Information fusion full-order Kalman filter for multisensor descriptor system

Abstract

For the multisensor linear stochastic descriptor system, the information fusion full-order descriptor Kalman filters are presented, which are different from the reduced-order Kalman filtering algorithms and can improve filtering accuracy. The centralized fusion full-order descriptor Kalman filter can obtain the globally optimal filter, by extending all measurement information. The weighted measurement fusion full-order descriptor Kalman filter also has global optimality and has less computational burden, based on the least square method. The state fusion and sequential covariance intersection full-order descriptor Kalman filters yield the suboptimal filters. The sequential covariance intersection full-order descriptor Kalman filter avoids computing the cross-variance of local full-order Kalman filters and can reduce the computational burden. The accuracy of the centralized fusion and the weighted measurement fusion full-order Kalman filter is higher than that of state fusion and the sequential covariance intersection fusion full-order Kalman filter, and the accuracy of these algorithms are all higher than that of local full-order Kalman filters. A simulation example verifies the effectiveness.