DeepParkour

Training a humanoid agent to efficiently apply parkour skills to an obstacle environment.

Dependencies

• requirements.txt includes all the dependecies required to run this project.

Trained Agent:

• You can watch a few of our agents in this video

Installation

• Clone the repo and cd into it:

  git clone https://github.com/aagrawal20/DeepParkour.git
  cd DeepParkour

• If have access to a CUDA-compatible gpu then install tensorflow gpu.

  pip install tensorflow-gpu 

  Refer to TensorFlow installation guide for more details.

• Install DeepParkour package.

  pip install -e .

Training Agent

• You can train an agent using the train_agent.py file.
• You can add specific flags to the argument parser.

  python src/main/train_agent.py

Rendering an Agent

• You can render an agent using the render_agent.py file.
• You can add specific flags to the argument parser.

  python src/main/render_agent.py

  Note: PyBullet only supports CPU rendering. Turn off render flag or manually turn off gpu.

Visualizing Agent training

• You can visualize different stastics for eg: loss vs timesteps or reward vs timesteps.

  jupyter notebook

• After loading the local host navigate to the visualization.ipynb in src/util.