Reproduce “ReGATE: Recurrent Gated Visual Fusion for Reactive Humanoid-Object Interaction” 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
Hao Zhu, Jorge Peña-Queralta
ReGATE controls humanoid box pickup by maintaining an interaction estimate from proprioception and selectively incorporating depth through learned gates. It predicts object state, geometry, and contact, then uses this representation for commanded interactions. In simulation, it succeeds in 281/330 static acquisition trials (85.2%), versus VisualMimic’s 160/330 (48.5%); this comparison does not isolate the gates’ effect.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.