RBF-ARX Model-Based Predictive Control Applied To Inverted Pendulum System

The RBF-ARX modeling is one method which combine ARX model with the RBF neural network and can descript the system character.It is a new branch of "the intelligent control",and opens a new way for the solution complex nonlinear control question.Predictive control based on RBF-ARX is a new forward problem on study in the field of complex nonlinear industrial process control.Firstly,The development of inverted pendulum is introduced.The background,the development and applications of RBF-ARX are discussed.The origin and applications of predictive control are also summarized briefly.Secondly,The structure,the identification and parameter optimization based on RBF-ARX model are discussed strongly. RBF-ARX model is used to build the global model of a single inverted pendulum system for representing the system's nonlinear dynamics.The structure is designed and the parameter is estimated off-line by SNPOM which is a fast-converging hybrid method.The predictive outputs and modeling errors between the RBF-ARX model and a global linear ARX model,as well as between a local model from estimated RBF-ARX model and a local linear ARX model of the inverted pendulum system are compared respectively.The results show the effectiveness of RBF-ARX model to the modeling of inverted pendulum system and also provide a foundation based on modeling optimized control.Thirdly,Aiming at the typical predictive control algorithm,the basic structure and theory of the predictive control are discussed.A detailed analysis of generalized predictive model including the basic principle and the advantages and disadvantages in the application is made.Finally,the predictive control based on the local RBF-ARX model are applied to the inverted pendulum control system successfully.The simulation results show that the predictive errors of RBF-ARX model of the system are so small that we can conclude that the RBF-ARX model of the system can describe the system's local dynamical characters fully.The control performance of predictive control based on RBF-ARX is much better than the control performance of predictive control based on local linear ARX model and LQR control.