ANTI-WINDUP FEEDFOWARD CONTROLLER DESIGN FOR REFERENCE INPUT EXPRESSED AS TIME POLYNOMIAL

摘要

In this paper, we propose an anti-windup method for feedforward controller of 2 degree of freedom (2D0F) control (Araki et. al, 1991) by which model matching is done. The proposed method replaces memorized variable of output of reference model in the feedfoward controller with controlled variable when input is saturated. Furthermore, we propose that model of controlled object used in the method is derived by using identification method (Kosaka et. al., 2000), (Kosaka et. al., 2002) for a reduced model with zeroing 0-N-tuple integral values of output error. By using the model, tracking error for reference input that consists of 0 ~ N-th power of time becomes zero even if controlled object has uncertainty. The resulting system has neither deterioration in the feedback characteristics nor increase in the tuning man-hour because feedforward controllers do not affect the feedback characteristics and the feedback controller can be designed with accumulated know-how. It is verified by numerical simulations that satisfactory reference input response is derived by proposed method even if input is saturated or controlled object has uncertainty.