Developing a Configurable Air Traffic Controller Agent for Fast-Time Simulation

Abstract

The Federal Aviation Administration continually pursues advances that improve operational efficiency while preserving passenger safety. Research of new concepts and technologies often requires the use of simulation to cost-effectively assess the impact of these envisioned improvements on an abstracted version of the real system. In an effort to improve our ability to conduct fast-time simulations on a wide range of concepts and scenarios, The MITRE Corporation has begun the development of a simulation platform that supports configurable air traffic control agents. The agents are designed to exercise a set of simple rules in a stimulus-response framework. Stimuli such as predicted conflicts and metering delays are sent to a controller agent, and the controller responds via flight commands such as speed, altitude, or heading changes. This controller agent is configurable such that analysts can define different response strategies and stimuli response prioritizations for different experiments and for different sectors within a given experiment. When integrated in a fast-time simulation environment, this agent enables users to run experiments with thousands of scenarios as no humans are needed to provide flight commands. This paper illustrates the design framework for the controller agent as well as the model validation and verification. MITRE validated and verified this model in two ways: first, by comparing the number of and type of commands that the controller agent issued with historical data. Second, by comparing the spacing conflicts in simulation runs with and without the controller agent.