Proceedings of the 2002 Winter Simulation Conference
E. Yücesan, C.-H. Chen, J. L. Snowdon, and J. M. Charnes, eds. ABSTRACT
This paper describes The MITRE Corporation Center for Advanced Aviation System Development (CAASD) re-arch towards simulation of advanced aviation concepts. Rearch activities are aimed toward improving tactical and strategic decision making methods in the near and long term. We describe how CAASD simulation capabilities assist in determining how to achieve our goals for improving tactical and strategic decision making. For the long term, our simu-lation capabilities are becoming more agent-bad. 1 INTRODUCTION
CAASD is involved in simulation modeling and rearch at the strategic and tactical levels. Tactical modeling address air traffic control and traffic flow management. Strategic modeling address permanent changes to the national air-space system. Near-term projects focus on decision-making strategies for the next few years. Long-term rearch focus on determining how tactical and strategic decisions will be made beyond veral years out. Table 1 categorizes some of the current CAASD rearch toward advanced aviation con-cepts that u CAASD simulation models.
Table 1: CAASD Near-term and long-term Rearch to-ward Advanced Aviation Concepts
Near-term projects: Decision-making Long-term rearch:
Advanced concepts
Tactical Traffic Flow Management Probabilistic Traffic
Flow Management
Strategic Infrastructure changes Institutional Change
2 NEAR-TERM TACTICAL
Tactical traffic flow management decision support requires a different kind of modeling and simulation capability than that ud for strategic applications. While strategic applica-
tions often require modeling the general statistical character-istics of traffic flows in order to predict the average or typi-cal impact of system changes, tactical applications require predicting the specific traffic which is currently airborne or planned to be airborne in the next few hours. Also, to pro-vide uful decision support, the capability needs to provide trial planning tools to project the impact of propod traffic flow management initiatives on the traffic flows.
The FAA/CAASD Collaborative Routing Coordina-tion Tools (CRCT) program is focud on developing such tools (Wanke 2000). The CRCT concept development pro-totype predicts future positions of all flights in the national airspace system bad on their currently-filed flight plans, radar surveillance data, and the current departure delay status of airports. The positions are ud to predict traffic levels at key national airspace system resources, namely enroute ctors and airports.
With this ba data t, urs can identify flow prob-lem areas due to congestion and vere weather. Once a problem has been identified, solutions can be developed and evaluated. For example, in Figure 1, a strategy is being developed to reroute aircraft around a large convective weather system. In this example, two reroutes have been specified, emphasized by two bold ries of line gments in the lower half of the figure; one for eastbound flights and one for westbound flights.
With the reroutes specified, the CRCT prototype evaluates the impact of propod strategies in two ways. First, the safety impact is shown in terms of predicted changes in peak en-route ctor aircraft counts. The matrix in Figure 2 contains predicted peak ctor counts over 15 minute intervals along the vertical axis for each ctor.
Each box along the horizontal axis in Figure 2 repre-nts a ctor (top number) and a peak count threshold (bottom number). Counts above the threshold produce yel-low or red alerts. Figure 2 shows some ctors had yellow alerts (pale boxes) during certain portions of the analysis period. In this example, ctors for which predicted peak
皂角米的功效DECISION SUPPORT FOR ADVANCED AVIATION CONCEPTS
Lisa A. Schaefer Leonard A. Wojcik Thomas P. Berry Craig R. Wanke
Center for Advanced Aviation System Development
The MITRE Corporation McLean, VA 22102, U.S.A.
Figure 1: Developing Reroutes around the Weather
Figure 2: Reroute Strategy Evaluation with CRCT counts will increa with the reroute in place are sur-rounded by dark, heavy outlines; ctors with decread peak counts are outlined in a paler shade. Using this data, the traffic manager can adjust the reroute strategy to dis-tribute flight loads better over the involved ctors.
The cond way in which CRCT evaluates propod strategy impacts is by calculating and displaying the time and distance added to involved flights. This data can be ud to choo strategies that have minimum economic impact on airspace urs.
