首页 | 官方网站   微博 | 高级检索  
     

基于强化学习的网络时延自动化控制数学建模研究
引用本文:荆荣丽,葛书荣,王鹏,宁玉文.基于强化学习的网络时延自动化控制数学建模研究[J].自动化与仪器仪表,2020(3):57-59.
作者姓名:荆荣丽  葛书荣  王鹏  宁玉文
作者单位:安康职业技术学院;空军军医大学网络中心
基金项目:国家自然科学基金(No.U1608253,91748208)。
摘    要:传统的网络时延控制模型在分析时延原因时,仅从宏观角度分析,缺少建立网络模型的过程,导致时延控制能力差、数据传输时间长、丢包率大的问题。为解决此问题,设计一种基于强化学习的网络时延自动化控制模型。该模型的构建主要分为两部分,先是确定网络模型,具体分析网络时延出现的原因,在此基础上,利用强化学习中的Q学习算法构建自动化控制模型,以解决网络时延问题。实验结果表明:与传统的基于均衡调度的网络时延控制模型相比,该模型对网络时延的控制性能更好,且数据包传输时间缩短3.7 s,数据包丢包率降低5%,应用优势明显。

关 键 词:强化学习  网络时延  自动化控制  Q学习算法

Research on mathematical modeling of network delay automation control based on reinforcement learning
JING Rongli,GE Shurong,WANG Peng,NING Yuwen.Research on mathematical modeling of network delay automation control based on reinforcement learning[J].Automation & Instrumentation,2020(3):57-59.
Authors:JING Rongli  GE Shurong  WANG Peng  NING Yuwen
Affiliation:(Ankang Vocational and Technical College,Ankang Shanxi 725000,China;Network Center of Air Force Military Medical University,Xi’an 710032,China)
Abstract:When analyzing the causes of delay,the traditional network delay control model only is analyzed from the macroscopic perspective,and it lacks the process of establishing network model,which leads to the problems of poor delay control ability,l ong data transmission time and high packet loss rate.To solve this problem,a model of network time delay automation control is designed based on reinforcement learning.The construction of this model is mainly divided into two parts.First,the network model is determined,and the causes of network delay are analyzed in detail.On this basis,the Q learning algorithm in reinforcement learning is used to build an automatic control model to solve the network delay problem.The experimental results show that compared with the traditional network delay control model based on balanced scheduling,this model has better performance in controlling network delay,and the transmission time of packet is shortened by 3.7s and packet loss rate is reduced by 5%.
Keywords:reinforcement learning  network delay  automatic control  Q learning algorithm
本文献已被 维普 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司    京ICP备09084417号-23

京公网安备 11010802026262号