Research Of Intelligent Control For Inverted Pendulum

The inverted pendulum system is a typical single-input and multiple-output, non- linear, complex and unstable system. The research on such a complex system involves many important theory problems about system control, such as nonlinear problems, robustness, ability, and tracking problems. At the same time, the control research of the inverted pendulum has important engineering background; its control methods have extensive use in military, aerospace, robotics and general industrial process areas.The paper is based on the digital inverted pendulum of Feedback Company. Simulations and experiments are done on the inverted pendulum using several kinds of control algorithms and draw the corresponding conclusions.A state feedback controller is designed based on LQR method. Matlab simulation and Real time experiment are done on the linear model of the inverted pendulum system, and the results indicate the controller can keep the pole of the inverted pendulum stable in a small vertical upward region.Genetic Algorithms is used for adjusting parameters of a PID controller. Crane mode simulation and Real time experiment of the inverted pendulum obtains satisfying control results.The paper also researches the fuzzy control method of the inverted pendulum system and designs a fuzzy inference system based on Mamdani type controller. The number of rules is greatly reduced by setting up the cart and pendulum rules separately, the results indicate that this controller not only stabilizes the system, but also has strong anti-interference capability.Considering NN fuzzy control idea, the paper designs an initial fuzzy inference system based on the T-S model using Matlab fuzzy toolbox. Membership functions of the variables are adjusted after the learning on the sample data. After the training, an effective fuzzy controller is obtained. Simulation achieves better control results.In addition, the paper uses a neural network identification software packages (nnsysid) to identify the inverted pendulum system. The structure of the various models and identification results are discussed. It proved feasible to build the neural network model of the inverted pendulum using the neural network structure of input and output delay.