A Comparison of Optimization Approaches for Nationwide Traffic Flow Management

Abstract

Given thousands of flights in a capacity-limited airspace, finding optimal scheduling strategies that minimize delay costs is a computationally difficult task. In this paper, stochastic search techniques are applied to this fundamental Traffic Flow Management problem. Genetic algorithms and simulated annealing rely on searching the solution space in a manner much different than traditional optimization methods. These stochastic search techniques are compared to an integer linear programming model previously described in the literature. The runtime and aircraft schedules resulting from each model are analyzed. Results indicate that the integer programming model finds the optimal schedule faster than the next best technique, simulated annealing. In the presented experiments, solving the same problem with different simulated annealing solvers in parallel, a quality solution (within 5% of optimality) can be found in most cases. With roughly the same number of computations, simulated annealing reached a better solution than the genetic algorithm in 6 out of the 9 scenarios tested.