Navigation and collision avoidance of underwater vehicles using sonar data

Collision avoidance is one of the main challenges in the field of autonomous underwater vehicles (AUV). In this paper a method for detecting obstacles is proposed, using a single-beam mechanically scanning sonar, including planning of an optimal path around the obstacles. Obstacle detection is achieved with an inverse-sonar model updating a vehicle-fixed occupancy grid. A new and obstacle-free path is planned using Voronoi diagrams and Dijkstras algorithm. The path is smoothed using Fermats spiral and a line of sight-guidance system with a time-varying lookahead-distance as guidance. The method is implemented and a full-scale test is performed from IKMs onshore control room on a remotely operated vehicle (ROV) operating at Statoils Snorre B oil field on the Norwegian Continental Shelf. The technology is applicable to ROVs and AUVs in underwater operations.