ma-gym provides a collection of multi-agent reinforcement learning (MARL) environments built upon the OpenAI Gym framework. It targets researchers and practitioners in MARL, offering a standardized interface for developing and testing multi-agent algorithms, with the benefit of easy integration with existing RL libraries.

How It Works

The library extends the OpenAI Gym API to support multiple agents, each with its own observation and action space. Environments are designed to simulate cooperative, competitive, or mixed-motive scenarios, facilitating the study of emergent behaviors and coordination strategies in decentralized systems.

Quick Start & Requirements

• Install: pip install ma-gym
• Source Install: git clone https://github.com/koulanurag/ma-gym.git && cd ma-gym && pip install -e .
• Dependencies: Requires specific older versions of pip, setuptools, and wheel for installation.
• Usage: See the provided Python snippet for basic environment interaction.
• More Details: Refer to the Wiki for comprehensive usage instructions.

Highlighted Details

• Includes environments like Checkers, Combat, PredatorPrey, Pong Duel, Switch, Lumberjacks, and TrafficJunction.
• Supports multi-agent wrappers for existing single-agent OpenAI Gym environments (e.g., ma_CartPole-v0).
• Can be used with minimal-marl for warm-starting agent training.
• Provides a citation in BibTeX format.

Maintenance & Community

• The project was initially developed during a research internship in Summer 2019.
• No explicit community channels (Discord/Slack) or roadmap are mentioned in the README.

Licensing & Compatibility

• The README does not explicitly state a license. However, the project structure and common practice suggest it is likely open-source. Further clarification on licensing is recommended for commercial use.

Limitations & Caveats

• The installation requires specific, older versions of pip, setuptools, and wheel, which may conflict with current package management best practices.
• Community support and maintenance activity are not clearly indicated.