Multi-agent Reinforcement Learning

WORK IN PROGRESS

What's Inside

- MADDPG

• Implementation of algorithm presented in OpenAI's publication "Multi-Agent Actor-Critic for Mixed Cooperative-Competitive Environments" (Lowe et al., https://arxiv.org/pdf/1706.02275.pdf)
• Does not include "Inferring policies of other agents" and "policy ensembles"
• multiagent-particle-environment: follow install instructions in https://github.com/openai/multiagent-particle-envs
• Tested on Spread environment.
• TODO:
  • Test on competitive tasks

Running the code & Dependencies

• Python 3.6
• PyTorch 0.4.1
• multiagent-particle-environment: follow install instructions in https://github.com/openai/multiagent-particle-envs