ABSTRACT
This chapter shows the relationship between the classical Generalized Predictive Control (GPC) and the special case of synthesis approach. In the 1980s, the adaptive control techniques, such as minimum variance adaptive control, had been widely recognized and developed in the process control area. GPC was developed along the investigation of adaptive control. GPC not only inherits the advantages of adaptive control for its applicability in stochastic systems, on-line identification etc., but preserves the advantages of predictive control for its receding-horizon optimization and lower requirement on the modelling accuracy. Model Algorithmic Control and Dynamic Matrix Control both apply non-parametric models, that is, impulse response model and step response model, respectively. The techniques of multi-step prediction, dynamic optimization and feedback correction are applied. In GPC, the modelling coefficients are on-line estimated continuously based on the real-time input/output data, and the control law is modified correspondingly. GPC with terminal equality constraint is a special synthesis approach of model predictive controls.
