Properties of generalized predictive control

Control design using long-range prediction based on a dynamic model of the plant has become an important contender for high-performance applications. The model can be derived analytically, but more typically it is provided by simple experiments, or in the adaptive context by recursive estimation. Many methods have been proposed in the literature depending on the assumed model structure and the choice of cost-function: here the approach called generalized predictive control or (GPC) is reviewed. Of special interest is the derivation of prediction equations using an observer polynomial and a demonstration that the selection of particular horizons (the “costing horizons' and the “control horizon”) leads to well-understood basic techniques such as dead-beat, pole-placement and LQ. Simulated examples show that GPC is suitable for controlling a complex plant such as unstable/inverse unstable systems and how the observer polynomial independently tailors the response to the disturbances. Preprogrammed set-points (as with robot trajectory control) and actuator nonlinearities can also be catered for. A discussion of adaptive applications of GPC to several industrial process concludes that the method is easy to use and effective.