American Institute of Aeronautics and Astronautics

1

The Effects of Future Vehicles on Controller and Pilot Workload

Kenneth Wright,1 Matt Blake, 2 Jim Smith,3 Ricky Mediavilla4

Sensis Corporation, Reston, VA, 20190

This paper describes the results of an ongoing study into the effects of new vehicle types on the National Airspace System (NAS). Notional air traffic schedules were developed that include each of five new vehicle types—ranging from small unmanned cargo aircraft to large supersonic transports—for the years 2025, 2040, and 2086. The schedules were flown using the Airspace Concept Evaluation System (ACES), a computer model that simulates a day in the NAS. The ACES output was analyzed to estimate the effects of the new vehicle types on controller and pilot workload. In particular, we wanted to find out if the new vehicle types produced substantially different controller and pilot workload than the baseline scenarios with no new vehicle types. For our baseline schedules we found that the number of aircraft conflicts increased linearly with the number of scheduled flights whereas delay increased exponentially. Overall, the effect of the new vehicles on workload was much smaller than the effect on delay. Additionally, we identified a number of limitations of the ACES model for this type of analysis.

I. Introduction HE Sensis Corporation is currently directing a large study—referred to here as the New Vehicle NRA (NASA Research Announcement)—of the potential impacts of future new vehicle types on the National Airspace

System (NAS) 1,2. Vehicle types being studied include cruise-efficient short-takeoff and landing (CESTOL) aircraft, very light jets (VLJs), Civil Tilt-Rotor aircraft (CTR), large supersonic transports (SSTs), and uninhabited aerial systems (UASs). The study looks at a range of impacts, including passenger delay, new procedures, and environmental impacts. Future scenarios were developed that reflect expert opinion on the expected number and most plausible uses of each vehicle, as well as their expected cost and aerodynamic performance.

The purpose of this paper is to analyze the modeling and simulation output from the New Vehicle NRA and make inferences about the effects of new vehicle types on controller and pilot workload. In particular, we seek to find out if the new vehicle scenarios produce substantially different controller and pilot workload than the baseline scenarios that have no new vehicle types.

Future scenarios in the New Vehicle NRA were modeled using the Airspace Concept Evaluation System (ACES). Developed by the NASA Ames Research Center, ACES is an agent-based fast-time simulation of the NAS. The first version of ACES, Build 1, was completed in March 20033, and the version used in this analysis, Build 5, was released in the summer of 2008. A physics-based model, ACES computes the trajectory of each flight by integrating aerodynamic energy-balance equations for each airframe. These physics-based computations along with elaborate traffic flow management (TFM) modeling enable ACES to differ substantially in capability and complexity from statistically-based models such as NASPAC4 or the MITRE Corporation’s systemwideModeler5.

The simplest measure of controller workload is the number of aircraft in a sector at a given time. In fact, current tools for managing controller workload measure workload relative to a sector’s Monitor Alert Parameter (MAP), an operationally defined upper limit on the number of flights a sector should contain under normal conditions. Still, numerous studies6,7,8 have pointed out that flight counts alone do not fully reflect controller workload. For example, managing traffic in a sector with intersecting flight paths is more difficult than managing traffic in a sector where aircraft flow along parallel lines. Several studies have published metrics that correlate better with controllers’

1 Senior Research Engineer, 11111 Sunset Hills Road, Ste 130, AIAA member. 2 Project Lead, 1700 Dell Avenue, Campbell, CA 95008, AIAA member. 3 Project Manager, 11111 Sunset Hills Road, Ste 130, AIAA member. 4 Project Manager, 11111 Sunset Hills Road, Ste 130, AIAA member.

T

9th AIAA Aviation Technology, Integration, and Operations Conference (ATIO) <br>and <br>Air21 - 23 September 2009, Hilton Head, South Carolina

AIAA 2009-7045

Copyright © 2009 by Sensis Corporation. Published by the American Institute of Aeronautics and Astronautics, Inc., with permission.

American Institute of Aeronautics and Astronautics

2

reported perceptions of workload than simple aircraft counts9,10,11,12. These metrics include the density of the aircraft in a sector, the number of altitude changes, and the number of expected sector conflicts.

