Reproduce “DMFC-GraspNet: Differentiable Multi-Fingered Robotic Grasp Generation in Cluttered Scenes” 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
Philipp Blättner, Johannes Brand, Gerhard Neumann, Ngo Anh Vien
DMFC-GraspNet predicts many point-associated grasp proposals for a multi-fingered hand in cluttered scenes. It predicts palm offsets, orientations, and finger-joint angles from scene points, then fine-tunes using labeled grasps and differentiable collision and grasp-quality losses. In MuJoCo simulations on 3–4-object scenes, overall success rose from 0.507 ± 0.001 to 0.71 ± 0.001 after fine-tuning with all losses.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.