DFWThe Automation System (CTAS) should yield a wide range of benefits, including reduced aircraft delays and controller workload.To determine the trafficflow benefits achievable from future terminal airspace automation, an analysis of aircraft landing rates and separations was performed for the Dallas/Fort Worth International Airport (DFW) using live radar inlormation.The primary goal was to obtain a reference baseline for the assessment of the CTAS as it is tested at the airport: a secondary goal was to aid in the further development of CTAS through an increased understanding of controller and pilot practices during the final approach segment of flight.This report describes the data-gathering and analysis procedure used, presents results, and makes recommendations [k)rcontinued study. The CTAS ConceptCTAS is a computer-based tool that is designed to relieve the worsening terminal-area delays caused by the continued growth of air traffic.It is intended to improve the efficiency of air traffic operations by optimizing traffic flow in terminal areas.CTAS is under development at the NASA Ames Research Center in cooperation with the Federal Aviation Administration (FAA).When complete, CTAS will consist of several integrated software tools that provide computer-generated advisories lor en-route and terminal-area controllers to guide them in managing and controlling arrival traffic.CTAS will provide accurate route projections for the efficient scheduling and sequencing of aircraft as they transition from en-route to terminal airspace.In addition, it will identify potential aircraft conflicts and present options for resolving them.One of the tools, known as the Final Approach Spacing Tool (FAST), is designed to aid terminal radar approach control (TRACON) controllers.It will generate advisories to produce optimally spaced aircraft on the final approach course, thereby maximizing runway efficiency.The FAST tool will work in conjunction with the Traffic Management Advisor (TMA) tool, which performs scheduling and sequencing of arriving aircraft in airspace controlled by the Air Route Traffic Control Center (hereafter called Center) betbre they enter terminal airspace.Reference I presents a detailed overview of the CTAS design.DFW, currently serving as a test site for CTAS, is scheduled to bc the first site for field testing of the FAST component.In the testing, controllers will provide clearanccs to rcvcnuc flights based on the displayed advisories generated by FAST.Real-time controller-in-the-loop simulations of FAST have demonstrated its potential to increase landing rates, thereby reducing arrival delays in the Center during rush periods fief.2). Study GoalsData from actual air traffic operations must be analyzed to determine the benefits potential of new automation concepts such as CTAS.Benefits analyses that rely on theory and modcls in lieu of data are open to challenge.Simulations relyonmodels thatmust bevalidated, so theymust beused inconjunction withdata collection.In addition, it isdifficulttorelate models based ontheory to actual trafficflow.Themodels require simplifications of theactual airtrafficenvironment, andoften they relyon assumptions regarding trafficflowandthepractices of controllers andpilots.Datacollection isalsoimportant forthedesign and optimization of such automation systems.A keyfunction ofCTASistomake accurate predictions oftimes of arrival tothethreshold.These arrival times areused to schedule andsequence aircraft.Accuracy ofthepredictionsisdependent onmany factors, several ofwhich can bedetermined onlythrough theanalysis ofactual data.Arrival trafficdataarealsouseful forunderstanding the practices of thecontrollers andpilots.Thisunderstanding canbeused tooptimize parameters inCTAS, such as acontroller's preferred spacings onfinal approach.Although FAST provides spacing andsequencing only tothepointofhand-off tothetower, thatspacing canbe done best if asmuch knowledge aspossible about the final phases offlightisincorporated intothetool.Theprimary goals oftheanalysis were tocharacterize DFWairtrafficflowwithout automation tools, toidentify some ofthepotential improvements ofFAST and TMA intheterminal area, andtoserve asabenchmark in analyzing thebenefits ofFAST inthefuture.Theanalysis obtained asample ofactual aircraft arrival spacings that resulted when controllers manually balanced theneed to followcomplex procedural constraints withtheneed to maintain highlanding rates. Assumptions and ScopePrevious findings based on simulation have indicated that the final approach segment is the critical point of constriction for arriving aircraft if the terminal airspace is managed effectively.Aircraft arrive at the runways in streams from several TRACON feeder gates; each of these streams may in turn be separated into two or more independent streams based on aircraft category.All streams must merge to land on three or fewer runways at DFW.If the number of arriving aircraft exceeds airport capacity, unusable time gaps between aircraft on landing can be eliminated through effective management of aircraft in the terminal airspace.Any delays incurred by aircraft were assumed to be caused by a constriction of traffic on the final approach segments of the active runways.If it was necessary to delay aircraft, the study assumed that they were delayed in en-route (Center) airspace rather than in terminal airspace.If the assumption is made that the airport acceptance rate was specified correctly, the buildup of Center delay in arriving aircraft indicated an arrival demand greater than the arrival capacity of the airport.If a traffic management system can achieve landings of consecutive aircraft as rapidly as is possible without violating FAA spacing minima, the study also assumed that the airport was handling arrivals at its maximum capacity.Although not investigated in this study, additional increases in landing rates may be achieved by resequencing aircraft to some optimal landing order.The characteristics of arrival traffic flow on final approach at DFW were identified based on radar track data, recorded for a selected set of arrival rushes over a six-month period.The results are useful in obtaining an approximation of savings achievable by optimizing traffic flow in terminal airspace.Observed trends in the utilization of runways and controller practices in spacing aircraft are also useful for the design of a terminal airspace automation tool and the tuning of its internal parameters.Because the results are not comprehensive, they should not be interpreted as an accurate statistical representation of conditions or practices at DFW. Precise dollar-value estimates of the benefits of automation must be obtained through a much larger study that incorporates the impacts of surface operations, gate availability, and air-carrier banking operations as well as a more comprehensive assessment of runway utilization. ApproachResults were based on empirical observation to the maximum extent possible.Traffic-flow data were collected from terminal and Center radar, from which threshold separations in time and distance were determined.Observed trends in the data were documented, and simple measures were used to characterize the potential benefits of a terminal-area air traffic automation tool.The analysis did not consider the possible limitations of existing automation tools.The procedures used were as follows: Radar track data and additional supporting information were recorded at DFW and The Fort Worth Air Route Traffic Control Center (ZFW), as described in section 4. Using these data, landing runways and threshold crossing times were determined for each landing aircraft.