Research On Adaptive Control Based On CMAC For Nonlinear Systems

The rapid development of computer and science technology has led to a close combination of artificial intelligence technology and automatic control theory,which has formed the research hotspot in the field of control--intelligent control theory.In practice,control system is a nonlinear system and there is a difference between mathematical model and practical system.Also the structure and parameters of the system itself is unknown or time-varying and stochastic disturbance,which cannot be measured in many cases.As a result the artificial intervention is not realistic in real time.Therefore,the study of adaptive control of nonlinear systems has great theoretical significance and application value.Cerebellum model(CMAC)neural network not only has the nonlinear approximation ability,adaptive and generalization ability,and associative memory ability,and it is a kind of fast convergence speed of neural network,which is very suitable for real-time nonlinear control system.Therefore,this research include CMAC neural network and its data classification,focusing on its application in nonlinear system modeling and robust adaptive control.The main research results of this paper are summarized as follows:1.In view of the high misdiagnosis and missed diagnosis in identification of a breast tumor using traditional method,put forward a kind of self-validation cerebellum model joint controller(SVCMAC)neural network,the diagnosis result has high prediction accuracy and low false negative rate.The SVCMAC neural network has better learning and prediction ability based on the combination of fast learning speed of CMAC network,strong generalization ability and fast response speed,and combined with self-calibration unit.2.Proposed a CMAC adaptive control(PICMAC)system based on PI.The PICMAC system consists of a neural controller and a robust compensation controller.The neural network controller USES a PI type CMAC to simulate an ideal controller to accelerate the convergence of the tracking error,and the robust compensation controller is used to eliminate the approximate error introduced by PI type CMAC.The stability of the closed-loop system and the control effect of the Gencesio chaotic system are derived.3.A robust adaptive TSK fuzzy small brain model joint controller(RATFC)is proposed,which is applied to the control of manipulator arm,and obtains high precision position and velocity tracking control.The TFCMAC is used to approximate the ideal controller,and the parameters are adjusted according to the adaptive rules derived from lyapunov function.In order to suppress the influence of approximate error,robust compensation controller is designed.4.This paper presents a fuzzy brain emotional cerebellum model joint controller(FBECMAC)for a multi-input multi-output(MIMO)nonlinear system,applied to the UAV(UAV)control,the state control to the ideal target,especially in the trajectory tracking problem of UAV is significant.Fuzzy inference system,brain emotional lcarning,and CMAC are combined,making the control system has the advantages of fuzzy system and the brain mood the advantages of fast learning,with robust compensation controller at the same time,to suppress approximation error and improve the system stability.