基于图神经网络的联合用户调度与波束成形优化算法  

GNN-based optimization algorithm for joint user scheduling and beamforming

作　　者：何世文[1,2,3] 袁军 安振宇[3] 张敏 黄永明 张尧学 HE Shiwen;YUAN Jun;AN Zhenyu;ZHANG Min;HUANG Yongming;ZHANG Yaoxue(School of Computer Science and Engineering,Central South University,Changsha 410083,China;National Mobile Communications Research Laboratory,Southeast University,Nanjing 210096,China;Purple Mountain Laboratories,Nanjing 211111,China;School of Information and Communication,Hunan Post and Telecommunication College,Changsha 410015,China;Department of Computer Science and Technology,Tsinghua University,Beijing 100084,China)

机构地区：[1]中南大学计算机学院,湖南长沙410083 [2]东南大学移动通信国家重点实验室,江苏南京210096 [3]紫金山实验室,江苏南京211111 [4]湖南邮电职业技术学院信息通信学院,湖南长沙410015 [5]清华大学计算机科学与技术系,北京100084

出　　处：《通信学报》2022年第7期73-84,共12页Journal on Communications

基　　金：国家自然科学基金资助项目(No.62171474,No.61720106003);东南大学移动通信国家重点实验室开放研究基金资助项目(No.2022D03);OPPO广东移动通信有限公司研究基金资助项目(No.CN05202112160224)。

摘　　要：协作多点(CoMP)传输技术具有降低同频干扰和提高频谱效率的特点。对于CoMP,用户调度与波束成形是2个分别位于媒体访问接入层和物理层的基本研究问题。在考虑用户服务质量需求下,重点研究用户调度与波束成形的联合优化问题,并以网络吞吐量最大化为目标。为了克服传统优化算法计算开销大且未有效利用网络历史数据信息的问题,提出了一种基于图神经网络联合用户调度与功率分配模型,并结合波束向量的解析公式,以实现联合用户调度与波束成形优化。仿真分析表明,所提算法能够以较低的计算开销实现与传统优化算法相匹配,甚至更优的性能表现。The coordinated multi-point(CoMP)transmission technology has the characteristics of reducing co-channel interference and improving spectral efficiency.For the CoMP technology,user scheduling(US)and beamforming(BF)design are two fundamental research problems located in the media access control layer and the physical layer,respec-tively.Under the consideration of user service quality requirements,the joint user US-BF optimization problem was investigated with the goal of maximizing network throughput.To overcome the problem that the traditional optimiza-tion algorithm had high computational cost and couldn’t effectively utilize the network historical data information,a joint US and power allocation(M-JEEPON)model based on graph neural network was proposed to realize joint US-BF optimization by combining the beam vector analytical solution.The simulation results show that the proposed algorithm can achieve the performance matching or even better than traditional optimization algorithms with lower computational overhead.

关 键 词：跨层优化 图神经网络 协作多点 用户调度 波束成形 

分 类 号：TN92[电子电信—通信与信息系统]