Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Quadrupedal Robot Skateboard Mounting via Reverse Curriculum Learning

Danil Belov, Artem Erkhov, Elizaveta Pestova, Ilya Osokin, Dzmitry Tsetserukou, Pavel Osinenko

TL;DR

This work tackles autonomous skateboard mounting for quadrupedal robots, a problem that existing approaches largely overlook by assuming the robot is already on the board. It adopts a reverse curriculum reinforcement learning framework, trained with PPO in a physics-based simulator, to bootstrap from mounted states and progressively handle less favorable initial conditions, including a moving skateboard. The authors augment a realistic skateboard model within an Isaac-based environment, define a sparse yet informative reward structure, and demonstrate mounting from diverse starting poses in simulation, achieving mounting in about 3 seconds after first contact. The study provides a promising pathway toward a full autonomous skateboarding capability, with future work aimed at integrating mounting and riding, validating on real robots, and enabling steering through body lean.

Abstract

The aim of this work is to enable quadrupedal robots to mount skateboards using Reverse Curriculum Reinforcement Learning. Although prior work has demonstrated skateboarding for quadrupeds that are already positioned on the board, the initial mounting phase still poses a significant challenge. A goal-oriented methodology was adopted, beginning with the terminal phases of the task and progressively increasing the complexity of the problem definition to approximate the desired objective. The learning process was initiated with the skateboard rigidly fixed within the global coordinate frame and the robot positioned directly above it. Through gradual relaxation of these initial conditions, the learned policy demonstrated robustness to variations in skateboard position and orientation, ultimately exhibiting a successful transfer to scenarios involving a mobile skateboard. The code, trained models, and reproducible examples are available at the following link: https://github.com/dancher00/quadruped-skateboard-mounting

Quadrupedal Robot Skateboard Mounting via Reverse Curriculum Learning

TL;DR

This work tackles autonomous skateboard mounting for quadrupedal robots, a problem that existing approaches largely overlook by assuming the robot is already on the board. It adopts a reverse curriculum reinforcement learning framework, trained with PPO in a physics-based simulator, to bootstrap from mounted states and progressively handle less favorable initial conditions, including a moving skateboard. The authors augment a realistic skateboard model within an Isaac-based environment, define a sparse yet informative reward structure, and demonstrate mounting from diverse starting poses in simulation, achieving mounting in about 3 seconds after first contact. The study provides a promising pathway toward a full autonomous skateboarding capability, with future work aimed at integrating mounting and riding, validating on real robots, and enabling steering through body lean.

Abstract

The aim of this work is to enable quadrupedal robots to mount skateboards using Reverse Curriculum Reinforcement Learning. Although prior work has demonstrated skateboarding for quadrupeds that are already positioned on the board, the initial mounting phase still poses a significant challenge. A goal-oriented methodology was adopted, beginning with the terminal phases of the task and progressively increasing the complexity of the problem definition to approximate the desired objective. The learning process was initiated with the skateboard rigidly fixed within the global coordinate frame and the robot positioned directly above it. Through gradual relaxation of these initial conditions, the learned policy demonstrated robustness to variations in skateboard position and orientation, ultimately exhibiting a successful transfer to scenarios involving a mobile skateboard. The code, trained models, and reproducible examples are available at the following link: https://github.com/dancher00/quadruped-skateboard-mounting
Paper Structure (13 sections, 6 equations, 3 figures, 2 tables)

This paper contains 13 sections, 6 equations, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Control Algorithm
  4. Experimental Design
  5. Environment
  6. Mathematical Model of the Skateboard
  7. Observation and Action Space
  8. Reward Function
  9. Baseline Training
  10. Reverse curriculum learning
  11. Implementation Details
  12. Results
  13. Conclusion and Future Work

Figures (3)

  • Figure 1: Unitree A1 quadrupedal robot on a skateboard. Autonomous mounting, i.e. climbing on the board, is a challenging problem, and was not yet considered.
  • Figure 2: Training performance curves across different curriculum stages for the quadruped skateboard mounting task, showing mean episode reward versus learning iterations
  • Figure 3: Sequence of the quadrupedal robot skateboard mounting process. Unitree A1 progresses from (a) initial approach to (f) fully mounted state.