(See section 5.)Estimates of optimal threshold separations were derived (section 6), and controller practices in spacing aircraft were investigated (section 7).All observations from data were followed up as much as possible through personal communication with active and former controllers at DFW. Trends in runway utilization were also observed and runway capacity was estimated (section 8).The buildup ofdelays ofarriving aircraft intheCenter was used asthemeasure ofthepotential forairport capacity improvement.TRACON-arrival-gate crossing predictions generated byCTAS were used toquantify thedelay for each aircraft; therefore, theaccuracy ofthese predictions was evaluated (section 9).A more detailed description of FAA separation regulations for radar arrivals may be obtained from the FAA air traffic control handbook (ref.3). Data RecordingRecordings of live traffic flow at DFW were made over a 6-month period during the winter of 1994-95.Radar track data were supplied using a direct feed from the ZFW radar and the DFW ASR-9 terminal radar.Center and terminal radar recordings were made simultaneously to identify arrival rush periods in terminal airspace based on delay buildup in the Center, and to provide Center delays of each aircraft to augment the landing separation analysis.The combined recordings, which started up to 30 minutes prior to crossing the meter gate and continued until the TRACON radar track dropped out near the runway threshold, provided position histories of each arriving aircraft. DatasetA dataset suitable for analysis was extracted from the complete set by eliminating recordings containing incomplete information and recordings that contained unusual situations.Recordings with winds greater than 15 knots were also eliminated because the analysis tools were not developed to account for the separation time expansion that occurs under such conditions.The resulting usable rush-period dataset was made up of landings toallactive runways.At least tworunways were active forallrecordings.Foreach recording, thefollowing additional information wasgathered tosupport theanalysis: 1)flightrules in effect, 2)airport visibility, 3)airport ceiling, 4)runway conditions, 5)windbearing andvelocity attheairport, 6)approach typeineffect (simultaneous orstaggered), 7)theaircraft acceptance rate, and8)special conditions orrestrictions ineffect. Rush-Period IdentificationRushes (See fig.6.)The position of this noreassign point was adjustable along the final approach course from the threshold to the beginning of the common approach path.A position of 0.5 nm before the threshold was established by determining the closest point on the approach prior to the point that most aircraft started their heading change lbr the inboard runway, or started a climb-out if executing a missed approach.To estimate the threshold crossing time, the algorithm interpolated between the two radar hits closest to the noreassign point to obtain a time of closest approach.The aircraft ground speed was then used to extrapolate to the threshold.A comparison with tower observations showed the threshold crossing times to be accurate to within about 10 seconds.Very few wrong estimates of aircraft landing runways could be tolerated.Each incorrect estimate that an aircraft did not land on a particular runway resulted in an incorrect unused slot in the arrival stream, a possibly incorrect lead/trail class combination, and an incorrect assessment of landing rate.Incorrect estimates that aircraft did land on a particular runway resulted in increased counts of negative excess separations.A validation was performed to ensure that these errors were small.Landing runways and threshold crossing times were recorded for 1135 aircraft by an observer located in the DFW tower.The validation was performed under visual approach conditions.The landing estimation errors were found to be low after a zero-phase-shift Butterworth-characteristic filter was added to attenuate noise in the radar data. Measured separation distance and time _',An average error rate of 0.9 percent was seen in the landing runway estimations.Data feed and radar dropout errors caused additional errors, resulting in a total error rate of 2.3 percent for the full validation sample.This rate was deemed acceptable for the analysis.Aircraft landing on Runway 31R sometimes used the stadium visual approach, which caused the landing runway function to fail.Therefore, Runway 31R was excluded from the analysis when this approach was being used.Other errors were caused by very large separations between leading and trailing aircraft, which invalidated the function's lineof-sight separation distance approximation.Because these cases were rare, they were not removed from the analysis.0.5 nm -.,,N Threshold _d a/cReference point for separation measurement Optimal Spacings EstimationTo predict a time of threshold crossing, an automation program must use a representative model of expected aircraft trajectories during the final approach segment.In CTAS, trajectories are generated by a process called the trajectory synthesizer (TS).The TS relies on a knowledge base consisting of aircraft performance models, aircraft physical characteristics such as weight, and pilot procedures for instrument and visual final approaches.The trajectories can be used to determine representative approach profiles, defined herein as the distance to the threshold as a function of time to threshold crossing.In this study, the profiles were investigated and modified to represent the observed trajectories more accurately.The modified approach profiles were then used as a basis for estimating optimal spacings between aircraft on the final approach segments.For VMC, the TS is designed to hold a constant speed to the FAF.At the FAF, the aircraft begins to decelerate to its final approach speed, which it typically captures about two nm later.Under weather-minimum condilions, pilots are trained to reduce high workload during the final portion of the approach by making all speed adjustments and configuration changes early.ThereR)re, for IMC, the TS initiates a deceleration to final approach speed two to three nm before the FAF, and acquires the final approach speed at the FAF.Without records of voice communication between the pilot and the radar controller, it is impossible to determine whether data for a given aircraft correspond to a visual or a weather-minimum instrument approach.As explained in section 3.2, during a visual approach, a pilot may choose not to follow the FAA separation minima if he feels it is sate to have less separation.Crosswind conditions and his altitude relative to the lead aircraft will impact the pilot's separation decisions.Pilots often try to hold maximum speed as long as possible, and, depending