Generalized Predictive Control: A Robust Self-Tuning Algorithm

Abstract

A novel algorithm call Generalized Predictive Control (GPC) is shown to be particularly effective for the self-tuning control of industrial processes. The method uses long-range predictive control (LRPC) ideas with a carefully-chosen CARIMA plant model and various `horizons' which allow the choice of a rich variety of control objectives. Stabilization theorems show that even unstable/inverse-unstable systems are regulated by GPC, and a multivariable version gives tight control of interacting plant without prior knowledge of the interactor matrix. The application of GPC to a cement mill, a spray-drying tower and a compliant robot arm show that performance better than fully-tuned PID regulators is achieved.