in their respective airport systems.To leverage the knowledge and experience gained by each side, DLR and NASA agreed to collaborate on airport surface management research.This paper evaluates the approaches and tactical scheduling algorithms developed by DLR and NASA using a common airport, Hamburg Airport (EDDH) in Germany.This paper is organized as follows.Section II provides a description of current day operations in Europe and U.S. airports.Section III presents the concepts, modeling and simulation tools independently developed by DLR and NASA.The simulation model development, common performance metrics, and evaluation of the two scheduling approaches and system setups for Hamburg airport surface operations are presented in Section IV.The paper ends with concluding remarks and suggestions for future work. +II. CURRENT DAY OPERATIONS AT EUROPEAN AND US AIRPORTSThis section provides a brief description of operations at US and European airports. +A. Definition of Airport Surface AreasIn a US airport, the movement area comprises runways, taxiways and other areas that are used for taxiing, takeoff, and landing.The FAA ATCT is in charge of traffic in the movement area.The terminal ramp area, called the nonmovement area, includes gates and parking areas [9].It is quite common that airlines lease gates and hence has exclusive control over all ground activities in the ramp area, including gate pushback.The transition of control between the ramp area and the movement area takes place at spots on the boundary of the two areas.The movement area in a European airport, as defined by the International Civil Aviation Organization (ICAO), is part of the airport surface for pushback, takeoff, landing, and taxiing.The movement area consists of the maneuvering area and the apron(s) [10].The maneuvering area, including taxiways and runways, is under the control of Air Traffic Control (ATC).The apron is either controlled by the airport operator, e.g., Frankfurt, Munich, and Hamburg Airports, or delegated to the ATC, e.g., Dusseldorf Airport. +B. Taxi InWhile arrival aircraft are taxiing in, the transfer of control from the ATCT to the apron/ramp control occurs near the boundary of the taxiway and the apron/ramp area.The taxi-in process ends when an aircraft completely stops at the gate or parking stand and its engines are off.In the US, the ATCT controller issues a taxi clearance to a spot after the aircraft lands and exits from the runway.The aircraft is then asked to contact the ramp controller before the aircraft reaches the spot.Once the aircraft crosses the spot, control is handed to the ramp controller.Similarly, in Europe, the control hand-over from the ATCT controller to the apron controller usually takes place before the aircraft enters the apron area to avoid stopping the aircraft. +C. TurnaroundThe flight turnaround is a complex process that spans the time between gate-in and gate-out, including all of the ground activities, such as passenger de-plane/boarding, baggage unloading/loading, fueling, and cleaning.The airline, the airport authority, or a third party handler, manages the turnaround process.Due to the uncertainties associated with this multi-step process, it is difficult to predict when the aircraft will be ready for departure from the gate.In the US, the flight operator does not have accurate flight ready time estimates.Therefore, the ramp controller relies on the pilot's request for pushback to make any planning decisions.Currently, John F. Kennedy International Airport (KJFK) has implemented the Ground Management Program (GMP) based on a CDM concept to help mitigate the situation by requiring participating airlines to provide an estimated gate departure time or Earliest Off-Block Times (EOBT) of each flight to the system for departure scheduling.In A-CDM equipped European airports (e.g., Munich and Zurich airports), the Target Off-Block Time (TOBT) is the gate departure time guaranteed by the airline/ground handler. +D. Taxi OutThe aircraft taxi out process begins at gate pushback and ends at runway takeoff.In the US, a Pre-Departure Clearance (PDC) is issued by the Clearance Delivery position to the pilot 30 minutes prior to the proposed departure time.The pilots call the ramp to get a pushback clearance when the aircraft is ready to depart.Once pushback is complete, the pilots call the ramp for a taxi clearance.Control is transferred from ramp to ATCT at the designated spot, and the ground controller will issue a clearance to taxi to the assigned runway.As the aircraft approaches the runway, control is transferred to the local controller who issues line-up and takeoff clearances.At A-CDM equipped airports in Europe, pilots are required to monitor Clearance Delivery and call in 5 minutes or less prior to TOBT to get an en-route and engine start-up clearance.Depending on the Target Start-up Approval Times (TSATs) and the traffic situation, the clearance is issued.The apron controller issues a pushback clearance at TSAT.The airplane is under the control of the apron controller until it reaches the border of maneuvering area, where the ground controller guides the aircraft to its assigned runway.The remaining taxi-out process is the same as that at the US airports. +E. Line-up and TakeoffThe procedures for line-up at the runway and takeoff are very similar in the US and Europe.It is common practice that aircraft receive takeoff clearances after line-up once the runway separation requirement is met.The runway separation requirement is determined by multiple criteria, including aircraft weight class and divergent heading based on Area Navigation (RNAV) routes. +F. Traffic Flow ManagementSurface operations are affected by Traffic Flow Management (TFM) decisions.TFM related decisions are made to balance demand and capacity of air traffic in the entire airspace to minimize system delay and maintain throughput.The NMOC and the Air Traffic Control System Command Center (ATCSCC) are the two central authorities responsible for implementing TFM procedures in Europe and the US, respectively.In the US, the ATCSCC assigns takeoff times to the flights bound for the airport or sector where arrival rates are degraded due to heavy traffic volume or by adverse conditions such as bad weather.The assigned departure time, called Expect Departure Clearance Time (EDCT), has a compliance window of ±5 min around the assigned takeoff time.Another TFM procedure called Milesin-Trail (MIT) can be negotiated between Air Route Traffic Control Centers (ARTCCs) to limit the en-route traffic flow that in turn could impact airport departure throughput.Similarly in Europe, NMOC publishes departure constraints by issuing Calculated Takeoff Time (CTOT).A compliance window of -5 min/+10 min around CTOT has to be adhered to by the flight. +III. OPERATION CONCEPTS AND TOOLS BY DLR AND NASA +A. Simulation and planning tools of DLRAccording to the Single European Sky ATM Research (SESAR) road map for modernizing air traffic management in Europe [11], a new generation of airport decision support systems has been introduced in the last decade.The Departure Management System (DMAN) and the Surface Management System (SMAN) are two of these airport surface systems that provide critical decision and execution support capabilities to controllers and flight deck.DMAN's responsibility is to improve departure flows on the airport surface by providing the Target Takeoff Time (TTOT) and the TSAT for each departure under multiple constraints and airline preferences [12][13].The Controller Assistance for Departure Optimization (CADEO) is DLR's implementation of DMAN [12].It is an ATC tool that optimizes the departure takeoff sequence by calculating the TTOTs for each departure.Working without being coupled with SMAN, CADEO estimates the TSATs from TTOTs using the Variable Taxi Times (VTTs) provided by A-CDM.CADEO is a generic tool that can be adapted to different airports.CADEO's departure sequence optimization takes into account the operational constraints at the airport, such as wake vortex separations based on aircraft weight category, runway occupancy times based on aircraft type, and miles-in-trail constraints for consecutive departures going to the same departure fix.Optimization objectives may include maximum runway throughput, departure slot time adherence, taxi-out delay reduction, and planning stability [13].Through a series of real-time simulations conducted for Prague Airport [13], it was found that CADEO advisories helped reduce taxi times and stops and avoided excessive queues at the runway holding points.The Taxi Routing for Aircraft: Creation and Controlling (TRACC) is DLR's prototype for SMAN [14].TRACC generates optimal conflict-free taxi trajectories for all aircraft surface movement, including gate pushback.In the context of SMAN, a taxi trajectory differs from a time-based taxi route such that in addition to position and time, the trajectory also includes speed and acceleration information.TRACC provides the taxi guidance instructions to ATC controllers and potentially to the flight deck as well.The concept of integrated CADEO-TRACC system was evaluated in recent studies [15][16].The vision is to use TRACC to generate conflict-free taxi trajectories that meet the TTOTs prescribed by CADEO.The expected outcome was to improve taxi operations and gain experience in the integrated concept.In this integrated CADEO-TRACC concept, CADEO calculates departure TTOTs based on the Earliest Line-up Times (RLUTs) estimated from TRACC and derives the Target Line-up Times (TLUTs) as the target times for TRACC to deliver departure aircraft to the runway.Then TRACC calculates the optimized taxi trajectories starting from pushback and sets TSATs.Based on these taxi trajectories, Estimated Line-up Times (ELUTs) are computed and compared with the TLUTs.If ELUT is later than TLUT, TRACC will adjust the trajectory to meet TLUT, and if RLUT is later than TLUT, CADEO will re-plan TTOTs.TRACC optimizes TSATs to hold departures at the gate as long as possible and to reach the departure runway in time.The integration concept was first tested with the Munich Airport via simulations.The results confirmed the feasibility and potential benefits of integrated departure and surface traffic management [16].Air traffic operations are simulated using the National Aerospace Laboratory (NLR) Air Traffic Control Research Simulator (NARSIM) [17].NARSIM consists of four simulation systems (i.e., the air system, the ground system, the air/ground communication system, and the "meteo" system) and one control system.The air system generates realistic flight trajectories and facilitates communications between the simulated pilot and the aircraft model.The ground system creates an experimental environment for air traffic controllers to evaluate human-machine interfaces of advanced air traffic management functions.The air/ground system provides the inter-communication and synchronization between the air and the ground systems.The meteo system presents meteorological data that contribute to the generation of realistic flight trajectories and weather radar observations.Under the supervision of an experiment leader, the control system supervises (i.e., initialize, reset, start, etc.) the entire experiment. +B. NASA's SARDA concept and experimentNASA has recently developed the Spot and Runway Departure Advisor (SARDA) tool [18][19] [20][21] [22] to provide advisories to controllers to help improve airport surface operations.The original SARDA concept was developed as an ATCT control decision support tool and tested in human-in-the-loop (HITL) simulations for Dallas/Fort Worth (Texas) International Airport (DFW) [18].It provided runway sequence advisories to the local controller and spot release advisories to the ground controller.It was a tactical tool aiming at reducing taxi delay by shifting delay from taxiways and runway queues to the ramp area without affecting airport throughput.The results from the initial simulations with SARDA showed reductions in taxi delay (45-60%) and fuel consumption (23% and 33%) in the movement area compared with non-advisory cases in both medium and heavy traffic scenarios.In a study conducted in 2012, the SARDA advisory tool was integrated with a strategic scheduling component to support the S-CDM concept [19].This strategic scheduling component provided a mechanism for the airline operator to share data and preferences with the SARDA system.The strategic scheduler received flight ready times within a predetermined planning window from the airline and then generated the target pushback times and spot release times.These times were communicated back to the airline for confirmation.The tactical scheduling component provided the advisories for spot release sequence to the ground controller as aircraft push back from the gates, and runway sequence advisories to the local controller as the aircraft join the runway queue.Integrating the strategic scheduling component in the overall scheduling process helps reduce the uncertainties of flight readiness and potential missed opportunities.The results from the study based on a real-time automated simulation for DFW showed reductions in both mean and variation in taxi delay under varying uncertainties in actual pushback times.The most recent SARDA HITL experiment was conducted for Charlotte Douglas (North Carolina) International Airport (KCLT) through collaboration with American Airlines [20], where SARDA was used as a ramp controller decision support tool.Note that at KCLT, American Airlines is the dominant air carrier whose