Based on three canonical forms of descriptor stochastic systems, applying the Kalman filtering and white noise estimation theory, the three reduced-order Kalman state estimators and Wiener state estimators of descriptor systems are presented respectively. They can handle the prediction ,filtering and smoothing problems in a unified framework, and have the symptotic stability. Otherwise, the three reduced-order pole-assignment fixed-lag Kalmam smoothers having a global asymptotic stability are presented. Compared with non-reduced-order estimators, they can obviously reduce the computational burden, and is suitable for real time appplycations. Many simulation examples show effectiveness of the proposed algorithms. |