Abstract
In this paper, a data-driven model-free design method of PID controller is proposed for a SISO linear timeinvariant plant. A problem setting for loop shaping is given where the integral gain of the PID controller is maximized under the maximum sensitivity constraint. This optimization problem is reduced to a linear programming problem whose constraint can be calculated from the input-output response data of the plant based on unfalsified control concept without assuming any mathematical model. In order to avoid many experiments to obtain many response data, a method of generating fictitious data by applying the wavelet transform to a single response data is given. Our design method does not require complicated iterations for plant identification, performance index refinement, and controller design.