operations account for 85% of the entire airport operations.In addition, American Airlines manages the ramp tower.A series of HITL simulations were conducted where the SARDA system provided pushback advisories to the ramp controller, while no additional SARDA guidance was given to the local or ground controllers.The simulation results showed that the SARDA tool helped reduce taxi times on average by one minute per flight without decreasing runway throughput.In addition, the advisories improved EDCT conformance and reduced ramp controller's workload.The core component of the SARDA tool is its scheduler called the Spot Release Planner (SRP) [21].It is a two-stage algorithm.The first stage is a runway scheduler.It takes a snapshot of the current surface traffic situation and calculates the optimized sequence and times for runway usage, including departure takeoff and aircraft waiting for crossing.The algorithm incorporates numerous constraints such as wake vortex separation and miles-in-trail restriction and can be solved for multiple objectives including maximum throughput and minimum system delay.The second stage determines times to release aircraft from gates or assigned spots to meet the previously calculated runway departure schedules.It uses predicted taxi times in the calculation.The simplest taxi time prediction can be based on a percentile of unimpeded taxi time distribution in historical operations.A recent study of taxi time prediction algorithms for KCLT operations revealed that fast-time simulation methods or machine learning techniques outperformed the method using unimpeded taxi times in terms of taxi time prediction accuracy [23]. +IV. THIS WORKThis section discusses and evaluates the concepts and approaches of the two different surface traffic management systems developed by DLR and NASA, i.e., CADEO-TRACC and SARDA.The analysis was based on simulations using Hamburg Airport (EDDH) in Germany.To provide a common basis for the evaluation both systems were setup using a common airport model and a common simulation scenario.A common set of metrics was also used for performance assessment.It should be noted that the selection of Hamburg Airport was primarily due to the availability of extensive operational data.It was not the objective of this work to address particular operational challenges at Hamburg Airport. +A. Airport and traffic scenarioIn this experiment, a two-hour traffic scenario that contained 35 departures and 34 arrivals for EDDH, as shown in Fig. 1, was created based on actual flight data on May 25, 2004.Out of the total 69 aircraft, 63 are of ICAO Medium wake category (e.g., A320 and B733).Calm winds and good visibility were assumed.EDDH has two intersecting runways: Runway 33 for departures and Runway 23 for arrivals.This is the most common runway configuration at EDDH.Arrivals exit from Runway 23 and cross the departure runway at a single location, entering the apron area to the gates.There are five runway exits on the left of Runway 23.The first one is a high-speed exit and the other four are standard ones.The gates and stands in Apron 1 (right of the runway crossing) are very close to Runway 33 entrances.For the purpose of the experiment, it was assumed that all flights were under Instrument Flight Rules (IFR).Each departure had a Scheduled Takeoff Time (STOT) or wheelsoff time, and each arrival had a Scheduled Landing Time (SLDT) or wheels-on time.The Scheduled Off-Block Time (SOBT) was estimated from STOT using unimpeded taxi speeds (more details in the following sections).Fig. 2 shows the STOT and SLDT time event count in the scenario.A minimum separation for departures from Runway 33 is set at two minutes because of the required 4nmi straight ascent after takeoff by the Standard Instrument Departure (SID) for IFR flights. +B. System setup of DLR simulationAn integrated system was setup and adapted for an automated real time (not HITL) simulation for EDDH (see Fig. 3).The system was comprised of CADEO, TRACC, NARSIM and SimNet.NARSIM generated the air traffic from the scenario, and was connected with CADEO and TRACC via SimNet to exchange planning information such as RLUT and TLUT.It executed the taxiing guidance received from TRACC.For this purpose, TRACC was equipped with an automated control interface to translate the commands into instructions for the simulator.In HITL simulations, TRACC commands would be shown to the air traffic controller who then provided the advisories to the pseudo-pilot via radiotelephony.Typical simulation events and the planning loop were as follows.When a departure flight approached its off-block time, it triggered the scheduling loop.First, TRACC calculates the RLUT using the standard taxi route and sends the RLUT to CADEO.Then CADEO computes optimal runway schedule with TTOT, considering overall runway traffic demands, and sends TLUT back to TRACC.According to the TLUT, TRACC attempts to create a conflict-free taxiing guidance to minimize stopping and waiting along the taxiways.CADEO's objective is to maximize the runway throughput by optimizing the TTOTs/TLUTs, and TRACC provides the advisory for aircraft movement to meet the runway schedule while minimizing taxi delays.Because of the dynamic bi-directional coordination of CADEO and TRACC, an aircraft is subject to schedule time change until it takes off or arrives at the gate.If TRACC receives new flight plans entering the planning horizon or there are aircraft that deviate from their assigned taxi trajectories, recalculations are necessary, and CADEO is notified and adjusts the runway schedule accordingly.Table 1 explains five schedule update stages.For all stages except for "standard routes,", optimization and conflict detection are performed.For the performance evaluation introduced in this paper, the schedule at "Last Optimization" was chosen for arriving and departing aircraft.It is the latest dataset for every flight because no optimization was performed beyond that and, therefore, represents the actual traffic situation influenced by the optimization of the planning tools.The nominal taxi speed (i.e., the speed at which aircraft taxi unimpeded by other traffic) is based on the airport area (high-speed and normal taxiway, apron, pushback at the gate).The nominal taxi speed was set to 50, 30, and 15 kts for highspeed taxiways (e.g., runway exit), regular taxiways and apron, respectively.Pushback was performed with 5 kts of taxi speed.The initial selection of runway exit for arriving aircraft was made based on the results of a probability calculator [29].There, the probabilities of landing distances are assigned to weight classes.In combination with the air traffic simulator NARSIM, runway exits can be changed occasionally and TRACC accepts the new runway exit. +C. SARDA system setupThe SARDA system setup for this study is shown in Fig. 4. NASA's fast-time simulator, called the Surface Operation Simulator and Scheduler (SOSS) [24][25], was used together with a SARDA scheduler implemented for the EDDH airport model.SOSS models surface operations of flights, including flight readiness, pushback, taxi, takeoff, and landing.It uses a nodelink graph to represent locations and connections of gates, ramps, taxiways, runway crossings and runways.EDDH nodelink graph in SOSS was adapted from the same EDDH nodelink model from DLR's Airport Data Editor for NARSIM.SOSS uses a set of rules to maintain appropriate separations of aircraft in runway operations such as takeoff, landing, and crossing.These rules are dependent on airport runway geometry.For EDDH operations in this study, the runway separation rules included: 1) consecutive departures at Runway 33, 2) departure and arrival at Runway 33 and Runway 23 (i.e., intersecting runway operations), and 3) taxiing arrivals that cross Runway 33 to enter the apron area.Consecutive arrivals at Runway 23 were spaced in the traffic scenarios.The separation parameters followed the standard ICAO separation requirements based on aircraft weight class (wake category).For instance, a departure following a heavy aircraft (e.g., Airbus A330) requires more separation than a lighter aircraft (e.g., E110).The SOSS landing and takeoff models produced runway events such as wheels on/off and threshold crossing.Separation in runway operations was based on these events in simulation.The landing model selected appropriate exits for arrivals based on their aircraft types and the distances of exits from the runway threshold.In this EDDH simulation, all arrivals made exits at either the third (M) or the fourth exit (N).On the movement area (i.e., taxiways and apron), SOSS uses its internal Conflict Detection & Resolution (CD&R) logic with a First-Come, First-Served (FCFS) rule to move aircraft and maintain a proper separation among them.When two aircraft are predicted to have a taxi conflict such as in lead-follow or crossing situation, the one ahead is allowed to continue and the aircraft behind has to slow down or wait.The logic is strictly within a smalllocalized area and is designed not to interfere with strategic taxi schedules.The nominal taxi speed (i.e., the speed at which aircraft taxi unimpeded by other traffic) is aircraft type based.Because in this simulation, almost all aircraft are in the Medium category, the nominal taxi speed was averaged at 15 kts, and 10 kts in taxiway and apron, respectively.Pushback speed was set at 5 kts.The SARDA Hamburg scheduler was an implementation of the SARDA concept for the study.It was a tactical planner and consisted of a runway scheduler and gate pushback planner.The runway scheduler created an optimized runway departure schedule for Runway 33 with maximum runway throughput as the objective.The constraints to the optimization algorithm were the arrival operations on Runway 23 that affect the departures from Runway 33 due to their intersecting runway geometry, as well as the departure wake vortex separation requirements.Arrivals crossing at Runway 33 were given a lower priority to help boost departure throughput.During simulation, SOSS sent to the scheduler the surface traffic situation information (e.g., the aircraft positions and the latest runway operations) and flight plans (e.g., STOT).The scheduler calculated the best runway departure schedule and pushback times, and sent them to the simulator.Standard taxi routes and nominal taxi speeds were used.One key difference from the CADEO and TRACC system is that the taxiing schedule was time-based only and not guaranteed conflict free.SOSS resolved potential taxi conflicts using the FCFS rule, which is similar to what a human controller does to resolve conflict situations [31]. +D. Performance metricsTo make a meaningful assessment of the two different concepts and tools, a set of common performance metrics were first defined and agreed upon.These metrics, shown in Table 2, were derived from the ICAO's Key Performance First optimization with a target takeoff time given by CADEO.Last Optimization before Pushback (Departures only)Optimization with the last adjustment of TSAT as response to a changed target takeoff time by CADEO.After that the TSAT would not change anymore. +Last OptimizationLast optimization done by TRACC for the observed aircraft.Areas (KPAs) [27] and the Civil Air Navigation Services Organization (CANSO) [28].Two KPAs were included in this study: capacity (one metric) and efficiency (four metrics). +E. Simulation Results and EvaluationsThe following results and evaluations are guided by the metrics outlined in Table 2.The terms CADEO-TRACC and SARDA thereafter are used to refer to the two simulation system setups, respectively.As discussed previously, the study included the same airport, the same simulation scenario, the same metrics but different simulation systems.The differences were not only in the scheduling and taxi concepts implemented, but in the simulators as well (real-time versus fast-time).Noticeably, CADEO-TRACC's nominal taxi speed was configured at 30 and 15 kts for taxiway and apron, whereas SARDA was at 15 and 10 kts.CADEO-TRACC used the first two arrival exits at Runway 23 while SARDA chose the third and fourth exits. +1) Departure runway throughputFig. 