Parameter Estimation Using Genetic Algorithms for Air Traffic Management Simulation Scenario Generation

Abstract

An optimal parameter estimation technique using nested genetic algorithms is presented for estimating flight planning cost functions from observed flight plans. An outer-loop is described that estimates flight plan cost function parameters for input to an inner-loop flight plan optimization given the estimated cost function. These estimated cost functions are designed to be suitable for the generation of optimal flight plans for use in simulations. Optimal flight planning necessitates tailoring of aircraft performance data to ensure suitability for flight plan optimization. A method is provided to tailor aircraft performance data for this type of optimization. Initial results indicate that observed flight plan data can easily be matched in one dimension (route, altitude or speed), but that further refinement of the outer-loop objective function is required to match observations in all dimensions.