Reproduce “EMS-Nav: Embodied Multi-Scale 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
Zenghui Yang, Xinan Wang, Danyang Li, Zhe Xu, Zheng Yang, Jingao Xu, Qiang Ma
EMS-Nav organizes maps around functional regions linking objects to local spaces and larger areas, then uses them for language-goal navigation. It builds regions from object, layout, and spatial-relation evidence, connecting them to details and a semantic hierarchy. Across 18,628 LangMap instructions in 36 HM3D scenes, it achieves 44.51% SR@1 versus DualMap’s 35.12%; this is a mapped-observation benchmark, not unknown-environment exploration.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.