A Student's T-Based Measurement Uncertainty Filter for SINS/USBL Tightly Integration Navigation System

In order to suppress the impact of some error factors of the existing strap-down inertial navigation system and ultra-short base line (SINS/USBL) position matching loosely integration approach on positioning accuracy, a tightly integration strategy is creatively proposed and designed relying on the derived state error equation and measurement equation, the relative measurement information are directly used as the observation information. The error factors such as attitude error and installation error are considered in the filtering model to avoid the influences on the matching position, and the filter algorithm can be also effectively designed according to the sensor parameters of USBL, which can improve the integration positioning accuracy. Meanwhile, in order to address the decreased positioning performance caused by the measurement uncertainty, a Student's t-based Kalman filter with adaptiveness and robustness is reorganized and derived for the proposed SINS/USBL tightly integration strategy, the adaptiveness can be obtained by estimating the unknown measurement noise statistics using variational Bayesian (VB) approximation, and the robustness can be achieved by dealing with the measurement outliers based on the Student's t distribution. Finally, the feasibility and the superiority of the proposed strategy are evidenced both through the simulations and the field tests in the river.