Reproduce “Compass 1: Towards Open-World and Persistent Embodied Navigation from Monocular RGB Videos” 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
Hai Zhang, Siqi Liang, Yanjun Guo, Shoufeng Zhang, Yunhao Wang, Hangyu Du, Haoguang Mai, Hongchen Li and 12 more
Compass 1 helps robots return to landmarks encountered earlier during navigation. It selects sparse past views and organizes their positions and headings in a map, then uses questions to identify which earlier view contains a requested target and which way to travel. On the tested real-world persistent-history-backtracking task, it succeeds in 58% of trials, while the strongest baseline scores 0%.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.