One-shot Adaptation of Humanoid Whole-body Motion with Walking Priors

TL;DR

This paper introduces a data-efficient method for humanoid motion adaptation that learns from a single target sample by leveraging walking priors and optimal transport, enabling realistic whole-body movements.

Contribution

It presents a novel one-shot adaptation technique combining optimal transport and reinforcement learning, reducing data requirements for humanoid motion synthesis.

Findings

Outperforms baseline methods on CMU MoCap dataset

Achieves realistic and collision-free humanoid motions

Effective with only a single target sample

Abstract

Whole-body humanoid motion represents a fundamental challenge in robotics, requiring balance, coordination, and adaptability to enable human-like behaviors. However, existing methods typically require multiple training samples per motion, rendering the collection of high-quality human motion datasets both labor-intensive and costly. To address this, we propose a data-efficient adaptation approach that learns a new humanoid motion from a single non-walking target sample together with auxiliary walking motions and a walking-trained base model. The core idea lies in leveraging order-preserving optimal transport to compute distances between walking and non-walking sequences, followed by interpolation along geodesics to generate new intermediate pose skeletons, which are then optimized for collision-free configurations and retargeted to the humanoid before integration into a simulated environment for policy adaptation via reinforcement learning. Experimental evaluations on the CMU MoCap dataset demonstrate that our method consistently outperforms baselines, achieving superior performance across metrics. Our code is available at: https://github.com/hhuang-code/One-shot-WBM.