Reproduce “SP-HLM: Sparse-Pose-Driven Motion Generation and Whole-Body Control for 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
Zhaohui Yang, Quanbao Lin, Chengqi Shi, Zhenglang Weng, Jicheng Lin, Boyu Chen, Tengjie Zhu, Guanzhu Ren and 4 more
SP-HLM combines five explicit body-pose targets for the pelvis, wrists, and ankles with a learned motion prior to coordinate whole-body movement. SP-VLA predicts sparse poses and finger targets; SP-MGM generates full-body motion, while SP-GMT combines its latent with explicit targets to control movement. Across eight simulated tasks, success was 52.75% versus 33.75% for the GR00T token-space baseline, for these tasks.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.