Reproduce “Unified Perceptive Fall Safety for Humanoids” 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

Osher Azulay, Zhengjie Xu, Andrew Scheffer, Stella X. Yu

VIGOR combines sparse human recovery motions with varied simulated terrain to train a terrain-aware humanoid fall and recovery policy. A privileged teacher learns goal–terrain context, while a student infers recovery goals from egocentric depth and proprioceptive history. In simulated Fall-Recovery, it achieves 91.3% success and 90.5% stable success, versus 89.3% and 85.3% for separate encodings; these are simulation results.

Summary by alphaXiv.

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