A Method of Avoiding Floating Obstacles During AUV Surfacing Using Sonar and Camera

Ice floes in polar regions make it difficult for Autonomous Underwater Vehicles (AUVs) to surface safely after missions. In this paper, we propose a method to find safe areas for AUVs to surface by using camera images and scanning sonar data. Although this paper assumes operations of AUVs in polar regions where there are ice floes, the proposed method can be applied to other drifting obstacles. The underwater space is divided into three-dimensional grids, and the grids in which the ice floe exists are probabilistically estimated by a binary Bayes filter. To detect ice floes, we use a scanning sonar mounted on the AUV and a camera on the mother ship. The velocity of the ice floe is estimated from the camera images. Then, using a transition matrix that considers the velocity vector of the ice floe, the future obstacle occupancy probability 3D grid map (3D OOP Map) is predicted. Finally, considering the AUV's travel time, safe surfacing areas are shown on the obstacle occupancy probability 2D grid map (2D OOP Map). The method is verified through a tank experiment.