Evaluating evolution and monte carlo for controlling air traffic flow

Abstract

The automated optimization of air traffic flow is a critical component of the next generation air traffic system, designed to facilitate the future expansion of air traffic with little increase in infrastructure. While many traditional optimization approaches have been applied to the air traffic flow problem, they have difficulty scaling to large problems and in handling the nonlinearities inherent in the air traffic flow patterns. As a solution, this paper shows how genetic algorithms can be successfully applied to this problem. With this approach, the airspace is broken up into separate control points, with a single gene within a chromosome controlling an individual point. A genetic algorithm can then be used to find a controller that maximizes the performance of the airspace. To validate this approach, we use FACET, an air traffic simulator developed at NASA and used extensively by the FAA and industry. On a scenario composed of one thousand aircraft and two points of congestion, our results show that the evolutionary method provides 60% higher performance than more traditional Monte Carlo methods