A new method of direct data-driven predictive controller design

In this paper, we try to find a direct path between data and predictive controller. A straightforward data-driven predictive controller for the linear multivariable systems is proposed, without identifying any representation of the system in an intermediate step. The minimal image representation is used to describe the controlled linear multivariable system instead of model or dynamical description matrix. Data-based prediction can be estimated directly from an input/output trajectory of the system and thus the computation of dynamic optimization. For the unconstrained condition, control laws can be analytically determined directly from the data Hankel matrices without model or any intermediate step to meet the given performance specifications. The proposed predictive controller is demonstrated on a multivariable system.