MAProj

This is a multi-agent project (commnet https://arxiv.org/pdf/1605.07736.pdf, bicnet https://arxiv.org/abs/1703.10069, maddpg https://arxiv.org/abs/1706.02275) in pytorch for the environment of Multi-Agent Particle Environment "simple_spread"(https://github.com/openai/multiagent-particle-envs)

INFERENCE:

• https://github.com/xuemei-ye/maddpg-mpe
• https://github.com/bic4907/BiCNet
• https://github.com/0b01/CommNet
• https://github.com/starry-sky6688/StarCraft

Commnet:

Bicnet:

Maddpg:

Training curves:

How to use

• pip install -r requirements.txt
• cd MAProj/algo
• python ma_main.py --algo maddpg --mode train

To do list

• trained in more maps
• fix graphics memory leak

Blog link

https://zhuanlan.zhihu.com/p/143776727