Study On Inverted Pendulum Controlmethod And Application Based On Neural Network

Inverted pendulum is a natural unstable system with multivariable, nonlinear and strongcoupling. It is as an ideal experimental platform for the research of control theory, which cantest the control system controllability, robustness, anti-jamming capability and convergencespeed etc. Inverted pendulum experiment could verify the new control theory or method. Itpromotes the application of inverted pendulum.It analyzes the physical model of linear pendulum inverted, to establish the mathematicalmodel of inverted pendulum. And it uses the SimMechannics toolbox to establish the physicalmodel of inverted pendulum. Comparing the two models, it chooses the simple and intuitivephysical model as the research object, and using modern control method validate the validityof inverted pendulum model.Based on discussing the advantages and disadvantages of neural network, it points outthat the self-learning ability of neural network to make up for the deficiency of traditionalcontrol method, and it provides a new way for solving control problems. It analyzes theprinciple of BP (Back Propagation) algorithm and based on the existing experimentalconditions, it proposes use parallel structure to process information. Due to the use of theminimum error method, it is easy to realize the network convergence.It uses multilayer feedforward neural network to control the inverted pendulum systemand verifies the defect of traditional BP algorithm, such as: slow convergence, poor stabilityof system. Levenberg-Marquardt (LM) algorithm and elastic gradient algorithm could solvethe bottleneck problem of traditional neural network; It guarantees the network in the wholelearning process has high learning efficiency, improving the convergence speed. LMalgorithm is the most used algorithm in neural network control; It is widely used in control field.We use BP neural network with the LM algorithm to realize the control of the invertedpendulum, and using the SimMechannics toolbox for simulation, through the invertedpendulum experiment platform to complete real-time control. The experimental results showthat it is a good way to realize the control of inverted pendulum, inverted pendulum systemcan quickly achieve stability, and it has strong anti-interference capability and adaptability.