marlbenchmark/on-policy
Multi-Agent PPO (MAPPO) implementation for cooperative multi-agent reinforcement learning.
Not currently ranked — collecting fresh signals.
star history
This repository provides the official implementation of MAPPO, a multi-agent variant of Proximal Policy Optimization. It implements the algorithm from the paper ‘The Surprising Effectiveness of PPO in Cooperative Multi-Agent Games’ and is built on a PyTorch A2C-PPO-ACKTR foundation. The implementation supports training across multiple multi-agent environments including StarCraftII (SMAC/SMACv2), Hanabi, Multiagent Particle-World Environments, and Google Research Football.
Frequently asked
- What is marlbenchmark/on-policy?
- Multi-Agent PPO (MAPPO) implementation for cooperative multi-agent reinforcement learning.
- Is on-policy open source?
- Yes — marlbenchmark/on-policy is open source, released under the MIT license.
- What language is on-policy written in?
- marlbenchmark/on-policy is primarily written in Python.
- How popular is on-policy?
- marlbenchmark/on-policy has 2k stars on GitHub.
- Where can I find on-policy?
- marlbenchmark/on-policy is on GitHub at https://github.com/marlbenchmark/on-policy.