5 shows the departure throughputs on Runway 33.The numbers of departures (i.e., wheels-offs) are counted in a 10minute interval during the less than 2.5 hour-long scenario.For reference, the STOTs are included.In the first 1.5 hours, both systems produced very similar throughput levels.The obvious difference occurred in the last 40 minutes, where CADEO-TRACC's throughput lagged.Although both schedulers of the two systems sought the maximum departure throughput, the means of delivering aircraft to the runway were different.SARDA in this experiment used an optimistic taxi time estimate from gate to runway to keep runway queue pressure.The planning process between CADEO and TRACC allowed negotiation between throughput and taxiing efficiency.By avoiding runway queues and taxi stops, TRACC might not be able to fill emerging gaps because the departures that could have been scheduled for these times were planned to later departure times by CADEO-TRACC in order to achieve better taxi efficiency. +2) Taxi timesTaxi time is an important performance metric to measure taxi efficiency.Fig. 6 displays the total taxi times of departures and arrivals.The total taxi time is the summation of all aircraft taxi times.It is evident CADEO-TRACC used less taxi times than SARDA, in both departure and arrival categories.However, comparing these numbers directly would be biased because of the different taxi speed setups.To get a more equal assessment, a normalized taxi time method is used.The normalized taxi time is the taxi time divided by its unimpeded taxi time.The unimpeded taxi time is the time an aircraft taxi at nominal speeds unimpeded.Unimpeded time is a lower bound taxi time.In SARDA where aircraft never taxi above their nominal speeds, unimpeded time is the lower bound taxi benchmark.Since the two systems had their own nominal taxi speeds, their unimpeded taxi times are different.Table 3 shows the total unimpeded and normalized taxi times.SARDA had the longer unimpeded times because of its lower nominal taxi speeds.The difference between the two simulations was less dramatic in departure than in arrival taxi times because the apron is very close to the runway entrances.The long taxiway for arrivals from Runway 23 combined with the fact that TRACC chose the first two runway exits (O, P) (i.e., shorter taxi distance to the apron) made a big difference in arrival unimpeded taxi times.Nonetheless, the normalized taxi times show that CADEO-TRACC performed closer to its unimpeded taxi times than SARDA.This indicates that TRACC's conflict-free taxi guidance resulted in efficient surface movement for both departures and arrivals.On the other hand, the SARDA system relied on the fast-time simulator to resolve taxi conflicts, thus leading to taxi slow downs and stops, particularly in the apron area.Fig. 7 shows the boxplots for both departure and arrival taxi times.CADEO-TRACC's taxi times showed less variation (i.e., the 1.Qu and 3.Qu were closer to the median) than SARDA's taxi times.It seems to indicate that the conflict-free taxiing sought by the CADEO-TRACC would lead to better taxi time predictability or less uncertainty.Predictability, not measured in this study, is one of the important ICAO KPAs. +3) Pushback delay and departure queue timeFigs. 8 and9 show the pushback delays (i.e., gate holding) and departure queue times.The CADEO-TRACC held departures at gate/stand much longer than the SARDA did, 149 (5218/35 departures) to 38 seconds on average.This supports the finding discussed in Section 1): the CADEO-TRACC tend to keep aircraft longer at gates/stands to achieve conflict-free taxiing, which benefits the taxi efficiency and, at the same time, reduces the departure queue time at the runway.The maximum of 946 seconds of gate holding by CADEO-TRACC that occurred during the last 30-40 min in simulation was probably due to the peak demand of arrivals taxiing into the apron at the time (see Fig. 2).The departure queue was defined as ~100 meters towards the runway line-up (takeoff) position.The average queue time was 17 (587/35 departures) for CADEO-TRACC and 23 seconds for SARDA.This is consistent with the pushback delay metric, meaning that more potential queue delay was shifted to gate holding by both systems.Moving taxi and queue delay to gate holding would reduce fuel burn and pollution (a metric in ICAO's environment impact KPAs). +F. Further Evaluations and DiscussionOther performance metrics can be derived and evaluated using the metrics already discussed.For departures, takeoff delay, which is defined as the difference between the actual takeoff time and the earliest possible takeoff time, can be considered.Since the earliest takeoff time can be calculated by adding the unimpeded taxi time to the scheduled pushback time (or flight readiness time), the takeoff delay is the same as the sum of pushback delay and taxi-out delay (between actual and unimpeded).From Figs. 6 and8, and Table 3, the takeoff delay values for CADEO-TRACC and SARDA are 5,331 (=5,218+113) and 1,566 (=1,328+238) seconds, respectively, shown in Fig. 10.Because the CADAO-TRACC tends to hold departures at gate/stand for a longer time in order to find conflict-free taxi trajectories, the takeoff times are also delayed, which is consistent with the departure throughput result shown in Fig. 5.From the perspective of both departure and arrival operations, total system delay can also be calculated.The total system delay is defined as the sum of takeoff delay and taxi-in delay.The system delays for CADEO-TRACC and SARDA are computed as 5,748 and 3,664 seconds, respectively (see Fig. 10).Since the two systems have different unimpeded taxi speeds and times, the emphasis of this metric is to show the balance between departures and arrivals when scheduling surface traffic having intersecting runways.In the CADEO-TRACC simulation, the system delay was dominated by gate holding and in the SARDA simulation both gate holding and arrival delay had more comparable contributions to the system delay.Considering the delay propagation, a long queue of arrivals crossing the active departure runway should be avoided, and excessive pushback delay is also undesirable.The numbers and analysis thus far indicate the different approaches and results of the two simulations.The CADEO-TRACC's conflict-free taxi solution aimed to push the taxi efficiency closer to unimpeded performance.This strategy appeared to associate with larger taxi speed range (i.e., 30/15 kts nominal speed setups) and longer gate holding.A larger taxi speed range made it easier for TRACC to find solutions in conflict-free taxiing resolution space.Longer gate holding seemed to happen at the peak of the arrivals, which led to longer takeoff delays and therefore impacted departure throughput (Figs. 5 and10).In addition to a good taxi time performance, the conflict-free taxi time distributions showed less variation, which would result in good taxi time prediction, another important performance metric for airport operation but not included in this study.The CADEO-TRACC's approach is a futuristic concept, and the SARDA's concept was based on current and near-term technologies.Both approaches focused on the throughput and efficiency improvement.The results showed a good departure throughput and taxi performance balance of SARDA but less efficiency in departure taxi time reduction compared to CADEO-TRACC. +V. SUMMARY AND FUTURE WORKThis work is the first attempt to evaluate two different airport traffic management concepts and tools developed by DLR and NASA research teams.Two independent simulations were conducted using a simulation model of Hamburg Airport in Germany.One was a real-time simulation of the integrated CADEO-TRACC system conducted by DLR.The other was a fast-time simulation using the SARDA planning concept by NASA.Both systems were developed for the improvement of taxi efficiency while maintaining the runway throughput.A main difference between the two approaches is that the CADEO-TRACC system employs conflict-free optimized taxi guidance to reduce taxi stops while SARDA calculates pushback schedules by subtracting nominal taxi time from the optimized runway schedule.The planning cycle of the CADEO-TRACC system involves iterations between the two components (i.e., departure management and surface management), and allows negotiation between runway target times and taxi efficiency.This approach effectively takes the taxi conflicts into consideration in the runway scheduling.The SARDA system consists of a runway scheduler and gate pushback or spot release advisory depending on whom the advisories are provided to.In this work, the gate pushback schedule was calculated from the runway schedule that optimized at maximum runway throughput and the simulator executed pushback advisories.SARDA's runway scheduler objective is similar to that of CADEO.The taxi guidance of the SARDA system is limited to the release time at gate in this study, where the taxi conflicts are resolved by the simulator rather than by the scheduling algorithm.The two simulation outputs are evaluated using a set of common performance metrics adapted from two of the ICAO's KPAs: capacity and efficiency.The results showed that both CADEO-TRACC and SARDA were able to improve taxi efficiency while maintaining runway throughput under a normal traffic conditions.The CADEO-TRACC had better taxi efficiency performance due to the conflict-free taxiing capability of TRACC.The strategy to incorporate taxi conflict avoidance into the runway and taxi planning might lead to longer gate holding in order to obtain conflict-free taxi solutions and, consequently, impacted runway throughput and cause potential gate conflicts with arrival aircraft.The SARDA system, on the other hand, allowed maintaining a small number of aircraft in the runway queue in order to prevent the runway from starving and reducing runway throughput.Future work under the research collaboration between DLR and NASA includes investigating the scalability of the two approaches, especially the integrated CADEO-TRACC system, by testing at a much busier hub airport than Hamburg.Also, robustness and resilience of the systems will be Fig. 10 Total system delay investigated under varying degrees of uncertainties and operational constraints.Lastly, the future research plan includes integration of the TRACC conflict-free taxi capability with the SARDA runway scheduler in SOSS and conducting fast-time simulations for a major US airport.This would help investigate the feasibility of introducing the trajectory-based taxi operation concept [30] and evaluate its benefit for US airport surface operations.This concept is a critical technology advance in not only addressing current surface traffic management problems, but also having potential applications in operations of unmanned vehicles on the airport surface.Fig. 1 Fig. 313Fig.1Hamburg Airport[26] +Fig. 44Fig. 4 Hamburg airport system setup for SARDA +Fig. 5 Fig. 656Fig. 5 Departure runway throughput +Fig. 7 Fig. 979Fig. 7 Taxi time quartiles +Table 1 . Schedule update stages of CADEO-TRACC coordination Schedule update stage Definition1Standard Routes (noPredefined route set based on assumptionsoptimization, no conflictlike fastest, shortest routes or standarddetection and resolution)routes in use at the observed airport.First OptimizationOptimization for getting the first possibleLine-up timeFirst Optimization withTLUT by CADEO(Departures only) +Table 3 . Unimpeded and normalized taxi times CADEO-TRACC SARDA3Departure unimpeded taxi time6,178 seconds6,640 secondsArrival unimpeded taxi time7,884 seconds12,877 secondsDeparture normalized taxi time 1.0181.036Arrival normalized taxi time1.061.16Fig. 8 +Total Pushback delay and departure queue time + + + + +ACKNOWLEDGMENT This work was conducted in a collaborative effort with equal contributions from DLR and NASA based on the Implementing Agreement for "Coordinated Arrival/Departure/Surface Operations Research".NASA's research efforts were funded by the Airspace Technology Demonstrations (ATD) project under the Airspace Operations and Safety Program (AOSP). + + + + + + + + + 2013 Comparison of Air Traffic Management-Related Operational Performance + + FaaEurocontrol + + + U.S./Europe + + + EUROCONTROL, FAA, "2013 Comparison of Air Traffic Management-Related Operational Performance: U.S./Europe." + + + + + Impact of Congestion on Taxi Times, Fuel Burn, and Emissions at Major Airports + + IoannisSimaiakis + + + HamsaBalakrishnan + + 10.3141/2184-03 + + + Transportation Research Record: Journal of the Transportation Research Board + Transportation Research Record + 0361-1981 + 2169-4052 + + 2184 + 1 + + 2010 + SAGE Publications + + + I. Simaiakis, H. Balakrishnan, "Analysis and control of airport departure processes to mitigate congestion impacts," Transportation Research Record: Journal of the Transportation Research Board, pp. 22-30, 2010. + + + + + Demonstration of reduced airport congestion through pushback rate control + + IoannisSimaiakis + + + HarshadKhadilkar + + + HamsaBalakrishnan + + + TomGReynolds + + + RJohnHansman + + 10.1016/j.tra.2014.05.014 + + + Transportation Research Part A: Policy and Practice + Transportation Research Part A: Policy and Practice + 0965-8564 + + 66 + + 2011 + Elsevier BV + Berlin, Germany + + + USA/Europe Air Traffic Management Research and Development Seminar + + + I. Simaiakis, et al "Demonstration of Reduced Airport Congestion Through Pushback Rate Control," 9 th USA/Europe Air Traffic Management Research and Development Seminar, Berlin, Germany, 2011. + + + + + + EuropeanEnvironment + + + Agency + + + European Aviation Enviornmental Report + + 2016 + + + European Environment Agency, "European Aviation Enviornmental Report 2016," http:// http://ec.europa.eu/transport/modes/air/aviation- strategy/documents/european-aviation-environmental-report-2016- 72dpi.pdf + + + + + Technology Pathways + + Nasa + + 10.17226/11420 + + 2011 + National Academies Press + + + NASA, "Next Generation Air Transportation System Concepts and Technology Development Project Plan," 2011. + + + + + Sesar, Alois + + Sesar + + 10.1093/benz/9780199773787.article.b00168032 + + + + Oxford University Press + + + SESAR, "SESAR Concept of Operations Step1," http://www.sesarju.eu/sites/default/files/documents/highlight/SESAR_C onOps_Document_Step_1.pdf + + + + + Airport CDM Implementation -The Manual + + ELagios + + + + + Last access: 2016/05/26 + E. Lagios, "Airport CDM Implementation -The Manual," available on: http://www.eurocontrol.int/publications/airport-cdm-implementation- manual-version-4 [Last access: 2016/05/26]. + + + + + Collaborative Decision Making (CDM) in Airport Surface: Europe vs usa implementations, challenges and best practices + + SimonOkwir + + + AntonioCorreas + + 10.1109/icnsurv.2014.6819989 + + + 2014 Integrated Communications, Navigation and Surveillance Conference (ICNS) Conference Proceedings + + IEEE + 2012. July 2013 + + + FAA Surface Collaborative Decision Making (CDM) Team (2012), "U.S. Airport