Research On Control Based On Linear Inverted Pendulum

Inverted pendulum system is an ideal platform to study and verify various control methods and control theories,and is widely used in industrial production and aerospace fields.Because of the nonlinear,strong coupling and uncertainty of the inverted pendulum system,the control difficulty of the system will be greatly increased.Based on the linear inverted pendulum system,this paper studies the stable pendulum control and self starting pendulum control of the system respectively.Firstly,the research and development status of inverted pendulum system and neural network are described.The simplified mathematical model of inverted pendulum is established by Newton Euler method and Lagrange method respectively,and the simplified mathematical model is verified.According to the relevant theorems of observability and controllability,the performance of the system is analyzed.Secondly,the pendulum stabilization control of inverted pendulum is carried out from the following two points: first,the classical PID controller is a single variable controller,which can not meet the control requirements of multivariable inverted pendulum system.In this paper,a double closed-loop PID controller is designed,and a single neuron is introduced to solve the problem that it is difficult to adjust the parameters of the PID system.Because the fixed gain K value in single neuron is difficult to achieve the ideal control effect,a single neuron adaptive control strategy based on PSD control algorithm is proposed;Second,a single neuron is used to construct a BP neural network controller,and the data set of the training network affects the accuracy of the network controller.In this paper,the inverted pendulum is simulated by display model predictive control(EMPC),and the constraints of the system are added to the controller to simulate the inverted pendulum system more close to the actual running state.The output data set of model predictive control is used as the "teacher" of neural network training.Finally,the control of the system self swing up is studied.Based on the basic principles of BangBang feedback control and energy feedback control,a self starting pendulum controller is designed.Combined with the pendulum stabilizing controller of BP neural network,the whole control process of inverted pendulum system from starting to stabilizing is simulated.In this paper,the PID parameters are optimized online by using the self-learning and adaptive characteristics of single neuron,and the model predictive control closer to the actual system is used to obtain the training set of BP neural network,and the swing stabilization control of inverted pendulum is realized by simulation.At the same time,the swing up controller designed by using Bang-Bang feedback control and energy feedback control principle can be simulated,and the system can swing up successfully in a short time.