A novel Gaussian Particle filter based on randomized Quasi Monte Carlo for initial alignment in SINS

The error model of marine strapdown inertial navigation system on the swaying base is nonlinear, while the azimuth angle is large. For the nonlinear error model, a new recursive Gaussian Particle filter based on randomized Quasi Monte Carlo is proposed. The randomized Quasi Monte Carlo methods use the weighted randomized low discrepancy particles to replace the weighted random samples, in order to avoid the possible gaps and clusters that arise from random sampling in Monte Carlo methods, and improve the sampling efficiency and calculation accuracy. The simulation experiment shows that the new approach obtains the better estimation performance in initial alignment of large azimuth misalignment on the swaying base of the marine strapdown inertial navigation system.