Reserch On Stability Control Of Triple Inverted Pendulum Via Fuzzy Neural Network

The inverted pendulum system is characterized as a typical high-order, multi-variable, unstable, strong-coupling nonlinear system. And it is well known as an effective experiment appliance for control theory and techniques. Fuzzy-neural network, which combine fuzzy logic inference with neural network appropriately, has advantages of both fuzzy logic and neural network. It is such an exceptional control strategy that it has become a hot research field in intelligent control. Stability control system of triple inverted pendulum based on fuzzy neural network is researched. Solutions of the rule number explosion in multi-variable system by using fuzzy neural network are researched in this paper.Firstly, this article introduces the structure and moving characters of triple inverted pendulum system. The mechanism and mathematical models of triple inverted pendulum are presented by employing lagrange equation method. A complete, accurate, non-linear simulation model is established under the Simulink environment.Secondly, fuzzy-neural network controller is designed for triple inverted pendulum. The rule numbers of fuzzy neural network controller is reduced by designing a fusion function. An algorithm of variable universe of input variables is designed for improve the control accuracy. Comparison of fuzzy neural network controller, LQR controller and fuzzy logic controller is done in Matlab/Simulink simulation environment. Contrast analysis proves that the fuzzy neural network controller has better control capability.Finally, this article analyzes the shortcomings and deficiencies of the method which reduces the rule numbers by using fusion function. Then two improved programs are given to solve the problem based on state priority. The first solution improves the structure of controller based on " seize the big and free the small" principle: a fuzzy-neural network by using feed-forward compensation. The second solution improves the design method of fuzzy rules combined theory of cloud model: a fuzzy-neural network by using rules priority. The improved controller is applicable to the multi-variable system that its state variables can be distinguished between the kind and importance. It can reduce the rule numbers of fuzzy neural network sharply and make good use of expert experience. Simulation results and contrast analysis prove that the improved controller has better robustness and stronger nonlinear adaptive ability.