ZICQ
中 Log in / Sign up
Newsroom Agentic #Hugging Face #Robotics #Human-Robot Interaction #Optimal Transport #Reinforcement Learning

OTRetarget Framework Released: Enabling Joint Human-Robot Motion Retargeting and Optimization

Avatar of Mr.Xu

By Mr.Xu Compiled & Reviewed by Editorial

Published: · 7 views

中文阅读 (Chinese) English Version
Co-reading Select any sentence in the article to highlight, annotate, or copy a quote card. Join discussion ↓

Summary:Hugging Face has introduced OTRetarget, a novel approach for joint retargeting of human and multi-object motion to humanoid robots. This method leverages entropic optimal transport to transfer surface interactions across human, robot, and object geometries while preserving contact consistency. Evaluations on the OMOMO dataset show that OTRetarget achieves a robot-object interaction Jaccard score of 87% and a depth error of 8.7 mm, outperforming the baseline OmniRetarget. The approach has been su


Background and Challenges

Transferring human motion to humanoid robots while maintaining interaction consistency with the environment is a significant challenge. Traditional methods struggle to balance motion style and contact preservation, especially in complex tasks involving ground contacts and object manipulation.

The OTRetarget Approach

OTRetarget addresses these challenges through the following steps:

  1. Surface Interaction Representation: Utilizes signed distances, closest surface points, and relative directions to describe surface interactions.
  2. Entropic Optimal Transport: Employs entropic optimal transport to transfer interaction targets across human, robot, and object geometries.
  3. Constrained Inverse Kinematics: Integrates interaction targets into a constrained inverse kinematics formulation, jointly optimizing robot and object poses at each frame to balance contact preservation and motion style.

Experiments and Results

Evaluations on the OMOMO dataset demonstrate that OTRetarget outperforms OmniRetarget in terms of robot-object interaction Jaccard score (87% vs. 28%) and depth error (8.7 mm vs. 29.3 mm). Additionally, the method has been successfully tested on a physical G1 humanoid robot, enabling complex motions such as two-handed box pick-and-place.

Industry Impact and Future Directions

OTRetarget offers a new solution for complex human-robot interaction tasks, particularly in scenarios requiring high-precision robot manipulation and object interaction. In the future, this method could be extended to more types of robotic platforms and further enhanced with reinforcement learning to improve adaptability and robustness.

Developer Recommendations

  • Application Scenarios: Ideal for industrial automation and service robotics where high-precision human-robot collaboration is required.
  • Technical Integration: Consider combining with reinforcement learning or imitation learning to further enhance motion retargeting.
  • Data Requirements: High-quality human motion and robot morphology data are crucial for applying this method.

Source: Hugging Face Daily Papers (2026-09-29)

— END —

Tags: #Hugging Face #Robotics #Human-Robot Interaction #Optimal Transport #Reinforcement Learning

Editorial & Fact-Checking Note: This article is compiled from primary research, official release documentation, and source papers by the ZICQ Newsroom pipeline with automated entity verification and human editorial review. If you notice any technical inaccuracy, please submit a correction via our corrections policy or email our editorial desk directly.

What's your take on this story?
🚀 Explore & Submit Works

Community Comments

Loading live comments and annotations…