Reproduce “CQDC-Warp: GPU-Accelerated Differentiable Quasi-Dynamic Simulation for Contact-Rich 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

Yanchen Ding, Jiayi Chen, Jianfeng Xu, Wei Zhao, Li Yi, He Wang

CQDC-Warp plans contact-rich robot motions by optimizing commands in batched differentiable simulations, given an initialization and intended contacting links but not exact contact points or a full contact schedule. On an Allegro in-hand rotation benchmark, it reports 79.8% mean success and 0.794 s mean trajectory-generation time across 17 objects. These comparisons use GPU-parallel planning versus CPU-sequential baselines.

Summary by alphaXiv.

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