Reproduce “Generating Grasp Poses for a High-DOF Gripper Using Neural Networks” 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

Min Liu, Zherong Pan, Kai Xu, Kanishka Ganguly, Dinesh Manocha

The method directly predicts a grasp for a high-DOF hand from an object occupancy grid. It trains on 100 valid poses per object and uses consistency loss to select the closest pose, with collision loss to reduce penetration. On the 25-DOF Shadow Hand, consistency residual falls from 2.630 to 0.043; hardware grasps succeed for 12 of 15 held-out YCB objects.

Summary by alphaXiv.

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