Model Predictive Control for Open Loop Unstable Process

The discrete models used in Predictive Controllers are usually in non-parametric step response or parametric ARMA form. This paper shows how both models can be combined into a state space formulation that will handle both stable and unstable processes. A feedback system based on the use of an observer with user specified poles is also developed which shows that control can be maintained even in the presence of model process mismatch.