This repository serves as a curated collection of Multi-Agent Reinforcement Learning (MARL) papers, categorizing key research by methodology and application, and linking to their associated code implementations. It is intended for researchers and practitioners in the MARL field seeking to explore foundational and recent advancements, providing a structured overview of the landscape.

How It Works

The repository organizes MARL papers into categories such as Independent Learning, Value Decomposition, Policy Gradient, and Communication, alongside a comprehensive list of MARL environments like StarCraft, Football, and PettingZoo. Each entry includes a link to the paper and, where available, a direct link to its open-source code implementation, facilitating practical exploration and reproduction of research.

Quick Start & Requirements

• Installation: No direct installation is required; this is a curated list of resources.
• Prerequisites: Access to the internet to view papers and code repositories. Some code repositories may have specific Python, deep learning framework (e.g., PyTorch, TensorFlow), or environment dependencies.
• Resources: Links to official documentation, demos, and related collections are provided within the README.

Highlighted Details

• Comprehensive categorization of MARL algorithms and environments.
• Links to code implementations for numerous seminal and recent MARL papers.
• Includes a dedicated section for MARL environments, crucial for benchmarking.
• Provides a citation for the repository itself.

Maintenance & Community

The repository is actively maintained by the author, Hao Chen, with an invitation for community suggestions and contributions to fill gaps in coverage. Contact information is provided via email.

Licensing & Compatibility

The repository itself is not licensed for commercial use. Individual code repositories linked within the collection will have their own licenses, which must be consulted for compatibility with commercial or closed-source projects.

Limitations & Caveats

The repository is described as a "first draft" with ongoing updates planned, acknowledging that important papers and categories may be missing. Some provided links may be invalid.