Input design for hybrid system identification for accurate estimation of submodel regions

This paper addresses about designing of input sequence for identification of hybrid systems as PWARX model. Identification of PWARX model requires estimation of parameter vectors and estimation of discriminant surfaces of submodels in regressor space. Since we especially focus on accuracy of discriminant surface, we propose a method to determine input which generates regressor vector data nearby the discriminant surfaces by optimizing a cost function. Since the cost function requires information about discriminant surfaces, we propose a sequential identification algorithm, which updates the model and the cost function alternatively. By this method, we can obtain a high accurate model with short input sequence. Validity of the proposed method is illustrated in numerical examples.