This thesis develops a Bayesian inversion algorithm for autonomous underwater vehicle localization, and carries out a study of several factors contributing to localization accuracy in an underwater acoustic positioning system. Specifically, a ray-based algorithm is described that estimates target position through the linearized inversion of transmission arrival time differences, and provides linearized uncertainty estimates for model parameters. Factors contributing to source localization uncertainty considered here included: (1) modelling transmission paths accounting for refraction due to a depth-varying SSP instead of using a constant sound-speed approximation and straight-line propagation, (2) inverting for a potential bias in the measured sound-speed profile, (3) accounting for errors in hydrophone position by including these positions as unknown parameters in the inversion, and (4) applying path-dependent timing correction factors to account for lateral variability in the sound-speed profile. In each case, nonlinear Monte Carlo analysis is applied in which a large number of noisy data sets are considered, to obtain statistical measures of the localization improvement that results by addressing these factors.
展开▼
