Research Of Fuzzy Neural Network Controller Based On Genetic Arithmetic

The modern industry systems control becomes more and more complex with the rapid development of science and technology. Conventional control science faces new challenges. However, the development of artificial intelligence, system science and information science provide an intelligentized road for the control science. At present, fuzzy logic, artificial neural network and genetic arithmetic are three important fields in artificial intelligence science. Researchers pay great attention to the three because of their great future. And it is one of the most important problems in control science to make use of their advantages for intelligent control field.This paper is to do research on fuzzy neural network controller based on genetic arithmetic. At first, this paper studies theory about genetic arithmetic, analyses shortcomings of the Standard GA (genetic arithmetic). And it designs a kind of Improved GA by making same improvement in the Standard GA. Using Standard GA and Improved GA in the Double Integrator optimized problem, the results of simulation prove that the Improved GA has solving some shortcomings of the Standard GA. Then, this paper studies theory foundation of fuzzy control and neural network control. And it analyses shortcomings of both each and the inevitability of their combination. On this foundation, designs a fuzzy neural network controller based on T-S fuzzy model. At the same time, presents using Improved GA and BP arithmetic optimize parameters of the controller together. Finally, this paper studies the structure and characteristics of the Cart & Pendulum system, designs a fuzzy control system of Cart & Pendulum, a neural network control system of Cart & Pendulum and a fuzzy neural network, based on control system GA of Cart & Pendulum. The results of simulation experiment of the three control systems prove that the fuzzy neural network controller based on genetic arithmetic designed in this paper could be used for non-linear object control, and showing good performance and robustness.