Self-tuning optimal predictors for singular discrete stochastic linear systems

Using the modern time series analysis method, this paper deals with the optimal and adaptive state estimation for the singular discrete stochastic linear systems. The optimal predictor is presented by converting the state estimation into the output prediction and noise estimation, and the asymptotic stability for the initial values of the optimal predictor is proved. The self-tuning predictor is also presented as the covariance matrices are unknown in this paper. A simulation example shows its usefulness.