cogrid 0.2.0

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

CoGrid

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.

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]

Citation

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

@article{mcdonald2024cogrid,
  author  = {McDonald, Chase and Gonzalez, Cleotilde},
  title   = {CoGrid and Interactive Gym: A Framework for Multi-Agent Experimentation},
  year    = {forthcoming},
}