Surface Collaborative Decision Making (CDM) Concept of Operations (ConOps) in the Near-Term: Application of Surface CDM at United States Airports," July 2013. + + + + + Federal Avation Regulation Part 139 -Certification of Airports + + Faa + + + + + FAA, "Federal Avation Regulation Part 139 -Certification of Airports," available online at: www.faa.gov/airports/airport_safety/part139_cert/. + + + + + Air Traffic Management + + Icao + + + 2010 + International Civil Aviation Organization + 9883 + Montréal, Canada + + + 15th Edition + ICAO, Air Traffic Management, Doc 9883, 15th Edition, Montréal, Canada, International Civil Aviation Organization, 2010. + + + + + + Sesar + + + European ATM Master Plan + + 2 + + + SESAR, "European ATM Master Plan," Edition 2, https://www.atmmasterplan.eu/ + + + + + Phase transitions and ordering in liquid crystals — a case study + + MSchaper + + + DBöhme + + 10.1007/978-3-540-85226-1_394 + + + EMC 2008 14th European Microscopy Congress 1–5 September 2008, Aachen, Germany + Darmstadt, Germany + + Springer Berlin Heidelberg + 2008, 23-25 September, 2008 + + + + M. Schaper, D. Böhme, "Improved Departure Management Through Integration of DMAN and A-SMGCS," DLRK (German Aerospace Congress) 2008, 23-25 September, 2008, Darmstadt, Germany. + + + + + Egypt in its African Context: Proceedings of the Conference held at The Manchester Museum, University of Manchester, 2-4 October 2009 + + MSchaper + + 10.30861/9781407307602 + + + CEAS 2009, European Air and Space Conference + Manchester, UK + + University of Michigan Press + 26-29 October, 2009 + + + M. Schaper, "Operational Improvements in The Context of DMAN, A- SMGCS and A-CDM," CEAS 2009, European Air and Space Conference, 26-29 October, 2009, Manchester, UK + + + + + Taxi routing for aircraft: Creation and Controlling -Ground Movements with time constraints + + IGerdes + + + + 2nd SESAR Innovation Days + Braunschweig, Germany + + 27-29 November, 2012 + + + I. Gerdes, "Taxi routing for aircraft: Creation and Controlling -Ground Movements with time constraints," 2nd SESAR Innovation Days, 27-29 November, 2012, Braunschweig, Germany. + + + + + Trajectory based ground movements and their coordination with departure management + + MeilinSchaper + + + IngridGerdes + + 10.1109/dasc.2013.6712662 + + + 2013 IEEE/AIAA 32nd Digital Avionics Systems Conference (DASC) + Syracuse, NY, USA + + IEEE + 6-10 October, 2013 + + + M. Schaper, I. Gerdes, "Trajectory Based Ground Movements and Their Coordination with Departure Management," 32nd Digital Avionics Systems Conference, 6-10 October, 2013, Syracuse, NY, USA. + + + + + Call for Papers + + IGerdes + + + MSchaper + + 10.1027/2192-0923/a000067 + + + Aviation Psychology and Applied Human Factors + Aviation Psychology and Applied Human Factors + 2192-0923 + 2192-0931 + + 4 + 2 + + 10-13 June, 2015 + Hogrefe Publishing Group + Lisbon, Portugal + + + I. Gerdes, M. Schaper, "Management of Time Based Taxi Trajectories coupling Departure and Surface Management Systems," Eleventh USA/Europe Air Traffic Management Research and Development Seminar, 10-13 June, 2015, Lisbon, Portugal. + + + + + The development of the NLR ATC Research Simulator (Narsim) + + JMTen Have + + 10.1016/0928-4869(93)90009-f + + + Simulation Practice and Theory + Simulation Practice and Theory + 0928-4869 + + 1 + 1 + + 1993. 15 July 1993 + Elsevier BV + + + Flight Division of National Aerospace Laboratory NLR + + + ten Have, J.M., 1993, The development of the NLR ATC Research Simulator (Narsim): Design philosophy and potential for ATM research, Flight Division of National Aerospace Laboratory NLR, Simulation Practice and Theory, Volume 1 Issue 1 p.31-39, 15 July 1993. + + + + + Performance evaluation of a surface traffic management tool for Dallas/For Worth International Airport + + YJung + + + 2011 + Berlin, Germany + + + USA/Europe Air Traffic Management Research and Development Seminar + + + Y. Jung et al, "Performance evaluation of a surface traffic management tool for Dallas/For Worth International Airport," 9 th USA/Europe Air Traffic Management Research and Development Seminar, Berlin, Germany, 2011. + + + + + An Integrated Collaborative Decision Making and Tactical Advisory Concept for Airport Surface Operations Management + + GautamGupta + + + WaqarMalik + + + YoonJung + + 10.2514/6.2012-5651 + + + 12th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference and 14th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference + Indianapolis, IN + + American Institute of Aeronautics and Astronautics + Sept 17-19, 2012 + + + G. Gupta, W. Malik, Y. Jung, "An integrated collaborative decision making and tactical advisory concept for airport surface operations management," 12 th AIAA Aviation Technology, Integration, and Operation (ATIO) Conference, Indianapolis, IN, Sept 17-19, 2012. + + + + + Call for Papers + + MHayashi + + 10.1027/2192-0923/a000067 + + + Aviation Psychology and Applied Human Factors + Aviation Psychology and Applied Human Factors + 2192-0923 + 2192-0931 + + 4 + 2 + + June 23-26, 2015 + Hogrefe Publishing Group + Lisbon, Portugal + + + 11 th USA/ + M. Hayashi et al, "Evaluation of Pushback Decision-Support Tool Concept for Charlotte Douglas International Airport Ramp Operations," 11 th USA/Europe Air Traffic Management Research and Development Seminar, June 23-26, 2015, Lisbon, Portugal. + + + + + Spot Release Planner: Efficient Solution for Detailed Airport Surface Traffic Optimization + + WaqarMalik + + + GautamGupta + + + YoonJung + + 10.2514/6.2012-5498 + + + 12th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference and 14th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference + Indianapolis, IN + + American Institute of Aeronautics and Astronautics + Sept 17-19, 2012 + + + W. Malik, G. Gupta, Y. Jung, "Spot release planner: efficient solution for detailed airport surface traffic optimization," the 12 th AIAA Aviation Technology, Integration, and Operation (ATIO) Conference, Indianapolis, IN, Sept 17-19, 2012. + + + + + Usability Evaluation of the Spot and Runway Departure Advisor (SARDA) Concept in a Dallas/Fort Worth Airport Tower Simulation + + MHayashi + + + 10-13 June, 2013 + Chicago, IL, USA + + + Tenth USA/Europe Air Traffic Management Research and Development Seminar + + + M. Hayashi et al, "Usability Evaluation of the Spot and Runway Departure Advisor (SARDA) Concept in a Dallas/Fort Worth Airport Tower Simulation, Tenth USA/Europe Air Traffic Management Research and Development Seminar, 10-13 June, 2013, Chicago, IL, USA + + + + + Taxi Time Prediction at Charlotte Airport Using Fast-Time Simulation and Machine Learning Techniques + + HanbongLee + + + WaqarMalik + + + BoZhang + + + BalajiNagarajan + + + YoonCJung + + 10.2514/6.2015-2272 + + + 15th AIAA Aviation Technology, Integration, and Operations Conference + Dallas, TX + + American Institute of Aeronautics and Astronautics + June 22-26, 2015 + + + H. Lee, W. Malik, B. Zhang, B. Nagarajan, Y. Jung, "Taxi time prediction at Charlotte Airport using fast-time simulation and machine learning techniques," 15 th AIAA Aviation Technology, Integration, and Operation (ATIO) Conference, Dallas, TX, June 22-26, 2015. + + + + + A Simulator for Modeling Aircraft Surface Operations at Airports + + ZacharyWood + + + SivakumarRathinam + + + YoonJung + + + MatthewKistler + + 10.2514/6.2009-5912 + AIAA 2009-5912 + + + AIAA Modeling and Simulation Technologies Conference + Chicago, IL + + American Institute of Aeronautics and Astronautics + 2009 + + + Z. Wood, M. Kistler, M. Rathinam, and Y. Jung, "A simulator for modeling aircraft surface operations at airports," AIAA Modeling and Simulation Technologies Conference, AIAA 2009-5912, Chicago, IL, 2009. + + + + + Validation of Simulations of Airport Surface Traffic with the Surface Operations Simulator and Scheduler + + RobertDWindhorst + + + JustinVMontoya + + + ZhifanZhu + + + SergeiGridnev + + + KatyGriffin + + + AdityaSaraf + + + SteveStroiney + + 10.2514/6.2013-4207 + + + 2013 Aviation Technology, Integration, and Operations Conference + Los Angeles, CA + + American Institute of Aeronautics and Astronautics + Aug 12-14, 2013 + + + 13 th AIAA Aviation Technology, Int Integration, and Operation (ATIO) Conference + R. Windhorst et al, "Validation of simulations of airport surface traffic with the surface operations simulator and scheduler," 13 th AIAA Aviation Technology, Int Integration, and Operation (ATIO) Conference, Los Angeles, CA, Aug 12-14, 2013. + + + + + Aerodrome Chart from the German Aeronautical Information Publication + + Dfs + + + + + German Air Navigation Service Provider + + 2 EDDH 2-5 + + + DFS (German Air Navigation Service Provider), Aerodrome Chart from the German Aeronautical Information Publication, AD 2 EDDH 2-5, available online at: http://www.opennav.com/pdf/EDDH/ED_AD_2_EDDH_2-5_en_2013- 12-12.pdf + + + + + Manual on global performance of the air navigation system + + Icao + + + 2009 + Montreal, Canada, ICAO + + + Doc 9883, 1 st Edition + ICAO, "Manual on global performance of the air navigation system," Doc 9883, 1 st Edition, Published in Montreal, Canada, ICAO, 2009. + + + + + Recommended key performance indicators for measuring ANSP operational performance + + Canso + + + + + CANSO, "Recommended key performance indicators for measuring ANSP operational performance," available on : https://www.canso.org/sites/default/files/RecommendedKPIforMeasurin gANSPOperationalPerformance.pdf. + + + + + J.P. Guilford, Fundamental Statistics in Psychology and Education. Third Edition. Mac Graw-Hill Series in Psychology. New York, Mc Graw-Hill Book Company, 1956, XI p. 565 p., 47/–. + + RHoronjeff + + + FMcKelvey + + 10.1017/s0770451800050569 + + + Bulletin de l'Institut de recherches économiques et sociales + Bulletin de l'Institut de recherches économiques et sociales + 1373-9719 + + 22 + 7 + + 1983 + Cambridge University Press (CUP) + New York + + + Third Edition + R. Horonjeff and F. Mc Kelvey, "Planning and Design of Airports", Mc Graw-Hill Book Company, Third Edition 1983, New York et al. + + + + + A Concept of Operations for Trajectory-based Taxi Operations + + JanNikolaiOkuniek + + + IngridGerdes + + + JoernJakobi + + + ThomasLudwig + + + BeckyLHooey + + + DavidFoyle + + + YoonCJung + + + ZhifanZhu + + 10.2514/6.2016-3753 + + + 16th AIAA Aviation Technology, Integration, and Operations Conference + Washington D.C, USA + + American Institute of Aeronautics and Astronautics + 2016. June 13-17, 2016 + + + N. Okuniek et al., "A Concept of Operation for Trajectory-based Taxi Operations," 16 th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference 2016, Washington D.C, USA, June 13- 17, 2016 + + + + + Microscopic Analysis of Airport Surface Sequencing (ALTERNATE PAPER) + + ChrisBrinton + + + BryanWood + + + ShawnEngelland + + 10.2514/6.2008-8961 + + + The 26th Congress of ICAS and 8th AIAA ATIO + Anchorage, Alaska + + American Institute of Aeronautics and Astronautics + 2008. 2016 + + + The 26 th Confress of ICAS and 8 th AIAA ATIO. 35th Digital Avionics Systems Conference September 25-29 + C. Brinton, B. Wood, S. Engelland, "Microscopic Analysis of Airport Surface Sequencing," The 26 th Confress of ICAS and 8 th AIAA ATIO, Anchorage, Alaska, 2008. 35th Digital Avionics Systems Conference September 25-29, 2016 + + + + + + diff --git a/file846.txt b/file846.txt new file mode 100644 index 0000000000000000000000000000000000000000..e6aeeedabc333407d4cf22f51fb216bccbcf02f5 --- /dev/null +++ b/file846.txt @@ -0,0 +1,571 @@ + + + + +IntroductionThe Virtual Flight Rapid Integration Test Environment (RITE) project was initiated to develop an information technology process to rapidly and easily merge data from computational fluid dynamics (CFD), wind tunnel, and/or flight test into a real-time, piloted flight simulation.The process then cycles the knowledge gained from the simulation back into the design process.To accomplish this, a new engineering design environment was constructed that combined these various data generation methods and test environments within one infrastructure.The goal of this project was to develop such a design environment to improve current design methodologies and to reduce design cycle time.Current design environments do not allow data transfer and integration to take place easily between the different technologies during the preliminary design.By providing an infrastructure that brings together all these technologies, designers will be better equipped with higherfidelity tools and methods, including simulation studies, which will lead to higher-fidelity preliminary designs.The main advantage of conducting piloted simulation studies early is to identify problems and deficiencies in aerodynamic performance, vehicle stability and control, and guidance and navigation, which can be addressed in the preliminary design phase.Simulation studies also allow for the opportunity to develop preliminary control systems early in the developmental phase.Historically, the outer mold lines of a design are defined before simulation studies and control system development can begin which may lead to expensive and complex control systems and less than optimal vehicle performance.The Space Shuttle Orbiter 1 , which was designed in the 1970's, is an example of such developmental problems.More recent design studies have used simulation tools in the design phase but without an infrastructure and process in place to facilitate and expedite its use. 2,3 sently, the Virtual Flight RITE project has demonstrated the rapid and seamless integration of CFD, flight, and wind-tunnel data into a simulation