Within the confines of my thesis, I have become more acquainted with the role and potential of using neural networks for nonlinear system controlling. During my task I managed to successfully process and improve the publication of Yi Zhang, Chun Feng and Bailin Li called „PID Control of Nonlinear Motor-Mechanism Coupling System Using Artifical Neural Network”. The examined thesis also can be found in the book called: „Advances in Neural Network ISNN 2006 Part 2”.
Foremost I dealt with the Neural Network PID controller described in the article. The NNPID controller essentially is a discrete time PID controller tuned by the neural network in an adaptive manner. The outputs of the NN provide the time constants and the gain for the PID controller.
During accomplishing my thesis I made the model of a linear system, then designed a traditional PID controller for the system. In course of time I managed to copy the behavior of the PID controller with the NNPID in Matlab environment. Finally I tested the adaptive capabilities of this type of control.
Finally I tried to improve the original NNPID controller, to achieve better performance in the case of nonlinear systems. I designed a state-feedback control for a nonlinear system and also managed to copy the behavior of the state-feedback controller.