CoDance: Learning Reactive and Compliant Human-Humanoid Interaction from Video
CoDance learns compliant two-hand humanoid dancing from one partnered-dance video, including on a physical robot.
CoDance learns reactive, compliant partnered dancing for a humanoid from a single video of two human dancers, retargeting their motion into a robot reference and a moving partner. A multi-link compliance augmentation adapts the kinematic demonstration under structured forces at both hands so training data is force-aware. In simulation, policies adapt footsteps to the partner and reproduce approximately 80% of the wrist displacement in the augmented demonstrations. On a physical humanoid, the system sustains two-hand contact with a person, including repeated transitions between forward and backward motion.
- One partnered-dance video is retargeted to a robot and a moving partner.
- Multi-link compliance turns the kinematic demo into force-aware training data.
- Simulation recovers about 80% of augmented wrist displacement.
- A physical humanoid sustains two-hand dancing with direction changes.
From the original article
1 min read at huggingface.co
Partnered human-humanoid interaction couples locomotion with continuous physical contact. A humanoid needs to coordinate with a person's motion while responding to interaction forces and maintaining stable and natural movement. We present CoDance, a framework for learning reactive and compliant human-humanoid interaction from video.
By Zhuoqun Chen, Shucheng Jia, Boyuan Chen · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.