In this thesis, the learning methods of the direct neural network controller having the multilayer perceptron are studied by giving the appropriate cost functions to control the system response. This study represents the feasibility of the direct neural network controller via the simulations.
Also, the stability problems are investigated in the sense of the Lyapunov stability conditions. And, a new learning algorithm for the multilayer perceptron is proposed by giving the stability of the new learning algorithm. Some simulation results are given to show the validation of this study.