Side scan sonar based self-localization for small Autonomous Underwater Vehicles

This paper presents the development and experimental validation of a side-scan sonar based self-localization algorithm for Autonomous Underwater Vehicles (AUVs). Different classes of visually salient features are continuously extracted from dual-frequency sonar imagery using machine perception and subsequently utilized to generate sparse bathymetric maps. The diversity within the feature set, i.e. features of different sizes and features extracted at different sonar operating frequencies, allows for robust multimodal feature matching. Highly robust navigation solutions are generated in real-time by matching newly acquired features to a previously generated reference map. For each feature class, feature matching is carried out in four dimensions correlating feature location as well as perceived intensity. Overall utility and technical claims have been validated using real-world sonar imagery collected using an Iver3 AUV platform.