cogrid 0.3.1

A library for creating multi-agent grid-world environments for reinforcement learning.

CoGrid is a library for creating multi-agent grid-world environments for reinforcement learning research. It features a functional array-based simulation core, pluggable components (rewards, features, objects), and dual NumPy/JAX backend support.

CoGrid utilizes the parallel PettingZoo API to standardize the multi-agent environment interface. The JAX API is similar to that of JaxMARL.

CoGrid is designed to offer an approachable API for environment customization, compatibility with standard tooling, and pre-build benchmark environments. Full documentation is available at cogrid.readthedocs.io.

Installation

Install from PyPI:

pip install cogrid

To install with JAX backend support:

pip install cogrid[jax]

For development (includes test, lint, and docs tools):

pip install cogrid[dev]

[!IMPORTANT] CoGrid has gone through a major overhaul and the API has changed significantly. If you need the previous version, you can install it with pip install cogrid==0.0.16.

Citation

If you use CoGrid in your research, please cite the following paper:

@article{mcdonald2026cogrid,
  title={CoGrid \& the Multi-User Gymnasium: A Framework for Multi-Agent Experimentation},
  author={McDonald, Chase and Gonzalez, Cleotilde},
  journal={arXiv preprint arXiv:2604.15044},
  year={2026}
}

Acknowledgements

This work builds on the invaluable efforts of many others:

@article{carroll2019utility,
  title={On the utility of learning about humans for human-ai coordination},
  author={Carroll, Micah and Shah, Rohin and Ho, Mark K and Griffiths, Tom and Seshia, Sanjit and Abbeel, Pieter and Dragan, Anca},
  journal={Advances in neural information processing systems},
  volume={32},
  year={2019}
}

@article{rutherford2024jaxmarl,
  title={Jaxmarl: Multi-agent rl environments and algorithms in jax},
  author={Rutherford, Alexander and Ellis, Benjamin and Gallici, Matteo and Cook, Jonathan and Lupu, Andrei and Ingvarsson, Gar{\dh}ar and Willi, Timon and Hammond, Ravi and Khan, Akbir and de Witt, Christian S and others},
  journal={Advances in Neural Information Processing Systems},
  volume={37},
  pages={50925--50951},
  year={2024}
}


@article{gessler2025overcookedv2,
  title={Overcookedv2: Rethinking overcooked for zero-shot coordination},
  author={Gessler, Tobias and Dizdarevic, Tin and Calinescu, Ani and Ellis, Benjamin and Lupu, Andrei and Foerster, Jakob Nicolaus},
  journal={arXiv preprint arXiv:2503.17821},
  year={2025}
}