A collision avoidance controller for autonomous underwater vehicles

A 3D collision avoidance controller for an autonomous underwater vehicle (AUV) that uses a forward-looking high-frequency active sonar is described. Multiple objects are differentiated by clustering sonar returns, and Kalman filters are then used to track both stationary and moving obstacles. A minimum safe distance must be maintained between the AUV and any obstacle, and a modified potential field is used to determine appropriate maneuvers. The merit function which defines this potential field is based on obstacle bearing, distance, and visit count, as well as the AUV heading, the depth, and the direction of the goal