database.During an early phase of the project, the Space Shuttle Orbiter was selected as the baseline configuration for this re-design demonstration.The radius and length of the nose of the Orbiter were altered as the design parameters.This phase of the project led to a successful demonstration of integrating CFD, flight and wind-tunnel data into a simulation database for rapid preliminary design work.More recently, an integrated design process augmented by the RITE process was used to develop a viable conceptual design for a Crew Transfer Vehicle (CTV).The goal of this work was to develop an optimal design for a CTV while improving the tools and techniques of the RITE process.CTV concepts are being studied as elements of various launch architectures under the 2 nd Generation Reusable Launch Vehicle program.NASA Ames Research Center personnel developed conceptual designs for a candidate CTV which is presented here and in Reference 4. The primary mission objectives of the CTV included orbit-to-orbit transfer and rendezvous with the International Space Station (ISS).The Ames preliminary designs included the ultrahigh temperature ceramic (UHTC) material 5 .This new UHTC material enabled the use of sharp leading edges and nose geometries during hypersonic flight.Historically, re-entry space vehicles have been designed using blunt-body concepts 6 to meet the temperature constraints of current thermal protection material.Because of unique structural, thermal and chemical properties, UHTC's are capable of nonablating operation approaching 5100 deg.F. Using this new material, designers were allowed to use sharpbodied concepts in the conceptual designs.As has been reported 4 , sharp-bodied designs for reentry can greatly improve the crossrange, allowing significantly greater flexibility in selecting re-entry trajectories and landing sites.However, achieving good transonic and subsonic flight characteristics for this class of vehicle presented a challenge to the designers, which warranted the study of the approach and landing characteristics.The RITE process was ideally suited for this design study because it provided the infrastructure needed for simulation studies to be integrated into this design process.For this reason, the CTV design was selected as the next case study under the RITE process.This paper will present the results of this study and process, which included vehicle design optimization, CFD, wind tunnel and a full-motion simulation experiment conducted in the NASA Ames Vertical Motion Simulator (VMS) facility.The objectives of this experiment were to evaluate the approach and landing flight characteristics of the CTV and to develop a control system optimization tool for use in the RITE process.Six astronaut pilots evaluated the handling qualities of the CTV configurations and for comparison purposes, also evaluated the handling qualities of the Space Shuttle Orbiter and NASA-Langley's HL-20 7 design.The details of this simulation experiment and significant results will be summarized in this paper.This paper will also contain a detailed summary of the overall integrated design process including the RITE process used in developing a design for the CTV vehicle concept. +Integrated Design ProcessVarious sharp-bodied design concepts for the CTV were developed for study using an integrated design framework developed under the High Performance Computing and Communication Program (HPCCP) and Information Technology (IT) Base Programs.This integrated design process was further augmented by the RITE process to evaluate concepts during approach and landing through piloted simulations.A summary of this integrated design process including the RITE process is outlined in Figure 1.Initially, the configurations were designed using a Newtonian-based aerodynamic method in the hypersonic speed regime.Surface geometry was defined using the Rapid Aircraft Modeler (RAM) 8 , Pro-Engineer and Gridgen computer programs.Volume grids for the new geometry were generated using CART3D 9 for the unstructured, cartesian Euler solutions, MESH3D Navier-Stokes methods.The Euler solutions were computed using CART3D and AIRPLANE [10][11] while the viscous solutions were computed using Overflow-D [12][13] .Stability and control dynamic derivatives were computed using VORVIEW 14 , a vortex lattice method.Wind-tunnel data were used for validating the lowspeed CFD results.Using these CFD data, a desktop simulation tool, using Matlab Simulink, was built and run to evaluate the trim and stability characteristics of the initial design configurations.In addition, a control system optimization tool, named CONDUIT 15 , was used to optimize the control system gains for each CTV configuration.From the results of these desktop simulation studies and CFD data analyses, a gradient-based design optimization method, QNMDIF 16 , coupled with AIRPLANE was used to improve the trim, stability and aerodynamic performance of the design in an iterative process.From the results of these data generation methods, new math models for these configurations were developed and integrated for use in full-motion, 6degree-of-freedom simulation studies in the NASA Ames VMS facility.Selection and integration of these different CFD datasets was initially done manually to determine all the steps involved in this process.These datasets contained different levels of fidelity, which warranted careful integration, and generation of the data.Information technologies were then applied to aid in the decision-making and automation of this data integration process.Once the new math models were integrated into the VMS simulator's database, the handling qualities of each CTV configuration were evaluated using Cooper-Harper 17 ratings during the approach and landing phase.As shown in Figure 1, results from the handling qualities evaluations were fed back to the design team.Using this new information, a new vehicle configuration was developed and cycled back through the rapid re-design process.To demonstrate the speed of this process, a new configuration was developed after initial pilot evaluations and was completed within 4 weeks. +Vehicle Design and Optimization Baseline DesignThe baseline design for this conceptual study was based on a sharp-bodied vehicle design, designated CTV-v7.The conceptual design study of CTV-v7 and its earlier version named CTV-v5, are described in Reference 4. The tools used in developing these vehicles are also described in Reference 4. A hypersonic aerospace vehicle synthesis code (HAVOC) was the main tool used in this design process.This tool used engineering analysis methods to compute vehicle performance and design characteristics, including aerodynamics and thermodynamics, propulsion, structures, trajectory and system cost.In the design process, the synthesis code converges the design to meet mission performance requirements.The CTV-v5 and CTV-v7 adopted the same mission requirements as the NASA-Langley HL-20 design to allow a direct comparison between the blunt-bodied and sharp-bodied designs.While Reference 4 details these requirements, the major requirements are listed below:1. Reusable 2. 8 passengers and crew 3. Launch on an expendable launch vehicle 4. Subsonic L/D greater than 4The CTV-v7 design was further modified under this current study to improve its trim and stability characteristics.The CTV-v7 was renamed as Concept Vehicle 0 or CV0 to represent the baseline design for this current study.Future design configurations were deemed Concept Vehicles and increased in number sequentially.Figure 2 presents a 3-view and an isometric view of CV0. +Aerodynamic Shape OptimizationThe Euler unstructured-tetrahedral-grid-based CFD code, AIRPLANE, was recently coupled to a gradient-based optimization algorithm, QNMDIF, to develop an aerodynamic shape optimization technique for the NASA Ames HPCCP.QNMDIF is an unconstrained quasi-Newton finite-difference optimization method.These codes were evaluated for the HPCCP program by applying the method to the current CTV conceptual design study.This method was used to optimize aerodynamic performance by varying selected geometry design variables for each Concept Vehicle design.In addition, manual design efforts were employed periodically for rapid exploration of design variables and to limit the extent of shape changes made during optimization.The resulting configurations (CV1 through CV5) from this optimization process were used for the approach and landing simulation database for the RITE project.The stability and control for this class of vehicle was particularly problematic since the vehicle needed to be trimmed from re-entry at hypersonic speeds to subsonic speeds for approach and landing.AIRPLANE was used for simultaneous optimization at Mach 6.0 (descent after re-entry) and Mach 0.3, (approach/landing speed).The design objective was to trim the vehicle at the two conditions simultaneously while achieving the best lift/drag (L/D) in the process. +CV1 DesignThe design objective for CV1 was to modify the longitudinal stability so that the vehicle would be trimmed or nearly trimmed using ideal elevator deflections while simultaneously improving the performance of the vehicle at both design points.Wing twist, camber and elevator (inboard wing flap) deflection angle were the design variables.The wing flap angle was determined independently of the camber and twist design variables at the two Mach numbers, thus different geometries were analyzed at each design point during optimization.The two design points used for this optimization were 1) Mach 0.3 at angle of attack, = 9°, and 2) Mach 6.0 at = 13°.The results showed substantial changes to the camber of the vehicle combined with -12.4° of twist (wash out) to maximize the trimmed aerodynamic performance of the model.The optimized elevator deflection angles were; 0.05° (flap down) and 5.2° (flap up) for the Mach 0.3 and Mach 6.0 design points, respec-tively.Figure 3 compares the twist about the trailing edge and airfoil modifications between CV0 and CV1.The AIRPLANE lift, drag and pitching moment coefficient (C L , C D and C m ) data for CTV configurations, CV0 through CV5, are shown in Figures 4 and5 for Mach 0.3 and Mach 6.0, respectively.CV5 was designed during the VMS simulation experiment using simulation data and pilot feedback and is, therefore, described in a later section. +CV2 DesignOne aspect of the CV1 design that was considered undesirable was an excessive amount of negative camber that leads to a concave upper surface.Therefore, the upper surface of the break section was modified manually to eliminate this concavity.As can be seen from Figures 4 and5, the effect on the force and moment coefficients was minimal as expected.The moment coefficient was shifted slightly to the right (increased nose down moment) at Mach 0.3, and almost imperceptible at Mach 6.0 (Figures 4 and5).At the hypersonic Mach numbers the pressures on the wing surfaces became nearly constant regardless of upper surface shape. +CV3 DesignThe design of the CV3 configuration was done manually.The CV3 configuration consisted of the CV2 configuration with positive wing dihedral applied to the outer wing panel beyond the trailing edge break.The inclusion of wing dihedral was deemed necessary to stabilize the lateral-directional stability modes.A linear analysis was conducted on the preliminary aerodynamic data to determine the longitudinal and lateral-directional stability modes for all the CV configurations to date.The results of this analysis predicted dutch-roll mode instabilities for CV0, CV1 and CV2 for the open-loop dynamic system.Figure 6 shows the open-loop lateral-directional poles for CTV configurations CV0 through CV2 and CV3 with 30° and 45° dihedral.This figure focuses in only on the poles near the imaginary axis where the problem occurred and does not show the poles to the far left.As can be seen from the figure, the open-loop poles of CV0 through CV2 are on the right-hand side of the real axis and, hence, are unstable.This instability increased from CV0 to CV2.With the addition of dihedral, the open-loop poles moved to the left or stable side of the real axis as shown in the figure.The 45° dihedral case resulted in complex poles more stable than the 30° dihedral case; however, its poles on the real axis were seen to be less stable than the 30° dihedral case.Figure 7 compares the untrimmed pitching moment characteristics at Mach 0.3 between the CV3 dihedral cases and CV0 through CV2.As can be seen from this figure, the effect of the 30° dihedral case caused the pitching moment curve to become less stable above a C L of 0.3 to a C L of 0.6.This effect was even more dramatic for the 45° dihedral case.Based on these two figures, it was decided that the 30° dihedralC m CL CV3 30° dihedral CV3 45° dihedral CV4 2.25° dihedral CV4 30° dihedral Figure 7. Untrimmed pitching moment characteristics at Mach 0.3 for 30°and 45° dihedral case would yield better results and, therefore, the CV3 configuration was defined to be CV2 with 30° dihedral.Figure 8 compares the outboard dihedral change between CV2 and CV3.The force and moment coefficients are compared in Figures 4 and5.A positive aspect to the addition of the wing dihedral is that the vehicle can easily be trimmed for a wider range of lift coefficients.The effect of the wing dihedral at Mach 6.0 shows a similar rotation of the moment curves.Neutral stability is also seen for fairly wide range of lift coefficients (0.075 to 0.225). +CV4 DesignThe design of the CV4 configuration was performed using AIRPLANE optimization of the CV0 configuration.This