Reproduce “OdorSpaces: Visual-Olfactory Embodied Navigation in 3D Environments” 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

Changyeon Lee, Kelin Yu, Zheng Wei, Shuna Ni, Ruohan Gao

OdorSpaces presents a simulated source-localization benchmark coupling time-varying gas-density fields with navigable indoor scenes. Its policy combines depth, odor measurements, and pose history through recurrent memory to predict navigation actions. In methanol-fire tests on Replica, it achieves 24.72 ± 3.69% SPL versus 17.58 ± 3.45% for DRT-RL, but episodes begin only after odor detection.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.