SLR: Learning Quadruped Locomotion without Privileged Information

Shiyi Chen1, Zeyu Wan1, Shiyang Yan1 Chun Zhang†,1, Weiyi Zhang1,
Qiang Li†,2, Debing Zhang1, Fasih Ud Din Farrukh1
1Tsinghua University, 2Shenzhen Technology University
Paper Code (Coming Soon) arXiv

Without relying on any privileged or visual information, we can train a robust locomotion policy. In real-world deployment, the quadruped robot is capable of traversing various challenging terrains.

Abstract

Traditional reinforcement learning control for quadruped robots often relies on privileged information, demanding meticulous selection and precise estimation, thereby imposing constraints on the development process. This work proposes a Self-learning Latent Representation (SLR) method, which achieves high-performance control policy learning without the need for privileged information. To enhance the credibility of our proposed method's evaluation, SLR is compared with open-source code repositories of state-of-the-art algorithms, retaining the original authors' configuration parameters. Across four repositories, SLR consistently outperforms the reference results. Ultimately, the trained policy and encoder empower the quadruped robot to navigate steps, climb stairs, ascend rocks, and traverse various challenging terrains.

Training Framework

The SLR training framework leverages the Markov Decision Process, guiding the latent's self-learning based on state transitions (transition model), state distinctions (random sampling), and cumulative rewards (critic), without relying on manually set privileged information constraints.

Ascend and descend mountain

Climb long stairs smoothly

Apply anti-disturbance to the robot

Navigate over challenging rocks

Tunnel through vegetation

BibTeX


        @misc{chen2024slr,
          title={SLR: Learning Quadruped Locomotion without Privileged Information}, 
          author={Shiyi Chen and Zeyu Wan and Shiyang Yan and Chun Zhang and Weiyi Zhang and Qiang Li and Debing Zhang and Fasih Ud Din Farrukh},
          year={2024},
          eprint={2406.04835},
          archivePrefix={arXiv},
          primaryClass={cs.RO}
    }