To compare controller workload for the different new vehicle scenarios, we computed a number of standard workload metrics, including flight counts, both absolute and relative to sector MAP values; number of ascending and descending flights; and number of conflicts at 3 and 5 nautical miles (nm). The results of this analysis are discussed later in this paper along with the limitations of the ACES output.

II. Baseline schedules Baseline schedules for the years 2025, 2040 and 2086 were provided by the Joint Planning and Development

Office (JPDO), which is responsible for coordinating research into the Next Generation Air Transportation System (NextGen). Government entities involved in JPDO include the Departments of Transportation, Defense, Homeland Security, Commerce, FAA, NASA, and the White House Office of Science and Technology Policy.

The baseline schedules provided by JPDO were developed starting from actual flight data from a day in the summer of 2006 and applying the growth rates in the FAA’s Terminal Area Forecast (TAF).13 The resulting 2025 demand set has 48% more flights than in 2006; the 2040 has 92% more than 2006; and the 2086 has roughly three times the number of flights as in 2006. These schedules are based on unconstrained consumer demand and do not account for the finite capacities of airports or airspace. When the schedules are flown using ACES (or any other simulator) the delays that result are much longer than would be tolerable in practice. To obtain schedules with reasonable delays, flights are removed from the unconstrained schedules using a process known as trimming. A list of specific flights to be deleted from the JPDO-provided schedules was generated by researchers at the Logistics Management Institute (LMI) using their LMInet14 simulator, a queuing network developed in the late 1990s. The LMI researchers identified flights having the highest congestion score; that is, the flights that flew through the most

crowded sectors and arrived or departed from the most crowded airports. Deleting all of the flights in the JPDO-provided list from the baseline schedule produces what we refer to here as the 100% trimmed schedule, which provides moderate system-wide delay; deleting half the flights in the list gives the 50% trimmed schedule. As shown in Fig. 1 above, the number of flights trimmed from the 2086 schedule is much greater than the number trimmed from the 2025 schedule. The airport capacities used in all of the New Vehicle NRA analyses were the JPDO-provided capacities for the year 2025; the sector capacities used were also provided by JPDO and are 1.7

Figure 1. Number of scheduled flights used in ACES simulations.

American Institute of Aeronautics and Astronautics

3

times higher than today’s sector MAP values.15 Trimming levels were varied during the New Vehicle NRA to illustrate the effects of different levels of demand.

Fig. 2 shows the average NAS-wide delay computed by ACES for the 2025 and 2040 schedules at five trim levels: 0%, 25%, 50%, 75%, and 100%. Note that delay is plotted on a logarithmic scale. For both years’ schedules, delay increases exponentially with the number of flights. The 2086 scenario (not shown) is also referred to as the 3X scenario because it has three times as many flights as today. For that schedule, average delay ranged from 9 minutes at 91 thousand flights (100% trimmed) to 411 minutes at 164 thousand flights (0% trimmed).

Figure 2. Average minutes of delay by year versus number of flights plotted on a logarithmic scale.

American Institute of Aeronautics and Astronautics

4

III. New Vehicle Schedules The five new vehicles in the study include a 100-seat cruise-efficient short-takeoff and landing (CESTOL)

vehicle, a 90-seat tiltrotor aircraft, a 90-seat supersonic transport (SST), a Cessna Caravan-sized uninhabited aerial system (UAS), and a 4-passenger very light jet (VLJ). Usage scenarios and schedules were created for each new vehicle based on expert opinion. The schedules contained just one of the new vehicle types along with conventional aircraft. In addition, a combined schedule was created that contained all five new vehicles. The number of flights by each type of new vehicle used in the 100% trimmed dataset is shown in Fig. 3. The CESTOL and tiltrotor scenarios reduce delay by replacing conventional aircraft flying into congested airports with new vehicles flying to nearby airports and vertiports respectively as well as less congested runways and helipads at some congested airports. The supersonic transport replaces conventional aircraft flying conventional long-distance routes. In the UAS and VLJ scenarios, new routes were added to those in the baseline scenarios.

