Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces a map-reference-aware conservative fusion framework that enhances the safety of flow field mapping in mobile robots by addressing issues of ambiguous onboard sensing and localization drift. The method predicts local velocity patches and generates a write-safety score to mitigate uncertain map updates, allowing for safe initialization in the absence of reliable references. Experimental results demonstrate a 42% reduction in ghost contamination compared to ungated fusion methods, with a further 39% reduction in real-world tests while maintaining 81% map coverage.
Safe flow mapping can be achieved even in the presence of ambiguous sensor data and localization errors, significantly reducing ghost structures in robot maps.
Mobile robots can infer local flow structure from onboard sensing, but a locally plausible estimate is not always safe to write into a global map. Similar flow structures may produce ambiguous observations, while localization drift causes predicted patches to be written at incorrect locations. Repeated misregistered updates then accumulate into persistent ghost structures. We address this failure mode with a map-reference-aware conservative fusion framework. The model predicts a local velocity patch and a learned write-safety score that continuously attenuates uncertain map updates while permitting initialization when no reliable map reference is available. Across synthetic jet and crossflow environments, the proposed method reduces average ghost contamination by 42% relative to ungated fusion. A zero-shot hardware replay using real pressure and optical-flow measurements from a thruster wake further reduces ghost contamination by 39% while retaining 81% map coverage. These results show that safe map writing is critical for flow mapping under ambiguous sensing and localization drift.