Research On Modeling And Reinforcement Learning Control Of Rotary Inverted Pendulum

Rotating inverted pendulum is a typical high-order,multi-variable and nonlinear complex system,and it is a classic model in the research and teaching of control theory.Researchers often verify advanced control methods based on the rotating inverted pendulum,and it is often used in teaching to explain control theory knowledge.An accurate mathematical model is the prerequisite for using the inverted pendulum in research and teaching.However,due to the complexity of the inverted pendulum,it is usually difficult to establish a mathematical model that matches the actual prototype.Therefore,it is of great significance to study how to establish an accurate model for the actual electromechanical system.The traditional control methods of rotating inverted pendulum,such as PID control and LQR control,need to constantly adjust the relevant parameters in order to achieve the ideal control effect.Adjusting the experimental parameters mostly depends on the experience of the researchers,which will greatly increase the workload of the researchers and the uncertainty of the control effect.Reinforcement learning continuously learns experience from the environment through the interaction between the agent and the environment,and automatically updates the control strategy to achieve the optimal control effect.Therefore,it greatly reduces the design difficulty and manpower input of the control system,and has broad application prospects.At the same time,the rotating inverted pendulum is used as a research platform for reinforcement learning algorithms,which can explore the practicality of reinforcement learning.First,the application of reinforcement learning to inverted pendulum control is highlighted.Introduces the theoretical basis of reinforcement learning and its related algorithms.Then,the working principle and characteristics of the rotating inverted pendulum are discussed.The research status of reinforcement learning algorithm in inverted pendulum control is further outlined,and the existing problems in the current research are pointed out.Aiming at these problems,an appropriate reinforcement learning algorithm and experimental platform are selected for research.Finally,a reinforcement learning method for the balance control of a rotating inverted pendulum is presented.Secondly,the dynamic equation of the rotating inverted pendulum is established by using the Lagrange method,and the parameters that cannot be directly measured in the equation are identified through carefully designed experiments,so as to improve the degree of agreement between the simulation model and the experimental prototype and the fidelity of the simulation experiment.Then,through the RL Toolbox and Simulink in MATLAB,the simulation environment is built,the DDPG agent is created,and the DDPG agent is trained.The design of the balance controller of the rotating inverted pendulum based on the DDPG algorithm is completed.Finally,the software and hardware system of the rotating inverted pendulum is built,and the simulation and physical experiment of the rotating inverted pendulum balance control are completed based on the DDPG algorithm,which verifies the feasibility and effectiveness of the reinforcement learning algorithm in the rotating inverted pendulum control.