Underwater active mappingIntermediate
Active Mapping of Underwater Litter Using Camera-Sonar Fusion
This preprint presents an active underwater mapping system that fuses forward-looking sonar and camera detections in a shared Bayesian occupancy map and chooses the next-best viewpoint by balancing exploration and object exploitation. In a realistic simulator, active dual-sensor search found all 11 litter objects after 27 m of travel, versus 110 m for a lawnmower pattern, and was roughly twice as fast as either sensor alone. The result matters for making litter surveys and underwater inspection more efficient.