Research On Stability Control Of Rectilineal Inverted Pendulum

Inverted pendulum is a typical model of multivariable, nonlinear, essentially unsteady system. It can effectively reflect many key problems such as stability, robustness and tracking. So it is a perfect model used to test various control strategies. This thesis builds a real-time control workspace utilizing the MATLAB RTW Real-Time Toolbox, which administrates the external hardware to control the inverted pendulum by downloading the simulation model generated by Simulink to the target real-time kernel. Based on the inverted pendulum of GUGAO Company in Shenzhen, Different kinds of control theories have been analyzed and their applications have been realized, corresponding conclusions are given simultaneously in the thesis. The main research of the thesis is as follows:At first, the background and research situation of the inverted pendulum are introduced, meanwhile, various kinds of control methods of the system are presented. The model of the system is established and the state space expression of the model is deduced.Secondly, this thesis carries out the optimal LQR control of inverted pendulum based on Elitist Preserving Genetic Algorithm. Also, Adaptive Neural-Fuzzy Inference System (ANFIS) is explained and used to the real time control of single inverted pendulum.Further more, based on the theory of Support Vector Regression(SVR), a fuzzy inference system based on support vector machines is proposed ,which is proved that it has better control results than neural network through inverted pendulum control experiment.It is proved that LQR Controller based on Elitist Preserving GA is realizable inEngineering practice, which has been widely used in inverted pendulum control. But this method depends on linearization of model which can not be easily obtained in complicated industrial process. Adaptive Neural-Fuzzy Inference System (ANFIS) is capable of nonlinear system identification. But the structure of the system must be determined in advance. Support Vector Machines (SVM), as a novel method to extract fuzzy if-then rules, automatically extracts the optimal number of support vectors for further application in the IF-THEN rules generation. It is applicable to many practical and complicated situations, where the number of fuzzy rules can not be determined in advance. Fuzzy rules extraction based on SVM will be interested in nonlinear system control in future.