A. Cruise-Efficient Short Take-Off and Landing (CESTOL) Aircraft The main advantage of CESTOL aircraft is that they can take off and land on runways as short as 3,000 feet.

When taking off from a short runway the range of the aircraft is limited to about 600 nautical miles (nm). The CESTOL concept of operations replaces excess flights at congested airports with flights operating out of nearby satellite airports. Fig. 4 shows the locations of 469 public-use airports with runways that could support CESTOL operations and that are within 70 nm of a major airport.

Figure 3. Number of scheduled flights by vehicle.

American Institute of Aeronautics and Astronautics

5

The CESTOL schedules were created by modifying the baseline schedules for 2025, 2040, and 2086.

Specifically, whenever the demand-capacity ratio within a 15-minute time-bin at a major airport exceeded 90% of the airport’s visual flight rules (VFR) capacity, the excess flights were shifted to a satellite airport that was within 70 nm of the original airport. The equipment type of the shifted flights, all of which had stage lengths less than 600 nm, was converted to CESTOL. Applying the CESTOL reassignment algorithm to the untrimmed 2025 baseline schedule caused 2,490 flights to be shifted, resulting in a 5% reduction in traffic at the busiest 35 airports. Using estimated CESTOL production rates, it was assumed that the 2025 schedule would have approximately 10,000 daily CESTOL flights. To make up the difference, a random number generator was used to substitute CESTOLs for 7,500 conventional aircraft flying routes with stage lengths less than 600 nm. The CESTOL schedules were created in the same manner for 2040 and 2086. It was assumed that there would be 50,000 daily CESTOL flights in 2040 and 80,000 daily CESTOL flights by 2086.

B. Tiltrotor As with the CESTOL, the tiltrotor would relieve congestion at major airports by shifting traffic away from

runways with high demand at congested airports. A short-haul aircraft with 90 seats and a 500 nm range, the tiltrotor can take off and land from a 600 ft. landing pad, freeing up an airport’s main runways for longer-haul transports. Because of high expected production and operations costs, it was assumed that tiltrotor operations would be confined to profitable high-frequency shuttle routes. It was also assumed that the 2025 scenario would have 300 tiltrotors, the 2040 scenario would have 800, and the 2086 scenario would have 2,300.

C. Uninhabited Aerial Systems The scenario created for the UAS involved forwarding freight using Cessna Caravan-sized aircraft piloted from

the ground. Today private air carriers make limited use of piloted Cessna Caravans to move mail from hub to spoke airports, although trucks are most often used for this task. The UAS schedules developed for the New Vehicle NRA assumed 4,000 daily flights between known cargo hubs and spoke airports located 30 to 150 nm away. See Fig 5. The impact of the UAS on delay is small because only small numbers of UAS flights were added at passenger hubs.

Figure 4. CESTOL-capable satellite airports within 70 nm of major airport.

American Institute of Aeronautics and Astronautics

6

D. Very Light Jets VLJs are in commercial use today. Expanded use of VLJs assumes decreasing operating costs relative to other

forms of transportation. The VLJ schedules used in the New Vehicle NRA were created by researchers at NASA’s Langley Research Center using Virginia Tech’s Transportation Systems Analysis Model (TSAM).16 Using projected demand between city-pairs, the model computes the number of flights that could be flown in an economically competitive manner by a 4-passenger VLJ given certain assumptions regarding airline operating costs, the price of jet fuel, and the cost of auto travel. Because few VLJs fly into major hubs, their impact on delay is small.

E. Large Supersonic Transport Two scenarios were modeled for the SST, a 90-seat aircraft with a maximum range of 4,000 nm. In the first

scenario, the SST replaced conventional aircraft on long commercial routes having stage lengths under 4,000 nm. In the second scenario, the SST randomly replaced conventional aircraft on routes with stage lengths between 2,000 and 4,000 nm. No SSTs were assumed to have been built by 2025; the 2040 and 2086 schedules assumed 400 and 800 SST flights respectively.

Figure 5: Cargo routes for the UAS.

