Offline Adaptation of Quadruped Locomotion using Diffusion Models

TL;DR

This paper introduces a novel diffusion-based framework for quadruped locomotion that enables offline skill adaptation and goal-conditioned behavior extraction directly on robot hardware, improving flexibility and efficiency.

Contribution

It is the first to apply classifier-free guided diffusion to quadruped locomotion, allowing offline skill adaptation and behavior extraction from unlabelled data with minimal computational overhead.

Findings

Successful hardware experiments on the ANYmal robot.

Effective extraction of goal-conditioned behaviors from unlabelled datasets.

Compatibility with multi-skill policies and low computational requirements.

Abstract

We present a diffusion-based approach to quadrupedal locomotion that simultaneously addresses the limitations of learning and interpolating between multiple skills and of (modes) offline adapting to new locomotion behaviours after training. This is the first framework to apply classifier-free guided diffusion to quadruped locomotion and demonstrate its efficacy by extracting goal-conditioned behaviour from an originally unlabelled dataset. We show that these capabilities are compatible with a multi-skill policy and can be applied with little modification and minimal compute overhead, i.e., running entirely on the robots onboard CPU. We verify the validity of our approach with hardware experiments on the ANYmal quadruped platform.