Utility-based adaptive path planning for subsea search

We present a decision-theoretic adaptive path planning approach to efficiently search for an unknown number of stationary objects distributed in a bounded search area subject to time or distance constraint. We find the optimal finite-length search trajectory by maximizing a decision-theoretic utility function, and we re-plan continually as new information is gathered. We assume that the performance of the search sensor varies with variations in the environment, and our planning approach incorporates stochastic predictions of the environment if available. Our work is inspired by subsea applications. We compare our approach with a naive mowing-the-law approach that is commonly used in subsea search. We also investigate how uncertain information affects search performance.