Online identification of switched linear systems using the Hough transform

This paper proposes a Hough transform (HT) based identification algorithm for the online identification of switched linear systems (SLS). The proposed algorithm utilizes the point-hyperplane duality to estimate parameters, which is applicable in different identification environments, e.g., in the presence of outliers, with an unknown number of modes, and without a priori information of parameters. The first part of the paper studies the well-known HT technique and develops a new implementation of the online HT-based estimator. Subsequently, an online clustering estimator and a feedback mechanism are integrated to further improve the estimation performance in terms of memory use and model fit. The effectiveness of the proposed algorithm is demonstrated via numerical examples.