design permitted small changes in the planform of the vehicle, whereas all previous designs did not.The design variables included wing twist about the leading edge rather than the trailing edge since the previous CV1 design developed substantial outboard leading edge droop from trailing edge twist, twisting about the leading edge would likely raise the trailing edge up.Other design variables included wing sweep and wing dihedral.The sweep was limited to ±2 inches while the dihedral was limited to ±2.2 in. of vertical movement during optimization.The rolling (C l ) and yawing (C n ) moment coefficients were assessed for this vehicle at Mach 0.3 and are presented in Figure 10.The magnitude of the rolling moment was determined to be too small for the optimized dihedral angle of 2.25°.Thus, to increase the magnitude of the rolling moment coefficient, the wing dihedral was increased to 30° to match that of the CV3.This increase in dihedral made the rolling and yawing moments equal and opposite which was desirable.However, the increase in dihedral had a negative effect on the pitching moment curves.As can be seen in Figures 7 and9, the Mach 0.3 moment curve was rotated so that the vehicle was no longer statically stable and the trimmed Mach 6.0 vehicle now had a positive, nose-up pitching moment, respectively.he design team decided that the rolling and yawing moments being equal and opposite with the 30° of dihedral outweighed the adverse affects on the trim and stability at both Mach numbers.In addition, the team was interested in the challenge of achieving a vehicle with good handling qualities via the control system optimization process for this design.Thus, the CV4 configuration included the 30° dihedral.The resulting changes to the planform as compared to the CV3 configuration are shown in Figure 11. Figure 12 shows the resulting changes to the wing design also compared to the CV3 wing design.Finally, comparison of the CV4 force and moment coeffi-cients to the other CV configurations is shown in Figures 4 and5.The wind-tunnel model was fabricated from a polyester resin using a stereo lithography technique.The +Data Generation and IntegrationOnce the CTV geometries (CV1 through CV4) were defined using the aerodynamic shape optimization technique described previously, various data generation methods were employed to develop an aerodynamic database for each CTV configuration.These databases were developed for integration into the full-motion, 6-degree-of-freedom simulation in the VMS for approach and landing studies.The data generation methods consisted of Euler solvers (AIR-PLANE and FlowCart), a Navier-Stokes solver (Overflow-D) and a vortex lattice method (VOR-VIEW). +Geometry Modeling and Grid GenerationEach CTV geometry was transferred to the RAM program at the end of the aerodynamic shape optimization process.The RAM program is a tool for quickly defining geometry for aerospace vehicles and exporting data for use in grid generation.At the end of the aerodynamic shape optimization process, the changed geometry sections were entered manually into the RAM program.RAM then re-lofted the wing with the new sections.Once in the RAM program, the geometry file was automatically transferred into a grid generation package.By the end of CV4, an automated process was developed to directly transfer the geometry files from AIRPLANE to the CAD packages.All CTV CAD databases were generated in Pro/Engineer (ProE) 2000i.ProE was the CAD system of choice because of its parametric design capability.After the geometry from RAM was imported into ProE, the gridding software, Gridgen was used to generate the surface grids.Gridgen read the CAD database in IGES format, which ProE converted to its own file format.Beginning with the baseline, CV0 model, a basic grid topology for Overflow-D was developed to simplify and expedite future geometry changes to the models.Figure 16 shows the symmetry plane of a typical unstructured Cartesian grid used in the Euler (FlowCart) calculations.Similarly, the symmetry plane and surface overset structured grids created for the Navier-Stokes solver are illustrated in Figure 17.Finally, the symmetry plane of the AIRPLANE tetrahedral grids is illustrated in Figure 18.In comparison to Figures 16 and17, the AIRPLANE grid showed a smooth gradation from the surface to the farfield boundary. +Data Generation using CFD codesThe first step in the data generation process was the selection of appropriate CFD codes.Ideally, all data generated would use the highest fidelity code available which implies that only Navier-Stokes (N-S) simulations would be computed.However, this is not feasible because these calculations are very time consuming.Hence, the choice of a particular solver is usually a tradeoff between accuracy and turnaround time.Under the RITE process, a combination of N-S and Euler simulations were used for the creation of the simulator aerodynamic database.The N-S solver Figure 19 compares the differences in the force and moment coefficients computed for the various CFD codes used.As mentioned previously, the Euler solver, AIRPLANE, was coupled with an optimizer to maximize the performance characteristics of the configurations.Once the configurations were optimized, the Euler solver, FlowCart, was used to expedite the data generation process.The figure shows good agreement between these two Euler solvers.Figure 19 also shows good agreement between the N-S, the Euler solvers and the wind-tunnel data for <10°, except for the pitching moment coefficients.At higher , the differences between the N-S and wind-tunnel data can be attributed to several factors including differences in Reynolds number, turbulence modeling, model sting correction and small-scale effects.The N-S solver was run using a Reynolds number of 0.5 million/ft.and the wind tunnel was run at a Reynolds number of 0.6 million/ft using a 5%scale model.These findings indicated that further study was needed to resolve the differences in the data.However, the wind-tunnel test was run very late in this integrated design process and no further studies could be undertaken prior to the simulation experiment.Therefore, the N-S data was used for most of the data integration process.In addition to static forces and moments, the motion simulator also needed dynamic coefficients to model the vehicle.VORVIEW was selected for this task because it was the only tool capable of computing these quantities in a reasonable amount of time.It is unknown how well VORVIEW predicted these coefficients since no wind tunnel data or other CFD results were available for a comparison.A check of the dynamic coefficients estimated using VORVIEW on the HL-20 with data from NASA Langley showed some differences in some of the quantities.Clearly, a study to assess the accuracy of VORVIEW and the accuracy required for the simulator is warranted.Alternative methods of computing dynamic derivatives (such as modifying FlowCart or running unsteady simulations using Overflow-D) are currently being evaluated in the next phase of the project. +Data IntegrationThe integration of data was very straightforward.Although many sources of data were available, the aerodynamic database was constructed mainly from the N-S results.No attempts were made to merge the wind tunnel data with CFD solutions.As discussed previously, there was insufficient time for careful study of these results to integrate into the database.The CTV configurations (CV0-CV4) without deflected control surfaces were modeled using Overflow-D.The N-S simulations of deflected control surfaces were computed only on the baseline vehicle (CV0).These results were then extrapolated to the other CTV configurations when constructing the aerodynamic database.The Euler results consisted of the deflected speedbrake (split rudder) and ground effect cases.Finally, the VORVIEW results were integrated into the database which consisted of the dynamic derivatives.Control System Design and Optimization A preliminary control system and linear aerodynamics model were initially developed and evaluated using desktop simulation tools (Matlab and Simulink).The HL-20 control system was used as a starting point for the CTV control system.This control system was modified to account for the control surfaces used on the CTV configurations.Figures 13 and20 illustrate the HL-20 and CV1 configurations, respectively, with the control surfaces accentuated in color.As can be seen from the figures, there were major differences between the HL-20 and CTV control surfaces.For the CTV configuration (Figure 20), the two inboard flap surfaces were designed for pitch control and trim.The size of these flap surfaces and alternate control surfaces were explored for best trim performance.The baseline vehicle, CV0, was originally designed with 20% chord flaps.In addition to increasing the size of this control surface, body flaps, leading edge slats and flaperons were all evaluated as Referring to Figure 20, the two outboard surfaces were used as ailerons for roll control and the rudder surface was used for directional control.A split rudder surface was used as a speedbrake for speed control on approach and landing.These control surfaces were used for all the CTV configurations.The CTV preliminary control system was implemented into a full-motion 6-degree-of-freedom simulation in the NASA Ames VMS facility.The control system, aerodynamics and trajectory performances were initially evaluated during the build-up period in fixed-base operation.The initial simulation studies, conducted on CV0, revealed the following deficiencies: inadequate trimming capability, insufficient longitudinal control power and lateral-directional Figures 21 and22 present the longitudinal and lateral-directional control system architectures used for all the CTV configurations.Initially, the directional control law consisted of a yaw-rate damper in the feedback loop of the yaw channel.From initial pilot evaluations, this control law was found to be inadequate and at times unstable for the approach and landing tasks.Alternate feedback control schemes were explored to improve the directional control.As a result, beta-dot feedback was determined to yield improved handling qualities for directional control.This feedback system is shown in Figure 22 and was used for all the CTV configurations.For each of the Concept Vehicles, new aerodynamics models were generated and implemented into the simulation.New control system gains were optimized for each new aerodynamics model, corresponding to CV0 through CV4, using CONDUIT.A linear model of the vehicle dynamics was developed and implemented for use with CONDUIT.The model was verified against the 6-degree-of-freedom simulation using dynamic checks.In addition to optimizing control system gains, the optimization tool also predicted handling qualities levels from userdefined handling qualities and flight control system specifications.These specifications can be defined by the user or selected from CONDUIT's libraries of standard fixed-and rotary-wing specifications.When optimizing in CONDUIT, the program tries to achieve the best or Level 1 handling qualities for the selected specifications by varying the user-defined design parameters.Since this was the first time using CONDUIT for this particular application, a major goal of the optimization process was to determine the best combination of handling qualities specifications and design parameters to be used consistently for all the CTV configurations.Initially, longitudinal and lateral-directional specifications for a fixed-wing vehicle of similar class and mission to the CTV were selected from the military specifications library included in the CONDUIT software.Although the results from CONDUIT showed Level 1 handling qualities for all the selected specifications, pilot evaluations from the full-motion simulation resulted in Level 2 and 3 handling qualities for all the CTV configurations.Much work was done during the VMS simulation to re-define the specifications used in the CONDUIT optimization process.These military specifications 19 were defined for large, heavy class vehicles similar in size to the CTV but for higher L/D performance than the CTV.Therefore, new specifications were needed for a reentry space vehicle like the CTV.After several iterations of obtaining pilot evaluations and re-defining the specifications, a good set of handling qualities and flight control system specifications were achieved.In order to obtain good results, the boundaries between the Level 1 and 2 regions for the longitudinal crossover frequency and damping ratio specifications were altered.As a result, much higher damping specifications and crossover frequencies than the military specifications were achieved as shown in Figures 23 and24.In addition, a good set of design parameters were also achieved.These design parameters, or in this case the control system gains are shown with prefixes dpp_ in Figures 21 and22 for the longitudinal and lateral-directional control systems.This final combination of specifications and design parameters were used for all the CTV configurations.In this way, each new CTV configuration was quickly evaluated during the full-motion simulation without laboriously fine-tuning each control system gain and conducting a time-and frequency-domain dynamic analysis.The results of the handling qualities evaluations are discussed in the next section. +VMS Simulation ExperimentA piloted simulation experiment of each of the CTV configurations in the approach and landing phase took place during a 4-week window in the NASA