Incremental predictive Kalman filter for alignment of inertial navigation system

Abstract

In this paper, a new filter called incremental predictive Kalman filter is developed and employed for the alignment of inertial navigation system using zero velocity updates method. Utilizing the incremental model error, a well posed cost function is presented for incremental predictive Kalman filter that leads to bias-free predictions. Besides, a weighted incremental term of past and present states is evident in the model error solution. This term, in conjunction with an integral action, has substantial effects on the robust performance of the alignment process against intense model uncertainty. Due to the horizon extension of the predictions and of the model errors to more than one-step ahead in the incremental predictive Kalman filter, this filter has a very flexible structure which enables it to implement the available references in the zero velocity updates method as a whole over its wide prediction horizon. The Monte Carlo simulations indicate that the alignment accuracy is noticeably affected due to use of the incremental predictive Kalman filter.