Pac-Man Artificial Intelligence

Description

The goal of this project was to develop a game-playing AI agent for Pac-Man using object oriented programming in Python. Throughout the course of the project, I developed algorithms for depth first search (DFS), breadth first search (BFS), and A* search with a custom-built admissible heuristic function. In order for Pac-Man to play well against multiple ghosts simultaneously, I also implemented multi-agent adverserial search algorithms with a custom evaluation function. Finally, I created a learning Pac-Man agent that leverages Q-Learning during a training phase to make more optimal decisions during the actual gameplay. Since this project was completed as part of my Duke coursework, the code cannot be publicly accessible, but can be provided on request.