Reproduce “AMB3R-SLAM: Kilometer-scale SLAM with Hierarchical Backend” 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

Hengyi Wang, Lourdes Agapito

AMB3R-SLAM tracks camera motion locally while building a map, then uses overlapping submaps, sparse long-context links, and verified loop closures in a pose graph to limit route-scale drift without pixel-reprojection bundle adjustment. On seven VBR sequences, its monocular average ATE is 7.42 m versus 26.65 m for LingBot-Map, a 72% reduction. The test reaches 10.2 FPS on an RTX 4090.

Summary by alphaXiv.

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