一回の閉ループ実験データを用いた最小二乗法に基づく制御器パラメータチューニング

Fictitious Reference Iterative Tuningの拡張

抄録

Recently, for the purpose of the minimization of the squared error between the desired and real experimental output of the plant, Iterative Feedback Tuning (IFT), Virtual Reference Feedback Tuning (VRFT), and Fictitous Reference Iterative Tuning (FRIT) were proposed as reasonable methods for tuning of the parameters of controllers. The first one requires many experiments for the updating step in Gauss-Newton method, while the later two methods requires only one-shot experimental data. Moreover, VRFT and FRIT have useful features in a complementary manner. From these backgrounds, this paper provides a new parameter tuning for controllers based on least-squares optimization by using one-shot closed loop experimental data. Here, we provide a basic idea of this approach and discuss the optimality of the obtained parameter. Moreover, we introduce the pre-filter so as to guarantee that the optimized parameter in the fictitious area corresponds to that in the real area. We also provide the algorithm of our method for enhancement of the use of this procedure. Finally we give an experimental example to illustrate the validity of our results.