Research on unscented particle filter for initial alignment of SINS

This paper derives nonlinear error model of strapdown inertial navigation system (SINS) in initial alignment. The traditional method of dealing with nonlinear system is extended kalman filter (EKF), namely, it is based on the principle of linearization the measurement and evolution model using Taylor series expansions. Unscented particle filter is based on Monte-Carlo method and Bayes estimation theory. It uses weighted particles to approximate probability density function and estimate the state value and covariance with such updated particles by measurement value, then combine unscented kalman filter (UKF) to iteratively calculate real time state value. In this paper, first we research the unscented particle filter algorithm, and apply it to attitude estimation in initial alignment of large azimuth misalignment of SINS. The computer simulation and experiment results prove the estimation accuracy and convergence rate of the azimuth estimation is greatly superior to the EKF.