on experience with the aircraft, will decelerate and capture final approach speed as late as one mile before the threshokt (ref.5).In addition, the aircraft landing weight can play a significant role in the pilot's speed decisions. Required Separations ModelMinimum separation constraints are most critical for the final approach flight segment, when all aircraft share a common approach path and the danger fi'om wake turbulence is highest (fig.6).Under VMC, the common path is typically about six nm long at DFW; under IMC, final-approach-course intercept requirements result in a common approach path that is approximately nine nm long.A slow leading aircraft may be overtaken by a faster trailing aircraft, so the aircraft must be spaced so that they will not violate the minimum requirements at the threshold.A fast leading aircraft will pull away from a slower trailing aircraft, so minimum separation occurs before crossing the threshold.A model was developed to convert required minimum separations that apply over the entire common path to actual required minima at the threshold.The model was also used to estimate the corresponding required threshold interarrival times.This "required separations model" is dependent on the length of the common approach path and the final approach trajectories of the leading and trailing aircraft.A simple model based on constant speeds during the entire final approach segment is not acceptable, since aircraft typically slow to a landing speed near the FAF, as explained in section 6.1.The threshold separation was defined as the separation when the leading aircraft crosses the threshold, determined such that no separation constraint was violated along the entire common path.The common path separation was defined similarly for the point when the trailing aircraft crosses the start of the common path.Using an iterative loop, an automated function compared representative approach profiles of leading and trailing aircraft from the start of the common path to the threshold, and adjusted the spacings between the profiles so that the separation constraint was not violated along the path.Figure 7 illustrates the threshold and common-path distance and time separations as defined for the study.The figure shows an example for which the minimum required separation occurred between the common-path start and the threshold.For the VMC cases, data and TS-generated approach profiles agreed well for all except one of the five evaluated aircraft speed/weight classes.Figure 9 compares the TS large-jet VMC profile with actual distance/time separation data for all VMC cases having large jcts in trail.In the figure, a linear fit of the data between 0 and 3 nm was used to approximate the average final approach speed.For this case, the observed data show good agreement with the TS-generated approach profile.The high variance in the data suggests that providing automated advisories for spacings that are accurate to within a few seconds requires the modeling of additional parameters.Some of the variance results from the fact that, under VMC, aircraft are not required to follow the final approach course through the approach gate.¥• o • • e • ,q= 1, • 5O • o0: •o • • _ °oO 0 2 I I 1 , .3 4 5Distance to threshold, nm Less overall agreement was found for the IMC comparisons, although the small number of data samples made conclusions difficult.Contrary to expectations, variation in the data was found to be of the same order of magnitude as that of the VMC data.Figure 1 I is an example of a data and TS comparison for large jets in trail under IMC.A linear fit of the data between 0 and 5 nm to the threshold was used to approximate the final approach speed.As seen in the figure, the data predict a speed that is lower than the TS prediction by approximately 10 knots, a difference that is within the range of wind conditions encountered in the data.Among the speed/ weight classes, the large turboprops and B757 comparisons showed the poorest agreement.Table 3 summarizes the comparison of TS-generated approach profiles with DFW data; small props were not included in the table because of their small sample size. Model Modifications Improved estimates of IMC and VMC approach profileswere developed by using the data.Linear fits of the data were combined with assumptions about speeds beyond the FAF and knowledge of pilot procedures.The required separations model was then modified to use these improved profile estimates to obtain optimal time separations for the aircraft lead/trail pair combinations.The results are shown in tables 4 and 5.Note that the observed results show lower differences in speed between many of the speed/weight class types than predicted by the TS profiles.Therefore, problems of widening separation gaps or overtaking aircraft were found to be smaller than anticipated."" " aBased on observed data approach profiles and assuming a common approach path length of 6 nm.final approach segment.The level of modeling accuracy needed by an automation tool to provide the appropriate level of benefit should also be investigated. Separations AnalysisThreshold separations associated with arrival rushes were analyzed statistically to document and understand current threshold spacing under high-demand arrival conditions.All records associated with weather conditions that required aircraft to follow the final approach course through the approach gate were included.Small props were not analyzed because the low frequency of smallprop landings at DFW resulted in a small sample size.In another study, several controller spacing aids were evaluated using simulations of final approach traffic during rush periods (ref.7).The subjects were responsible for all the major tasks of approach controllers, including spacing, sequencing, and issuing vectors.The simulations were conducted for instrument conditions.Frequency aBased on observed data approach profiles and assuming a common approach path length of 9 nm.With these assumptions, the controller/pilot spacings accuracy was obtained by filtering out all data points that have spacings greater than the distribution maximum, and then obtaining the root mean square (rms) deviation of the samples with respect to the maximum point.Figure 14 shows this filtered separation distance half-distribution for the 2.5-nm required separations case.The standard deviation was found to be approximately 0.6 ran.Note that, since the maximum point was based on the smoothed density function, it does not necessarily agree with the maximum population class shown by the histogram. Statistical CharacteristicsThe full symmetric distribution that would occur for cases with no gaps was obtained by adding the left side hall'distribution to its mirror image.For this distribution, nine percent of the cases had smaller separations than the required instrument approach minimum.This metric is referred to as the minimum separation fraction, FM, in the discussion that follows.The time separations for the samples of figure 14 For the 2.5-nm required separations case of the figure, the standard deviation is about 20 seconds.Figure 16 shows the distance separation distributions for all lead/trail combinations with 3-nm required minimum