Kalman/particle filter for integrated navigation system

Particle filter is a best way for the state estimation of highly nonlinear integrated navigation systems with non-Gaussian uncertainties. Since the error model of SINS has high dimensions, traditional particle filter would bring hard computation. A new Kalman/Particle mixed filter used on SINS/GPS integrated navigation system was proposed. The new method divides the system into two sub-models, one is linear, the other one is nonlinear, and then implement Kalman filter and particle filter separately. Residual-systematic resampling and regularized algorithms were involved to decrease particle filter' particles collapse weakness and hard computation. The simulation results show that their performance is almost equal, but the computation complexity of the Kalman/particle filter is much lower than traditional particle filter.