Design of an iterative learning controller for nonlinear systems with time-varying using ANN and GA

Finding an effective neuron topology method and an optimal algorithm to adjust the parameters in using artificial neural network (ANN) is the key process of practical nonlinear system control. This paper presents the new control structure that iterative learning of neural network by genetic algorithm (GA) is possible. GA is used to optimize neural network topology and connection weights. Our method is different from those using supervised learning algorithms, such as the back-propagation (BP) algorithm, that needs training information in each step. Simulation results verify excellences of the proposed iterative learning controller using neural network and GA. The contributions of this paper are the new approach method to construct neural network architecture and its training for nonlinear system control.