A generalized fuzzy adaptive control method

This paper presents and discusses the architecture and learning process of an adaptive fuzzy control methodology. This methodology combines fuzzy decision implementation in the form of linguistic rules and a mechanism to fine tune the initial fuzzy plant identifier and fuzzy controller linguistic rules simultaneously using a gradient descent method. The non-optimal linguistic rules are refined online by the adaptation and learning mechanism to maintain a consistent desired optimal control performance. An analytic dynamic plant Jacobian is estimated via a parallel forward fuzzy plant identifier model of the plant because the plant in this control scheme is situated between the controller and the error to be fed back. The use of an analytic Jacobian matrix gives additional robustness to this control scheme. The computer simulation results have shown that the designed fuzzy controller using this methodology is capable of providing good control system performance and effective control of nonlinear dynamic systems