Multi-Agent Generative Adversarial Imitation Learning

Source code for our paper: Multi-Agent Generative Adversarial Imitation Learning

By Jiaming Song, Hongyu Ren, Dorsa Sadigh, Stefano Ermon

Running the Code

• For code implementing MAGAIL, please visit multiagent-gail folder.
• For the OpenAI particle environment code, please visit multiagent-particle-envs folder.

How to run

Run Multi-Agent ACKTR:

python -m sandbox.mack.run_simple

Run MAGAIL with Multi-Agent ACKTR:

python -m irl.mack.run_mack_gail [discrete]

Render results:

python -m irl.render

Citation

If you find this code useful, please consider citing our paper:

@article{song2018multi,
  title={Multi-agent generative adversarial imitation learning},
  author={Song, Jiaming and Ren, Hongyu and Sadigh, Dorsa and Ermon, Stefano},
  year={2018}
}