OmniInteract: Towards Unified Whole-Body Interaction Control for Dual Humanoids

Key points

• Physical interaction is a key capability for humanoid collaboration, requiring coordinated whole-body motion and precise contact.1
• However, learning diverse dual-humanoid interactions faces two major challenges: (1) human-robot morphological differences disrupt contact interfaces and spatial relationships during retargeting, and (2) jointly learning motor skills and partner coordination makes it difficult to retain control fidelity as interaction diversity grows.1
• In this paper, we propose OmniInteract to address these challenges through surface-aware modeling and scalable residual policy learning.1
• Second, we propose a Scalable Flow-based Interaction Policy for partner-conditioned residual generation.1

Sources (1)

• [1]OmniInteract: Towards Unified Whole-Body Interaction Control for Dual Humanoids
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 19, 09:03 AM

  First, we introduce Surface Contact-Aware Retargeting, which leverages dense human-robot surface correspondences to jointly retarget both humanoids, preserving explicit interaction contact interfaces and their near-field geometric context.

  Physical interaction is a key capability for humanoid collaboration, requiring coordinated whole-body motion and precise contact.

Extractive summary: sentences quoted from the sources.