Multiple model adaptive estimation for the celestial navigation system

Abstract

This paper describes the application of multiple model adaptive unscented Kalman filters (MMAUKF) algorithm to celestial navigation system. Unscented Kalman filters are utilized to estimate the terminal system state of each model and to generate residual signals. In the multiple-model adaptive estimation technique, the residual signals are used to generate probabilities, which determine the correctness of state estimation of each unscented Kalman filter. This algorithm demonstrates better precision than UKF, compared with single model celestial navigation system by simulation. Simulation results show that introducing multiple model adaptive estimation theory into celestial navigation system combined with UKF can enhance the adaptability of system scheme to the environment, meanwhile, greatly improve the accuracy of celestial navigation system.