Robust search for structured object placement using unmanned vehicles

We develop a robust method for planning an undersea search by unmanned vehicles; the goal of which is to find groups of objects that may be placed randomly or arranged in predefined patterns on the ocean floor. This approach revolves around the computation of target pattern priors that describe the underlying geometrical structure of the expected group of objects. These priors are updated based on limited search observations, followed by a search performance assessment in the remaining region utilizing these observed target pattern distributions. We outline the derivation of this approach to assessing the performance of unmanned searches. The algorithms performance is tested in a simulation example to illustrate the improvement in search evaluation accuracy provided by the consideration of object structure. We conclude with a discussion of how this search planning strategy can be employed in realistic scenarios, searching for objects that are expected to appear in more general patterns.