A novel initial alignment algorithm based on the interacting multiple model and the Huber methods

Initial alignment is one of the key technologies in Strapdown inertial navigation system (SINS). It is divided into coarse alignment and fine alignment. The conventional method of fine alignment is to adopt Kalman filtering and uses single filter to estimate system states. In Kalman filtering, it is known that system model should be matched with actual system and the statistical characteristics of noise are supposed to be Gaussian. However, single model cannot describe the unknown filtering parameters in practical application. Moreover, noise may be contaminated and present a non-Gaussian form. This paper is devoted to solve these problems, presenting a new alignment method based on interacting multiple model (IMM) algorithm, in which sub-filters are designed to be Huber-based Kalman filters. Uncertain parameters can be depicted by a set of switching sub-models and Huber-based filters can deal with the problem of contaminated noise. Finally, simulations show that the result of this proposed method performs a higher accuracy than conventional method's.