Abstract
The multivariate generalized predictive controller (GPC) is studied (Clarke 1987a, Shah 1987). The multivariable GPC algorithm is presented in a stochastic framework, and a model-following capability is introduced in the control law. The method is described for a controlled autoregressive integrated moving average (CARIMA) model, and for the equivalent state-space representation in its innovation form. The closed-loop system equations are derived and analyzed, using the state-space approach.