Semi-Structured Reinforcement Learning (SSRL)

Code for our paper "Learning to Walk from Three Minutes of Real-World Data with Semi-structured Dynamics Models".

@inproceedings{levy2024learning,
  title={Learning to Walk from Three Minutes of Real-World Data with Semi-structured Dynamics Models},
  author={Levy, Jacob and Westenbroek, Tyler and Fridovich-Keil, David},
  booktitle={8th Annual Conference on Robot Learning},
  year={2024}
}

• For a brief video overview of the project, see our website.

Instructions

• To run training in a simulated Go1 environment, see ssrl.
• For code to interface with the Go1 robot and perform real-world experiments, see ssrl_hardware.