An adaptive square-root unscented Kalman filter for underwater Vehicle navigation

Abstract

In order to increase the approximation accuracy of the state estimate of nonlinear systems and to guarantee numerical stability of the unscented Kalman filter (UKF), a novel adaptive square-root unscented Kalman filter (ASRUKF) based on modified Sage-Husa noise statistics estimator is proposed. The new adaptive filter method with adaptability to statistical characteristic of noise is able to compensate the lack of a priori knowledge of the system's noise statistics. A six-degree-of-freedom dynamic model is introduced to denote the motion model of Human Occupied Vehicle (HOV) in the water, while the adaptive SRUKF is employed for off-line estimation of the state of HOV. Tests are conducted with respect to the data obtained from previous sea trial, and the results are compared with those obtained by normal UKF and SRUKF to indicate its effectiveness and improvements.