Classification and statistical learning for detecting of switching time for switched linear systems
Abstract In this paper, a new method for the detection of switching time is proposed for discrete-time linear switched systems, whose switching mechanism is unknown. The switching instant estimation problem consists to predict the mode switching for discrete behavior from a finite set of input–output data. First, the proposed method use a clustering and classification approach define the number of submodels and the data repartition. Then, by the use of statistical learning approach, we define the linear boundary separator of each validity region. Finally, a technique of detection given an explicitly estimation of switching time. A numerical example was reported to evaluate the proposed method.
Paper
Full text
Classification and statistical learning for detecting of switching time for switched linear systems
Semantic Scholar · Computer Science · 2015
Abstract
Abstract In this paper, a new method for the detection of switching time is proposed for discrete-time linear switched systems, whose switching mechanism is unknown. The switching instant estimation problem consists to predict the mode switching for discrete behavior from a finite set of input–output data. First, the proposed method use a clustering and classification approach define the number of submodels and the data repartition. Then, by the use of statistical learning approach, we define the linear boundary separator of each validity region. Finally, a technique of detection given an explicitly estimation of switching time. A numerical example was reported to evaluate the proposed method.