Advances in precision navigation of oceanographic submersibles

At present, few three-dimensional navigation techniques possess the update rate and precision necessary to achieve the scientific and engineering tasks commonly required of underwater vehicles. This vitiates our ability to fully exploit the potential uses of underwater vehicles in oceanographic research. The goal of this thesis is to develop new methodologies for improving navigation of underwater vehicles, with an emphasis on position estimation in the XY horizontal plane. Improved navigation of underwater vehicles benefits both the control of these vehicles, and the utility of data obtained by these vehicles for oceanographic science.; Two specific problems are addressed in this thesis. The first problem is the estimation of the alignment calibration of Doppler sonars typically used in underwater vehicle navigation. Previously reported studies have identified poor estimates of the Doppler alignment as a significant source of error in Doppler navigation, and this thesis proposes three techniques for estimating the alignment using data collected by underwater vehicles during normal vehicle operations. Two of these techniques constrain the alignment estimate to the group of rigid body rotations---the first employs a previously reported least-squares methodology, and the second is the first reported adaptive identifier on the group of rigid body rotations. The derivation of the adaptive identifier is reported, along with a proof of asymptotic stability. The performance of these methodologies is evaluated on data collected from a laboratory remotely operated vehicle (ROV) and a field-deployed autonomous underwater vehicle. The experimental data demonstrates that the techniques provide calibration estimates that improve Doppler navigation precision not only on the calibration data set itself, but also provide improved precision over a wide variety of vehicle trajectories other than the calibration data set. The adaptive identifier may be of broader interest because of its applicability to more general problems in the identification, control, and dynamics of rigid body motion.; The second problem is the development of exact nonlinear model-based observers for underwater vehicle navigation. This class of observers exploits exact knowledge of the vehicle's nonlinear dynamics, the forces and moments acting on the vehicle, and disparate data from navigation sensors. The stability of the observer is shown using Lyapunov techniques and the Kalman-Yakubovich-Popov Lemma. The observer is evaluated using data from single degree-of-freedom experiments with a laboratory ROV, with a 300kHz Long Baseline (LBL) acoustic positioning system providing high-precision position measurements. The observer provides position estimates whose errors possess a standard deviation significantly lower than the those for 12kHz LBL positioning systems. The performance of the observer when position measurements are provided at update rates varying from 1 Hz to 0.1 Hz is examined. The observer position estimates are compared to those computed by an Extended Kalman Filter. This implementation is the first known reporting of the application of exact nonlinear dynamic model-based observers in underwater vehicle navigation, and is, to the best of the Author's knowledge, one of only a few published results discussing the implementation and testing of exact nonlinear observers.; Antecedent to the results reported within is the development of two tools for underwater vehicle research. The first is the development and deployment of DVLNAV, a Doppler based navigation program for oceanographic submersibles. The program has been deployed on numerous submersibles and results from its deployment aboard an at-sea ROV are reported. The second tool is the building of a test tank facility for underwater vehicle research at The Johns Hopkins University. In addition, this thesis contains tutorial appendices on background material in control and identification on the group of rigid b...