Navigation and mapping with distributed mmWave radars in inland waterways

Kurzfassung

This work investigates whether distributed automotive mmWave radars can support navigation and mapping services in inland waterways, where automation is motivated by outdated maps, constrained infrastructure, bridge-collision risks, and operator shortages. It compares LiDAR and mmWave radar, highlighting LiDAR’s dense geometric reconstruction and radar’s robustness in degraded visibility, direct target-velocity measurement, and lower hardware cost. Using the Berlin IWT dataset collected across multiple measurement campaigns with GNSS, LiDAR, stereo camera, and four mmWave radars, this work evaluates radar-based reconstruction of waterway layouts, riverbank contours, quay walls, and bridge structures. The results show that radar can recover the main environmental structure and provide coarse bridge-clearance information, although LiDAR remains superior in spatial density and geometric detail. Radar odometry preserves the overall trajectory shape in bridge-passing and island-loop scenarios, but local motion is underestimated due to corridor effects, with improved stability achieved through multi-frame aggregation and coarser voxelization. Overall, this work demonstrates that distributed mmWave radar is a feasible and promising sensing modality for inland-waterway automation.