In addition to reroute strategies, a similar “what-if” ca-pability has been developed to asss imposition of miles-in-trail restrictions. During miles-in-trail restrictions, aircraft are slowed and maneuvered to cross a ctor boundary at some minimum spacing (e.g. “20 miles-in-trail”) behind the preceding aircraft. This is done to control ctor volume, of-ten in combination with a weather-induced reroute. Thus the prototype allows traffic managers to superimpo multiple reroutes and restrictions and develop comprehensive flow management strategies. Further rearch is underway in pro-viding automation assistance in developing solutions to such complex flow problems.
CAASD has been working cloly with the FAA and the Volpe National Transportation Systems Center, on a strategy and supporting plan for deploying CRCT capabili-ties as part of the operational traffic flow management sys-tem. The first CRCT-derived flow problem recognition tools were deployed two years ago in the Enhanced Traffic Management System, the FAA’s operational traffic flow management decision support system for the national air-space system, and the rerouting evaluation capability is scheduled for deployment in the Enhanced Traffic Man-agement System this
year.
3 NEAR-TERM STRATEGIC
CAASD is responsible for modeling how different kinds of technologies, infrastructure enhancements, and procedural changes could improve airport capacity, increa ctor throughput, and reduce delays. Recently the FAA devel-oped the Operational Evolution Plan to meet the air trans-portation needs of the United States for the next ten years with a focus on maintaining safety, increasing capacity, and managing delays. Over the next few years, CAASD expects to address the performance of the national airspace system with various combinations of improvements.
We analyze national airspace system performance with an aggregate simulation model called the Detailed Policy Asssment Tool (DPAT). DPAT simulates the air traffic system as a network of queues. DPAT can be ud to ana-lyze how congestion and delays result from the limited ca-pacities of airports and air ctors, and to forecast future congestion to inform policy-makers about airport and air-space improvement needs. DPAT also models the propaga-tion of delay throughout a system of airports and ctors. DPAT models the flow of approximately 50,000 flights per day throughout the airports and airspace of the U. S. na-tional airspace system and can simulate flights to analyze delays at airports around the world.
We u the Future Demand Generator, a data preproc-essor developed by CAASD, to estimate where future flights will fly and connect quences of flights into itiner-aries. Each simulated aircraft flies to the airports listed in its itinerary. Linked itineraries are necessary for simulating delay propagation from departure and arrival queues at air-ports to other airports. General aviation flights are also in-cluded in DPAT analys since they account for some de-mand at airports and in the airspace. Weather at airports is modeled by reducing the capacities, or increasing the run-way rvice rates, at the affected airports. Airport capacitiy
analysis for 31 of the busiest airports in the United States was performed by The MITRE Corporation (2001). Ca-pacities were derived for both good and bad weather situa-tions. For more details about DPAT inputs and architec-ture, e Wieland (1999). The results of a typical national airspace system simu-lation analysis are in terms of schedule delays and queue-ing delays. Queueing delays are an indication of capacity problems at an airport. Schedule delays include propagated delays by an aircraft from one airport to subquent air-ports in its itinerary. Several combinations of itineraries and capacities are run to reprent different demand growths and different national airspace system enhance-ments, such as adding runways or improved technology to airports. Results from each t of runs are compared to as-sist in deciding which improvements have the best impact on the national airspace system. Figure 3 shows an exam-ple of the reduction of delays at an airport for a ten year period as the airport capacity increas by 10%, 30%, and 80% more than year 2000 capacity, for a particular type of enhancement. The delays are plotted for increasing air travel demand predictions for year 2000 through 2010.
1234562000
2001
不伦不什么2002
可能拼音
2003
2004
2005
2006
2007
2008
2009
2010
Demand Scenario
A v g . Q u e u e i n g D e l a y p e r A r r i v a l (m i n u t e s )
Figure 3: Delays at an Airport for Years 2000 through 2010 with Different Airport Capacity Enhancements
The character of air travel demand is expected to change in upcoming years due to reduced costs of indi-vidualized air travel. Predictions (Airbus 2000, Boeing 2001, Holmes 2000, Huettner 2001) state that demand for small aircraft will grow much more than demand for air travel by the traditional hub-spoke system. Figure 4 speci-fies broad categories of aviation demand, as well as other (non-air) modes of transportation. As transportation gets farther from the hub, it becomes less stable and less pre-
dictable. In general, larger growth rates are expected for the less predictable types of demand than for hub rvice, which today predominates in terms of number of pasn-gers. CAASD is working on developing new strategic
modeling capabilities within its aggregate simulation mod-els to address different types of demand.
