柔性双工网络中基于图神经网络的能源效率联合优化方法
Joint energy efficiency optimization based on graph neural network approach in flexible duplex networks
王子威 1李晖 2史振婷 1张见 3陶旭3
作者信息
- 1. 南京信息工程大学电子与信息工程学院 南京 210044
- 2. 南京信息工程大学电子与信息工程学院 南京 210044;无锡学院江苏省集成电路可靠性技术及检测系统工程研究中心 无锡 214105
- 3. 无锡学院江苏省集成电路可靠性技术及检测系统工程研究中心 无锡 214105
- 折叠
摘要
在绿色通信和双碳目标的背景下,提升网络能源效率(energy efficiency,EE)是当前无线通信系统设计和可靠运行的关键技术之一.首先,针对柔性双工网络(flexible duplex networks,Flex-Net)的"和能源效率"最大化的问题,构建了一个具有成对固定通信链路的网络;接着,鉴于图神经网络在通信网络资源优化中的优势,提出了一种新的基于图神经网络(GNN)架构的柔性双工网络(GFlex-Net),以联合优化"通信方向"和"传输功率",达到网络能效最大化.仿真结果表明,与传统的算法相比,所提出的架构有近乎最优的性能表现,达到穷举法性能的95%,但保持了较低的计算复杂度O(n2),同时算法揭示了GNN在资源优化中面对样本复杂性、可扩展性和泛化能力方面的优势.
Abstract
In the context of green communication and the carbon peaking&carbon neutrality goals,enhancing network energy efficiency(EE)is one of the key technologies for the design and reliable operation of wireless communication systems.Firstly,this article presents the construction of a network consisting of paired fixed communication links to address the problem of maximizing energy efficiency in flexible duplex networks(Flex-Net).Subsequently,taking into account the benefits of graph neural networks in optimizing communication network resources,we propose a new architecture called GFlex-Net based on the graph neural network(GNN)framework.This architecture aims to jointly optimize communication direction and transmission power,thereby achieving maximum network energy efficiency.Simulation results demonstrate that,when compared to traditional algorithms,the proposed architecture achieves nearly optimal performance,reaching 95%of the performance of exhaustive methods,while maintaining a lower computational complexity of O(n2).Furthermore,the algorithm showcases the advantages of GNN in resource optimization,including reduced sample complexity,improved scalability,and enhanced.
关键词
柔性双工网络/能源效率/无线通信/图神经网络/联合优化Key words
flexible duplex networks/energy efficiency/wireless communication/GNN/joint optimization引用本文复制引用
基金项目
国家自然科学基金(61661018)
江苏省基础研究计划青年基金(BK20210064)
无锡市科技创新创业资金(WX03-02B0137-022200-34)
出版年
2024
NETL
NSTL
维普
万方数据