A horizon-recursive form for predictors and their computation

Concerns the design of self-tuning controllers using output predictors. A simple, horizon-recursive expression for output predictors is obtained for both SISO and MIMO systems described by ARMAX models. This expression reveals that the predictors for the different prediction horizons are recursively related by a key parameter vector. Thus, by recursively computing this key parameter vector, an efficient algorithm for generating predictors is obtained. It shows certain advantages over methods involving recursive solving of Diophantine equations