NEAT-Pacman

NEAT-Pacman is a Pacman agent trained with NEAT (NeuroEvolution of Augmenting Topologies). It learns to move through the maze and eat dots over many generations.

Demo

Trained agent replay:


Fitness history:

Project structure

• src/Pacman.py: game simulation, training, and replay
• src/optimize.py: Optuna hyperparameter tuning
• src/config/neat_config.txt: NEAT settings
• src/outputs/: generated training and replay files
• assets/: images and gif used in this README

Getting started

Requirements:

• Python 3.7+
• Windows or macOS for GIF export with PIL.ImageGrab

Install dependencies:

pip install -r requirements.txt

Usage

Run commands from src/:

cd src

Train an agent:

python Pacman.py

Then select option 1.

Replay the best agent:

python Pacman.py

Then select option 2.

Replay a specific generation:

python Pacman.py

Then select option 3 and enter a generation (example: 042).

Tune hyperparameters:

python optimize.py

Configuration

Main config file:

• src/config/neat_config.txt

Important constants are in:

• src/Pacman.py

References

• https://neat-python.readthedocs.io/en/latest/
• https://optuna.org/
• https://pypi.org/project/freegames/
• https://github.com/grantjenks/free-python-games

License

MIT License: LICENSE