Evolutionary fuzzy systems for generating better Ms.PacMan players

PacMan is one of the most popular video games played in all over the world. Ms.PacMan, a mutant of PacMan, employs probabilistic transitions of ghosts. Therefore, there is no deterministic (surefire) routes to solve for each stage in Ms.PacMan. That is, real time control mechanisms are needed to realize auto-play for Ms.PacMan by computers. This paper proposes evolutionary fuzzy systems for playing Ms.PacMan. The proposed method employs fuzzy logic and the evolution of their fuzzy rule parameters. That is, basic structure of fuzzy rule is given in advance. Then, the rule parameter is evolved by using (1 + 1)ES.