Skip to content

Latest commit

 

History

History
106 lines (72 loc) · 5.35 KB

README.md

File metadata and controls

106 lines (72 loc) · 5.35 KB
Raw

AI-Powered Snake Game with Search Algorithm Visualizations

A Python-based implementation of the classic Snake game, featuring AI agents powered by various search algorithms and interactive visualizations of their exploration process.

Implementation Details

The AI agent can be controlled using different search algorithms, each with its own characteristics:

AI Play Algorithms:

  • Online Search (Reinforcement Learning with Danger/Success Memory): This agent learns by experience, remembering dangerous locations and successful paths. It uses a custom heuristic incorporating learned danger to guide its decisions, dynamically adapting to the game environment. This exemplifies a simple reinforcement learning approach.
  • A Search (with Dynamic Difficulty Adjustment):* Employs the A* search algorithm with Manhattan distance as the heuristic. To manage search complexity and adapt to the game's dynamics, it incorporates a depth-first search (DFS) component. Parameters like DFS depth and a randomization factor are adjusted based on the snake's performance (e.g., increasing DFS depth if the snake consistently performs well).
  • Breadth-First Search (BFS): A systematic algorithm that guarantees finding the shortest path to the food but can be computationally expensive in larger or complex game states.
  • Depth-First Search (DFS): Explores a single branch as deeply as possible before backtracking. While potentially faster than BFS, it doesn't guarantee finding the shortest path.
  • Greedy Best-First Search: A heuristic-driven approach that prioritizes the move seemingly closest to the food. While efficient, it might miss optimal solutions due to its lack of broader exploration.
  • Hamiltonian Cycle (Demonstration): This agent showcases a deterministic approach by following a pre-calculated Hamiltonian cycle, a path that visits every grid cell exactly once. This strategy provides a safe but not always the most efficient way to reach the food.

AI Vision Algorithms (Visualization):

These algorithms focus on visualizing the search process itself, providing insights into how different strategies explore the game board to find a path to the food:

  • A Search:* Visualizes A* using a combined heuristic considering distance to food, distance to walls (avoiding collisions), and available space to provide a more comprehensive evaluation of possible moves.
  • Breadth-First Search (BFS): Visualizes the layer-by-layer exploration of BFS.
  • Depth-First Search (DFS): Visualizes the deep exploration and backtracking behavior of DFS.
  • Greedy Best-First Search: Visualizes the heuristic-driven, locally optimal choices of Greedy Best-First Search.
  • Online Search: Visualizes the real-time learning process of the online search agent as it explores, encounters dangers, and refines its internal map of the environment.

Technical Details:

  • Programming Language: Python
  • Libraries: Pygame (for game logic and rendering), NumPy (for numerical operations), Matplotlib (for visualizations) , Pyinstrument (used for optional code profiling during development.)
  • Game State Representation: Utilizes a grid-based representation of the game board to facilitate search algorithms.

How to Run

  1. Prerequisites: Python 3.7+ (recommended to use a virtual environment)

  2. Installation:

    git clone https://github.com/Ta-Pc/Snake-Game.git 
cd ai-snake-game
python3 -m venv .venv
source .venv/bin/activate  # Activate virtual environment (Linux/macOS)
.venv\Scripts\activate      # Activate virtual environment (Windows)
pip install -r requirements.txt

  3. Running the Game:

    python main.py

  4. Running the Visualization:

    python visualization.py

Controls

  • Arrow Keys: Control the snake's direction (in human play mode).
  • SPACE: Pause/Resume the game.
  • ESC: Return to the main menu. (Exits the visualiser)
  • Visualizer Mouse Interactions:
    • Left click + drag: Rotate the 3D view.
    • Right click + drag: Zoom in/out.
    • Middle click + drag: Pan the view.
    • Space: Pause/Resume the simulation.
    • R: Reset the view to the default angle.

Features

  • AI agents using various search algorithms.
  • Real-time visualization of AI search processes.
  • Human playable mode.
  • Multiple difficulty levels corresponding to AI speed.
  • Big food for bonus points and AI triggering events.
  • Pause/resume functionality.
  • High score tracking.
  • Main menu for selecting game modes and options.

Future Enhancements

  • Implementing more advanced AI techniques (e.g., Q-learning, Deep Reinforcement Learning).
  • Adding more sophisticated game elements and levels.
  • Improving the user interface and visualization tools.
  • Incorporating sound effects and music.

Contributing

Contributions are welcome! Feel free to submit pull requests for bug fixes, new features, or improved documentation.

License

This project is licensed under the Apache License, Version 2.0. See the LICENSE file for details.

You may obtain a copy of the License at:

License

Code Style: black