Fuzzy Clustering for Multiple-Model Approaches in System Identification and Control
Robert Babuška, M. Oosterom · Studies in fuzziness and soft computing · 2001
A review of fuzzy clustering and its use in the data-driven construction of nonlinear models and controllers is given. The focus is on algorithms of the fuzzy c -means type. Two application examples are presented: automated design of operating points for gain scheduling in flight control systems and nonlinear black-box identification. In the latter case, a comparison with an alternative technique is given. It is shown that fuzzy clustering is an effective technique for the decomposition of a complex nonlinear problem into a set of simpler local problems.