Session:

MSA - Military and Security Applications of Evolutionary Computation

Title:

Playing to train: Case injected genetic algorithms for strategic computer gaming

Authors:

Sushil J. Louis
Chris Miles
Nicholas Cole
John McDonnell

Abstract:

We use case injected genetic algorithms to learn how to competently play computer strategy games that involve long range planning and complex dynamics. Such games inspire, and are inspired by, military training simulations on which trainees spend considerable time honing their skills. By instrumenting the game interface, we unobtrusively acquire knowledge in the form of cases from human experts playing the game and use case-injected genetic algorithms to incorporate this knowledge in evolving competent game players. The games we have focused on have two sides, Blue and Red, and an evolved player can thus serve two purposes. A competent player for Blue serves as a decision aid for a Blue trainee. At the same time, a competent Blue player serves as a training opponent for a Red trainee. Results in the context of a strike force planning game show that with an appropriate representation, case injection is effective at biasing the genetic algorithm towards producing competent plans that contain important strategic elements used by human players.