Reproduce “FORGE: Force-Aware Multi-Granularity Control for Efficient Humanoid Loco-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
Liangjun Xing, Ke Fan, Kailin Li, Bin Wang, Xiaojie Niu, Qingwei Ben, Jiafei Cao, Wei Yao and 3 more
FORGE coordinates humanoid manipulation using sparse commands for movement, foot placement, wrist pose, and optional force. A frozen locomotion policy anchors a learned residual; a momentum observer plus learned correction estimates low-frequency wrist-equivalent force for target adaptation. On staged-load x-axis traces, hybrid RMSE was 0.81 N, versus 2.83 N observer-only and 2.43 N learned-only; held-out vertical loads had higher errors.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.