Ames VMS facility. +Experiment SetupFull 6-degree-of-freedom simulations for each of the 6 CTV configurations (CV0 through CV5) were built and validated for this experiment.The CV5 configuration was designed, integrated and flown during the simulation experiment.The rapid re-design process of the CV5 configuration is described in a later section.The HL-20 simulation originated from NASA-Langley and was transferred and integrated into the VMS simulation.The Langley simulation was also a 6-degree-of-freedom, full-motion simulation.In addition to dynamic check comparisons made between the HL-20 simulations, the simulation in the VMS was checked out and verified by NASA-Langley HL-20 pilot, Robert Rivers 3 .With the help of Rivers, the VMS HL-20 simulation was tuned and adjusted to match the Langley HL-20 simulation.The Space Shuttle Orbiter simulation was taken from the existing VMS simulation of the Orbiter and was used without modification.This VMS simulation of the Orbiter is considered a high-fidelity simulation and is currently used for astronaut training and space shuttle engineering studies.In addition, the same cab and hardware from the Orbiter simulation was used for the HL-20 and CTV simulations.Only the software was changed between each simulation.It took approximately 5 to 10 minutes to make the changes between simulations.Each simulation consisted of an aerodynamic model, trajectory and guidance algorithms, control system architectures, HUD and graphics displays.For each CTV configuration, a function table was generated consisting of the aerodynamics data.The experiment also took advantage of the virtual laboratory (VLAB) capability available in the VMS facility.This VLAB was developed to provide realtime transmittal of data, voice and graphics displays throughout the simulation to researchers off-site.The VLAB was used throughout the simulation to facilitate the exchange of information and data between the various technology groups at different sites at NASA-Ames and for demonstration purposes at NASA-Johnson and Marshall Space Flight Centers. +Piloted TasksThe main task during the simulation was to approach and land from the Heading Alignment Cone (HAC) and 10K ft.initial conditions on to Kennedy Space Center (KSC) runways in wind and turbulent weather conditions.Other tasks included pitch, roll and rudder maneuvering tasks in air and during rollout.Three-axis doublets were also performed to evaluate the open-loop and closed-loop dynamic response of the vehicle.For the handling qualities evaluations, the test matrix was narrowed down to three piloted tasks.These tasks included: 1) a nominal landing, 2) a lateral offset landing, and 3) a crosswind landing.All tasks were initiated at 10K ft.After the first landing, the pilot had the option to request a 5K ft.initial condition.For the nominal landings, the pilot was instructed to follow the guidance marker all the way down to touchdown.The guidance marker followed a 17° glideslope or 300 knots approach speed for the Orbiter and HL-20.The target touchdown speeds for the Orbiter and HL-20 landings were 195 kts.and 200 kts.respectively.The approach and target touchdown speeds for the CTV configurations were 200 kts.and 160 kts.respectively.This resulted in a 12° glideslope angle on final.Because of the higher L/D of the CTV as shown in figure 14, the pilot was capable of approaching and touching down at desirable slower speeds.Several other touchdown speeds were evaluated for the CTV configurations.It was determined that 160 kts.was the slowest speed that could be achieved without affecting the handling qualities ratings.During the lateral offset tasks, the pilot was instructed to follow the guidance marker all the way down to 3000 ft.altitude.At this point, the pilot was instructed to make a sharp, abrupt right turn to a landmark approximately 400 ft.away from the centerline and follow this new alignment down to 300 ft.altitude.Then the pilot turned back towards the runway and proceeded to land on the centerline.This task was designed to be an aggressive maneuver to excite the lateral-directional modes to uncover any deficiencies in the control and dynamic systems.The crosswind landings were evaluated in 20 kt.crosswinds.These were realistic winds taken from the Orbiter's simulation database of wind profiles measured at KSC and Edwards Air Force Base landing sites.These winds were then extrapolated to give 20 kts. of crosswind.The components of the winds used for the crosswind landing evaluations were 15 kts.tailwind and 20 kts.left crosswind. +Piloted EvaluationsSix astronaut pilots, including three flightexperienced Shuttle commanders, and one NASA-Langley HL-20 pilot participated in the experiment and handling qualities evaluations.The pilots provided valuable information and insight to the development of the simulation and design modifications to the CTV.In addition to the piloted tasks described previously, the pilots were instructed to land within desired criteria when assigning Cooper-Harper ratings.The landing criteria are outlined in Table 2 below.Based on these criteria and the Cooper-Harper scale, the pilots were instructed to rate the handling qualities of each vehicle.In addition, the pilots were instructed to perform at least 2 repeat runs before assigning their ratings.These guidelines were followed by all the pilots for all the simulation evaluations.Figure 25 presents the final Cooper-Harper ratings (CHR) for the Orbiter, HL-20 and the CTV Concept Vehicle configurations (CV0 through CV5).This figure shows averaged piloted ratings for each of the 3-piloted tasks described previously in the last sub-section.The ratings were averaged over all the pilot ratings for the specified piloted task and vehicle configuration.Using the CHR scale, ratings 1 through 3 (with CHR of 1 being the best rating) are defined as Level 1 handling qualities.Ratings 4 through 6 are defined as Level 2 handling qualities while ratings 7 through 9 are defined as Level 3 handling qualities.The handling qualities ratings for the CTV configurations in Figure 25 were taken using the final set of control system gains obtained with CONDUIT.As described in an earlier section, a good combination control system and handling qualities specifications along with selected design parameters were required for CONDUIT to achieve optimal control system gains and Level 1 handling qualities.Throughout the simulation, the pilot handling qualities were compared to the predicted handling qualities of CON-DUIT.Initial comparisons resulted in poor correlations between the pilot ratings and CONDUIT.Although the CONDUIT program predicted Level 1 handling qualities, the pilot handling qualities for the CTV configurations ranged from Level 2 to Level 3 handling qualities.Throughout the simulation experiment, pilot feedback was used to make changes in the control system architecture, trajectory and guidance algorithms, HUD displays and control system optimization specifications for improved performance in handling qualities ratings.As a result, good correlations were finally obtained between pilot ratings and CONDUIT for the CTV configurations.These pilot ratings improved from Level 3 to Level 1 ratings for 5 of the 6 CTV configurations (CV0 through CV3 and CV5) as shown in Figure 25.The handling qualities for CV4 as shown in the figure did not use this final combination of specifications in CONDUIT.Due to the time constraints of the pilots' schedules, the simulation experiment ended before this final evaluation could be made.The CTV pilot ratings of Figure 25 were much lower than the pilot ratings of the Orbiter and HL-20 for similar tasks except for CV4.The pilots' comments also reflected and confirmed these good ratings.From a comparison of the CTV ratings, CV3 is shown to have the best overall ratings.However, one cannot immediately conclude that the CV3 configuration was the best overall configuration.More analysis is required to determine the control power used in achieving each task. +Rapid Redesign of CV5The simulation experiment was also successful in demonstrating the capability to rapidly re-design a configuration using pilot feedback.During this rapid redesign process, a new aerodynamics model was generated and implemented into the full-motion simulation, control system gains were re-optimized and new pilot evaluations were obtained on the new design.A sixth configuration (CV5) was generated and evaluated using this rapid process during the simulation experiment.From pilot feedback and recommendations, fuselage geometry variables were optimized to increase fuselage width and nose droop.Increasing these variables would improve the design by increasing fuselage volume and improving pilot visibility.Poor pilot visibility on this vehicle was a limiting factor in selecting a touchdown speed and contributed to the handling qualities ratings on landing.As mentioned previously, the touchdown speed for the CTV configurations was 160 kts.which required an angle-ofattack of approximately 12°.Preliminary evaluations of CV0 through CV3 led the pilots to choose CV3 as the one with the best handling qualities of the 4 designs.Thus, CV3 was used as the baseline configuration for the rapid redesign case.The goal was to further improve the performance and handling characteristics of the CV3 configuration by applying small shape changes to the body, and allowing the wing to be re-twisted.Seven design variables were used.These included wing twist about the leading edge applied at the tip and lofted to the side of body.The remaining 6 design variables were applied to the fuselage.These variables were body leading and trailing droop, forebody camber, and forebody upper surface thickness.The resulting fuselage change to CV5 is shown in Figure 26.The difference in the fuselage can be easily seen by this side view.The nose was moved up 1.2 percent, and the aft body was drooped 3.2 percent.The wing twist about the leading edge was -2.293 degrees (further washout).The forebody thickness was increased at 25% and reduced at 60% of the forebody.These forebody thickness changes were the most pronounced changes seen in the model.From the results of the computational simulations, a new function table for CV5 was generated and implemented into the simulation.In addition, the control system was optimized using the same specifications as the previous 5 CTV configurations.The pilot evaluations yielded Level 1 handling qualities ratings for CV5 for all specified tasks during the approach and landing as shown in Figure 25.In conclusion, the VMS simulation experiment was successful in obtaining Level 1 handling qualities ratings for 5 of the 6 CTV configurations (CV0 through CV5 except CV4).The experiment was also successful in obtaining handling qualities ratings for the Space Shuttle Orbiter and HL-20 simulations for comparison purposes.The experiment also provided enough pilot feedback to help establish good design specifications for the control system optimization tool.Finally, the re-design case was successful in demonstrating that the RITE tools and processes are capable of rapidly changing the vehicle design and obtaining handling qualities from a piloted simulation. +SummaryThe Virtual Flight Rapid Integration Test Environment project has demonstrated the capability of integrating CFD, flight and wind-tunnel data into a simulation rapidly and seamlessly.The goal of this project was to develop a design environment that merges these technologies and data to meet the challenges of designing air and space vehicles.The objectives, to reduce the design cycle time, and to maximize the performance and utilization of these current resources, were met.A design of a Crew Transfer Vehicle concept was developed using this process.An information technology process and infrastructure was created to facilitate the integration and selection of the data.A simulation experiment, conducted in the NASA Ames VMS facility, evaluated the lowspeed handling qualities of the various configurations in the approach and landing phase.Six astronaut pilots evaluated each of the configurations using Cooper-Harper ratings.The knowledge gained from the simulation data and pilot evaluations were quickly returned to the design team.From these findings, a new configuration was developed and cycled back through the simulation evaluation.The results and findings from this simulation experiment were presented.The details of this integrated design process along with the six resulting CTV configurations (CV0 through CV5) were also presented.Figure 2 .2Figure 2. Baseline Vehicle Design, CTV-v7 (or CV0) +Figure 3 .Figure 4 .Figure 5 .345Figure 3.Comparison of CV0 and CV1 +6 .6Lateral-directional poles for open-loop system for CV0 through CV3 +Figure 8 .8Figure 8. Wing dihedral change between CV2 and CV3 +Figure 9 .9Figure 9. Untrimmed pitching moment characteristics for CV4 at Mach 6.0 at 30°and 45° dihedral +Figure 10 .10Figure 10.CV4 rolling and yawing moment coefficients at Mach 0.3 for angle of attack of 10°T +Figure 11 .Figure 12 .1112Figure 11.Wing planform changes to CV4 +Figure 13 .13Figure 13.NASA Langley's HL20 vehicle Wind-Tunnel Test As part of the design process, low-speed wind tunnel tests of the CV0 and CV2 configurations were performed in the NASA Ames Fluid Mechanics Lab 3foot by 4-foot indraft wind tunnel.A detailed de- +Figure 14 .14Figure 14.Comparison of untrimmed aerodynamic data at Mach 0.3 for Space Shuttle Orbiter, HL-20 and CTV configurations wings of the CV0 and CV2 were attached to a common fuselage.Inboard and outboard elevons and rudder