Active sonar-based perception and wall following for AUV operations in enclosed waters

This study presents the development and validation of active sonar-based perception and wall-following methods for autonomous underwater vehicle (AUV) inspection in enclosed aquatic environments such as reservoirs, dams, and water tanks. A hardware-in-the-loop (HIL) simulation framework was established to reduce development risk, integrating a ray-based rangefinder simulator with a mechanically scanned imaging sonar (MSIS) generator that includes noise and side-lobe modeling. Two complementary methods were implemented: a dual-rangefinder approach estimating lateral offset and heading deviation from paired starboard rangefinders, and an MSIS-based approach extracting wall geometry using median filtering and Hough transform, fused with Doppler velocity log (DVL) and gyro data for state estimation between image updates. Simulation and towing-tank experiments demonstrated consistent performance. In HIL trials, the rangefinder method maintained a 1.2 m stand-off with mean deviations of 0.02 m and 0.69°, while the MSIS method held a 1.5 m offset with errors of 0.06 m and 1.66°. Experimental results confirmed these trends: the rangefinder method ensured precise tracking, whereas the MSIS method maintained 1.45–1.56 m altitude and robust wall following under magnetic disturbances. Overall, the two methods exhibit complementary strengths, collectively demonstrating a validated perception–control framework for AUV wall following and inspection in confined underwater environments.