Worst -Case l1 Identification Based on Correlation Analysis

Abstract

In this paper, we propose a new model set identification method which determines both nominal models and uncertainty bounds in l1 norm. This method, first, obtains estimates of the plant impulse response and their error bounds based on the correlation analysis using given experimental data and a low-correlated noise model. Since we consider the low-correlated input and noise, the error bounds of the impulse response estimates tend to zero as the number of data is increased. And then, a low order model and the nonparametric error bound in l1 norm are obtained by using the obtained impulse response estimates and the error bounds. Therefore, this method gives less conservative model sets when we have more experimental data, which is one of the distinguished features compared with most of the existing model set identification methods. Numerical examples show the effectiveness of the proposed method.