Mechanical Sonar Scanning Strategies for Fast Underwater Obstacle Detection on an Autonomous Surface Vessel

Underwater obstacle detection is an increasingly necessary function of modern marine robotics, due to its important role in sub-surface collision avoidance in unknown environments. Mechanical scanning sonars (MSS) are affordable and accurate underwater sensors, that provide precisely controllable 360° coverage. However, MSS suffers from a slow sampling speed due to its single beam and mechanical actuation. To address this, this paper develops adaptive scanning strategies coupled tightly with a custom mapping and planning system to maximise information gain and collision response time on an autonomous surface vessel. The mapping system is demonstrated in field trials, and the scanning strategies are evaluated in simulation based on metrics of collision avoidance rate and smallest obstacle clearance. Results show our adaptive strategy increases the rate of intervention-free autonomous navigation by 70% in dense obstacle environments compared to an adaptation of a current state-of-the-art speed-adjusted strategy. Secondary findings suggest the adaptive strategy is also more robust across a range of obstacle reflection intensities and traversal speeds, and gathers more collision-relevant information from its surroundings than alternatives.