Position information fusion using particle filter for underwater vehicle navigation

Abstract

Navigation (or localization) is a basic and crucial capability for all types of manned or unmanned vehicles, including aerial vehicles, ground vehicles, surface ships, and underwater vehicles. Integrated INS/GPS systems have been commonly used and successfully applied to vehicle navigation applications in open outdoor environments. However, the use of GPS is not possible underwater because high-frequency electromagnetic waves cannot pass through sea-water. The use of acoustic systems is a practical option for underwater navigation. However, acoustic systems also have restrictions such as limited coverage and the need of pre-deployment. This dissertation presents two position information fusion application studies for underwater navigation that can be used in underwater environments. One is a localization algorithm that utilizes the seafloor topography, and the other is a navigation algorithm that uses weld line grids on the bottom surface of a ship. Because both applications deal with highly nonlinear information, they are formulated as nonlinear estimation problems. This suggests the use of a particle filter for improving the overall estimation performance. The particle filter algorithm, which is one of the nonlinear filtering methods, requires virtually no restrictive assumptions about the system dynamics or uncertainty distributions. Consequently, it often provides better results for highly nonlinear or non-Gaussian systems than the standard Kalman filter algorithm and its variations. Terrain-based localization for underwater vehicles, which is the first application of this study utilizes subsea terrain information to minimize drift errors caused by dead-reckoning or inertial navigation by using drift-free position fixes with respect to the terrain. A mathematical model is derived to design an estimator, and numerical simulations using both artificial and actual seafloor topography maps are implemented to demonstrate the validity and feasibility of terrai...