dwaynee174/ghost-hunter-pacman

A variation of Pacman arcade game designed to train Pacman agents that use sensors to locate and eat invisible ghosts with phenomenal efficiency. Used Joint Particle Filter algorithm in AI to get 30% optimized results.

Ghost-Hunter-Pacman

Abstract

A variation of the standard Pac-man game, the problem is Pac-man has always spent his life running from ghosts. But we need to train Pac-man to track the ghosts and hunt them down by their banging and clanging. The focus of this project is to design Pac-man agents that use sensors to locate and eat invisible ghosts. This can be done by advancing from locating single, stationary ghosts to hunting packs of multiple moving ghosts with phenomenal efficiency. The goal is to implement joint particle filters in Bayes’ Nets to track the ghosts.

Index Terms—Pac-man, Joint Particle Filters, Inference, Monte Carlo Tree Search (MCTS), Sequential Monte Carlo (SMC), Bayes’ Nets