separations, which usually resulted from separation restrictions caused by wet runways or fog.A comparison with figure 12 shows the distributions to be almost identical, indicating that controllers do not distinguish between 2.5-nm and 3-nm separations.Controllers appear to be aiming for separations slightly over 3 nm for both cases.Figure 17 presents the distributions for five-nm required minimum separations.The small number of samples makes conclusions difficult, but the spacing precision appears to be quite low.The distribution also appears to have a significant left tail, with some separations actually lower than three nm.No excess buffer is evident from the distribution.A controller suggested that the low separations seen in this case may have resulted from trailing turboprops, which are able to land well past the runway marker.By landing long on visual approaches, turboprops are able to stay above the leading-aircraft tlightpath, thereby avoiding its wake.However, examination of the data showed no such correlation between turboprops and the low separations.Table 7 compares results of the rush-period analysis, broken out by minimum required separation, and results of the simulations of reference 7 are also shown.The live data results tend to support the earlier simulation findings:for manually controlled traffic, standard deviations of interarrival time separation distributions are approximately 19 to 20 seconds for an arrival mix made up mostly of 2.5-rim required separations.The table also shows that both spacing precision and controller buffers tend to decrease with increasing required minimum separations.The results were found not to correlate with weight class or separation restrictions; perhaps some controllers were aiming lor one average separation distance that was adequate for most cases. ControllerTarget SeparationsFinal approach target separations at the threshold used by the controllers were estimated by using results presented Buffer Reduction PotentialThe controller target separations were also used to develop a simple method for estimating the potential for an automated spacing tool to reduce spacings, thereby increasing arrival capacity.As previously explained, the target separations were assumed to include the minimum required separation and a buffer.Because lower separalion variance is anticipated when using an automated tool, some reduction of the buffer can be achieved without changing the minimum separation fraction.The amount Given the assumption of normal distributions, the buffer reduction potential was a linear function 6A.Table 11 shows the resulting buffer reduction slopes for the four required separation cases.For the 2.5-nm cases, each second of reduction in standard deviation resalts in a buffer compression of about 1 sec.Ifa value ofc_A of l0 sec can be achieved through automation, a buffer reduction of approximately 7.5 sec can be obtained, which corresponds to a 7-to 9-percent time reduction between landing aircraft.This estimate of buffer reduction potential is strongly dependent on the minimum separation fraction and the size of the excess buffer observed in the data.For example, the excess buffer is slightly negative for the Interarrival timeFigure 18.Method for estimating the potential of active FAST to reduce the controller buffer.five-rim minimum required separations case, resulting in no potential for buffer reduction, as seen in table 11.The results may not be representative because of the small sample size for the five-nm case, but they indicate that the improvement possible for this case is small at best, using the buffer reduction performance metric as defined.Further study is needed to identify and develop performance metrics that capture the full value of more accurate spacing. Runway Utilization and CapacityAn analysis of runway utilization and capacity was performed for all runways that were active during each of the 30 recorded rush periods.Spacings of individual aircraft pairs and the relationship of these spacings to each other were examined to identify usage trends.For each aircraft pair, interarrival distance and time separations were compared with the required threshold minima, which were computed using the required separations model. Threshold Spacing PlotsA graphical representation of landing spacing as a function of time was developed to facilitate the analysis.An example of the graphic, referred to herein as a "threshold interarrival spacing plot," is shown inN-I L = (a/c/hr) t N -t1where N is the number of aircraft that landed over the rush period and t is the time.The average required minimum separation (SR....) that included a controller t_¢g separation buffer b was also determined.The buffer was used to keep all capacity estimates conservative.N 1 Sravg=-_--(Z(Sri +b) (nm)i=2An approximation was also made of the number of additional aircraft that could have landed per hour.The positive and negative excess separations were totaled to obtain a net excess for the rush period, which was adjusted to units of aircraft per hour.This result was divided by the average required minimum separation and limited to values greater than or equal to zero:lI = max 0, 1 SEi (a/c/hr) SR.v (iN-tj) =where excess separation (SE) was defined as the actual longitudinal separation minus the required separation for each in-trail aircraft. S E = S -S R (nm)A more commonly used alternative estimate of additional runway capacity was obtained by calculating the maximum capacity with no excess separations and subtracting the actual landing rate from it:12 = max[0, C-L] (dc/hr)The maximum capacity was found by inverting the time separation corresponding to the average required minimum separation: The estimate displayed on the threshold spacing plots, II, yields a higher result than the alternative, 12. Assuming that negative excess separations seen in the data are acceptable, the/l estimate may be a better representation of the maximum potential of each runway.1 C = --(hrInputs to the analysis included meteorological conditions and known separation restrictions.A no-reassign point of 0.5 nm was used for all cases.To obtain an approximation of the upper bound of delay reduction potential, the total of all positive delays incurred in the Center for all displayed aircraft is also provided in the figure.This value is also adjusted to show seconds of delay per hour of rush period. Plot Analysis ExamplesFigure 19 corresponds to a rush period referred to as the "Noon Balloon" by DFW personnel; it lasted approximately 1 hour, with an arrival type mix resulting in an average required minimum separation (with buffer) of about 3.6 nm.The landing rate during this period was about 32 aircraft per hour, and an additional 14 aircraft could have landed with no increase in negative excess separations.These values are typical of results seen for the north/south runways during IMC.The correlation between this threshold interarrival spacing plot and the interarrival time plot of figure 20 is good: similar trends are observed in both plots, an indication that the approach profiles extracted from the data are representative of flight times to the threshold for the various aircraft types.An exception is aircraft LSS 