American Institute of Aeronautics and Astronautics

7

IV. Simulation Results Results from the ACES simulations are shown in Fig. 6 below. For each new vehicle type, average delay is

plotted against the number of scheduled operations for five trim levels: 0%, 25%, 50%, 75%, and 100%. The all-vehicle scenario is plotted for the trim levels 0%, 50%, and 100%. In all scenarios, delay increases exponentially with the number of flights. Although the untrimmed baseline schedule has just 14% more flights than the 100% trimmed baseline—79,000 flights compared to 69,000 flights—it has ten times as much delay. Of all the new vehicle scenarios, the CESTOL scenario shows the largest decrease in delay compared to the baseline. The untrimmed CESTOL scenario reduced delay by 46% compared to the untrimmed baseline. This drop in delay is not surprising given that the CESTOL scenario shifts flights from the busiest hubs to uncongested satellite airports.

The tiltrotor scenario also decreased delay compared to the baseline, but the decrease was smaller than in the

CESTOL scenario. One reason for this difference is that fewer tiltrotors are expected to be available for service by 2025 than CESTOLs. Also, whereas CESTOLs were substituted for aircraft flying during highly congested time periods, tiltrotors were substituted for aircraft flying the most popular shuttle routes, regardless of whether the flights on these routes were delayed. The UAS and VLJ scenarios both had 4,000 more flights than the baseline scenario; however, neither caused a substantial increase in delay because few of the added flights were at major hubs or flying through congested airspace. Due to the additional UAS and VLJ flights, the all-vehicle scenario had about 8,000 more flights than the baseline. Delays were lower than in the baseline due to CESTOL and tiltrotor replacements.

Figure 6. Average delay by scenario and number of flights.

American Institute of Aeronautics and Astronautics

8

V. Workload

A. Aircraft Conflicts Using the aircraft tracks produced by the ACES simulation we experimented with computing a number of

metrics related to sector workload. Because ACES does not model the aircraft’s flight path within 40-nm of their origin or destination airports, the analysis was restricted to en route airspace. One of the most interesting metrics that was computed from the ACES output was the number of aircraft conflicts. A conflict occurs when two aircraft flying at the same altitude are separated by less than 3 nm in terminal airspace or 5 nm in en route airspace. To avoid potential conflicts, an air traffic controller will take action to adjust the speed or heading of either or both aircraft well in advance of the conflict. Therefore the number of conflicts that a scenario produces should be a good indicator of workload for controllers and pilots. When we ran our ACES simulation we specified that the software not maneuver the aircraft to avoid conflicts. From the simulation output we tabulated the number of conflicts at both 3 and 5 nm.

Figure 7 shows a common type of conflict that controllers must frequently resolve. An aircraft takes off from Newark International Airport (EWR) heading to Dallas Fort-Worth Airport (DFW). About ten minutes later a second aircraft takes off from Baltimore-Washington International Airport (BWI) heading for DFW along the same jet route as the first. The first aircraft overtakes the second aircraft causing a conflict.

Figure 7. Conflict between aircraft departing from EWR and BWI Airports.

American Institute of Aeronautics and Astronautics

9

Although many conflicts occur between aircraft following the same jet route, they can also occur at points where

jet routes intersect. Figure 8 shows the unusual case of a flight (indicated by arrow) from EWR to Phoenix Sky Harbor International Airport (PHX) that had 3-nm conflicts with seven other aircraft. The locations of the conflicts are marked by circles.

Figures 9 and 10 show the locations of all 13,000 3-nm conflicts in the 2040 trimmed dataset. The conflicts

appear to be evenly distributed over the most heavily travelled jet routes. The red dots in Fig. 10 show the high concentration of conflicts along routes into and out of New York.

Figure 8. Flight from EWR to PHX with seven 3-nm conflicts.

American Institute of Aeronautics and Astronautics

10

Figure 9. Locations of 3-nm aircraft conflicts.

Figure 10. Locations of 3-nm conflicts (zoomed).

