Intelligent Software Controller for Ms. Pacman

Abstract

In this project, a game tree idea is being implemented for playing Ms. Pac-man. Comparing to rules base method used by the current controller, using a game tree is a totally different approach and much more complicated. The game tree structure is also designed into a more complex way so that it is able to handle the followings: 1. Simulate the speed differences between Ms. Pac-man, edible ghost and normal ghosts. 2. Simulate the behaviours of edible ghost and make a better decision according to their direction and position. (In another word, Ms. Pac-man can trap the edible ghosts rather than chasing them) 3. Maximize the chance of eating more edible ghosts 4. Bet the correct direction that provides Ms. Pac-man with more opportunities of catching an edible ghost in a critical situation The game tree searching method implemented is based on "Monte-Carlo" method. This searching method is more flexibility on calculation time in order to solve the time consuming problem when facing Ms. Pac-man (a video game).