基于深度学习的控制器设计研究

Research on Controller Design Based on Deep Learning

深度学习目前在学术和工业领域中具有非常重要的地位,深度学习在特征提取与模型拟合方面存在相当的优势。对于存在高维数据的控制系统,引入深度学习具有一定的意义。论文介绍了使用深度学习算法来设计控制器,探索将深度学习应用于控制领域的情况。论文是通过让深度学习控制器来模拟控制领域中经典的PID控制器,来控制带负载的直流电机。用PID控制器的输入/输出用作深度学习控制器的训练数据集,训练了一个基于深度置信网络(DBN)算法的深度学习控制器,并对两种控制器进行了详细的比较,论文使用Matlab/Simlink进行了仿真,证明了深度学习控制器可以控制带负载的直流电机并且与PID控制器相比,控制效果良好。

At present,deep learning plays an important role in the academic and industrial fields.Deep learning has considerable advantages in feature extraction and model fitting.For control systems with high-dimensional data,the introduction of deep learning has certain significance.This paper introduces how to use deep learning algorithm to design controller and how to apply deep learning to control field.In this paper,the deep learning controller is used to simulate the classical PID controller in the control field to control the DC motor with load.Using the input/output of the PID controller as a training data set for a deep learning controller,a deep learning controller based on a deep belief network(DBN)algorithm is trained,and a detailed comparison of the two controllers is performed.Matlab/Simlink has carried out simulations and proved that the deep learning controller can control a DC motor with a load and the control effect is good compared with the PID controller.