面向集成相对能效的6G网络动态资源配置  

Dynamic Resource Allocation for 6G Networks Based on Integrated Relative Energy Efficiency

作　　者：王思敏 余涛 郑嘉怡 张舜卿 黄朋博 WANG Simin;YU Tao;ZHENG Jiayi;ZHANG Shunqin;HUANG Pengbo(School of Communication and Information Engineering,Shanghai University,Shanghai 200444,China)

机构地区：[1]上海大学通信与信息工程学院,上海200444

出　　处：《移动通信》2024年第7期80-86,共7页Mobile Communications

基　　金：国家重点研发计划项目“新能源驱动的6G跨制式融合组网方法”(2022YFB2902304);国家自然科学基金项目“面向非平稳环境的高能效无线传输关键技术研究”(62071284)。

摘　　要：随着6G网络技术的发展,组网区域内流量需求与网络容量分布之间的不匹配问题日益凸显,导致资源浪费现象严重。为了应对这一挑战,使用可衡量空间匹配程度的集成相对能效作为优化目标,通过结合提取动态时间特征的FourierGNN时序预测网络与提取空间特征的RBF网络,提出了一种动态资源配置算法优化基站的带宽与功率配置,通过设计两阶段运行机制解决RBF提取动态时间特征有限的问题,有效提高动态流量需求与网络容量之间的空间匹配性,实现集成相对能效的优化。通过数值实验,首先验证了所提算法比起基线算法与端到端算法分别有12.52%与13.47%的性能提升。其次,通过算法子模块的对比,验证了使用基于奇异值分解的稀疏压缩算法虽然损失了1.1%的性能,但是算法运行时间提升60%,实时性更好,同时相较于LSTM网络以及CNNLSTM网络,基于FourierGNN网络的算法有最大2.3%的性能提升。最后,根据所提算法运行机制对实际组网中的资源调度策略做出指导:流量快变时频繁使用两阶段运行机制消除RBF网络的累计误差,流量慢变时可以减少使用两阶段运行机制的频次,增强算法实时性。As 6G network technology evolves,the mismatch between traffic demand and network capacity distribution in network areas becomes increasingly prominent,leading to significant resource wastage.To address this challenge,we propose using Integrated Relative Energy Efficiency(IREE)as an optimization target.By combining the FourierGNN time-series prediction network,which extracts dynamic temporal features,with the RBF network,which captures spatial features,we introduce a dynamic resource allocation algorithm that optimizes bandwidth and power configurations for base stations.To overcome the limited dynamic temporal feature extraction of the RBF network,we design a two-stage operation mechanism.This approach effectively enhances the spatial matching between dynamic traffic demand and network capacity,achieving optimal IREE.Numerical experiments demonstrate that the proposed algorithm outperforms baseline and end-to-end algorithms by 12.52%and 13.47%,respectively.Furthermore,a comparison of algorithm submodules shows that while using a sparse compression algorithm based on singular value decomposition sacrifices 1.1%of performance,it improves runtime by 60%,enhancing realtime capabilities.Additionally,compared to LSTM and CNNLSTM networks,the FourierGNN-based algorithm achieves up to a 2.3%performance improvement.Finally,the proposed algorithm’s operational mechanism provides practical guidance for resource scheduling strategies in real network deployments:during rapid traffic changes,frequent use of the two-stage operation mechanism mitigates RBF network cumulative errors,while during slow traffic changes,reducing the frequency of this mechanism enhances real-time performance.

关 键 词：集成相对能效 FourierGNN网络 动态资源配置 

分 类 号：TN929.5[电子电信—通信与信息系统]