Learning Dexterous Manipulation Using Contact Wrench Guidance From Human Demonstration
Learning Dexterous Manipulation Using Contact Wrench Guidance From Human Demonstration
This paper presents CHORD, a framework for long-horizon dexterous manipulation of rigid and articulated objects using reinforcement learning with object-centric contact wrench guidance. The method represents human and robot motions via forces and torques on the object to measure similarity by induced instantaneous motions, improving scalability for contact-rich tasks. A large-scale simulation benchmark comprising 4,739 bimanual dexterous manipulation tasks was constructed from motion-capture datasets and in-house videos; evaluation on 1,831 tasks shows an average success rate of 82.12%. CHORD generalizes to whole-body manipulation from hand-only and third-person demonstrations with a 90.77% success rate and transfers to real-world open-loop and closed-loop settings.
The authors introduce CHORD, a reinforcement learning framework for dexterous manipulation that uses object-centric contact wrench guidance derived from human demonstrations, evaluated on a large-scale simulation benchmark of 4,739 tasks with reported success rates of 82.12% and 90.77% under specified conditions.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
Reported success rate
82.12%
Reported trials
View original evidence
ation tasks, constructed from motion-capture datasets and reconstructed in-house videos. Evaluated on 1,831 benchmark tasks, CHORD achieves an average success rate of 82.12%, demonstrating strong scalability. CHORD also generalizes to whole-body manipulation from hand-only and third-person demonstrations, achieving a 90.77% success rate, and the learned policies transfer to the real world in both open-loop and closed
Open source S4Reported success rate
90.77%
Reported trials
View original evidence
ation tasks, constructed from motion-capture datasets and reconstructed in-house videos. Evaluated on 1,831 benchmark tasks, CHORD achieves an average success rate of 82.12%, demonstrating strong scalability. CHORD also generalizes to whole-body manipulation from hand-only and third-person demonstrations, achieving a 90.77% success rate, and the learned policies transfer to the real world in both open-loop and closed
Open source S4
- paper: not_reported
What remains unknown
Not established in the collected evidence: Environment, Control, Data origin.
Reported performance applies to the described task. It does not establish general autonomy or deployment readiness.
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
ation tasks, constructed from motion-capture datasets and reconstructed in-house videos. Evaluated on 1,831 benchmark tasks, CHORD achieves an average success rate of 82.12%, demonstrating strong scalability. CHORD also generalizes to whole-body manipulation from hand-only and third-person demonstrations, achieving a 90.77% success rate, and the learned policies transfer to the real world in both open-loop and closed
Open source S4
The key idea is object-centric contact wrench space guidance: we represent human and robot motions by the forces and torques they can induce on the object, enabling similarity to be measured by the induced instantaneous motions. This guidance makes reinforcement learning more scalable for contact-rich dexterous manipulation. We further introduce a large-scale simulation benchmark with 4,739 bimanual dexterous manipul
Open source S3
Learning Dexterous Manipulation Using Contact Wrench Guidance From Human Demonstration
Open source S1
arXiv:2607.00033v3 Announce Type: replace Abstract: Dexterous robot manipulation can benefit from the abundance of human demonstrations, but transferring such demonstrations to robot policies remains challenging. We present Contact Wrench Guidance from Human Demonstration in Robotic Dexterous Manipulation (CHORD), a framework for long-horizon manipulation of rigid and articulated objects with reinforcement learning.
Open source S2
Source：arXiv Robotics — research abstracts · arxiv.org