OPTIMIZATION OF THE GENETIC ALGORITHM IN THE SHEHERAZADE WARGAMING SIMULATOR

Abstract

Stability and Support Operations (SASO) continue to play an important role in modern military exercises. The Sheherazade simulation system was designed to facilitate SASO-type mission planning exercises by rapidly generating and evaluating hundreds of thousands of alternative courses-of-action (COAs). The system is comprised of a coevolution engine that employs a Genetic Algorithm (GA) to generate the COAs for each side in a multi-sided conflict and a wargamer that models various subjective factors such as regional attitudes and faction animosities to evaluate their effectiveness. This dissertation extends earlier work on Sheherazade, in the following ways: 1) The GA and coevolution framework have been parallelized for improved performance on current multi-core platforms 2) the effects of various algorithm parameters, both general and specific to Sheherazade, were analyzed 3) alternative search techniques reflecting recent developments in the field have been evaluated for their capacity to improve the quality of the results.