MAPPO Implementation for Pistonball Environment

This project implements the MAPPO (Multi-Agent Proximal Policy Optimization) algorithm using the SKRL library (https://skrl.readthedocs.io/en/latest/). It is based on the Pistonball environment from PettingZoo (https://pettingzoo.farama.org/environments/butterfly/pistonball/).

Installation

To get started, clone the repository and install the required dependencies: It is recommended to use a virtual environment, such as conda.

git clone <repository_url>
cd <repository_directory>
pip install skrl["torch"]

Train

To train the model, run the following command with your preferred algorithm (either ppo or mappo):

python train.py --alg {ALG_NAME:ppo or mappo}

Play

To test the trained model, run the following command with your preferred algorithm (either ppo or mappo):

python play.py --alg {ALG_NAME:ppo or mappo}