Abstract
This paper presents a method for Takagi-Sugeno fuzzy modeling. This method updates on line both the structure and the parameters of the model by combining a new on line clustering algorithm with least squares techniques. The proposed clustering algorithm, that generates clusters that are used to form the fuzzy rule antecedents, is used for model structure identification. The update of consequent parameters is achieved by least squares estimators. Copyright (c) 2008 CEA.