Particle filter of measurement information fusion through Cokriging method

For the higher nonlinear or non-Gauss dynamic system, this paper put forward a new centralized fusion particle filter (PF) method which uses the Cokriging method to fuse the measurement information of different precision sensors. It effectively makes full use of the statistical property of all the sensors’ noises, and the property of which can be calculated out of line, and the calculated form of measurements weighted summation of particle filter state estimation under Linear Minimum Mean Square Error (LMMSE) estimation with the new Cokriging method. Then the particle filter of all the measurements weighted summation is formed. Compared with the traditional particle filter fusion algorithm, the fusion particle filter in this paper not only has the better optimal fusion estimation accuracy, but also can save lots of running time through calculating weighted coefficient out of line. Both the theoretical analysis and computer simulation indicate that this algorithm is valid and advantageous.