1262, which has a lower excess time separation than would be expected from the distance plot.This aircraft, a large turboprop, had a much higher speed than was predicted by the required separations model.Although further refinement of the required separations model may be needed, the results shown are greatly improved over initial results that were based on the original TS-generated approach profiles.Both recordings were made under IMC, with 3-nm separation restrictions and wet runways.Both sets of results are shown for Runway 18.The latter rush occurred during two-runway arrivals operations, whereas the former rush occurred during normal three-runway operations.For both rushes, there was significant delay buildup in the Center.The large differences in the two sets of results can possibly be attributed to differences between the TRACON radar controllers.If so, a possible measure of the effectiveness of automation is the reduction of these differences. Impact of meteorological conditions-Asexplained in section 6. I, meteorological conditions on approach impact the practices of pilots and procedures /bllowed by pilots and controllers.Some of these differences in the procedures can be seen by comparing radar tracks of rush periods for conditions better than 3000-ft ceiling/5-mi visibility with those under poorer conditions.Figure 31 is an example of an unusually long rush period, lasting 111 minutes.Two runways were in operation, with a ceiling of 300 ft AGL and 1.5-mi visibility.This rush probably corresponded to a delayed arrival rush from the east combined with a rush from the west at 18:30 CST.I _ i _ I i 1 _ I 1 1 . n . .. I 1 i 1 1 _ a . l, i 1 1 _ ¢ , I . 1 , I i . I , n11 n I 1 1 _ & ) I l 1 1 1 I I I I I I I I I I I I I I l l _c_ © _ = _- = _ _ c_ d _ © T N _ Ir "d c_ i q_ t_ E .c p; 3] -5 _o o E -_ "_ r1 E_ _'" E _-_ > -_ O_ ,,_ -- .. _ _>-_o _ ._ '_ -,_ _.8. _-_._ .o _E'_ ,1._ ..._ .o_.Controller practices and runway utilization seem fairly consistent over the duration of the rush period.The potential Center delay reduction of 761 sec/hr is very large, as would be expected under such poor conditions.It should be noted that, even under these conditions, there was potential for reducing Center delay and for landing additional aircraft.Another characteristic that was consistently observed in the data can be seen in figures 27 and 28.Aircraft that follow B757s tend to cross the threshold with negative excess separations more frequently than for other types.An example is flight AAL201, which has a negative excess separation of about 0.8 nm.This trend highlights the complexity and difficulty of the controller's task in achieving different required spacings lbr the various combinations of aircraft weight classes. Combined Data AnalysisAll the rush-period recordings were combined so that approximations of utilization and capacity could be made for each runway and the airport.Table 12 summarizes the number of rush periods used for each runway in the combined dataset.As can be seen, IMC rush periods were difficult to obtain, especially for the diagonal runways.These numbers should be kept in mind when interpreting the combined results.13; landing rates were found to be greater under VMC than under IMC.This trend probably resulted from the pilot discretion issucs discussed in section 3.2.Landing rates were also observed to be lower for the diagonal runways <_[ into terminal airspace is performed with respect to the four arrival meter gates, which are equally spaced 1400@Arrival rush from east; lasts approx. 50 minutes; FEL is beneficial to flow.approximatelyabout the 40-nm radius. They can be seenin the figure as the four points of convergence tracks. The recording corresponds to a severe rush period, of aircraft so many arriving aircraft were required to wait in a I Arrival rush from the west and BUJ; lasts approx. 50 minutes. 1420 1530East departure push; half go through the Lake sector; lasts approx. 25 minutes,I Iholding pattern before crossing the meter gate.Center delay buildup was identified by using the Center recordings as input to CTAS, which estimated undelayed times of arrival at the meter fix points (ETAff) up to I West departure push; lasts approx. 30 minutes. 1550 1700[_]East departure push; half go through the Lake sector; lasts approx. 50 minutes.I I30 minutes ahead of time. The ETAff values are com-1700Arrival rush from east; lastsputed by the CTAS TMA using information such as the aircraft type, its flight plan, and its position, ground Major arrival rush from the west and BUJ; lasts approx. 55 minutes. 1830<_approx. 50 minutes; FEL is beneficial to flow.Ispeed, and altitude. Weather is also normally used by the TMA to compute ETAff values, but weather information gets bulk of traffic; lasts 50 Westbound departure push; MOP 1830was not available for the recordings of this study. In minutes.2000most of traffic; several northcomputing these values, TMA assumes that undelayed[Z_East departure push; Lake gets departures; lasts approx. 1 hour.Idirect routing is used between the measuring point andthe meter fix. For each aircraft, the ETAff value was subtracted from the actual meter fix crossing time to obtain an estimate of delay incurred by each aircraft in West departure push over MOP; several north and south departures in push; lasts approx. 30 minutes. 2010<_Arrival rush from east; numerous freighters arrive in the I I rush; lasts approx. 50 minutes.the Center. 1 CTAS Aircraft Approach Profiles Thepilot procedures components of the TS final approach trajectories were based on formal training courses of the U.S. Air Force, United Airlines, and America WestAirlines (ref. 5). All approacheswere categorized asbeing one of two types: VMC, correspondingto probablevisual approaches;and IMC, for probable weather-minimum instrument approaches.Common approach path-0XTurn on ,_ finalFigure 6. Threshold spacing measurements for landing sequences.I06. Table 66contains standard and maximum deviations inthe flight times from the FAF to the threshold. The varia-bility of results for the heavy, small-turboprop,and B757VMC cases is low enough to be accounted for largely bywinds. The 15-knot wind variability in the data wouldaccount for a 12-to 25-second difference betweenminimum and maximum values. In contrast, the large-jetclass, which has aircraft types with a wide range ofweights and final approach speeds, has very large varia-bility. A comprehensiveerror-sourceanalysis is neededto determine the causes of flying time differences on the Figure 9. Distance and time separation comparisons for all VMC cases with large jets in trail.VMC TS approach profile ........._, ....... Landing speed linear fit ..>._E_ • : I,,_.:." -._ ._."• • ". ,,,, • ..,, .. -Figure 10.D/stance and time separation comparisons for all VMC cases with large turboprops in trail. Table 3 .3Summary comparison of TS approach profiles with DFW data••••oeee•',o'_,,_"A-'/ .o•" a_" ." _ • ••-. • .: ¢_-j¢_--_........:-...._;.....,_.,*,,."• ,;.: "; o.,.°e oe Q •e.,'"••_'"':Probable instrument approachesI.IJ,._II|I•.I•--I..I0123456789Aircraft speed/weightclassFinal approach ground speed, knotsTime to threshold at 5 nm, secand conditionsTS profileData linearData-TSTS profileData linearData-TSfitfitHeavy, VMC153136-17116127l 1Large jet, VMC13313301271270Large turboprop, VMC10012121! 