Study On Adaptive Pid Control Strategy Based On Actor-critic Learning

PID controller is the most commonly used controller in modern industrial control.It has the characteristics of simple structure,easy to implement,good control effect and strong robustness.It can meet the general industrial control requirements.However,with the progress of technology,demand for control accuracy is constantly improved.The electro-hydraulic servo system is nonlinear,time-varying,uncertain and disturbing.In order to achieve high precision control,the parameters of the controller should be adjusted automatically as the system changes.But the conventional PID controller can't be changed once the parameters are determined in the control process.Therefore,the PID controller is often difficult to achieve the ideal control effect for high demand systems.Due to the rapid development of artificial intelligence technology,the researchers combined artificial intelligence algorithm with PID control to create the adaptive PID controller.It is not dependent on precise system model,and has good adaptability to system changes.As a method of machine learning,reinforcement learning emphasizes the valuation of the return signal in interaction with the environment,and aims to maximize future returns.It does not need the teacher's signal,and it is widely used in solving the complex decision problem with less prior information.In many areas,its performance advantages have been demonstrated.In this paper,an adaptive PID control strategy based on Actor-Critic structure and RBF network structure is proposed.It doesn't require prior knowledge,and can adjust parameters online.Through the simulation analysis of Simulink and the experimental verification in Yanshan University,the results show that compared with the traditional PID controller and other algorithms,our controller system has better dynamic response characteristics and anti-disturbance capability.