Human2Humanoid: Physics-Aware Cross-Morphology Motion Retargeting for Humanoid Robots
Tianchen Huang, Feiyang Yuan, Junchi Gu, Shurui Fang, Xiaohu Zhang, Yu Wang, Wei Gao, Shiwu Zhang · 2026
Read Human2Humanoid: Physics-Aware Cross-Morphology Motion Retargeting for Humanoid Robots by Tianchen Huang, Feiyang Yuan, Junchi Gu, Shurui Fang, Xiaohu Zhang, Yu Wang, Wei Gao, Shiwu Zhang (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2606-03476.txt.
Description
Huang and co-authors transfer human motion onto humanoid robots without paired training data. Retargeting matters for teleoperation, imitation learning, and human-robot contact. Bodies differ in skeleton shape, limb proportions, and joint counts. Paired examples of the same motion on both bodies are scarce. Their framework, Human2Humanoid, works in that unpaired setting.
A CycleGAN design carries the transfer between domains. A skeleton-aware graph network reads features that depend on body topology. A special loss aligns end-effector paths after scale is factored out. That keeps the sense of a motion across very different bodies. Physics-based limits push the output to copy the source's contact patterns. Tests send human motion to a Unitree G1 humanoid robot. The method beats prior work on later controllability and on physical feasibility. Robotics teams get a path to reuse human motion libraries at scale.
Catalog record
| CREATOR | Tianchen Huang, Feiyang Yuan, Junchi Gu, Shurui Fang, Xiaohu Zhang, Yu Wang, Wei Gao, Shiwu Zhang |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 38 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2606.03476