Localization of Autonomous Underwater Vehicles Incorporating Flow Models and Acoustic Detection

The localization of autonomous underwater vehicles (AUVs) during subsurface travel is a challenging issue in that ocean flows are unknown and complex. This paper studies the localization problem of AUVs under ocean flows with spatial and temporal variability. To predict trajectories of AUVs, we develop an odometry model that employs the framework of controlled Lagrangian particle tracking. Odometry error, which is distinct from the classic odometry error of ground mobile robots, is affected by controllers of AUVs. We design a waypoint controller and then analytically derive the deterministic and stochastic error growth of odometry. This derivation ultimately allows us to develop discrete state and measurement equations for a particle filter algorithm that combines infrequent acoustic measurements and the odometry model. On-off acoustic measurements are provided by one receiver equipped in each AUV. Generating a graph from knowledge about non-disjoint regions composed of transmitters and detection ranges, we develop the particle filter algorithm with a special type of likelihood. This likelihood determines the weights of particles on the graph where real and predicted measurements are compared. The algorithms are verified by simulation results.