Using modem time-series analysis method, based on the autoregressive moving average(ARMA)innovation model and white noise estimators, White state estimators with application to bias estimation are presented for non- square descriptor system. Two-stage decoupled Wiener filters are presented for system with state observation and input observation, two-stage decoupled Wiener filters are presented for descriptor and non-descriptor systems with stochastic bias. Compared to the classical two-stage Kalman filter, they can handle the filtering, smoothing and prediclion problems in a united framework; the computation of the Riccati equations is avoided, they have the optimality and asymptotic stability and the complete decoupled is implemented. lots of simulation examples show the validity of the new algorithms proposed here. |