The Effects of Airline Adaptive Route Selection on NAS-wide Performance

NAS-wide simulations are used to evaluate concepts of operations, technologies, and their cost/benefits. When modeling stakeholder behavior, users of state-of-the-art NAS-wide simulators must encode decision-making in fixed sets of rules before executing the simulations. Adaptability and evolution in stakeholder decision-making is a next step to accommodate progressively realistic modeling of the NAS. This paper introduces adaptable behavior for airlines into a NAS-wide simulation tool. Adaptability is achieved by the application of Reinforcement Learning and the concept of domination to compare performance of multi-valued functions. The paper studies the effects on the NAS of adaptable route selection behavior of airlines in the presence of local and system-wide historic performance information. The results show the effectiveness of adaptive route selection in improving NAS-wide performance (even) when all agents seek use of the same resources simultaneously. It is also effective when the airlines have only information about their own past and current performance and do not have access to system-wide information. This effectiveness results in benefits for multiple stakeholders. The implications of these results are discussed.