基于深度强化学习的物联网动态聚类方案  

Dynamic Clustering Scheme for Internet of Things Based on Deep Reinforcement Learning

作　　者：吴宗卓 WU Zongzhuo(School of Computer Science and Software, Shaanxi Institute of Technology, Xi’an 710300, China)

机构地区：[1]陕西国防工业职业技术学院,计算机与软件学院,陕西西安710300

出　　处：《微型电脑应用》2022年第6期31-33,46,共4页Microcomputer Applications

基　　金：陕西省教育厅2019年度专项科学(19JK0085)。

摘　　要：为了提高物联网(IOT)对大型数据的采集性能,人们提出了多种网络聚类方法。然而,它们大多侧重于用静态拓扑划分网络,因此在处理网络中移动对象的情况时,它们不是最优的,并且都没有考虑过边缘服务器的计算性能。针对这些问题,提出了一种基于深度强化学习(DRL)边缘计算的高效物联网动态聚类解决方案。这种方法既能满足物联网的数据通信需求,又能满足边缘服务器的负载平衡需求。本文使用一个深度Q学习网络(DQN)模型来实现该方法。初步的实验结果表明,与目前的静态基准解决方案相比,DQN解决方案可以获得更高的聚类划分分数。In order to improve the performance of large-scale data collection in the Internet of Things(IOT),many network clustering methods have been proposed.However,most of them focus on dividing the network with a static topology,so they are not optimal when dealing with moving objects in the network.And they have not considered the computing performance of the edge server.In response to these problems,a dynamic clustering solution for efficient Internet of Things in edge computing based on deep reinforcement learning(DRL)is proposed.The method can not only meet the data communication requirements of the Internet of Things,but also meet the load balancing requirements of the edge server.A deep Q-learning network(DQN)model is used to implement this method.Preliminary experimental results show that the DQN solution can obtain a higher clustering division score than the current static benchmark solution.

关 键 词：边缘计算 深度强化学习 动态集群 物联网 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]