An Approach to On-line Design of Fuzzy Controllers with Evolving Structure

摘要

The problem of on-line design of Takagi-Sugeno fuzzy controllers, which have evolving structure, is treated in the paper. An approach for generating the fuzzy rules, representing the structure of the controller based on the data, which are collected during the process of control, is proposed. It is based on the recently introduced technique for on-line identification of Takagi-Sugeno type of fuzzy models. The controlled plant's output and the control signal are used to train on-line in a non-iterative, recursive way the Takagi-Sugeno fuzzy controller. The control rule-base structure is gradually evolving taking into account the new data and exploiting the quasi-linear nature of Takagi-Sugeno models. This approach is combined with the indirect adaptive control approach. A practical engineering example of an air-conditioning system control is given with illustrative purposes.