Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Robust Robot Walker: Learning Agile Locomotion over Tiny Traps

Shaoting Zhu, Runhan Huang, Linzhan Mou, Hang Zhao

TL;DR

A novel approach that enables quadruped robots to pass various small obstacles, or “tiny traps”, by focusing solely on proprioceptive inputs and introducing a two-stage training framework incorporating a contact encoder and a classification head to learn implicit representations of different traps.

Abstract

Quadruped robots must exhibit robust walking capabilities in practical applications. In this work, we propose a novel approach that enables quadruped robots to pass various small obstacles, or "tiny traps". Existing methods often rely on exteroceptive sensors, which can be unreliable for detecting such tiny traps. To overcome this limitation, our approach focuses solely on proprioceptive inputs. We introduce a two-stage training framework incorporating a contact encoder and a classification head to learn implicit representations of different traps. Additionally, we design a set of tailored reward functions to improve both the stability of training and the ease of deployment for goal-tracking tasks. To benefit further research, we design a new benchmark for tiny trap task. Extensive experiments in both simulation and real-world settings demonstrate the effectiveness and robustness of our method. Project Page: https://robust-robot-walker.github.io/

Robust Robot Walker: Learning Agile Locomotion over Tiny Traps

TL;DR

A novel approach that enables quadruped robots to pass various small obstacles, or “tiny traps”, by focusing solely on proprioceptive inputs and introducing a two-stage training framework incorporating a contact encoder and a classification head to learn implicit representations of different traps.

Abstract

Quadruped robots must exhibit robust walking capabilities in practical applications. In this work, we propose a novel approach that enables quadruped robots to pass various small obstacles, or "tiny traps". Existing methods often rely on exteroceptive sensors, which can be unreliable for detecting such tiny traps. To overcome this limitation, our approach focuses solely on proprioceptive inputs. We introduce a two-stage training framework incorporating a contact encoder and a classification head to learn implicit representations of different traps. Additionally, we design a set of tailored reward functions to improve both the stability of training and the ease of deployment for goal-tracking tasks. To benefit further research, we design a new benchmark for tiny trap task. Extensive experiments in both simulation and real-world settings demonstrate the effectiveness and robustness of our method. Project Page: https://robust-robot-walker.github.io/
Paper Structure (23 sections, 16 equations, 17 figures, 8 tables)

This paper contains 23 sections, 16 equations, 17 figures, 8 tables.

Table of Contents

  1. INTRODUCTION
  2. RELATED WORKS
  3. Learning-Based Locomotion
  4. Collision Detection
  5. METHOD
  6. Task Definition
  7. Reinforcement Learning Setting
  8. Reward Function
  9. Training and Deployment
  10. EXPERIMENTS
  11. Experiment Setup
  12. Tiny Trap Benchmark
  13. Simulation Experiments
  14. Real-world Experiments
  15. Additional Experiments
  16. ...and 8 more sections

Figures (17)

  • Figure 1: Our Robust Robot Walker is passing through various challenging "tiny traps" including Pit, Bar, and Pole solely relying on its proprioception.
  • Figure 2: Camera unreliability in fine-grained trap scenarios.
  • Figure 3: The training and deployment overview of Robust Robot Walker. Our method achieves explicit-implicit dual-state estimation and approximate omnidirectional movement. Each color on the quadruped robot corresponds to a type of joint link: Orange-Base, Yellow-Hip, Blue-Thigh, Purple-Calf, and Foot (Unseen).
  • Figure 4: Three categories of tiny traps: Bar, Pit, and Pole.
  • Figure 5: Tiny Trap Benchmark. Robots begin on the left side and must pass through tiny traps to reach the goal on the right side.
  • ...and 12 more figures