Constrained Control Using Model Predictive Control

Abstract

The most common way of dealing with constraints in control systems is to ignore them, pretend that the system is linear, and fix things up in a more-or-less ad-hoc fashion after performing a linear design. There are some rather systematic ways of ‘fixing things up’, including certain anti-windup techniques, but logically they are an afterthought, a way of dealing with the nuisance of constraints after the central work has been done. Until recently, things could not be done differently, because almost all the control design techniques available to us were linear techniques.³