A New Type of Adaptive Neural Network Fuzzy Controller in the Double Inverted Pendulum System

摘要

A new type of adaptive neural network fuzzy controller based on the stability for the double inverted pendulum control problem is introduced. The method uses a fusion function to reduce the dimension of the system, reducing the number of input variables to solve the fuzzy rule explosion problem. In order to optimize and amend the front-part and later-part parameter of Takagi-Sugeno fuzzy model, a mixed algorithm of backward propagation (BP) and least square method (LSE) algorithm are used. Using the collected original input and output data to establish adaptive neural network fuzzy inference system (ANFIS), and to control the double inverted pendulum system. Simulation results show that the controller is better than the other controller.