Write-Safe Flow Field Mapping under Ambiguous Onboard Sensing and Localization Drift

2026-07-30

AI Summary

This work addresses the challenge of spurious structure accumulation in global maps caused by mobile robots erroneously writing local flow fields under perceptual ambiguity and localization drift. To mitigate this issue, the authors propose a map-reference-aware conservative fusion framework featuring a learnable write-safety scoring mechanism that enables safe map initialization in the absence of reliable references and dynamically attenuates uncertain updates. By integrating predictions of local velocity fields with pressure and optical flow sensor data within a conservative fusion strategy, the method achieves a 42% reduction in spurious structure contamination compared to non-gated approaches in simulated jet and cross-flow environments, while maintaining 81% map coverage. Further validation on real-world thruster wake replay sequences demonstrates a 39% reduction in map pollution.

Abstract

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.

Problem

• flow field mapping
• ambiguous sensing
• localization drift
• ghost structures
• map fusion

Innovation

• write-safety score
• conservative fusion
• localization drift
• flow field mapping
• ghost structure