Awesome
<h1 align="center">Learning Smooth Humanoid Locomotion <br> through Lipschitz-Constrained Policies</h1> <p align="center"> <a href="https://zixuan417.github.io/"><strong>Zixuan Chen<sup>*,1</sup></strong></a> | <a href="https://xialin-he.github.io/"><strong>Xialin He<sup>*,2</sup></strong></a> | <a href="https://wangyenjen.github.io/"><strong>Yen-Jen Wang<sup>*,3</sup></strong></a> | <a href="https://qiayuanl.netlify.app/"><strong>Qiayuan Liao<sup>3</sup></strong></a> | <a href="https://yanjieze.com/"><strong>Yanjie Ze<sup>4</sup></strong></a> | <a href="https://zyliatzju.github.io/"><strong>Zhongyu Li<sup>3</sup></strong></a> <br> <a href="https://people.eecs.berkeley.edu/~sastry/"><strong>S. Shankar Sastry<sup>3</sup></strong></a> | <a href="https://jiajunwu.com/"><strong>Jiajun Wu<sup>4</sup></strong></a> | <a href="https://me.berkeley.edu/people/koushil-sreenath/"><strong>Koushil Sreenath<sup>3</sup></strong></a> | <a href="https://saurabhg.web.illinois.edu/"><strong>Saurabh Gupta<sup>2</sup></strong></a> | <a href="https://xbpeng.github.io/"><strong>Xue Bin Peng<sup>1,5</sup></strong></a> <br> <sup>1</sup> Simon Fraser University   <sup>2</sup> UIUC   <sup>3</sup> UC Berkeley   <sup>4</sup> Stanford University   <sup>5</sup> NVIDIA <br> * Equal Contribution </p> <p align="center"> <h3 align="center"><a href="https://lipschitz-constrained-policy.github.io/">Website</a> | <a href="https://arxiv.org/abs/2410.11825">arXiv</a> | <a href="https://youtu.be/GWeI48QK6B8?si=qi0vgf1Yg4v5qbjD">Video</a> </h3> <div align="center"></div> </p> <p align="center"> <img src="./cover.png" width="80%"/> </p>Codebase Overview
This codebase supports simulation training of five different humanoid robots: Fourier GR1T1, Fourier GR1T2, Unitree H1, Berkeley Humanoid, and Unitree G1. The simulation training is based on Isaac Gym. We also provide a sim-to-sim pipeline for these robots. The sim-to-sim is performed in Mujoco. Please find the detailed instructions for simulation training in simulation.
Our codebase currently provides the real-world deployment code of Fourier GR1T2. Please find the detailed instructions for real-world deployment in deployment.
Acknowledgements
- We would like to thank all the authors in this project, this project cannot be finished without your efforts!
- Our simulation environment implementation is based on legged_gym, and the rl algorithm implementation is based on rsl_rl.
- Humanoid-Gym and Expressive-Humanoid also provide lots of insights.
Citation
If you find this work useful, please consider citing:
@article{chen2024lcp,
title={Learning Smooth Humanoid Locomotion through Lipschitz-Constrained Policies},
author={Zixuan Chen and Xialin He and Yen-Jen Wang and Qiayuan Liao and Yanjie Ze and Zhongyu Li and S. Shankar Sastry and Jiajun Wu and Koushil Sreenath and Saurabh Gupta and Xue Bin Peng},
journal={arxiv Preprint arXiv:2410.11825},
year={2024}
}