Diffusion Based Robust LiDAR Place Recognition
Benjamin Krummenacher, Jonas Frey, Turcan Tuna, Olga Vysotska, Marco Hutter · 2025
Read Diffusion Based Robust LiDAR Place Recognition by Benjamin Krummenacher, Jonas Frey, Turcan Tuna, Olga Vysotska, Marco Hutter (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-12412.txt.
Description
Benjamin Krummenacher and co-authors use a diffusion model to help robots find their place on building sites. Their 2025 paper deals with global localization from LiDAR scans alone. Building sites are hard: flat walls repeat, and flats on different floors look alike. The robot must match its scan against an accurate mesh of the whole building. Training used synthetic scans made by simulating a sensor inside a real large scale mesh.
The model, built on a point cloud network, outputs several candidate positions for one scan. That spread captures genuine ambiguity instead of forcing one guess. Across five real data sets, 77 percent of positions landed within two meters. Mean error was half that of baseline methods. The approach targets survey and inspection robots that must reposition without external aids.
Catalog record
| CREATOR | Benjamin Krummenacher, Jonas Frey, Turcan Tuna, Olga Vysotska, Marco Hutter |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 30 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2504-12412