Optimization-based fuzzy iterative learning control

In this paper a new approach to fuzzy iterative learning control is presented. In the proposed approach, coefficients of fuzzy system and learning rate of ILC are calculated using optimization algorithms such as steepest descent and genetic algorithm. The optimization algorithm must be applied to a predetermined number of iterations to determine unknown coefficients, offline. Calculating fuzzy coefficients and learning rates, the controller performs with optimum coefficients for any required number of iterations. Therefore one of the main advantages is that learning rates can be determined in an optimal manner and the controller operates completely non-model based. These features make the method appropriate for highly nonlinear systems with model uncertainties. The overall developed fuzzy iterative learning controller is called FILC that not only keeps the advantage of ILC, but also creates appropriate updating law by using fuzzy TSK system. The controller is applied to electro hydraulic servo actuator as a highly nonlinear and uncertain case study and simulation results show superior advantages of the method.