Underwater Dredging Robot Navigation in Structured Environments Based on Imaging Sonar Features

To achieve precise localization of dredging robots in sewage tank environments, an underwater matching correction navigation algorithm based on Mechanical Scanning Sonar (MSIS) point cloud data is proposed. This paper extracts line features from sonar scan data using adaptive Hough transform and constructs geometric feature maps. Utilizing a priori map of sewage tanks of known dimensions, the algorithm compares the feature maps to determine the accurate position and orientation of the vehicle. The Extended Kalman Filter (EKF) algorithm estimates the underwater robot's pose state by fusing data from multiple sensors and extracted line features. To validate the effectiveness of the algorithm, simulations were conducted in a sewage tank environment using Gazebo, demonstrating the algorithm's ability to effectively suppress divergence and accurately localize the vehicle.