Research Of Adaptive Neural Network Study Arithmetic On Inverted Pendulum System

As a typical experiment device, inverted pendulum system has the performances of high-order,nonlinear unstable,multi-variable,strong-coupling and so on. Its structure is similar as control object in engineering. Therefore, the research on inverted pendulum control is very important in the theory and the aerial application. For all kinds of complex objects under control, the precision control on inverted pendulum system has broad instruction meaning.In this paper, the control objects are linear one-stage,two-stage and three-stage inverted pendulum provided by Googol Technology of Shenzhen. Based on establishing the mathematical model, we use LQR control method and adaptive neural network control method to implement the simulation and real-time control research on one-stage and two-stage inverted pendulum. In this paper, we focus on the following problems:First, we adopt the Newton mechanics method and the Lagrange equation to establish the mathematical model of one-stage,two-stage and three-stage inverted pendulum,then take the qualitative analysis. The LQR and PID control algorithm for the one-stage,two-stage and three-stage inverted pendulum were researched. The method of how to select the Q and R weighting matrix was discussed. Then we performed simulation and real-time control on one-stage and two-stage inverted pendulum, got the response curve when it was stable. The results showed that the LQR control method was better than other methods.Second, we adopt the adaptive neural network based fuzzy inference method to control the inverted pendulum, combined the fuzzy control into the neural control.This method can improve the capability of the fuzzy controller,self-correcting the membership function and the control rules.We can reduce the input dimension and the control rules to avoid the "rule explosion"by integrate modulus. When the model parameters were changed, the adaptive neural network based fuzzy inference system had good adopt ability to anti-interfere. The car can go to the destine position exactly.At last, the anti-interfere capability for one-stage and two-stage inverted pendulum was researched. The results showed that adapt the adaptive neural network based fuzzy inference system not only can control the inverted pendulum system stably but also had strong ability to anti-interfere.