A Novel Distributed Cubature Kalman Filter and Its Application to Target Tracking

Abstract

This paper studies the problem of distributed state estimation of nonlinear systems. Due to the increase of state components or dimensions, the unscented Kalman filtering algorithm cant solve this large-scale system with higher accuracy. We propose the cubature Kalman filtering strategy to deal with this case in microsystem. For wireless sensor networks, we extend the cubature Kalman filter to the distributed form which can greatly reduce the computation burden of single node. Compared with the previous distributed algorithms, this paper can achieve the exact consensus via the maximum consensus technique. Through the simulation, the maximum consensus cubature Kalman filter with higher estimation accuracy compare to the centralized algorithm and the average consensus cubature Kalman filter. Because of its higher estimation accuracy, the centralized algorithm can be replaced in the state estimation problem.