Reproduce “SAKI: Skill Assembly and Kinematic Imitation from Human Videos for Long-Horizon Mobile Manipulation” 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

Yijie Lu, James Zhao, Weiming Zhi

SAKI preserves interaction requirements while letting the robot adapt its route to the scene. It represents contact and motion constraints as reusable skills and generates coordinated base-and-arm motion. Across three continuous sequences, task-conditioned reference preparation raised complete-sequence success from 7/45 (15.6%) to 31/45 (68.9%) versus direct interpolation, with downstream components held fixed. This evidence is limited to these sequences.

Summary by alphaXiv.

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