A New Particle Filter Algorithm with Correlative Noises

Abstract

The standard particle filter (SPF) requirements system noise and measurement noise must be independent. In order to overcome this limit, a new kind of correlative noise particle filter (CN-PF) algorithm is proposed. In this new algorithm, system state model with correlative noise is established, and the noise related proposal distribution function characteristics were analyzed in detail. At last, the concrete form of the best proposal distribution function is derived based on the condition of the minimum variance of importance weight with the assumption of gaussian noise. Theoretical analysis and experimental results show the effectiveness of the proposed new algorithm.