Tracking Control Of Inverted Pendulum System Via Improved Neural Network Control Algorithm

The inverted pendulum system is a kind of classical plant in the control theory research and has been widely used in the fields of aerospace and robotics.Because the inverted pendulum system has unstable,multivariable,strongly coupled and nonlinear characteristic,its trajectory tracking problem is challenging.Considering that most controllers are based on digital realization,it is more practical to study the trajectory tracking of inverted pendulum system in discrete time framework.Recently,some researches have applied discrete time neural network method based on output regulation theory to solve the trajectory tracking problem of some typical inverted pendulum systems.The key of this method is to solve the discrete regulator equation composed of several nonlinear algebraic functional equations.However,the calculation process of the approximate solution of the discrete regulator equation is tedious and depends on the system parameters.Therefore,in order to reduce the computational complexity and achieve higher accuracy of position control,this thesis directly approximates the feedforward function by neural network,and then proposes an improved discrete time neural network control algorithm.In addition,although the discrete time nonlinear output regulation theory has achieved prosperous development,the experimental results of the discrete time nonlinear output regulation theory are rarely reported.Therefore,the experimental study of improved discrete time neural network control algorithm also has great significance.This thesis mainly studied the position tracking problem of two typical inverted pendulum systems,including the spherical inverted pendulum system with multi-inputs multi-outputs and the linear motor inverted pendulum system with single input single output,and carries out relevant experimental research on the linear motor inverted pendulum system.The main research work is as follows:(1)The position tracking problem of the spherical inverted pendulum system is studied.In this thesis,it is first described as a discrete time nonlinear output regulation problem,and then an improved discrete time neural network controller is proposed based on the output regulation theory.This controller applies the neural network approximation of feedforward function instead of the neural network approximation of the solution of the regulator equation,which greatly reduces the complexity of calculation.Simulation results show the effectiveness and superiority of the method.(2)The position tracking problem of the inverted pendulum system of linear motor is studied.In this thesis,it is first described as a discrete time nonlinear output regulation problem,and then an improved discrete time neural network controller is proposed based on the output regulation theory.In addition,considering that the friction of linear guide rail is inevitable in the experiment,an improved discrete time neural network controller with adaptive friction compensation is proposed to improve the tracking performance of the system.Finally,the experiment is carried out on the experimental platform of the linear motor inverted pendulum system based on c SPACE.The experimental results verify the effectiveness of the method and bridge the gap between the theoretical research and the practical application of discrete time nonlinear output regulation.