Simultaneous AUV Localization and Lagrangian Particle Tracking

A simultaneous autonomous underwater vehicle (AUV) localization and Lagrangian particle (LP) tracking (SLLPT) problem is proposed, and the solution to a simplified case is discussed. An AUV navigates along with the LP swarm to track its ego-position and the position of the LPs by utilizing ocean general circulation models and relative range and bearing measurements. SL-LPT is formulated as a stochastic inference problem where the target probability distribution can be tracked through Bayesian recursion. A Markovian random-flight model is adopted to describe the motion of the LPs. Based on this model, motions of the LPs are assumed to be dominated by large-scale mean flows, and the effect of eddy diffusivity on LPs' motions is assumed negligible. The AUV predicts LPs' positions through mean flow estimation with Gaussian noise. An extended Kalman filter (EKF) is applied to correct the dead-reckoning position tracking error by incorporating relative range and bearing measurements between the AUV and LPs. Simulation results in an artificial double-gyre flow field suggest that the tracking errors become correlated as the AUV navigations. It is revealed that the rate of error divergence decreases dramatically as the number of LPs increases.