Reproduce “Evolving Dexterous Robots from Scratch” 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

Zihan Guo, Shuzhe Zhang, Muhan Li, Peiyang Li, Sam Kriegman

The paper co-optimizes robot bodies and controllers, searching beyond fixed hand templates within a bounded space: one root, up to five unbranched appendages, and 25 joints. A learned genetic representation guides evolution. On 15 unseen objects, the didactyl robot averaged 1.39 rad/s, versus a reported 0.74 rad/s prior result; this was not a matched comparison.

Summary by alphaXiv.

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