Simulation Optimization of Airline Delay with Constraints and Multiple Objectives

Air traffic delay is a growing and expensive problem.We investigated ways to reduce the cost and magnitude ofsuch delays by means of several control strategies. Airmanagement and planning at this level can be facilitatedby simulation, especially for strategies that alter controlson the system. We used the SIMMOD air traffic simulationto model the system. The goal was to determine a set ofcontrol measures that achieve the best system performancesubject to restrictions on the decision parameters and selectedoutput measurements. Because observed system performanceis "noisy," the problem is a constrained stochasticoptimization problem with multiple nonlinear objectivefunctions and nonlinear, stochastic constraints, which requiresefficient stochastic optimization methods for its solution.Our approach used simultaneous perturbation stochasticapproximation (SPSA) with a penalty function tohandle the difficult constraints. The results are illustratedin simulation experiments.