50128-22Small turboprop, VMC108I 168145138-7B757, VMC146128-1812013111Heavy, IMC150134-16i 1713417Large jet, IMC138127-l I13014212Large turboprop, IMC10412319174146-28Small turboprop, IMC10512217163147-16B757, IMC146123-2312114625Distance to threshold, nmFigure 11.Distance and time separation comparisons for all IMC cases with large jets in trail. Table 4 .4VMCseparation timeestimates a(leading aircraft down, trailing aircraft across)Separations,secondsHeavyLarge jetLargeSmallB757turbopropturbopropHeavyCommon path103125126157129Threshold103125128157130LargeCommon path66666710866jetThreshold67667611572LargeCommon path76767511475turbopropThreshold67667511571SmallCommon path7878787878turbopropThreshold6766757871B757Common path106106107136108Threshold103103!09138108 Table 5 .5IMCseparation timeestimates a (leading aircraft down, trailing aircraft across)Separations,secondsHeavyLarge jetLargeSmallB757turbopropturbopropHeavyCommon path107137140171138Threshold107142149177146LargeCommon path74716211569jetThreshold67717612075LargeCommon path817774I 1776turbopropThreshold67717411773SmallCommon path8177757476turbopropThreshold6771747473B757Common path116116117146116Threshold107I 13109147116 Table 6 .6Deviations in times of flight from the FAF to the thresholdStandard deviation, secMax value -min value, secAircraft speed/weightVMC casesIMC casesVMC casesIMC casesclassHeavy9223645Large jet202014286Large turboprop18216163Small turboprop13Insufficient data27InsufficientdataB75712Insufficient data4326Combined data192215695distributionsof excess time and distance separations werethe minimum to the distributionmaximum;i.e., gapsfound to be symmetrical,and the authors presented theirassociated with missed slots are assumed not to signifi-results using measures of standard deviation. Thesecantly affect the left side of the distribution.The impactresults tend to support the busy-period component of theof the gaps was removed as much as possible by assum-reference 6 model.ing that the left side of the distributionis equal to the leftside of a symmetrical distributionthat represents trafficThe parametric model of refcrence 6 was used in thatflow with no gaps.study as a basis lot a maximum likelihood estimationof runway utilization and controller/pilotaccuracy.Since the maximum point of the distributionis to theHowever, the model is a simplified representationofright of the required minimum separation, controllers mayactual traffic flow; it uses only one required minimumhave been (intentionallyor unintentionally)adding extraseparation value and one controller buffer value forseparation buffers to account for spacing imprecision.Inall arrival traffic. Therefore, the model was deemedthe discussion that follows, the distributionmaximumunacceptabletot analyzing DFW traffic. Instead, anpoint was used as a measure of this aim point and as theassumption was made that the Poisson distributionmean of a symmetricaldistributionthat would representassociated with gaps does not significantlyimpact thetraffic flow with no gaps.observed distributionsin the range of separationsfromFigure 12.Distribution of distance separations for all cases wtih 2.5-nm required minimum separations.o o Figure 13.Distribution of time separations for a// cases with 2.5-nm required minimum separations. Table 7 .7Rush-period arrival spacing precisionMinimum required Samples Maxpoint,Symmetric Symmetric Minimumseparation, nmexcess nmseparation separation separationstddev, nmstddev, sec fraction, FM2.54700.70.6419.60.093.03230.10.5319.60.384.01120.40.8625.50.335.050-0.20.8338.20.56Simulation, ref. 7 a5140.31 (mean)0.6519.49N/AaLumpedsubject data interarrival error for 170-knot manual pattern procedure. Arrival traffic weightclass mix was made up of 87.5-percentlarge and 12.5-percent heavy, resulting in minimum requiredseparations as follows: 87.5 percent--2.5nm; 1.5 percent--4nm; 11 percent--5nm. Table 8 .8Estimated controller target threshold crossing separations a in nm(leading aircraft down, trailing aircraft across)AircraftHeavyLarge jetLargeSmallB757speed/weightclassturbopropturbopropHeavy4.25.25.66.25.2Large jet3.23.23.64.23.1Large turboprop3.13.13.24.23.1Small turboprop3.13.13.13.13.1B7574.24.24.25.24.2aFor conditions having a ceiling less than 3000 ft AGL or visibility less than 5 mi.Table 9. VMC controller target separation time estimates, based on live data approach trajectories andassuming a common approach path length of 6 nm(leading aircraft down, trailing aircraft across)Separations,HeavyLarge jetLargeSmallB757secondsturbopropturbopropHeavyCommon path108130139161133Threshold108130140161134LargeCommon path84859511483jetThreshold858510112087LargeCommon path89899411891turbopropThreshold82829412087SmallCommon path9191949593turbopropThreshold8282829587B757Common path110110I I I140112Threshold108108113141! 12equal to cyA and a mean equal to zero.The mean, P-A, was then determined by subtracting the obtained fractile value from the required minimum time separation.The buffer reduction potential was equal to I.tM-P-A. Table 10 .10IMC controller target separation time estimates, based on live data approach trajectories andassuming a common approach path length of 9 nm(leading aircraft down, trailing aircraft across)Separations,HeavyLarge jetLargeSmallB757secondsturbopropturbopropHeavyCommon pathI 12141157177143Threshold112146165! 83150LargeCommon path939110212188jetThreshold869110712592LargeCommon path95939412392turbopropThreshold83889412391SmallCommon path9593919192turbopropThreshold8388919191B757Common path121122123152122Threshold112119123153! 22Table 11. Buffer reduction potential, assuming the observedminimum separation fraction is maintained by an automatedsystemMinimum requiredReduction slope, secR, sec forseparation, nmR per sec _ACYA = 10 sec2.5-1.007.53.0-0.251.54.0-0.254.55.02.000 Table 12 .12Runway rush periodsLandingNumber of rush periodsrunwayIMCVMC13R261766186631R163531536315Total21548.5.1 Runway landing rates-The combined datasetlanding rates for each runway are shown in table Table 13 .13Runway landing rates during rush periodsConditionLandingLanding runwayrates, a/c/hr13R171831R3536IMCMean30.435.431.7! 6.333.134.1Minimum27.532.223.816.326.832.8Maximum33.340.737.116.336.535.4VMCMean30.940.735.931.537.737.9Minimum20.831.325.023.327.329.2Maximum35.046.548.038.647.042.2Table 14. Potential runway capacity increasesConditionCapacityLanding runwayincrease, l 117183536IMCMean, a/c/hr7.79.218.716.3Min, a/c/hr321412Max, a/c/hr13142419Percent21.728.956.447.8increaseVMCMean, a/c/hr12.315.213.510.3Min, a/c/hr4372Max, a/c/hr23223227Percent30.342.335.727.1increaseunder VMC. Arrival loads were more similar among theTable 15. Potential maximum runway capacities (allactive runways under IMC, although there were insuf-conditions)ficient data to draw conclusionsabout Runway 31R. Theresults also showed that there were wide ranges of runwayLanding runwayMaximumZero-excessutilization during rush periods, with landing rates as lowcapacity, a/c/hrmaximumas 16 aircraft/hourand as high as 48 aircraft/hour.