Figure 4: Levels of Long-distance Travel Demand Ex-pected to Change in Upcoming Years 4 LONG-TERM TACTICAL
CAASD strives to improve decision making regarding air traffic flow management, which is the daily process of managing flows of traffic to airports and ctors with lim-ited capacity. Prently, decision-making by the FAA at the Air Traffic Control System Command Center does not fully account for information uncertainty, especially re-garding weather impact predictions. Often, decisions are bad on what emed to work or did not work in recent operational experience, rather than on a probabilistic un-derstanding of weather predictions. The FAA has recog-nized the need for greater attention to incorporating an un-derstanding of uncertainty in its operations, and has termed this area “probabilistic traffic flow management.” The initial objective for improving tactical modeling is to develop a model to aid traffic flow management post-event analysis with uncertain weather forecasts, bad heavily upon actual experience with weather events in the past. The final objective is to change decision-making policies to reflect information uncertainty. If successful, the tool has the potential to fundamentally change deci-sion-making strategy, and to benefit the flying public dur-ing aviation schedule disruptions caud by weather.
Decision analysis is a well-understood framework for asssment of decision making with uncertain i
nformation. To begin our study of probabilistic traffic flow manage-ment, we have applied the decision analysis framework in the context of an agent-bad model to show how FAA traffic flow management decisions could be assd on the basis of the distribution of possible outcomes following an uncertain weather forecast (Wojcik 2001a). The agent-bad model, called Intelligent agent-bad Model for Pol-icy Asssment of Collaborative Traffic flow management
(IMPACT), reprents individual airlines and the FAA as independent, lf-interested agents within a simulated traf-fic flow management event.
IMPACT models arriving flights to an airport who capacity is limited. Weather effects are reprented by re-duced capacity for some period of time. Airline and FAA agents make decisions bad on specified decision criteria and available information. Airline agents make decisions heuristically bad upon anticipated future costs. As in real traffic flow management operations, their decisions include whether or not to cancel, delay or exchange the arrival times of their scheduled flights. Each airline is lf-interested, so airlines’ goals often conflict, and the effect of this conflict on aviation operations has been modeled with IMPACT as well as other more aggregated models (Camp-bell et al. 2001 and Wojcik 2001b).
温柔的诗
The FAA agent in IMPACT makes decisions about whether and when to exerci system-level actions called ground delay programs and ground stops. Ground delay pro-grams and ground stops are demand management options exercid by the FAA in real traffic flow management operations. A ground delay program controls arrival demand at the affected airport by revising the departure times of flights to reduce the arrival rate to a desired level. Typically a ground delay program is declared hours in advance of an anticipated capacity reduction due to weather. A ground stop controls demand at the affected airport by holding all sched-uled arrival flights on the ground for some period of time. A ground stop is a more drastic action than a ground delay program, and it is ud when the FAA determines that a -rious demand problem warrants such an action.
Sets of IMPACT scenarios were created to illustrate the tradeoffs inherent in decision making with imperfect weather forecasts at an airport. Each scenario t corre-sponds to a different strategy option to respond to a four-hour forecast of weather at the airport. In strategy option 1, the FAA made a decision about whether to declare a ground delay program, and the characteristics of the ground delay program, at the time of the four-hour fore-cast. In strategy option 2, the FAA made a ground delay program decision after waiting two hours after the initial forecast. In strategy option 3, the FAA took no actions. For all three strategy options, the airline agents took decisions in respon t
o the FAA’s decision and how the weather and other airlines’ decisions evolved. Finally, in strategy option 4, neither FAA nor the airlines took any actions to modify the original schedule of arrival flights. Strategy option 4 is not realistic, but was included for comparison against the other, more plausible scenarios.
Figure 5 shows the distribution of cost across 100 IMPACT runs for the four strategy options with perfect weather information. The number of weather scenarios for which average cost per flight fell into each $500 cost bin is shown. Although the variance is large in all cas, on aver-age the lowest cost strategy is when a ground delay pro-gram is declared four hours in advance of the event. This is expected, becau with perfect information, there is no benefit in waiting for a better forecast, and there is advan-tage in acting early before flights depart for the affected airport. However, the cost difference between waiting and not waiting was small.
