Introduction

This repository provides a simple, distributed and asynchronous multi-agent reinforcement learning (MARL) framework for the Google Research Football environment, along with research tools and results for benchmarking. In particular, it includes:

1. A distributed and asynchronous MARL framework

2. Implementations of algorithm IPPO, MAPPO, HAPPO, A2PO, MAT

3. Ready-to-run experiment configuration

4. Population-based training pipline, such as PSRO

5. Pre-trained GRF policies in both 5-vs-5 and 11-vs-11 full-game scenarios

6. Single-step match replay debugger

7. Tutorial for GRF online ranking

The goal is to provide standard benchmark results on Google Research Football scenarios which has been investigated in many MARL works, as well as to further boost the studies of Multi-agent reinforcement learning on GRF by offering advanced research tools targeting at generally stronger football AI.