Mason Archival Repository Service

An Analysis of the E ects of Net-Centric Operations Using Multi-Agent Adaptive Behavior

Show simple item record

dc.contributor.advisor Sherry, Lance
dc.contributor.author Calderon-Meza, Guillermo
dc.creator Calderon-Meza, Guillermo
dc.date 2011-04-27
dc.date.accessioned 2011-05-25T17:03:26Z
dc.date.available NO_RESTRICTION en_US
dc.date.available 2011-05-25T17:03:26Z
dc.date.issued 2011-05-25
dc.identifier.uri https://hdl.handle.net/1920/6362
dc.description.abstract The National Airspace System (NAS) is a resource managed in the public good. Equity in NAS access, and use for private, commercial and government purposes is coordinated by regulations and made possible by procedures, and technology. Researchers have documented scenarios in which the introduction of new concepts-of-operations and technologies has resulted in unintended consequences, including gaming. Concerns over unintended consequences are a signi cant issue for modernization initiatives and have historically been a roadblock for innovation and productivity improvement in the NAS. To support the development and evaluation of the Next Generation Air Transportation System (NextGen) and the Single European Air Tra c Management Research Programme (SESAR) concepts-of-operations and technologies, analysis methodologies and simulation infrastructure are required to evaluate the feasibility and estimate the bene ts. State-ofthe- art NAS-wide simulations, capable of modeling 60,000 ights per day, do not include decision-making. A few recent studies have added algorithms to these simulations to perform decision-making based on static rules that yield deterministic outcomes. In the real-world NAS, however, autonomous agents (e.g. airlines, air tra c control) are continuously adapting their decision-making strategies to achieve their enterprise objectives (i.e., minimize costs of operations). Further, analysis of an inventory of \gaming" scenarios in the NAS identi ed \adaptation" by agents as the underlying mechanism for taking advantage of opportunities to increase productivity in the NAS and unintended consequences. This dissertation describes: (1) the design, implementation, and integration of adaptive agent behavior in NAS-wide simulations, and (2) the use of quantitative methods to analyze the e ects of adaptive behavior on the bene ts of new concepts-of-operations and technology, and unintended consequences. The application of this approach is demonstrated in a case study evaluation of adaptive ightplan route selection and System-wide Information Management (SWIM) technologies using NASAs Future Air Tra c Management Concepts Evaluation Tool (FACET). The simulation results for 60,000 ights per day for more than 80 days can be summarized as follows: 1. Adaptation in ightplan route selection in the presence of SWIM resulted in a \steadystate" of the NAS that was not generated through collusion, but through self organization. 2. The steady-state in the ightplan route selection was achieved within 17 simulated days for a 60,000 ight per day NAS when global (i.e. airlines have access to data from other airlines and their own data), accurate, and real-time (i.e. no communication delay) SWIM information was available. Steady-state was achieved in 32 simulated days when the information was local (i.e. airlines have access only to their own data), real-time, and inaccurate (i.e. noisy). 3. The steady-state yielded a system-wide reduction in fuel burn (i.e. distance), departure delays, arrival delays, and airborne con icts compared to the random selection of routes. 4. When SWIM provided global information instead of local, there was no signi cant e ect on overall NAS performance (i.e. changes were marginal). The steady-state was reached in one additional day. Total number of airborne con icts experienced a decrease of 2.8%, but the variability of number of con icts was 270% higher. The variability of the total arrival delay decreased 38%, but the variability of fuel burn, departure delay, sector congestion, and arrival airport congestion did not change signi cantly. 5. With one day of latency in SWIM data steady-state was reached in 4 additional days with global data and 8 additional days with local data. Fuel burn did not change signi cantly. The total arrival delay increased 0.3% and the total departure delay increased 2.0% with global data. The total arrival delay increased 0.1%, the total airborne con icts increased 0.7%, and the total departure delay increased 0.5% with local data. The variance decreased with global information. With local information, variance only decreased for the delays, but increased or was equal for the other metrics. 6. Inaccuracy of +/-30% in the SWIM data decreased 3.7% (2,247) the airborne con- icts with global data, and 0.9% (583) with local data. The arrival delay decreased 1.0% with global data and 1.3% with local data. The departure delay and the %OL descreased marginally too. The fuel burn increased about 0.12% (410,362 to 506,895 kg/day). The variance of the airborne con icts increased 394%, and the arrival delay increased 103% with global data, but the variance of the departure delay and of %OL decreased 72% and 59%. With local data the variance for the total airborne con icts increased 79%, for fuel burn increased 71%, and for arrival delay increased 51%. The bene ts of this research are: (1) the establishment of architecture and algorithms for the analysis of adaptive behavior in NAS-wide simulations (such as FACET and Airspace Concept Evaluation System (ACES)), (2) methodology for analysis of the results of adaptive behaviors in the NAS, and (3) analysis robustness to degradation of SWIM functionality of adaptive ightplan route selection. This provides the capability for researchers, analysts, and policy-makers to evaluate proposed concepts-of-operations and technologies in the presence of adaptive behavior.
dc.language.iso en_US en_US
dc.subject Simulation en_US
dc.subject NextGen en_US
dc.subject Multi-agent en_US
dc.subject National Airspace System en_US
dc.subject Adaptive behavior en_US
dc.subject Reinforcement Learning en_US
dc.title An Analysis of the E ects of Net-Centric Operations Using Multi-Agent Adaptive Behavior en_US
dc.type Dissertation en
thesis.degree.name PhD in Information Technology en_US
thesis.degree.level Doctoral en
thesis.degree.discipline Information Technology en
thesis.degree.grantor George Mason University en


Files in this item

This item appears in the following Collection(s)

Show simple item record

Search MARS


Advanced Search

Browse

My Account

Statistics