Data-driven Update of Non-parametric Controller by Input-oriented Virtual Internal Model Tuning

Abstract

Input-oriented virtual internal model tuning can tune the feedback controller by using one-shot experiment data This method can realize desired closed-loop responses when the orders of numerator and denominator of the feedback controller is adequate. However, it is difficult to determine the orders of numerator and denominator of the feedback controller when the controlled object is unknown. From this reason, input-oriented virtual internal model tuning is expanded to non-parametric controllers. Proposed methods can obtain the feedback controller which is parametrized by the impulse response of the controlled object. The impulse response is estimated by Ridge regression. The proposed method can realize good controllers because over-learning is not occurred by Ridge regression. The validity of the proposed method is verified via numerical simulation and experiment verification. From verification results, the input-oriented VIMT based on least square methods can not realize desired closed-loop response because of over-learning. Proposed method can realize desired control response even when the controlled object is unknown.