Hybrid System Identification by Incremental Fuzzy C-regression Clustering

In this paper, an approach to the identification of hybrid systems is discussed. It is based on the incremental fuzzy C-regression clustering. Based on the distance between the current measurement and the hyperplane of the local model, local models are updated. If necessary, a new local model is constructed. To increase the robustness and prevent false local models, the data are kept in the buffer temporarily. The approach produces good results as shown in two examples. The first example can be modelled as a piecewise affine dynamical system and the second one as a switched dynamical system.