A novel nonlinear filter for initial alignment in strapdown inertial navigation system

The error model is nonlinear when the azimuth angle of strapdown inertial navigation system (SINS) on stable base is large, and a new filter results from using Unscented Kalman filter for proposal distribution generation imbedding latest observed measurements in importance sampling step, and combining Gaussian mixture model and weighted expectation maximization (EM) algorithm to replace the traditional resampling step. And the “sample depletion” problem was lessened. It is demonstrated by simulation that this new approach has an improved estimation performance in Initial Alignment of Large Azimuth Misalignment on Static Base of SINS.