Reproduce “Kinematic Motion Retargeting for Contact-Rich Anthropomorphic Manipulations” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Arjun S. Lakshmipathy, Jessica K. Hodgins, Nancy S. Pollard

A framework for kinematically retargeting complex human hand-object manipulations to diverse target hands is presented, leveraging dense contact areas and a novel non-isometric shape matching process. This approach produced high-quality, contact-preserving trajectories for a range of human and robotic hands, with consistent performance across various manipulation tasks.

Key insights

  • Formulating dense contact transfer using an atlas of coordinate charts with intrinsic properties (Laplacians, geodesics) enables robust non-isometric shape matching between highly varied hand morphologies.
  • Dense contact areas implicitly encode natural grasp states, resulting in low intersection volumes and plausible retargeted motions even without explicit physics simulation or intersection resolution.
  • The framework provides a powerful tool for 'design choice visualization,' allowing rapid prototyping and evaluation of robot hand designs by demonstrating the impact of kinematic variations (e.g., finger lengths, DOFs) on full manipulation tasks.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.