基于图卷积神经网络的超密集物联网资源分配策略  

Resource allocation strategy for ultra-dense Internet of things based on graph convolutional neural network

作　　者：黄杰 李幸星 杨凡 丁睿杰 蔡杰良 姚凤航 张鑫 HUANG Jie;LI Xingxing;YANG Fan;DING Ruijie;CAI Jieliang;YAO Fenghang;ZHANG Xin(School of Electrical and Electronic Engineering,Chongqing University of Technology,Chongqing 400054,China)

机构地区：[1]重庆理工大学电气与电子工程学院,重庆400054

出　　处：《通信学报》2024年第10期243-252,共10页Journal on Communications

基　　金：国家自然科学基金资助项目(No.62301094);重庆市教育委员会科技研究基金资助项目(No.KJQN202201157,No.KJQN202301135);重庆理工大学科研创新团队培育计划基金资助项目(No.2023TDZ003)。

摘　　要：针对超密集物联网(UD-IoT)中存在大量隐藏终端干扰严重影响资源管理问题,提出了一种基于图卷积神经网络的深度确定性梯度的超密集物联网资源分配策略。通过矩阵变换构建冲突图模型,采用极大团和超图理论将冲突图模型转化为冲突超图模型,进而将无冲突资源分配问题转化为超图顶点着色问题,并提出了一种基于图卷积神经网络的深度确定性梯度的超密集物联网资源分配算法,采用图卷积强化学习实现无冲突资源分配和资源复用率最大化。仿真实验表明,所提算法具有更高的资源复用率和吞吐量,可以在超密集物联网中提供更好的性能。To address the significant issue of hidden terminal interference that severely impacted resource management in ultra-dense Internet of things(UD-IoT)environments,a deep deterministic gradient-based conflict-free resource allocation strategy using graph convolution neural network was proposed.The conflict graph model was constructed by employing matrix transformations to represent potential hidden terminal interference among devices.Then,using the concepts of maximal cliques and hypergraph theory,the conflict graph model was transformed into a conflict hypergraph model.This transformation allowed the conflict-free resource allocation problem to be formulated as a hypergraph vertex coloring problem.A deep deterministic gradient-based conflict-free resource allocation algorithm,leveraging graph convolutional neural network reinforcement learning,was developed to achieve conflict-free resource allocation and maximize resource reuse.Simulation results demonstrated that the proposed algorithm achieved higher resource reuse rates and throughput compared to existing methods,providing superior performance in ultra-dense IoT.

关 键 词：超密集物联网 资源分配 深度强化学习 图卷积神经网络 

分 类 号：TN929.5[电子电信—通信与信息系统]