American Institute of Aeronautics and Astronautics

11

About 7% of the conflicts shown in Figs. 9 and 10 are byproducts of the cloning process that was used to “grow”

the 2025 schedule starting from a current-day schedule. The cloning process adds flights to the current schedule by exactly copying existing flights and shifting their departure times by a minute or so, which in turn generates an artificial conflict. Most conflicts appear to be legitimate, however, and are due to aircraft sharing jet routes, intersecting flight paths, or flights converging to the same merge fix. Figure 11 gives a histogram of the number of 3-nm conflicts per aircraft. The distribution resembles a Poisson distribution, which is associated with random processes. For our ACES data, about 17% of aircraft would experience a 3-nm conflict without controller intervention; roughly twice this number would experience a 5-nm conflict. The untrimmed baseline scenario had 15% more flights than the 100% trimmed baseline, however the conflict rate was 30% greater.

Figure 11. Histogram of the number of 3-nm conflicts per aircraft in the 2025 trimmed baseline.

American Institute of Aeronautics and Astronautics

12

Figure 12 shows the number of 5-nm conflicts for the different new vehicle scenarios versus the number of

flights in the 2025 scenario for the trim levels 0%, 25%, 50%, 75%, and 100%. The all-vehicle schedules shown are for trim levels 0%, 50%, and 100%. In all scenarios, conflicts increase linearly with the number of scheduled flights. The number of conflicts in the CESTOL and tiltrotor scenarios differs very little from the number in the baseline scenario. The UAS and VLJ schedules both have about 4,000 more flights than the baseline; however, neither shows a substantial increase in conflicts. The all-vehicle scenarios have about 6% more conflicts than the baseline schedules at the same trim level.

Figure 12. Number of 5-nm conflicts by scenario.

American Institute of Aeronautics and Astronautics

13

B. Sector Occupancy

Controller workload is considered high if the number of aircraft in a sector is at or above the sector’s MAP value, an operationally defined upper limit on sector occupancy that traffic managers try not to exceed. We examined sector occupancy using the ACES output for the baseline scenarios for 2025, 2040 and 2086. In the 2086 scenario, a number of sectors stood out for having large numbers of conflicts as well as for greatly exceeding their MAP values. The four sectors shown in Fig. 13 exceeded their MAP values more than any other sectors in the simulation. Together they accounted for 41% of the difference between the peak occupancy and sector MAP values for all sectors. These four sectors sometimes contained twice the number of aircraft as their MAP values. For example, at one point in the simulation sector ZBW01, which has a MAP value of 35, contained 87 aircraft.

The high sector occupancies for the sectors in Fig. 13 are surprising because the ACES simulation is designed to hold aircraft on the ground if the occupancy of a sector is expected to exceed the sector’s MAP value. Because it is difficult to precisely determine a sector’s occupancy in the far future the ACES model relies on estimates of future occupancy. To ensure that sectors never exceed their MAP values users can enable a “perform delay maneuvers” option when running the model. This will cause ACES to make small adjustments to individual flight paths. We did not enable the “perform delay maneuvers” option for the New Vehicle NRA, because doing so appeared to slow the simulation too much, especially with our largest datasets. When we experimented with this option we also found that the delay maneuvers option had a negligible effect on delay.

The high sector occupancies in the offshore sectors shown in Fig. 13 come about because ACES lacks the ability to delay international departures to the United States, even when the “perform delay maneuvers” option is enabled. Because the 2086 schedule has roughly three times the number of flights as today, the international arrivals overwhelm these gateway sectors. ACES responds to the flood of international arrivals by holding outbound aircraft until the flow of international arrivals has stopped. A time series showing the occupancy of sector ZHU87 by origin is shown in Fig 14. During the first 18 hours of the simulation, aircraft originating at international airports enter the

Figure 13. Sectors that most exceeded their MAP value capacities.

American Institute of Aeronautics and Astronautics

14

sector, causing the sector to exceed its MAP value of 25; in the second 18 hours of the simulation, aircraft depart through the sector for overseas destinations.

