Design of OBF-TS Fuzzy Models Based on Multiple Clustering Validity Criteria

Takagi-Sugeno Fuzzy Models within the framework of Orthonormal Basis Functions (OBF-TS Fuzzy Models) have shown to be an effective approach to nonlinear system identification and control due to several advantages they exhibit over those dynamic model topologies most commonly adopted in the literature. Despite all the theoretical advances and encouraging application results obtained so far, the automatic determination of the number of local OBF models remains an issue. This paper elaborates on the use of a mixture of clustering validity criteria to automatically determine the number of local models based on product space fuzzy clustering of I/O data.