A Novel Particle Filter for Nonlinear Non-Gaussian Estimation

A novel improved particle filter, cubature particle filter, is proposed for the estimation of nonlinear non-Gaussian system. Each particle is estimated by means of cubature kalman filter. The importance density function gets closer to the real posterior after taking the current observation into consideration on the basis of state transition. Both theoretical analysis and simulation experiment show that the cubature particle filter performs much better than the other parallel filters.