(/1 method)capacity, a/c/hr(12 method)8.5.2 Potentialrunway capacity increases-Table 14summarizes capacity increases possible, based on the l113R50.339.9measure described in section 8.1. Because the diagonal1748.040.3runways were often not operated at capacity, they were1846.039.1not included in the table. Average increases ranged between 8 and 19 aircraft/hour for IMC, and between 1031R60.541.2and 15 aircraft/hourfor VMC. Some potential increases3551.341.9were possible for all recorded rushes. The VMC potential3648.540.8increase values were unexpected;they resulted from theassumption that negative excess separationsseen in thedata are acceptable. VMC caused many excess separations, The low spacing consistency under thereby resulting in large potential increases. Table 15 summarizes thepotential maximum capacities for all meteorological conditions. Given the arrival mix at DFW, the results indicate that the potential maximum landing rates are In the figure, the 19-min TMA predictions are seen to have a very low bias of 15 sec and a standard deviation of 103 sec.Since no weather information was used, the low bias may indicate that the average wind velocity for the complete sample was close to zero, although it may instead indicate that the bias caused by no weather predictions is canceled by other biases in the TMA models.O CTAS/TMA,no weather information,Ifull datasetIII1:3 Operationalmetering,full dataset•CTAS/TMA,including weather, high-wind conditions• Operationalmetering,high-wind conditions200-A CTAS/DAfield test, September1994 (each point/The distribution is one aircraft) consist of the convolution of a normal distribution com-was assumed to ponent that represents ETAff prediction accuracy and some other distribution caused by delays. To estimate J U prediction accuracy, it was necessary to remove the effects-of-the-delays component. An approximation ETAffI9 prediction accuracy was made by constructing e i of I....... I II/ The undelayed meter fix time prediction (U_MFTI9) distribution from the ASP is compared with the ETAffI9 distribution in figure 35. The ASP distribution was found CTAS ETAff probability density ASP U_MFT probability density I I Assumed .I top of I descent / standard deviation of 125 sec. The figure indicates that f to predict crossing time with a bias of 76 sec and a ..or"a symmetrical I/)distributionbased on the left side of the /the ASP may be less accurate in predicting crossingobserved distribution; maximum corresponded ty-the location of the distribution to an approximate predictor bias. /times• Both predictors were impacted similarly by the presence of delays.100_ _':"od toIIII-5000500100015002000oI.ICenter delay, sec A ASamples: 4455 Median: 69!°o Dataset to .reMean: 187 of CTAS and ASP 19-min ETA predictions. Max point: 15 Prediction accuracy Std dev: 103 RMS error: 104 -A A 10 20 Minutes to meter fix crossing Symmetric distribution Symmetric distribution Smoothed probability density 18. Meter fix crossing Figure 35. Comparison .... Minutes to \ Table prediction analysis results Figure 36. Estimated accuracy of CTAS and ASP meter fix crossing predictions.meter fixmean, secstd dev, seco o o All data/19TMA 15ASP 76TMA 103ASP 125301119124157Jets only019_0 27841_01091_01212000303131129151Center delay, seeTurbopropsonly19-134890126Figure 34. Distribution of CTAS 19-min ETA predictions. 30 -3 60 117148! / ._,.-,--T'T , ,,At , , , I , , , , , ,.. ,...,...,...,...,...,...,.._,...,...,...,...,...,...,...,_..,...,...,...,...,TS profile........Empirically determined profileo°o"0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20Distance to threshold, nm Table B-6.Empirically derived separations torVMC4.0 167.9 187.4 4.1 4.1 1.1 2.5 95.2 83.2 145.1 4.8 4.0 0 2.5 83.2 3.5 2.5 2.5 3.2 0 3.5 2.5 2.5 i 05.0 83.0 115.0 115.0 4.2 2.9 2.5 2.5 0 Small prop B757 6.2 5.2 244.3 143.1 298.7 150.1 6.2 6.1 6.7 5.2 1.9 4.2 3.1 157.9 88.4 238.4 92.3 4.2 4.2 5.3 3.2 3.9 1.9 4.2 3.1 159.1 91.7 234.6 90.7 4.3 4.3 5.2 3.1 3.9 3.1 110.8 202.7 0 3.1 91.7 90.7 3.2 4.3 4.4 3.1 5.0 0 Small turbo-prop Small prop 6.2 6.2 108.0 129.6 139.5 160.9 262.2 133.5 B757 5.0 118.8 119.7 5.1 5.0 2.5 66.8 62.3 3.1 2.5 2.5 85.5 62.0 3.8 3.7 3.7 3.8 2.5 2.5 2.5 2.5 0 0 0 0 2.5 2.5 2.5 2.5 81.6 82.6 83.3 82.0 58.6 66.3 88.0 62.0 3.7 3.6 3.5 3.7 2.5 2.5 2.5 2.5 0 0 0 0 2.5 99.2 2.5 95.9 58.6 65.9 2.5 106.1 88.0 2.5 97.7 62.0 4.3 4.2 4.2 4.3 2.5 2.5 2.5 2.5 0 0 0 Large jet Large turbo-prop Small turbo-prop 5.2 5.6 6.2 141.2 157.3 177.1 145.7 164.6 182.7 6.2 6.2 6.8 5.2 5.6 6.2 0 0.3 0.9 3.2 3.6 4.2 91.2 102.3 121.4 91.2 107.0 125.4 4.4 4.3 4.9 3.2 3.6 4.3 1.8 1.4 2.3 3.1 3.2 4.2 92.8 94.3 123.2 88.3 94.3 123.2 4.5 4.0 5.0 3.1 3.2 4.2 1.1 3.1 3.1 92.8 88.3 0.8 3.1 91.4 91.4 91.4 91.4 4.5 3.9 3.9 3.1 3.1 3.1 1.9 1.9 Heavy Large jet Large turbo-prop B757 4.2 5.2 5.6 5.2 108.0 4.6 4.2 3.2 84.4 84.8 3.7 3.2 3.1 89.5 82.4 3.9 3.1 3.1 91.4 82.4 3.9 3.1 129.8 5.4 5.2 3.2 84.6 84.6 3.7 3.2 3.1 89.4 81.9 3.9 3.1 3.1 91.1 81.9 3.9 3.1 !.8 Table B-5. Empirically derived separations torIMC Common pathlength: 9.0nm Extraseparation buffers included Heavy Large jet Large turbo-prop Small turbo-prop Leading aircraft down, trailing aircraft across Req'd minsep, nm Cornpathsep, sec Thresh sep, sec Cornpath sep, nm Thresh sep, nm Lead a/crainseppos, nm Req'dminsep, nm Compath sep, sec Thresh sep, sec Compathsep, nm Thresh sep, nm Lead a/cminseppos, nm Req'd minsep, nm Compathsep, sec Thresh sep, sec Compathsep, nm Thresh sep, nm Lead a/cminseppos, nm Req'd minsep, nm Compathsep, sec Thresh sep, sec Corn path sep, nm Thresh sep, nm Lead a/c rain sep pos, nm Heavy 4.2 112.4 112.4 5.1 4.2 0.8 3.2 93.4 85.9 4.3 3.2 3.1 95.1 83.3 4.4 3.1 3.1 95.1 83.3 4.4 3.1 Common pathlength: 6.0nm Extra separation buffers included Leading aircraft down, trailingaircraft across Heavy Req'd rainsep, nm Cornpathsep, sec Thresh sep, sec Corn pathsep, nm Thresh sep, nm Lead a/crainseppos, nm Req'd rainsep, nm Compathsep, sec Thresh sep, sec Corn pathsep, nm Thresh sep, nm Lead a/crainseppos, nm Req'd minsep, nm Corn pathsep, sec Thresh sep, sec Corn pathsep, nm Thresh sep, nm Lead a/cminseppos, nm Req'd minsep, nm Corn pathsep, sec Thresh sep, sec Compathsep, nm Thresh sep, nm Lead a/cminseppos, nm Large jet Large turbo-prop Small prop Req'd min sep, nm Com path sep, sec Thresh sep, sec Corn path sep, nm Thresh sep, nm Lead a/c rain sep pos, nm 3.1 3.1 3.1 3.1 3.1 3.1 139.7 139.3 145.3 145.6 142.9 141.6 83.0 88.0 91.4 91.4 142.9 90.8 6.1 6.1 5.8 5.8 3.9 6.0 3.1 3.1 3.1 3.1 3.1 3.1 3.1 3.1 Req'dminsep, nm 108.9 108.2 Cornpathsep, sec 82.5 81.9 Thresh sep, sec 4.6 4.6 Corn pathsep, nm 3.1 3.1 Thresh sep, nm Lead a/cminsep pos, nm 0B757Req'd min sep, nm Req'd minsep, nm4.24.24.25.25.24.2Com path sep, sec Compathsep, sec121.0121.8122.6152.2201.7122.4Thresh sep, sec Thresh sep, sec112.4119.0123.4152.8264.6122.4Corn path sep, nm Compathsep, nm5.45.55.06.05.35.3Thresh sep, nm Thresh sep, nm4.24.24.25.25.94.2Lead a/c min sep pos, nm Lead a/crainseppos, nm0.81.32.90.8AllTrail a/c thresh time, sec Traila/cthresh time, sec 218.0221.0240.0241.0374.0229.0 Appendix A -Runway Selection LogicThe runway selection logic identifies the landing runway for each aircraft from a provided set of runway candidates.For each runway, the two aircraft radar hits that are closest to