Figure 5: Distribution of Cost with Perfect Weather Infor-mation
Figure 6 shows the distribution across the four strategy options with imperfect information. Note that there is a spike in the distribution of cost when a ground delay pro-gram is declared 4 hours in advance. This is becau the initial forecast is vere, and it turns out that actual weather rarely turns out to be wor than the initial forecast. Thus, a ground delay program bad on the initial forecast tends to clamp down demand to a level that is almost always manageable. H owever, the expected cost for the strategy option of waiting two hours is less. Thus, there is a tradeoff between predictability and expected efficiency.
This simple analysis illustrates how the decision analysis perspective can be applied to understand the effect of decision making strategies over the long term, i.e., many traffic flow management events. H owever, more work is needed to bring this work into practical u in actual traffic flow man
agement events. The basic limitation of the IMPACT work to date is that it has proven difficult to validate quantitatively against actual events. CAASD is in-volved in rearch to attempt to bridge the gap between the theoretical perspective of decision analysis and the com-plexity of the real world.
5 LONG-TERM STRATEGIC
CAASD is exploring the u of agent-bad modeling to analyze the behavior of airlines when faced with changes in the capacity of the national airspace system. The current modeling effort, called Jet:Wi, attempts to define the
Figure 6: Distribution of Cost with Imperfect Information characteristics of airlines bad on simple rules to deal with incread delay or cost of flight, or reduced pasnger demand, for example. Jet:Wi agents address the changes by varying fares, adjusting schedule, changing air-craft size, etc. Agent-bad modeling eks to simulate the interactions of the agents (flights, airline schedulers, etc.) by directing the agents to ek an objective function. In do-ing so, the model’s behavior exhibit emergent behavior, such as incread or decread airline hubbing. Jet:Wi does this in veral phas, as shown in Figure 7.
Figure 7: Phas of the JetWi Model
Jet:Wi is suited to rearching questions about likely outcomes for broad possibilities such as:
• Airline reaction to a reduction in flight times
• Demographic shifts and their effect on airline r-vice
• How airlines would rvice demand increas
• New airport stimulatory effect on airline rvice
• Fleet changes or utilization impact caud by in-cread or decread congestion.
Jet:Wi was a spin-off from work on the agent-bad IMPACT model, described in Section 4. It is the first at-tempt by CAASD to apply agent bad modeling to the en-tire air traffic system on a long-term time scale. Much re-mains to be learned as CAASD continues to explore the application of this exciting technique in the air traffic do-main. We expect Jet:Wi to improve our ability to make decisions on infrastructure changes while considering pos-sible changes to aviation business models in the future.
To date, little rearch has been done to examine the ways in which current and future information technologies will change aviation business models, or the future trav-elscape. The aviation travelscape can be describes as the sum of products and rvices that enable travel by air, in-cludin
g lodging, food, and multi-modal transportation. The travelscape is not limited to physical products and rvices, however. Information technology will play a primary role in enabling the traveler of the future to navigate the trav-elscape. Figure 8 shows an example of information that could be communicated to a pasnger through a personal digital assistant someday, assuming the future travelscape provides a method for this type of information to be trans-mitted to pasngers in real time. The pasnger would be able to u the information to lect alternate travel options if his flight were canceled. Technological trends in compu-tational speed, wireless connectivity, information exchange and intelligent systems could enable a shift in the industry from supplier-driven to pasnger-driven and enable more efficient u of existing asts.
人的生殖教案
Figure 8: Information Received on a Per-
sonal Digital Assistant as a part of a Pos-
sible Travelscape Scenario
The objective of the travelscape effort is to create and refine a vision of the future travelscape from a pasnger perspective and to relate it to the demand and supply of air traffic rvices. This project eks to define the likely areas of demand and examine them for future rearch consideration.
The demand scenarios can be explored with the Jet:Wi model. For example, to simulate a fleet change
scenario where the average aircraft size in airline fleets de-clines over time, Jet:Wi’s aircraft schedules can be ad-justed so that they are flown with smaller airplanes than currently exist. Other types of simulation from travelscape findings might include increas in operational costs be-cau of curity, avionics requirements, or incread land-ing fees at lected airports.
