Coup RL

Coup RL is a reinforcement learning agent that learns to play the deception-based board game Coup.

Learn the rules of Coup here.

Coup RL consists of several submodules:

• OpenAI Gym Environment: Contains the Coup game logic and the environment to simulate a complete 2-player game.
• Desktop app: Humans can play against an RL agent with this simple app.
• Reinforcement learning agent & algorithms: Contains the structures and algorithms required to train an agent, as well as save and load agents.

See the results of my training and testing in my Senior Design repo.

Requirements

The project can run on any OS with Python 3 installed that also supports OpenAI Gym and PyQt6. Developed and tested on macOS with Python 3.9.

Installation

First create a virtual environment:

$ cd coup-rl
$ virtualenv env
$ source env/bin/activate

Then install the dependencies:

$ pip install -r app/requirements.txt
$ pip install -r gym-coup/requirements.txt

Build and install the gym environment and RL agents:

$ pip install -e ./gym-coup
$ pip install -e ./rl

Usage

Run the Desktop App

$ python app/main.py

For further usage documentation see the individual modules:

• Gym Environment - gym-coup
• Desktop app - app
• Reinforcement learning agent & algorithms - rl

FAQ

Do I need to install everything?

No. If you aren't going to use the desktop app, you don't need to install PyQt6 and its dependencies. You can ignore the command:

$ pip install -r app/requirements.txt