UC Riverside

Reliable and high accuracy (decimeter level) localization of a rover relative to a defined frame is an enabling technology for numerous Intelligent Transportation Systems (ITS) applications (e.g., automotive guidance, routing, lane departure warning). The goal of localization is to compute the navigation state of the rover in some defined frame of reference such that the expected errors in the estimate is within a given performance specification.

Inertial navigation is a popular navigation technique since it provides full six Degree-Of-Freedom (DOF) navigation information. Further inertial sensors have been studied for decades and have well understood error models. This dissertation discusses the theoretical and implementation aspects of certain sensor aided Inertial Navigation Systems (INS). Though the presentation can be easily generalized to all forms of INS, the primary focus of this dissertation will be on automotive INS.

This dissertation formulates the localization problem in a mathematically rigorous fashion and poses it as a nonlinear Bayesian estimation problem. The INS kinematic equations and linearized error state equations required by the Bayesian estimation solution are derived. Aiding techniques like GPS, Vision and stationary aiding are described and mathematically formulated. Observability and performance analysis are presented for each of these aiding scenarios. The last part of the dissertation defines and formulates the Near Real Time estimation problem.