Pac-Man Q-Learning Agent

Welcome to the Pac-Man Q-Learning Agent repository! This project demonstrates the implementation of a reinforcement learning algorithm, specifically Q-Learning, to enable Pac-Man to learn and navigate its environment. Developed in Python, this project leverages a simplified version of Pac-Man game to illustrate how Q-Learning can be used for decision-making in an AI agent.

Project Overview

• Objective: Implement and evaluate a Q-Learning algorithm for controlling Pac-Man in the classic game.
• Language: Python 3
• Environment: Pac-Man game framework provided by UC Berkeley's AI course.
• Algorithm: Q-Learning, a reinforcement learning technique where the agent learns the value of actions based on rewards received from the environment.

Key Features

• Agent: Pac-Man, controlled by the QLearnAgent.
• Learning: Pac-Man learns to navigate the game environment, avoid ghosts, and collect food using Q-Learning.
• Parameters:
  • -n: Number of game iterations to run.
  • -x: Number of training episodes.

Usage

To run the Q-Learning agent for Pac-Man, use the following command:

python3 pacman.py -p QLearnAgent -x 2000 -n 2010 -l smallGrid