A Rule-Based AUV Path Planner with a Hybrid Image Denoising Filter Using Forward-Looking Sonar

Forward-looking sonar is the principal sensor for autonomous underwater vehicles (AUVs) operating in uncharted environments; however, sonar imagery is frequently compromised by speckle noise and multipath-induced ring artifacts, which degrade obstacle detection and path-planning performance. In this paper, we present a hybrid denoising methodology that combines a polar-coordinate transform with bilateral filtering to attenuate speckle noise, and further employs an adaptive mean filter to suppress ring artifacts. Following denoising, obstacle contours are extracted and a rulebased avoidance algorithm is executed to generate collisionfree trajectories. Comprehensive numerical simulations and sea trials confirm that the proposed approach yields high-fidelity obstacle representations and enables reliable, efficient path planning under realistic underwater conditions.