Study on Neural Network Controller Based on Embedded System

As traditional PID controller has mature technique, it is applied widely. But the design of it depends on mathematical model of the controlled object. But the parameters of PID controller use engineering tuning method, it requires much time and effort and the parameters are only effect in a specific range. It is not suitable for complicated nonlinear and time-variable system. In recent year, people have begun to combine neural network and PID controller, and used neural network to improve traditional PID controller. The paper analyzes and studies PID controller based on neural network setting parameters and adaptive PID controller based on single neuron, and uses inverted pendulum as nonlinear object for simulation study. The paper compares and analyzes the control performance of neural network PID controller and traditional PID controller. The designed embedded neural network controller has low cost, fast response speed, good control performance and better operability and portability. The controller can be improved and designed to be a functional and practical industrial controller.