Reproduce “HuG: Learning Humanoid Whole-Body Affordance-Aware Grasping with Active Perception” 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
Xuanbin Peng, Lai Wei, Yuze Zhai, Qingyuan Yu, Keyi Wang, Jianglong Ye, Xiaolong Wang
HuG combines active viewpoint search, whole-body alignment, and grasping for humanoid object retrieval. It trains on 600,000 synthetic trajectories retained only after replay with the deployment controller confirms stable execution, rather than relying on geometric reachability alone. HuG succeeds in 87.5% of 40 real-robot shelf trials; its 81.1% result below training heights is simulation-only.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.