VI. Conclusion This paper addressed pilot and controller workload by analyzing the modeling and simulation results from the

New Vehicle NRA. Although a number of researchers have identified metrics that correlate better with controller perceptions of workload than sector occupancy, many of the published metrics lack units and thus are difficult to calculate. It is also difficult to infer the relative importance of these metrics from the published studies. Therefore this study focused on the number of sector conflicts as the main indicator of controller and pilot workload. It is noteworthy that the number of conflicts in each scenario increased linearly with the number of scheduled flights, whereas delay increased exponentially. It is likely that in delaying aircraft, airports are reducing the number of en- route conflicts by limiting the number of aircraft that are airborne at a given time. It is also noteworthy that a substantial fraction of aircraft—between 20 and 40%—experience conflicts that would require controller intervention when flying their normal flight plans.

Acknowledgements The work contained herein is supported by NASA under contract NNA08BA64C. The authors wish to thank Harry Swenson, NASA COTR, and Phil Arcara NASA Technical Monitor.

Figure 14. Occupancy of sector ZHU87 by quarter hour in the 2086 untrimmed schedule.

American Institute of Aeronautics and Astronautics

15

References 1NASA Awards Sensis NextGen Research Contract to Model Effects of New Aircraft on the NAS,

http://www.sensis.com/docs/621/ 2 Wieland, F., Smith, J. A., and Clarke, J. P., “Implications of New Aircraft Technologies on the Next Generation Air

Transportation System,” Proceedings of the 2009 USA/Europe Air Traffic Research and Development Seminar, Napa, California, July 2009.

3Sweet, Douglas N. et. al., “Fast-Time Simulation System For Analysis of Advanced Air Transportation Concepts,” AIAA 2002-4593, AIAA Modeling and Simulation Technologies Conference and Exhibit, 5-8 August 2002, Monterey, California.

4Post, J. A Validation of Three Fast-Time Air Traffic Control Models, AIAA 5th Aviation, Technology, Integration, and Operations Conference (ATIO) 26 - 28 September 2005, Arlington, Virginia

5Kuzminksi, P., “MITRE-CAASD’s systemwideModeler, State and Near-Term Plans,” The MITRE Corporation, 10 December 2008.

6Kopardekar, P. and Magyarits, S. “Airspace Complexity Measurement: An Air Traffic Control Simulation Analysis,” Digital Avionics Systems Conference 2002. Proceedings, the 21st Volume1, 27-31, October 2002

7Sridhar, B., et. al. “Airspace Complexity and its Application in Air Traffic Management,” 2nd USA/Europe Air Traffic Management R&D Seminar, Orlando, 1-4 December 1998.

8Masalonis, A. et. al. “Dynamic Density and Complexity Metrics for Real-time Traffic Flow Management,” The MITRE Corporation, 2003

9Majumdar, A. & Ochieng, W.Y. “The factors affecting air traffic controller workload: a multivariate analysis based upon simulation modeling of controller workload,” Paper presented at the 81st Annual Meeting of the Transportation Research Board, Washington, D.C. 2002

10Athénes, S., Averty, P., Puechmorel, S., Delahaye, D., & Collet, C. (2002). “ATC complexity and controller workload: Trying to bridge the gap,” Proceedings of the International Conference on HCI in Aeronautics, Cambridge, 56-60.

11Laudeman, I., S. Sheldon, R. Branstrom, and C. Brasil (1998). Dynamic Density: An Air Traffic Management Metric. Report No. NASA/TM-1998-112226. Moffett Field, CA: NASA.

12Kopardekar, P. (2000). “Dynamic Density, A Review of Proposed Variables,” Egg Harbor TWP., NJ: FAA NAS Advanced Concepts Branch, ACT-540.

13http://aspm.faa.gov/main/taf.asp 14Kostiuk, P., Gaier, E., Long, D. “The Economic Impacts of Air Traffic Congestion,” Boeing Research Document IR811S1,

April 1998, http://www.boeing.com/commercial/caft/reference/documents/lmi_econ.pdf 15Gawdiak, Y., “IPSA NextGen Portfolio Analysis: Benefits Assessment Overview,” Joint Planning and Development Office,

May 2009 (unpublished). 16Trani, A., Baik, H. et.al. “Nationwide Impacts of Very Light Jet Traffic in the Future Next Generation Air Transportation

System (NGATS),” 6th AIAA Aviation Technology, Integration and Operations Conference (ATIO), Wichita, Kansas, Sep. 25-27, 2006.