Performance Evaluation of Visual and LiDAR SLAM for Inland Waterways

Kurzfassung

Inland waterway (IW) transport offers an eco-friendly and cost-effective alternative for heavy and bulky shipments. However, navigation in IWs presents challenges due to limited maneuverability, shallow depths, and the need for accurate data. Thus, there is a drive for a high level of autonomy for water surface vehicles (WSVs). This work reviews the state-of-the-art on visual and LiDAR Simultaneous Localization and Mapping (SLAM) solutions and addresses their performance for IW scenarios, where there is additional complexity due to the unstructured surroundings or the lack of planar features. To that end, data collection in the IWs of the Berlin metropolitan area is employed, comprising stereo camera, inertial, and 3D LiDAR data. The SLAM-based navigation performance evaluation is based upon the ground truth obtained from a post-processed, multi-antenna GNSS-based solution. Our experimentation concludes that LiDAR-based SLAM exhibits good horizontal positioning accuracy, with rapidly growing errors in the vertical direction. In contrast, visual SLAM provides positioning results with greater accuracy in the vertical component than in the horizontal one. This SLAM evaluation serves to understand the drawbacks of current solutions and to develop new ones accordingly, with the goal of increasing the autonomy levels for WSVs.