This serves as a repository for a python program that can play pacman all by itself.
01...is the code used from a group project in my undergraduate artificial intelligence course. This, agent was provided the least amount of information about the game. Each game, the breeder assigns a neural network to dictate Pacman's actions.02...uses the class Minimax approach. This agent proved to be good at avoiding the ghosts when a depth of 3 or 4 was used.03...uses another machine learning technique, Q-Learning. This agent was provided info about the game state (pellet locations, ghost locations, and legal actions), and quickly learns which actions will provide a better score. This agent learned how to beat the classic map fairly quickly (10-20 games).
- Huge credit to UC Berkeley for providing Intro to AI course materials online. Pacman files for the game itself were take from UC Berkeley's cs188 site.