CoGrid
A multi-agent grid-world research framework with NumPy and JAX backends.
CoGrid is a flexible framework for building multi-agent grid-world environments for reinforcement learning research. It supports both NumPy and JAX backends, enabling hardware-accelerated simulation with JIT compilation. CoGrid is compatible with the PettingZoo API, making it easy to integrate with standard RL libraries.