Control Performance Prediction of V-Tiger by High-order ARX Identification and Confidence Interval Analysis

Abstract

FRIT and VRFT are widely applied as representatives of Data-driven control, but the actual closed-loop response cannot be predicted. Recently, Data-driven prediction methods including V-Tiger that predict the actual closed-loop response have been proposed, but assume that there is no noise in one-shot experimental data. Because the actual experimental data contains noise, we do not know how much the predicted response will deviate from the true response. This is a serious problem from a practical standpoint. Therefore, to obtain a Confidence interval for the closed-loop response predicted by V-Tiger, we consider using High-order ARX identification, which does not require accurate information such as model order. The objectives of this paper are (1) to propose a metric for obtaining good Confidence interval, (2) to estimate the Confidence interval of the open-/closed-loop noisy response, and (3) to improve the prediction accuracy of the closed-loop response.