机器学习在信道建模中的应用综述  

Survey of application of machine learning in wireless channel modeling

作　　者：刘留[1] 张建华 樊圆圆 于力 张嘉驰 LIU Liu;ZHANG Jianhua;FAN Yuanyuan;YU Li;ZHANG Jiachi(School of Electronic and Information Engineering,Beijing Jiaotong University,Beijing 100044,China;State Key Laboratory of Networking and Switching Technology,Beijing University of Posts and Telecommunications,Beijing 100876,China)

机构地区：[1]北京交通大学电子信息工程学院,北京100044 [2]北京邮电大学网络与交换技术国家重点实验室,北京100876

出　　处：《通信学报》2021年第2期134-153,共20页Journal on Communications

基　　金：国家重点研发计划基金资助项目(No.2018YFB1801101);国家自然科学基金杰出青年基金资助项目(No.61925102);北京市自然科学基金–海淀原始创新联合基金资助项目(No.L172030);国家自然科学基金重点资助项目(No.61931001)。

摘　　要：信道建模是设计无线通信系统的基础,传统的信道建模方法无法自动学习特定类型信道的规律,特别是在针对特殊应用场景,如物联网、毫米波通信、车联网等,存在一定的局限性。此外,机器学习具有有效处理大数据、创建模型的能力,基于此,探讨了机器学习如何与信道建模进行有机融合,分别从信道多径分簇、参数估计、模型的构造及信道的场景识别展开了讨论,对当前该领域的重要研究成果进行了阐述,并对未来发展提出了展望。Channel characterization is primary to the design of the wireless communication system.The conventional channel characterization method cannot learn the law of certain types of channels by itself,which limits its application in several special scenarios,such as Internet of things,millimeter wave communication and Internet of vehicles.Machine learning was able to process the big data and establish the model.Based on this,the cooperation between the machine learning and channel characterization was investigated.The channel multipath clustering,parameter estimation,model construction and wireless channel scene recognition were discussed,and recent significant research results in this field were provided.Finally,the future direction of the machine learning in wireless channel modeling was proposed.

关 键 词：信道建模 机器学习 神经网络 聚类 场景识别 

分 类 号：TN929.5[电子电信—通信与信息系统]