Particle Filter Algorithm with Correlative Noises

The standard particle filter needs to meet the requirement of noise independent. In order to overcome this limitation, this paper proposes a correlative noise particle filter (CN-PF) algorithm. The method analyzes the characteristic of noise time correlation, and derives the joint probability density function of correlative noise based on the given nonlinear system model. The concrete implementation method of noise de-correlation is analyzed based on the Gaussian noise assumption. The optimal proposal distribution function is deduced in the condition of the importance weight variance minimum. The CN-PF algorithm compensates the shortage of the traditional particle filter algorithm effectively, and expands the application range of the PF algorithm. The theoretical analysis and simulation results show the effectiveness of the propose method.