6 DISCUSSION
Weather is a major factor in all analys. There is a need to better understand and model the effects of imperfect weather forecasts on decision-making and system perform-ance. CAASD is working on a tactical modeling of the ef-fect of imperfect weather information in a decision analysis framework. It is an open rearch question as to how to roll up this understanding to a more aggregated level for strate-gic analysis
Conventional, non-adaptive modeling is available for our near-term strategic modeling needs, but agent-bad modeling will be needed to fulfill many of our long term needs. CAASD is working toward developing better agent-bad models to improve our modeling capabilities for fu-ture national airspace system analys. H owever conven-tional modeling will still be required in the future; there-fore we are still working toward improving our conventional modeling capabilities.
Agent-bad simulation can be ud to help fill in the gap between infrastructure and institutional changes. Tra-ditional simulation analys consider only the effects of new technology or new runways on delay. Agent-bad models can show the effects of new business models: such as incread fractional ownership of aircraft or aircraft that can be chartered on short notice.
REFERENCES
Airbus. 2000. Global Market Forecast 2000-2019.
</pdfs/media/GM
F2000.pdf> [accesd April 12, 2002].
Boeing. 2001. Current Market Outlook. < /commercial/cmo/1eo00.
html> [accesd April 12, 2002].
Campbell, K. C., W. W. Cooper, D. P. Greenbaum, and L.
A. Wojcik. 2001. Modeling distributed human deci-爱心蛋糕
sion-making in traffic flow management operations.
Chapter 15 in Air Transportation Systems Engineer-
ing, ed. G. L. Donohue and A. G. Zellweger. 227-237.
Reston, Virginia: American Institute of Aeronautics and Astronautics (AIAA).
Holmes, B. 2000. National Aeronautics and Space Admini-stration Small Aircraft Transportation System (SATS) Program Planning White Paper. <sats.
v/downloads/SATS_White_P
表示快的词语>召开读音aper.pdf> [accesd April 12, 2002]. H uettner, C. H. 2001. Making the Opportunity of the
American Dream Accessible to Every American. Un-
published briefing.
The MITRE Corporation 2001. Airport Capacity Bench-mark Report 2001. Available online via <http: //www.caasd/library/general>[accesd April 12, 2002].
Wanke, C. 2000. Collaborative Routing Coordination Tools (CRCT): Developing traffic flow management decision support concepts. In Proceedings of the 2000 Air Traffic Control Association Conference. Wieland, F. 1999. The Detailed Policy Asssment Tool (DPAT) U r’s Manual, MITRE Technical Report 99W00000012.
Wojcik, L. A. 2001a. Three principles of decision-making interactions in traffic flow management oper
ations. In the 4th U SA/Europe Air Traffic Management R&D Seminar, Santa Fe, New Mexico, December 3-7. Wojcik, L. A. 2001b. Models to understand airline and ATM authority decision-making interactions in sched-
ule disruptions: from simple games to agent-bad models. chapter 37 in Handbook of Airline Strategy, ed. G. F. Butler and M. R. Keller. 549-575. New York: McGraw-Hill.
ACKNOWLEDGMENTS
The authors would like to thank Urmila Hiremath for help-ful comments toward this paper, and Glenn Roberts and Fran H oover for locating references for this paper. We would like to thank The MITRE Corporation Center for Advanced Aviation Systems Development for sponsoring this work. The contents of this material reflect the views of the authors. Neither the Federal Aviation Administra-tion nor the Department of Defen makes any warranty or guarantee or promi, express or implied, concerning the content or accuracy of the views expresd herein. AUTHOR BIOGRAPHIES
LISA A. SCHAEFER is currently a Senior Simulation Rearch Engineer at the MITRE Corporation in McLean, Virginia working on air traffic simulation validation and as the PI for a rearch project to generate future aircraft de-mand schedules. She received her B.S.E. and M.S. in transportation engi
neering and Ph.D. in industrial systems engineering at Arizona State University. Previous positions include simulation rearch at the Federal H ighway Ad-ministration and Wright Patterson Air Force Ba. Her in-terests include simulation of systems with distributed intel-ligence and concept development for future transportation systems. H er email and web address are: <mailto: LisaAnn@mitre> and <pwolfe.eas.asu. edu/lisa>.
