A decision-theoretic approach to acquire environmental information for improved subsea search performance

Abstract

This study addresses subsea search applications where an autonomous underwater vehicle (AUV) is tasked with finding the target density in a given search region within finite time. We assume that AUV is equipped with a side-scan sonar sensor that detects the targets at the sampled location. We consider that sensor performance is dependent on local environmental conditions (e.g., clutter density, sediment type) that vary throughout the search region, and we presume that environmental conditions are unknown or partially known. Due to uncertain and varying environmental conditions, resulting search performance is also uncertain and it varies by location. This paper specifically considers the cases where environmental information can be acquired either by a separate vehicle or by the same vehicle that performs the search task. Our main contribution is to formally derive a decision-theoretic cost function to compute the locations where the environmental information should be acquired so that the performance of the search task can be improved. For the cases where computing the optimal locations to sample the environment is computationally expensive, we offer an approximation approach that yields provable near-optimal paths. We show that our decision-theoretic cost function outperforms the information-maximization approach, which is often employed in similar applications.