Reproduce “One Scan is Enough: Demonstration-Free Adaptation for Language-Guided Navigation” 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
Zhuoyuan Yu, Yuxing Long, Du Wang, Junyang Wang, Hongwei Fan, Hao Dong
RealNav + PRPT-KS adapts language-guided robot navigation to scanned buildings without human-provided task routes. It turns one iPhone scan into rendered scenes, planned paths, generated instructions, and geometry-based training signals. Across five buildings, it achieves 85% real-robot success in scanned regions, versus 49% for human-demonstration RealNav, and 82% in unscanned regions; this does not establish transfer to arbitrary buildings.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.