DAG-NEG: Simulating human behavior to safely test far-reaching improvements to US air traffic management systems
The Situation
NASA is evaluating a number of potential new concepts and technologies for managing air traffic in the US National Airspace System, the FAA system that covers every aspect of aviation in the United States including aircraft, air traffic control, airlines, and airports. One advanced concept currently under evaluation, called a distributed decision-making environment (a DDM environment), will allow pilots and airline operations centers more freedom in changing and optimizing flight plans. By giving pilots and airline operations centers (airspace users) the flexibility to reroute planes in the air, operations can potentially be considerably streamlined, and significant improvements may be realized in fuel efficiency, schedule maintenance and safety.
The Charles River Analytics Solution
The introduction of this envisioned DDM environment, however, will have profound implications on the daily, hourly and minute-by-minute tasks of pilots, airline dispatchers, and air traffic controllers throughout the air traffic management system. These changes will affect tasks that range from gathering information, to communication, assigning responsibilities, allocating work, and making decisions. To make this transition safely, NASA is sponsoring a research program to define the best ways for these tasks to be accomplished prior to deployment in operational airspace. The program calls for software models to simulate the procedures and technologies that make up the proposed DDM environment.
With NASA sponsorship, Charles River Analytics is addressing this need. The project consists of three components. First, we are using our general behavior modeling software, to create models, called agents, of how individual pilots, air traffic controllers, and airline dispatchers make decisions. Next, these agents are embedded within a simulated air traffic environment; each agent within this simulation implements proposed versions of the rules for the distributed decision management system. Charles River Analytics specifically designed these agents to produce a wide range of behaviors and coordination strategies in response to different air traffic management concepts. Finally, to find the behaviors and decision-making strategies that result in the best performance for the entire air traffic management system, we applied a sophisticated tool-a genetic algorithm-to comb through the thousands of possible behaviors and distributed decision-making interactions to find those that offer the best results for safety, fuel savings, scheduling, and other factors.
The Benefit
The result is a software toolkit that allows NASA researchers to safely test proposed changes and improvements for the air traffic management system within a simulated environment. With this capability, NASA is able to identify the combined set of individual agent behaviors, procedures and protocols that optimize air traffic management system performance, with respect to safety and efficiency.
Related papers:
Harper, K. A., Mulgund, S. M., Zacharias, G. L., & Kuchar, J. (1998). "Agent-Based Performance Assessment Tool for General Aviation Operations Under Free Flight". AIAA Guidance, Navigation and Control Conference, Boston, MA. Vol.1. pp.1-10.
Harper, K. A., Mulgund, S. S., Guarino, S. L., Mehta, A. V., & Zacharias, G. L. (1999). "Air Traffic Controller Agent Model for Free Flight". Proceedings of the 1999 AIAA Guidance, Navigation and Control, Portland, OR.
Harper, K., Guarino, S., White, A., Hanson, M. L., Bilimoria, K., & Mulfinger, D. (2002). "An Agent-Based Approach to Aircraft Conflict Resolution with Constraints". AIAA Guidance, Navigation, and Control, Monterey, CA.
