Sequential covariance intersection fusion Kalman filter for multiple time-delay sensor network systems with colored noises

Abstract

This paper is concerned with the fusion estimation problem for multi-sensor discrete time-invariant linear systems with multiple time delays and colored measurement noise. In order to transform those systems into systems with correlated white noise, a system transformation method is introduced. A sequential covariance intersection (SCI) fusion Kalman filter is given based on the local optimal recursive Kalman filter in the linear minimum variance sense, which avoids the calculation of the cross covariance matrices between local sensors. It is proved that the presented fused Kalman filter has higher accuracy than those local filters. The simulation result reveals that the actual accuracy of the SCI fusion Kalman filter approximates to distributed fusion Kalman filter weighted by matrices, and based on the covariance ellipse, the geometric interpretation with respect to accuracy relation of the local and the iused Kalman filter is shown.