This repository serves as a curated, categorized collection of academic papers on Multi-Agent Reinforcement Learning (MARL). It aims to provide researchers and practitioners with a structured starting point for exploring various MARL subfields, including credit assignment, communication, game theory, and specific environments.

How It Works

The collection is organized by MARL research themes, with papers cross-listed if they fit multiple categories. Each entry typically includes the paper title, a link to its code (if available), and the venue/year of publication. The repository also highlights key MARL environments and foundational papers.

Quick Start & Requirements

This repository is a collection of links and does not require installation or execution. It serves as a reference guide.

Highlighted Details

• Comprehensive categorization of MARL research topics.
• Links to code implementations for many papers.
• Inclusion of prominent MARL environments like StarCraft (SMAC), Football, PettingZoo, and Melting Pot.
• Covers a wide range of MARL algorithms and challenges, from value decomposition to communication and opponent modeling.

Maintenance & Community

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

Licensing & Compatibility

The repository itself is not for commercial purposes. Individual papers retain their original licensing.

Limitations & Caveats

The collection is described as a "first draft" and may be incomplete, with potential for missing papers, categories, or broken links. The author welcomes corrections and additions.