Learning Transferability: A Two-Stage Reinforcement Learning Approach for Enhancing Quadruped Robots' Performance in U-Shaped Stair Climbing

TL;DR

This paper presents a two-stage deep reinforcement learning approach that enables quadruped robots to autonomously climb U-shaped stairs and transfer learned skills across different stair types, improving their versatility in construction scenarios.

Contribution

The study introduces a novel two-stage RL method that enhances transferability of stair-climbing policies for quadruped robots across various stair geometries.

Findings

Successful climbing of U-shaped stairs with stall penalty

Transferability of policies to different stair types and models

Improved autonomous stair navigation for quadruped robots

Abstract

Quadruped robots are employed in various scenarios in building construction. However, autonomous stair climbing across different indoor staircases remains a major challenge for robot dogs to complete building construction tasks. In this project, we employed a two-stage end-to-end deep reinforcement learning (RL) approach to optimize a robot's performance on U-shaped stairs. The training robot-dog modality, Unitree Go2, was first trained to climb stairs on Isaac Lab's pyramid-stair terrain, and then to climb a U-shaped indoor staircase using the learned policies. This project explores end-to-end RL methods that enable robot dogs to autonomously climb stairs. The results showed (1) the successful goal reached for robot dogs climbing U-shaped stairs with a stall penalty, and (2) the transferability from the policy trained on U-shaped stairs to deployment on straight, L-shaped, and spiral stair terrains, and transferability from other stair models to deployment on U-shaped terrain.