SwarmRL is a Python package aimed at providing a simple and flexible framework for developing and testing reinforcement learning and mathematical optimization algorithms for multi-agent systems. Some key features of SwarmRL include:
- Integration with EspressoMD simulation engine
- Support for experimental setups
- Actor-critic reinforcement learning strategies
Currently, SwamrRL is only available from source so it must be installed within the local directory.
git clone https://github.com/SwarmRL/SwarmRL.git
cd SwarmRL
pip install .
Our documentation is a work in progress but can be found here. If you have questions about the code or find any problems, please create an issue so we can work on it as soon as possible. If you're feeling adventurous, you can check out our custom-built Swarm GPT, here which has been conditioned on the SwarmRL repository and will be updated as more resources become available. Be careful though! It isn't perfect but not a bad place to start for general principles of reinforcement learning and pieces of the software.