Drift Error Reduction Based on the Sonar Image Prediction and Matching for Underwater Hovering

In this paper, we propose a sonar-image-based underwater-hovering algorithm for autonomous underwater vehicles (AUVs). Unlike the conventional methods that apply vision-based algorithms, the proposed algorithm predicts the sonar image of the target position in advance, and then compares the sonar image of the current position of the AUV to the predicted sonar image in order to derive the current position error. By feeding back the error, the AUV can compensate for its position error. We verify the accuracy and tolerance of the proposed algorithm by performing an experiment using a forward-looking imaging sonar.