Pac-Man AI Project

Description: This project is a mini-laboratory assignment for the 4ALGL4A course where students implement rational agents to play Pac-Man. The goal is to explore artificial intelligence concepts such as reflex agents, adversarial search (Minimax), alpha-beta pruning, and Expectimax in a controlled Python 3 environment using Tkinter for the graphical interface.

---

Repository Structure

.
├── pacman.py              # Main game script
├── graphicsDisplay.py     # GUI display
├── graphicsUtils.py       # GUI helper functions
├── textDisplay.py         # Text-based display
├── game.py                # Game state definitions
├── util.py                # Utility functions
├── projectParams.py       # Program parameters
├── layouts/               # Maze layout text files
├── layout.py              # Layout loader
├── keyboardAgents.py      # Human-controlled agent
├── ghostAgents.py         # Ghost AI agents
├── pacmanAgents.py        # Sample agents: LeftTurnAgent, GreedyAgent
├── multiAgents.py         # Your rational agents: ReflexAgent, MultiAgentSearchAgent, MinimaxAgent, AlphaBetaAgent, ExpectimaxAgent

---

Agents Overview

1. ReflexAgent – Evaluates the immediate effect of each action and chooses the best.
2. MultiAgentSearchAgent – Base class for agents considering other agents' potential moves.
3. MinimaxAgent – Implements the Minimax algorithm for adversarial search.
4. AlphaBetaAgent – Optimized Minimax with alpha-beta pruning.
5. ExpectimaxAgent – Uses expected values for ghost actions instead of worst-case scenarios.

Each agent has two main methods: get_action(self, game_state) and evaluation_function(self, current_game_state, action).

---

Usage

1. Run the game with a human agent:

python pacman.py

2. Run with ReflexAgent:

python pacman.py -p ReflexAgent

3. Run with MinimaxAgent (depth 4):

python pacman.py -p MinimaxAgent -l minimaxClassic -a depth=4

4. Run with AlphaBetaAgent (depth 3):

python pacman.py -p AlphaBetaAgent -l smallClassic -a depth=3

5. Run with ExpectimaxAgent (depth 3):

python pacman.py -p ExpectimaxAgent -l minimaxClassic -a depth=3

Use --help to see all options:

python pacman.py --help

---

Goals

• Implement rational agents for Pac-Man.
• Apply Minimax, Alpha-Beta pruning, and Expectimax algorithms.
• Improve evaluation functions to enhance agent performance.
• Test agents on various maze layouts and ghost configurations.

---

References

• UC Berkeley Pac-Man AI project: https://ai.berkeley.edu/multiagent.html
• Minimax algorithm: https://en.wikipedia.org/wiki/Minimax
• Alpha-beta pruning: https://en.wikipedia.org/wiki/Alpha-beta_pruning

This repository is for educational purposes and course submission only.