Reproduce “UniWAM Technical Report: Unified Mobile Manipulation via Mixed-Stream World-Action Modeling and Manipulation Anchor Pose Supervision” 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
Wei Xue, Keliang Liu, Mingzhang Cui, Jinhua Xie, Jinjie Wei, Jianan Hou, Jingcheng Lu, Lintao Wang and 9 more
UniWAM combines navigation and manipulation in one model, with separate inputs and action interfaces. Navigation learns an object-specific Manipulation Anchor Pose (MAP), the base position and heading for stopping to manipulate, using generated approach examples and image-plane supervision. On MAP-Bench, UniWAM-Scaled reports 9.85 cm position error and 2.30° heading error. Its real-robot mobile-manipulation success exceeds π₀.₅’s, but manipulation-only performance does not.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.