Pac-Man AI Game

A simple text-based Pac-Man game implemented in Python using NumPy. This project demonstrates basic game AI concepts, featuring a Pac-Man agent controlled by the Minimax algorithm (with Alpha-Beta Pruning) and Ghost agents using A* search to chase Pac-Man.

Features

• Text-Based Interface: Renders the game state directly in the terminal.
• AI-Controlled Pac-Man: Uses the Minimax algorithm with Alpha-Beta pruning to determine the optimal move based on a heuristic evaluation function.
• AI-Controlled Ghosts: Employ A* search algorithm to find the shortest path to Pac-Man's current location.
• Customizable Layout: Game board layout is loaded from an external text file (layout.txt).
• Configurable Parameters: Easily adjust the number of ghosts and the search depth for Pac-Man's AI.
• Scoring System: Includes scores for eating dots, penalties for moving, and significant scores for winning or losing.
• Game State Evaluation: A heuristic function evaluates game states considering score, distance to ghosts, and remaining dots.

Requirements

• Python 3.x
• NumPy library

1. Install NumPy:

   pip install numpy

Usage

1. Prepare the Layout File:

   • The game requires a layout.txt file in the same directory as the script.
   • The script includes a default layout and will automatically create layout.txt with this content if it doesn't exist.
   • Layout Characters:
     • o: Wall
     • .: Dot (Food)
     • : Empty Space
     • P: Pac-Man starting position (only one allowed)
     • G: Ghost starting position (optional, ghosts can also start on empty spaces)
     • *: Power Pellet (Note: Power pellet logic is not currently implemented in the provided code, but the constant exists).
   • Example layout.txt:

     oooooooooooooooooooo
     oP.................o
     o.ooooooooo.oooooo.o
     o.o.......o..o.....o
     o.o.ooooo.o.o.oooo.o
     o......G...o..o.o..o
     ooooo.o.ooo.oooo.o.o
     oG....o............o
     oooooo.oooooo.oooo.o
     o..................o
     oooooooooooooooooooo

   • Important: The layout must be rectangular (all rows must have the same length).

2. Run the Game: Execute the Python script from your terminal:

   python pacman_game.py

   (Replace pacman_game.py with the actual name you saved the script as).

3. Gameplay: The game will run automatically. Pac-Man and the Ghosts will move based on their respective AI algorithms. The game ends when Pac-Man eats all the dots (Win) or gets caught by a ghost (Lose).

Configuration

You can configure the game by modifying the if __name__ == "__main__": block at the end of the script (pacman_game.py):

if __name__ == "__main__":
    # --- snip ---

    try:
        # Change number of ghosts and Pac-Man AI depth here
        game = PacmanGame(layout_filename, num_ghosts=2, pacman_ai_depth=4)
        game.run()

    # --- snip ---

• num_ghosts: Set the desired number of ghosts. If the layout doesn't contain enough G characters, ghosts will be placed on available empty spaces ( ).
• pacman_ai_depth: Controls how many moves ahead the Pac-Man AI looks (using Minimax). Higher values lead to potentially better decisions but significantly increase computation time per move.

Code Overview

• pacman_game.py (Main Script): Contains all the classes and functions.
  • Board Class: Represents the game grid, handles layout loading, character access, valid moves, and drawing. Uses NumPy array for the layout.
  • Pacman Class: Represents the Pac-Man agent, manages its position and score.
  • Ghost Class: Represents a Ghost agent, manages its position and the character it's currently hiding.
  • PacmanGame Class: Orchestrates the game flow, initializes agents, runs the game loop, and handles game state updates.
  • AI Functions:
    • heuristic(): Manhattan distance heuristic used by A*.
    • a_star_search(): Implements A* pathfinding for ghosts.
    • evaluate_state(): Heuristic function to evaluate the "goodness" of a game state for Pac-Man.
    • simulate_move(): Creates a hypothetical future game state after a move.
    • minimax(): Implements the Minimax algorithm with Alpha-Beta pruning.
    • get_best_pacman_action(): Top-level function to invoke Minimax and select Pac-Man's next move.
  • Constants: Define scoring values, penalties, and character representations.

Future Improvements

• Implement Power Pellet logic (scare ghosts, allow eating ghosts).
• More sophisticated Ghost AI (e.g., Scatter/Chase modes, targeting specific areas, considering other ghosts).
• Graphical interface using libraries like Pygame or Tkinter.
• Different map layouts.
• Implement different AI search algorithms for comparison (e.g., Expectimax for handling probabilistic ghost moves).
• Add sound effects.

License

MIT License