Graph-theoretic sonar detection performance of maneuverable vehicles in variable propagation environments

Abstract

This work addresses the problem of passive acoustic target detection using a maneuverable unmanned underwater vehicle (UUV) equipped with a towed line array. The objective is to assess target detection performance as a function of vehicle maneuverability and environmental variability. Detection performance is evaluated by mapping the search region onto a graph with nodes that constrain the vehicle trajectory and edge weights determined by the local detection probability. Vehicle maneuverability is constrained by limiting the maximum turn angle at each node. The UUV trajectory that maximizes the probability of at least one target detection can then be determined using the computationally efficient Dijkstra algorithm. Increased maneuverability improves detection performance given known environmental variability since the optimal trajectory is able to visit more locations with favorable environmental conditions. The optimal detection performance of a maneuverable array with relatively short array lengths is compared to that of a nonmaneuverable vehicle with a longer array. It is shown that maneuverability can be even more advantageous than array length given sufficiently variable environmental conditions. The impact of uncertainty and/or mismatch of the assumed environmental variability is also studied as a function of vehicle maneuverability and towed-array length. [This work support by ONR 321SP.]