Towards tracking a row of discrete targets with AUVs using Markov Decision Processes and probabilistic techniques

While complete searches of an area are the standard method currently used by the U.S. Navy for detecting and determining the extent of a field of discrete underwater targets of interest, there are situations where time constraints or mission objectives prohibit a complete search. Furthermore, in many situations it is unnecessary to detect all targets within a field of targets of interest, but is sufficient to determine the area over which the field extends. In this paper we present a novel search procedure for determining the extent of a line of targets within an area for use with AUVs. This method uses a Markov decision process coupled with belief propagation. Monte Carlo simulations indicate that this algorithm out-performs the standard search techniques for determining the endpoints of a line of targets when the metric is time.