control surfaces were modeled as separate, removable components to represent various control surface deflections.Figure 15 below illustrates the wind-tunnel model of CV0. +Figure 15 .15Figure 15.Wind-tunnel model of CV0 The model was mounted on a sting-type support through the flat base of the model.A six-component force balance and 3-axis accelerometer was mounted inside the model for force and moment and angle-ofattack measurements, respectively.For these tests, the wind tunnel was operated at a dynamic pressure of 20 psf, which resulted in a Mach number of 0.11 and Reynolds number of 1.1 million based on a body length of 22 inches.Data were taken over a range of angle of attack from -3° to 22°, and a range of sideslip angles from 0° to 5° at 0° angle of attack.During the testing, each configuration was tested inverted as well as upright to determine wind tunnel stream angle correction.No blockage or wall-effects correc- +Figure 16 .16Figure 16.Unstructured Cartesian Grid +Figure 19 .19Figure 19.Baseline CTV (CV0) configuration, Mach< 0.3 used for these computations was Overflow-D and the Euler solvers used were FlowCart and AIRPLANE. +Figure 20 .20Figure 20.Control surfaces of CV1 configuration trimming devices.As a result, the 30% chord inboard flaps were determined to give the best trim performance.These 30% flaps were used on CV1 as shown in Figure 20 and all subsequent Concept Vehicles. +Figure 21 .21Figure 21.CTV Longitudinal Pitch Control Laws +FiguresFigures 23 and 24 show the longitudinal and lateraldirectional handling qualities specifications used in the final evaluation of each CTV configuration.Each specification or block has 3 levels of compliance, as shown.The blue region defines the Level 1 handling qualities.The magenta region defines the Level 2 and the red region defines the Level 3 handling qualities.The different symbols in Figure 23 represent the different size input steps used for the pitch loop.The arrows in figures indicate the values are beyond the scale of the figure. +Figure 23 .23Figure 23.Longitudinal Handling Qualities Specifications from CONDUIT +Figure 25 .25Figure 25.Summary of Cooper-Harper Ratings for all Vehicles +Figure 26 .26Figure 26.Fuselage changes to CV5 compared to CV3 The CV0 through CV5 Mach 0.3 AIRPLANE solutions were shown in Figure 4, and the Mach 6.0 AIRPLANE solutions were shown in Figure 5. Figure 27 shows a 3-view and an isometric view of the CV5 configuration. +Figure 27 .27Figure 27.Rapid redesign configuration, CV5 +10for the unstructured tetrahedral Euler-flow solutions, and Gridgen for the structured Navier-Stokes flow solutions.Data generation methods included methods of various levels of fidelity including vortex lattice methods, Euler methods andGeometry Modeling Geometry Modelingand andGrid Generation Grid GenerationVehicle Design And Optimization Vehicle Design And OptimizationIntegrated Design Process Integrated Design ProcessData Generation And Data Generation AndCFD Analysis CFD AnalysisDesktop Simulation Desktop SimulationAssessment of Assessment ofDesign DesignWind Tunnel Wind TunnelValidation ValidationSimulator SimulatorDatabase DatabaseCreation CreationControl System Control SystemOptimization OptimizationRITE RITEPiloted PilotedProcess ProcessPiloted PilotedEvaluations EvaluationsSimulations SimulationsFigure 1.Flowchart of integrated design process American Institute of Aeronautics and Astronautics +Table 11lists the range of flowconditions and control surface deflections computedfor this database.Table 1.Range of conditions for computational dataMach number0.3 to 0.9Angle of attack-10° to 50°Sideslip angle0 to 20°Inboard flaps-20° to 20°Outboard flaps-20° to 20°Rudder0 to 20°Speedbrake0 to 30° +Table 2 .2Landing criteria for Cooper-Harper RatingsParameterDesiredAdequateY touchdown+/-10 ft.+/-20 ft.Touchdown speed+/-7 kts.+/-12 kts.Altitude rate<3.5 ft/sec <5.0 ft/sec + + + + +AcknowledgementsThe authors would like to thank the following NASA Johnson Space Center astronauts and Langley pilot for their support and participation in the VMS experiment.Special thanks to Kenneth Ham, Terrence Wilcutt, Steve Lindsey, Scott Horowitz, George Zamka, Chris Ferguson and Robert Rivers.The authors would also like to thank the members of the Systems Analysis Branch and the Computational Sciences Division for their many contributions to the design and information technology process.Special thanks to Alex Te, Arsenio Dimanlig, Scott Thomas and Ray Hicks, all of the ELORET Corp., and Jorge Bardina of NASA Ames Research Center.The authors would also like to thank Mark Tischler, Jack Franklin, Sean Swei and Kenny Cheung for their assistance with the CONDUIT optimization process.Finally, special thanks to the members of the Simulation Division for their many contributions and operation of the VMS simulation experiment.Special thanks to Joseph Ogwell, John Bunnell, Dan Wilkins and Christopher Sweeney, all of the Logicon Corp. + + + + + + + + + Research analysis of Space Shuttle Orbiter entry aerothermodynamic flight data at the NASA Langley Research Center + + DThrockmorton + + 10.2514/6.1981-2429 + + + 1st Flight Test Conference + + American Institute of Aeronautics and Astronautics + March 1983 + + + Shuttle Performance: Lessons Learned, NASA Langley Research Center, NASA Conference Publication 2283, March 1983. + + + + + Investigation of the launch pad abort capabilities of the HL-20 lifting body + + EBJackson + + + RobertRivers + + + RajivChowdhry + + + WARagsdale + + + DavidGeyer + + 10.2514/6.1993-3690 + AIAA Paper 91- 2955 + + + Flight Simulation and Technologies + + American Institute of Aeronautics and Astronautics + August 1991 + + + Jackson, E. B., Ragsdale, W.A., Powell,R.W., Utilization of Simulation Tools in the HL-20 Conceptual Design Process, AIAA Paper 91- 2955, August 1991. + + + + + Preliminary piloted simulation studies of the HL-20 lifting body + + RobertARivers + + + EBruceJackson + + + WARagsdale + + 10.2514/3.46530 + + + Journal of Aircraft + Journal of Aircraft + 0021-8669 + 1533-3868 + + 31 + 3 + + May-June 1994 + American Institute of Aeronautics and Astronautics (AIAA) + + + Rivers, R. A., Jackson, E.B., Ragsdale, W.A., Preliminary Piloted Simulation Studies of the HL-20 Lifting Body, Journal or Aircraft Vol. 31, No. 3, pp. 556-563, May-June 1994. + + + + + Conceptual design of a 'SHARP' - CTV + + DavidKinney + + + JeffBowles + + + LilyYang + + + CathyRoberts + + 10.2514/6.2001-2887 + + + 35th AIAA Thermophysics Conference + + American Institute of Aeronautics and Astronautics + 2001-2887. June 2001 + + + Kinney, D. J., Bowles, J.V., Yang, L.H. and Roberts, C.D., Conceptual Design of a 'SHARP'- CTV, AIAA Paper 2001-2887, June 2001. + + + + + First Flight Demonstration of a Sharp Ultra-High Temperature Ceramic Nosetip, NASA TM 112215 + + PKolodziej + + + JBull + + + JSalute + + + DLKeese + + + December 1997 + + + Kolodziej, P., Bull, J., Salute, J. and Keese, D.L., First Flight Demonstration of a Sharp Ultra- High Temperature Ceramic Nosetip, NASA TM 112215, December 1997. + + + + + Flight-determined subsonic lift and drag characteristics of seven lifting-body and wing-body reentry vehicle configurations with truncated bases + + EdwinSaltzman + + + KCWang + + + KennethIliff + + 10.2514/6.1999-383 + AIAA Paper 99-0383 + + + 37th Aerospace Sciences Meeting and Exhibit + + American Institute of Aeronautics and Astronautics + January 1999 + + + Saltzman, E. J., Wang, K.C. and Iliff, K.W., Flight-Determined Subsonic Lift and Drag Characteristics of Seven Lifting-Body and Wing- Body Reentry Vehicle Configurations with Trun- cated Bases, AIAA Paper 99-0383, January 1999. + + + + + 21st century space transportation system design approach - HL-20 personnel launch system + + HowardWStone + + + WilliamMPiland + + 10.2514/3.25561 + + + Journal of Spacecraft and Rockets + Journal of Spacecraft and Rockets + 0022-4650 + 1533-6794 + + 30 + 5 + + September-October 1993 + American Institute of Aeronautics and Astronautics (AIAA) + + + Stone, H. W. and Piland, W. M., 21 st Century Space Transportation System Design Approach: HL-20 Personnel Launch System, AIAA Journal of SpaceCraft and Rockets, 30(5):521-528, Sep- tember-October 1993. + + + + + An integrated vehicle modeling environment + + JosephTotah + + + DavidKinney + + + JohnKaneshige + + + ShaneAgabon + + 10.2514/6.1999-4106 + + + 24th Atmospheric Flight Mechanics Conference + + American Institute of Aeronautics and Astronautics + 1999 + + + + Totah, J. J., Kinney, D.J., Kaneshige, J.T., and Agabon, S., An Integrated Vehicle Modeling Environment, AIAA Paper 99-4106, 1999. + + + + + Robust and efficient Cartesian mesh generation for component-based geometry + + MJAftosmis + + + MJBerger + + + JEMelton + + + MJAftosmis + + + MJBerger + + + JEMelton + + 10.2514/6.1997-196 + AIAA Paper 97-0196 + + + 35th Aerospace Sciences Meeting and Exhibit + + American Institute of Aeronautics and Astronautics + January 1997 + + + Aftosmis, M. J., Berger, M. J., and Melton, J. E., Robust and Efficient Cartesian Mesh Generation for Component-Based Geometry, AIAA Paper 97-0196, January 1997. + + + + + Tetrahedral Mesh Generation and Optimization + + TJBaker + + + JCVassberg + + + + Proceedings of the 6 th International Conference on Numerical Grid Generation, International Society of Grid Generation (ISGG), Mississippi State Univ + the 6 th International Conference on Numerical Grid Generation, International Society of Grid Generation (ISGG), Mississippi State Univ + + 1998 + + + + Baker, T. J., and Vassberg, J. C., Tetrahedral Mesh Generation and Optimization, Proceedings of the 6 th International Conference on Numerical Grid Generation, International Society of Grid Generation (ISGG), Mississippi State Univ., 1998, pp. 337-349. + + + + + Improvements to the aircraft Euler method + + AJameson + + + TJBaker + + 10.2514/6.1987-452 + AIAA Paper 87-0452 + + + 25th AIAA Aerospace Sciences Meeting + + American Institute of Aeronautics and Astronautics + January 1987 + + + Jameson, A., Baker, T. J., Improvements to the Aircraft Euler Method, AIAA Paper 87-0452, January 1987. + + + + + Recent enhancements to OVERFLOW + + DennisJespersen + + + ThomasPulliam + + + PieterBuning + + + DennisJespersen + + + ThomasPulliam + + + PieterBuning + + 10.2514/6.1997-644 + + + 35th Aerospace Sciences Meeting and Exhibit + + American Institute of Aeronautics and Astronautics + January 1997 + + + Recent Enhancements to OVERFLOW, AIAA Paper 97-0644 + Jesperson, D.C., Pulliam, T.H., and Buning, P.G., Recent Enhancements to OVERFLOW, AIAA Paper 97-0644, January 1997. + + + + + Unsteady aerodynamic simulation of static and moving bodies using scalable computers + + RobertMeakin + + + AndrewWissink + + 10.2514/6.1999-3302 + + + 14th Computational Fluid Dynamics Conference + + American Institute of Aeronautics and Astronautics + + + + + Meakin, R. L. and Wissink, A.M., Unsteady Aerodynamic Simulation of Static and Moving Bodies Using Scalable Computers, AIAA Paper 99-3302. + + + + + A Generalized Vortex Lattice Method for Subsonic and Supersonic Flow Applications + + LRMiranda + + + RDElliot + + + WMBaker + + Contract NAS1-12972 + + + NASA Contractor Report + + 2865 + 1977 + December + + + Miranda, L.R., Elliot, R.D. and Baker, W.M., A Generalized Vortex Lattice Method for Subsonic and Supersonic Flow Applications, NASA Con- tractor Report 2865, Contract NAS1-12972, De- cember 1977. + + + + + CONDUIT - A new multidisciplinary integration environment for flight control development + + MarkTischler + + + JasonColbourne + + + MarkMorel + + + DanielBiezad + + + WilliamLevine + + + VeronicaMoldoveanu + + + MarkTischler + + + JasonColbourne + + + MarkMorel + + + DanielBiezad + + + WilliamLevine + + + VeronicaMoldoveanu + + 10.2514/6.1997-3773 + AIAA Paper 97-3773 + + + Guidance, Navigation, and Control Conference + + American Institute of Aeronautics and Astronautics + August 1997 + + + Tischler, M.B., Colbourne, J. D., Morel, M. R., Biezad, D. J., Levine, W. S., and Moldoveanu, V., CONDUIT -A New Multidisciplinary Inte- gration Environment for Flight Control Devel- opment, AIAA Paper 97-3773, August 1997. + + + + + A Review of: “PRACTICAL OPTIMIZATION” + + PEGill + + + WMurray + + + MHWright + + 10.1080/03052158208928041 + + + Engineering Optimization + Engineering Optimization + 0305-215X + 1029-0273 + + 6 + 2 + + 1982 + Informa UK Limited + San Diego + + + th ed. + Gill, P.E. Murray, W., and Wright, M.H., Practi- cal Optimization, 4 th ed., Academic Press, San Diego, 1982. + + + + + The Use of Pilot Rating in the Evaluation of Aircraft Handling Qualities, NASA TN D-5153 + + GECooper + + + RPHarper + + + Jr + + + April 1969 + + + Cooper, G.E., and Harper, R.P., Jr., The Use of Pilot Rating in the Evaluation of Aircraft Han- dling Qualities, NASA TN D-5153, April 1969. + + + + + Wind-tunnel study of an observatory dome with a circular aperture + + GregoryGZilliac + + + EthanWCliffton + + 10.1086/132941 + + + Publications of the Astronomical Society of the Pacific + PASP + 0004-6280 + 1538-3873 + + 103 + 1211 + November, 1990 + IOP Publishing + + + Zilliac, G. G., and Clifton, E.W., Wind Tunnel Study of an Observatory Dome with a Circular Aperture, NASA TM 102888, November, 1990. + + + + + Relative evaluation of MIL-STD 1797 longitudinal flying qualities criteria applicable to flared landing and approach + + JBooz + + 10.2514/6.1988-4363 + + + 15th Atmospheric Flight Mechanics Conference + + American Institute of Aeronautics and Astronautics + 1987 + + + United States Department of Defense, Flying Qualities of Piloted Vehicles, MIL-STD-1797, 1987. + + + + + +