A unifying framework of steady-state Kalman filtering, smoothing and predic
tion for descriptor systems is presented by using the innovation analysis m
ethod in the time domain. The descriptor Kalman estimators ave presented on
the basis of the autoregressive moving-average innovation model and white-
noise estimators. The new algorithms of steady-state descriptor Kalman esti
mators gains ave given. The solution of the Riccati equation is avoided. To
ensure the asymptotic stability of descriptor Kalman estimators with respe
ct to the initial values of innovation process, formulae for selecting thei
r initial values are given. A simulation example shows the usefulness of th
e proposed results.