Nonlinear adaptive control using recurrent fuzzy models with stability analysis

In this paper, an indirect adaptive control scheme based on Takagi-Sugeno (TS)-type recurrent fuzzy models is proposed for nonlinear plants with unmeasurable states. The TS-type recurrent fuzzy model is used as the dynamic model of the nonlinear plant. Its recurrent property comes from that it can memorize temporal information with the feedback connections between its states layer and inputs layer, which makes it capable of more powerful learning ability compared with ordinary TS fuzzy models. The parameters of the model are adapted online by using gradient based neural network learning methods to allow for partially unknown or time-varying plants. The controller is designed completely based on the model structure, parameters and states. Comprehensive convergence analysis of the proposed adaptive nonlinear control schemes is studied and stability conditions are given. The effectiveness of the proposed control scheme is finally demonstrated by simulation examples.