MULTIVARIABLE CONTINUOUS-TIME GENERALIZED PREDICTIVE CONTROL (MCGPC)

This paper presents a multivariable generalization of the scalar continuous-time generalized predictive control (CGPC) and examines the resulting closed-loop system in some detail. The conditions for decoupling and model-following are established. The relation with the LQ control is illustrated by example. It is also shown that for the control of non-minimum phase systems there is no need for explicit control weighting. The presented method can be used both in the adaptive and the non-adaptive context. In the adaptive case, the only prior knowledge needed is an upper bound on the order of the common denominator of each row of the system transfer matrix. The method does not require a minimal realization of the system; therefore any non-minimal realization, which may arise from the estimation, does not create any problem. Effects of the design parameters and the properties of the algorithm are illustrated by simulations.