Self-Tuning Generalized Minimum Variance Control Using Update Rate Based on Control Deviation

Generalized Minimum Variance Control (GMVC) yields a control law by minimizing a cost function based on a generalized output composed of reference, output, and input signals. Unlike Generalized Predictive Control (GPC), GMVC allows the designer to directly design the closed-loop characteristics through the generalized output. Although GMVC has been studied from various perspectives, its performance depends on the accuracy of the plant model. In practice, model uncertainty due to parameter variations or aging is inevitable. Therefore, a self-tuning controller is employed to adapt the control law based on parameter estimation. However, even if the estimated parameters converge to the true values, the derived model to be controlled does not always represent the true plant. Conversely, control objectives may be achieved without exact parameter identification. Therefore, this research investigates a method for calculating the estimation error, in which the parameter updates are stopped not only when the estimation error becomes zero but also when the update rate, defined as a value based on the control deviation, becomes zero. A numerical simulation is shown to validate the effectiveness of the proposed method.