Coevolutionary optimization of fuzzy logic intelligence for strategic decision support

Abstract

We present a description and initial results of a computer code that coevolves fuzzy logic rules to play a two-sided zero-sum competitive game. It is based on the TEMPO Military Planning Game that has been used to teach resource allocation to over 20 000 students over the past 40 years. No feasible algorithm for optimal play is known. The coevolved rules, when pitted against human players, usually win the first few competitions. For reasons not yet understood, the evolved rules (found in a symmetrical competition) place little value on information concerning the play of the opponent.