humanoid-gail

Humanoid behavior imitation using Generative Adversarial Imitation Learning (GAIL).

An accompanying research blog post with details and references: https://uscresl.github.io/humanoid-gail/

If you find this code/work useful, consider citing the following:

@misc{DDHJK2017,
  author = {Debnath, Shoubhik and Devos, Arnout and Heiden, Eric and Julian, Ryan and Khatana, Fiona},
  title = {Humanoid Imitation Learning from Diverse Sources},
  year = {2017},
  publisher = {GitHub},
  journal = {GitHub repository},
  howpublished = {\url{https://github.com/uscresl/humanoid-gail}},
  commit = {d27232df8e3fd94948e5f70360c43e098265ac62}
}

Installation

1. Install Docker (We recommend the free CE edition): for Ubuntu, for Mac, for Windows

2. Build the Docker file:

   If your system does not have a NVIDIA® GPU that meets the prerequisites:

   docker build -f Dockerfile -t uscresl/deep-rl-docker:tf1.3.0-gym0.9.3-baselines0.1.4-py3 .
   

   Else:

   docker build -f Dockerfile.gpu -t uscresl/deep-rl-docker:tf1.3.0-gym0.9.3-baselines0.1.4-gpu-py3 .
   

3. Execute

   On macOS in terminal depending on whether you have a GPU version or not:

   sh run.sh
   

   or

   sh run_gpu.sh
   

Style

Please ensure that all your Python code conforms to the PEP8 standard. If your question is not answered by PEP8, please revert to the Google Python Style Guide. Similarly, shell scripts should use the Google Shell Style Guide.