Reproduce “PASS-VLN: Passability-Aware Local-Command Vision-and-Language 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

Cong Zou, Xiangsheng Shi, Xuanhui Wang, Fubin Wang, Guangjun Ge, Chao Wang, Yu Wang

PASS-VLN tests whether robots can execute navigation commands while accounting for body width. It projects depth into a local overhead map, expands obstacles by robot width, and predicts short motions in a closed loop. On map-informed commands at the seen 0.35-meter width, it achieves 76.1% collision-free success, versus 64.4% for fine-tuned StreamVLN. The benchmark allows at most one major turn.

Summary by alphaXiv.

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