a user-assigned point on the final-approach course are determined.The logic then uses a process of elimination to identify the most likely landing runway.If all candidate runways are eliminated, the aircraft is identified as not having landed.The logic also identifies a timc of threshold crossing for each landing aircraft.The radar track data are filtered using a tburth-order Butterworth-characteristic filter before application of the logic.A. 1 Aircraft approach flightpath angle elimination parameter.Used to compute altitude AGL maximum limit.If the maximum limit is exceeded at the noreassign point, runway is eliminated as a landing candidate.Aircraft altitude above the maximum approach path elimination parameter.Used to compute altitude AGL maximum limit.If the maximum limit is exceeded at the no-reassign point, runway is eliminated as a landing candidate.Aircraft heading difference elimination parameter.If the difference between the aircraft heading and the runway heading exceeds the value of this parameter at the no-reassign point, runway is eliminated as a landing candidate.Aircraft distance elimination parameter.If the distance between the noreassign point and the closest radar hit exceeds the value of this parameter, runway is eliminated as a landing candidate. A.2 Selection Logic1.For each runway, the radar hit closest to the runway no-reassign point is identified.2. The previous radar hit is identified.If there is no previous radar hit or if the data record is too short for aircraft speed data to be reliable, the aircraft is identified as not having landed.3. The aircraft true heading and bearing to the runway zero point, the aircraft climb rate, and its rate of distance closure to the runway zero point are determined based on the x and y positions of the two radar hits.4. The runway candidate tests are performed for each runway.All the following tests must be passed: a.The aircraft distance to the no-reassign point must be less than the distance limit.b.The distance closure to the runway must be positive.. . c.The difference between the aircraft heading and the runway heading must be less than the heading difference limit.d.The aircraft climb rate must be less than the climb rate limit.e.The aircraft must be below the altitude maximum limit.Of the remaining runway candidates, the most likely runway is selected.The closest runway is identified, and all runways with a distance greater than the bounds defined by the radar error are eliminated.If two or more runways remain as viable candidates, the candidate having the lowest difference between runway true bearing and aircraft true heading is selected.The threshold crossing time is estimated by computing the distance between the aircraft and the identified runway.The aircraft ground speed, computed by CTAS from radar hits using a Kalman filter, is used to extrapolate to the threshold.Advanced air traffic management systems such as the Center/TRACON Automation System (CTAS) should yield a wide range of benefits, including reduced aircraft delays and controller workload.To determine the trafficflow benefits achievable from future terminal airspace automation, live radar information was used to perform an analysis of current aircraft landing rates and separations at the Dallas/Fort Worth International Airport.Separation statistics that result when controllers balance complex control procedural constraints in order to maintain high landing rates are presented.In addition, the analysis estimates the potential for airport capacity improvements by determining the unused landing opportunities that occur during rush traffic periods.Results suggest a large potential for improving the accuracy and consistency of spacing between arrivals on final approach, and they support earlier simulation findings that improved air traffic management would increase capacity and reduce delays. 14.SUBJECT TERMS Design of Center-TRACON Automation System. AGARD-CP-538, paper no. 11, presented at the AGARD Guidance and Control Symposium on Machine Intelligence in Air Traffic Management HErzberger TJDavis SMGreen May 11-14, 1993 Berlin, Germany Erzberger, H.; Davis, T. J.; and Green, S. M.: Design of Center-TRACON Automation System. AGARD-CP-538, paper no. 11, presented at the AGARD Guidance and Control Symposium on Machine Intelligence in Air Traffic Manag- ement, Berlin, Germany, May 11-14, 1993. Design and evaluation of an air traffic control Final Approach Spacing Tool ThomasJDavis HeinzErzberger StevenMGreen WilliamNedell 10.2514/3.20721 Journal of Guidance, Control, and Dynamics Journal of Guidance, Control, and Dynamics 0731-5090 1533-3884 14 4 July-Aug. 1991 American Institute of Aeronautics and Astronautics (AIAA) Davis, T. J.; Erzberger, H.; Green, S. M.; and Nedell, W.: Design and Evaluation of an Air Traffic Control Final Approach Spacing Tool. Journal of Guidance, Control, and Dynamics, vol. 14, no. 4, July-Aug. 1991, pp. 848-854. Validation of the Federal Aviation Administration Air Traffic Control Specialist Pre-Training Screen DanaBroach JanBrecht-Clark 10.2514/atcq.1.2.115 Air Traffic Control Quarterly Air Traffic Control Quarterly 1064-3818 2472-5757 1 2 Sept. 1993 American Institute of Aeronautics and Astronautics (AIAA) FAA Order 7110.65J, Air Traffic Control. (Air Traffic Control Handbook), Federal Aviation Administration, Sept. 1993. FNeuman HErzberger MSSchueller CTAS Data Analysis Program. NASA TM-108842 Oct. 1994 Neuman, F.; Erzberger, H.; and Schueller, M. S.: CTAS Data Analysis Program. NASA TM-108842, Oct. 1994. Using Maximum Likelihood Estimation to Determine Statistical Model Parameters for Landing Time Separations HFVandevenne MALippert 41L-0400 Apr. 1992 MIT Lincoln Laboratory Technical Memorandum Vandevenne, H. F.; and Lippert, M. A.: Using Maximum Likelihood Estimation to Determine Statistical Model Parameters for Landing Time Separations. Technical Memorandum 41L-0400, MIT Lincoln Laboratory, Apr. 1992. A Comparison of Final Approach Spacing Aids for Terminal ATC Automation LeonardCredeur WilliamRCapron DanielJCrawford WilliamGRodgers GaryWLohr DershuenATang 10.2514/atcq.1.2.135 NASA TP-3399 Air Traffic Control Quarterly Air Traffic Control Quarterly 1064-3818 2472-5757 1 2 Dec. 1993 American Institute of Aeronautics and Astronautics (AIAA) Credeur, L.; Capron, W. R.; Lohr, G. W.; Crawford, D. J.; Tang, D. A.; and Rodgers, W. G., Jr.: Final-Approach Spacing Aids (FASA) Evaluation for Terminal-Area, Time- Based Air Traffic Control. NASA TP-3399, Dec. 1993 GHunter Estimating CTAS Benefits Nationwide. Presented to the CTAS Benefits Analysis Technical Information Meeting July 13, 1995 Hunter, G.: Estimating CTAS Benefits Nationwide. Presented to the CTAS Benefits Analysis Technical Information Meeting, NASA Ames Research Center, July 13, 1995. Descent Advisor preliminary field test StevenGreen RobertVivona BeverlySanford 10.2514/6.1995-3368 Guidance, Navigation, and Control Conference Baltimore, Md American Institute of Aeronautics and Astronautics Aug. 1995 3368 Green, S. M.; Vivona, R. A.; and Sanford, B.: Descent Advisor Preliminary Field Test. AIAA Paper 3368, Proceedings of the 1995 AIAA Conference on Guidance, Navigation, and Control, Baltimore, Md., Aug. 1995.