\The Computational Intelligence of the Game Pac-Man

The Game Pac-Man is a both challengeable and satisfactory video game that has been the focus of some important AI (Artificial Intelligence) research. The goal for using the Game Pac-Man as a test bed in our experiment is that the Pac-Man Game provides a sufficiently rich and challengeable platform for studying the AI in the computer game, and that it is simple enough to permit understanding of its characteristics. In this paper, we use AMAF (All-Moves-As-First) algorithm to manage the search tree for implementing the Pac-Man AI. Moreover, we also introduce the rule-based policy with the domain knowledge to improve the AI of Ghosts (or opponent). Finally, the experimental result from the Game Pac-Man is presented to demonstrate the effectiveness and efficiency of these algorithms.