Font Size: a A A

Design Of Intelligent Control Algorithm And Realization In Inverted Pendulum

Posted on:2019-09-29Degree:MasterType:Thesis
Country:ChinaCandidate:X S QiFull Text:PDF
GTID:2428330563957069Subject:Control engineering
Abstract/Summary:PDF Full Text Request
Reinforcement learning algorithm and Variable universe fuzzy algorithm are emerging intelligent control algorithm.The learning system in the reinforcement learning algorithm is learned by interacting with the environment and does not require or less require a prior knowledge,to solve the uncertainty of unknown variety which the nonlinearity of the controlled object and system modeling imperfect.Because the high precision control needs a lot of control rules,the traditional fuzzy control is not suitable for high precision occasions.And the variable universe fuzzy control algorithm dynamically increases the number of control rules,not only control increased number of fuzzy rules but also can achieve high precision control.The inverted pendulum is a typical multivariable,nonlinear,strongly coupled and unstable high order system that can be used to test the nature of the algorithm,control performance and feasibility.Using MATLAB / SIMULINK as the software development platform,with the straight inverted pendulum as the hardware experimental equipment,designed and completed the reinforcement learning algorithm and variable universe fuzzy control algorithm simulation.The followings are main works:1.Write the physical model program of linear inverted pendulum.Using the M file program to write the Q-learning algorithm process and the nonlinear dynamic equation code,which the experimental model is straight line inverted pendulum.2.Design Q-learning reinforcement learning algorithm.Design of Q-learning reinforcement learning algorithm for the return function,reward and punishment functions and value functions,which as a controller for linear first-order inverted pendulum simulation design,verify the characteristics and feasibility of the algorithm.3.Design variable universe fuzzy control algorithm.Using Newton mechanics and Lagrange method to construct a physical model of linear inverted pendulum as a control object.Designing controller based on fusion dimensionality reduction and variable universe fuzzy control algorithm,achieve simulation program of linear triple inverted pendulum,get the feasible conclusion of the algorithm.4.Design real time control program of three-stage inverted pendulum.Using real-time control software module of Googol company,design Real-time software control program of three-step inverted pendulum.Using S-function of the c language to design the variable universe fuzzy control algorithm,and the algorithm controller is used as a straight-line three-stage inverted pendulum for the real-time control program.This article is based on the mathematical model of a straight inverted pendulum,and the vertical position is a controllable condition.After the reinforcement learning algorithm controller and variable universe fuzzy controller were designed,simulation experiments were conducted.The results show that both control algorithms are effective;When the experimental reinforcement learning algorithm is used,the reinforcement learning system knows nothing about the environment,and only through the(action-return)function to judge the next behavioral action,the self-study characteristics of the straight inverted pendulum are obtained.The variable universe fuzzy controller highlights the characteristics of real-time,more suitable for high-order unstable systems such as linear three-stage inverted pendulum.
Keywords/Search Tags:reinforcement learning algorithm, variable universe fuzzy algorithm, linear inverted pendulum, real-time control, S function
PDF Full Text Request
Related items