Information fusion reduced-order Kalman filters for descriptor systems with delayed measurements

The filtering fusion problem of a descriptor system with delayed measurements is transferred to the different-step prediction fusion problem of two reduced-order normal subsystems without delayed measurements and with correlated noises. Using projection theory, the cross-covariance matrix of different-step prediction errors between any two sensor subsystems is derived. Based on the fusion algorithm weighted by scalars in the linear minimum variance sense, the distributed information fusion reduced-order Kalman filters are given. The proposed filters avoid the high-dimensional computation resorting to state augmentation, and can reduce the computation burden. The simulation research verifies the effectiveness.