Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Agile Continuous Jumping in Discontinuous Terrains

Yuxiang Yang, Guanya Shi, Changyi Lin, Xiangyun Meng, Rosario Scalise, Mateo Guaman Castro, Wenhao Yu, Tingnan Zhang, Ding Zhao, Jie Tan, Byron Boots

TL;DR

A hierarchical learning and control framework is designed, which consists of a learned heightmap predictor for robust terrain perception, a reinforcement-learning-based centroidal-level motion policy for versatile and terrain-adaptive planning, and a low-level model-based leg controller for accurate motion tracking.

Abstract

We focus on agile, continuous, and terrain-adaptive jumping of quadrupedal robots in discontinuous terrains such as stairs and stepping stones. Unlike single-step jumping, continuous jumping requires accurately executing highly dynamic motions over long horizons, which is challenging for existing approaches. To accomplish this task, we design a hierarchical learning and control framework, which consists of a learned heightmap predictor for robust terrain perception, a reinforcement-learning-based centroidal-level motion policy for versatile and terrain-adaptive planning, and a low-level model-based leg controller for accurate motion tracking. In addition, we minimize the sim-to-real gap by accurately modeling the hardware characteristics. Our framework enables a Unitree Go1 robot to perform agile and continuous jumps on human-sized stairs and sparse stepping stones, for the first time to the best of our knowledge. In particular, the robot can cross two stair steps in each jump and completes a 3.5m long, 2.8m high, 14-step staircase in 4.5 seconds. Moreover, the same policy outperforms baselines in various other parkour tasks, such as jumping over single horizontal or vertical discontinuities. Experiment videos can be found at https://yxyang.github.io/jumping_cod/

Agile Continuous Jumping in Discontinuous Terrains

TL;DR

A hierarchical learning and control framework is designed, which consists of a learned heightmap predictor for robust terrain perception, a reinforcement-learning-based centroidal-level motion policy for versatile and terrain-adaptive planning, and a low-level model-based leg controller for accurate motion tracking.

Abstract

We focus on agile, continuous, and terrain-adaptive jumping of quadrupedal robots in discontinuous terrains such as stairs and stepping stones. Unlike single-step jumping, continuous jumping requires accurately executing highly dynamic motions over long horizons, which is challenging for existing approaches. To accomplish this task, we design a hierarchical learning and control framework, which consists of a learned heightmap predictor for robust terrain perception, a reinforcement-learning-based centroidal-level motion policy for versatile and terrain-adaptive planning, and a low-level model-based leg controller for accurate motion tracking. In addition, we minimize the sim-to-real gap by accurately modeling the hardware characteristics. Our framework enables a Unitree Go1 robot to perform agile and continuous jumps on human-sized stairs and sparse stepping stones, for the first time to the best of our knowledge. In particular, the robot can cross two stair steps in each jump and completes a 3.5m long, 2.8m high, 14-step staircase in 4.5 seconds. Moreover, the same policy outperforms baselines in various other parkour tasks, such as jumping over single horizontal or vertical discontinuities. Experiment videos can be found at https://yxyang.github.io/jumping_cod/
Paper Structure (27 sections, 4 equations, 10 figures, 1 table)

This paper contains 27 sections, 4 equations, 10 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. Optimal Control for Dynamic Legged Locomotion
  4. Learning Perceptive Locomotion
  5. Hierarchical Frameworks
  6. Hierarchical Perceptive Jumping Framework
  7. Heightmap-Centered Terrain Perception
  8. Iterative Training for Heightmap Prediction
  9. Robust Perception with Heightmap Randomization
  10. Learning Versatile Motion Planning
  11. Environment Setup
  12. Motion Tracking Cost as Policy Reward
  13. Terrain Curriculum for Skill Training
  14. Accurate Motion Tracking with Feedback
  15. Gait Generation and Swing Leg Control
  16. ...and 12 more sections

Figures (10)

  • Figure 1: Our framework enables a quadrupedal robot to jump continuously over real-world stairs and steps.
  • Figure 2: Our hierarchical learning-control framework consists of a heightmap predictor, a motion policy, and a low-level leg controller. We use the heightmap as the intermediate representation for perception and motion planning (Section. \ref{['sec:heightmap']}), train high-performance motion planning with reward to encourage accurate tracking (Section. \ref{['sec:motion_policy']}), and combine a feedforward and a feedback controller for robust tracking of body orientations (Section. \ref{['sec:feedback']}). In addition, we reduce the sim-to-real gap by accurately identifying key hardware characteristics and reproducing them in simulation (Section. \ref{['sec:real-to-sim']}).
  • Figure 3: The environment consists of 4 terrain types.
  • Figure 4: We adopt the bounding gait with 4 contact modes. Red dot indicates foot contact.
  • Figure 5: The feedback controller improves the tracking accuracy of body angular velocities.
  • ...and 5 more figures