Multivariable Constrained Adaptive Predictive Control Based on Pole Placement Design

Abstract

A multivariable linear controller is based upon pole placement design developed from the family of all stabilizing controllers. In addition a long-range prediction objective function is minimized to satisfy constraints on any signal or linear combinations thereof. The controller is designed to cope with constant disturbances and step-wise setpoint changes. Simulations of single- and multivariable control problems illustrate superiority of the proposed algorithm over one-step strategies. The algorithm is expected to be superior to previous constrained predictive algorithms since it is closed loop in design. Finally it is demonstrated that the algorithm can directly be extended to handle varying plant parameters by incorporating parameter adaptivity.