Checkers-RL

A reinforcement learning agent capable of solving checkers.

Table of contents

Section	Description
Project overview	General description of the reinforcement learning checkers project
Author	Main contributors information
Documentation	Links to detailed documentation and presentation materials
How to use	Instructions for installation and command-line usage
Contributions	How to contribute to the repository
References	Scientific references
License	Licensing information

Project overview

This repository proposes a reinforcement learning-based approach to train an agent capable of playing checkers. The goal is to develop a model that can adapt to the game complexity by using some advanced reinforcement learning algorithms. The project provides tools to train, evaluate and visualize the agent performance, as well as to allow human to play against the trained agent [1].

Author

@Mowibox Ousmane THIONGANE

Documentation

The environment details are specified in the documentation wiki. A short presentation of the RL approaches used is also available in the 'docs/' folder.

How to use

Download the repository:

git clone https://github.com/Mowibox/CheckersRL.git

Download the necessary packages:

pip install -r requirements.txt

Run inside the repository:

python3 main.py [options]

    usage: main.py [-h] [--render] [-t TRAIN] [-e] [--human [HUMAN]]

    options:
        -h, --help            show this help message and exit
        --render              Enable rendering
        -t TRAIN, --train TRAIN
                                Train the RL model
        -e EVALUATE, --evaluate EVALUATE
                                Evaluate the provided RL model (Use 'random'/'mcts'/'model
                                filepath')
        --human [HUMAN]       Allows human to play against computer [w, b] (default: w)

Command-line examples

Train a TD(λ) Linear Value Function Approximation model named model.pkl:

python main.py --train model.pkl

Evaluate a random agent:

python main.py --evaluate random

Evaluate a TD(λ) LVFA model:

python main.py --evaluate model.pkl

Evaluate a Monte-Carlo Tree Search (MCTS) agent:

python main.py --evaluate mcts

See the evaluation episode:

python main.py --evaluate model.pkl --render

Play against the agent (white pawns by default):

python main.py --evaluate model.pkl --human

Contributions

Contributions are always welcome!

• Report Issues: Found a bug or have a feature request? Create a new issue here.
• Fix Bugs & Add Features: Find out where you can lend a hand by checking out existing issues.

References

• [1] Neto, H.C., Julia, R.M.S., Caexeta, G.S. et al. LS-VisionDraughts: improving the performance of an agent for checkers by integrating computational intelligence, reinforcement learning and a powerful search method. Appl Intell 41, 525–550 (2014). https://doi.org/10.1007/s10489-014-0536-y

License

This project is licensed under the MIT License. See the LICENSE file for more details.