Event-driven game theory for aviation safety the primary goal of this project is to develop a set of methodologies for making quantitative predictions about the safety impact of introducing nextgen…

Event-driven game theory for aviation safety the primary goal of this project is to develop a set of methodologies for making quantitative predictions about the safety impact of introducing nextgen automation, UAS's and other novel technologies into the NAS. In particular we are interested in making such predictions that account for the behavior of humans in the NAS. We will focus on airspace scenarios including both EN route and terminal operations that will be modeled after a major region of the NAS. The scenarios will include a variety of human roles, e.g., commercial pilots in different aircraft types; pilots engaged in takeoff, landing, and EN route operations; controllers in different sectors with different characteristics and goals; and participants with different information as a result of mixed equipage. The analysis will focus on interaction between commercial aircraft, controllers, and both autonomous and human-controlled UAS, with a particular focus on the safety impact of both autonomous and human-controlled uass. Important questions include whether the strategic response of the human participants to uass creates safety issues (or vice versa) and whether NAS safety is robust to equipment failure. To achieve this goal we will further develop a new class of game theory models called event-driven games. Game theory is the most developed field for predicting behavior of interacting humans. However conventional game theory requires strong assumptions about the timing and sequence of both player moves and natural events. NAS scenarios are complex and highly stochastic, and so do not readily conform to such assumptions. For example, there might be uncertainty about which aircraft will arrive at a merge point first, or there might be uncertainty about the time between aircraft maneuvers in a tcas scenario. Event-driven games formulated in terms of stochastic processes. This allows them to relax assumptions about the timing of player moves and sequences, treating them probabilistically. As such, they are perfectly suited to modeling human behavior in the NAS. An additional advantage of a stochastic process framework is that there are well-established techniques to optimally approximate a computationally intractable "fine-grained" stochastic process model of a system involving small-level details (e.g., an interlinked set of highly-detailed models of simultaneously occurring activity in the NAS) with a computationally tractable "coarse-grained" model of the same system, in which such small-level details are approximated away. We expect to use these approximation techniques in eventdriven game theory to scale up our models from the fine-grained level of details, involving individual aircraft, pilots, controllers, etc, to coarse-grained models involving the entire NAS. Developing such coarse-graining methods will be a focus of this research effort. The key elements of our approach include: 1.an event-driven framework to model the stochastic dynamics of large complex systems like those seen in air traffic safety situations 2.the ability of the framework model to include one or more humans interacting both with the physical subsystems and with one another 3.the ability of the framework to scale up in a principled manner 4.the potential to use the framework to assess safety issues and to provide associated engineering design guidance in addition to the specific objectives mentioned above, a more general objective of this research effort is to provide NASA with a set of tools for analyzing air safety issues, including issues not covered by our airspace model. Both the airspace model and the event-driven games framework will be contributions to that more general objective. Furthermore, the project is expected to release open-source software to implement the event-driven airspace models and to carry out the analysis. Software developed by the authors, for a closely related project in air safety can be accessed at https