清华大学电子工程系,北京 100084
王昭诚(1968年生),男;研究方向:无线通信;E-mail:zcwang@tsinghua.edu.cn
马可(1998年生),男;研究方向:无线通信、人工智能; E-mail:make15@tsinghua.org.cn
网络出版日期:2024-09-26,
收稿日期:2024-06-27,
录用日期:2024-08-15
移动端阅览
王昭诚, 马可. 深度学习赋能波束管理:现状、挑战与机遇[J/OL]. 中山大学学报(自然科学版)(中英文), 2024,1-11.
WANG Zhaocheng, MA Ke. Deep learning empowered beam management: State-of-the-art, challenges and opportunities[J/OL]. Acta Scientiarum Naturalium Universitatis Sunyatseni, 2024,1-11.
王昭诚, 马可. 深度学习赋能波束管理:现状、挑战与机遇[J/OL]. 中山大学学报(自然科学版)(中英文), 2024,1-11. DOI: 10.13471/j.cnki.acta.snus.ZR20240214.
WANG Zhaocheng, MA Ke. Deep learning empowered beam management: State-of-the-art, challenges and opportunities[J/OL]. Acta Scientiarum Naturalium Universitatis Sunyatseni, 2024,1-11. DOI: 10.13471/j.cnki.acta.snus.ZR20240214.
随着载波频率的不断提高和大规模天线阵列的广泛部署,基于模拟移相器的波束赋形成为下一代无线通信的标志性技术之一。此时,波束管理被用于获取和维护基站和用户端具有最大接收功率的最优波束对,以保障可靠的无线通信服务。传统波束管理方法往往依赖于海量搜索。同时,传统数学模型无法全面的、准确刻画非线性的波束的内在关联和高维无线环境特征,因而难以取得令人满意的波束增益性能。近年来,得益于深度学习强大的自适应拟合能力,深度学习赋能波束管理得到了国内外广泛关注。本文总结了深度学习赋能波束管理的研究进展,并展望了未来的研究方向。首先,阐述了深度学习应用于波束管理的典型场景和潜在优势;随后,从空/时/频域切入,讨论当前深度学习赋能波束管理的主要研究路线和代表性工作;最后,面向更大规模的无线网络、更多元的波束管理功能和更鲁棒的深度学习模型,阐述未来的研究挑战与机遇。
With the increase of carrier frequencies and the widespread deployment of large-scale antenna arrays, the analog phase shifter based beamforming has become one of the key technologies for the next-generation wireless communications. To ensure reliable wireless services, beam management is usually adopted to acquire and maintain the optimal beam pair with the maximum received power between the base station and the user equipment. However, traditional beam management methods generally rely on large-scale searching, which could bring huge overhead. Additionally, traditional mathematical models cannot comprehensively and accurately model the nonlinear intrinsic characteristics in the received signals from the beam and the high-dimensional features of wireless environments,making it difficult to achieve satisfactory beamforming gain performance. Thanks to its strong adaptive fitting capabilities, deep learning empowered beam management has drawn much attention recently. To this end, this paper reviews the state-of-the-art literatures of deep learning empowered beam management and discusses future research directions. Firstly, the typical scenarios and advantages of applying deep learning to beam management are elucidated. Subsequently, the main techniques and representative works in current deep learning empowered beam management are summarized from space/time/frequency domains. Finally, this paper addresses future research challenges and opportunities from the perspective of larger-scale wireless networks, more diverse beam management functions, and more robust deep learning models.
深度学习波束管理空域时域频域
deep learningbeam managementspace domainstime domainsfrequency domains
ALI A,GONZÁLEZ-PRELCIC N, HEATH R W,2017. Millimeter wave beam-selection using out-of-band spatial information[J]. IEEE Trans Wirel Commun,17(2): 1038-1052.
ALKHATEEB A, 2019. DeepMIMO: A generic deep learning dataset for millimeter wave and massive MIMO applications[C]//IEEE Information Theory and Applications Workshop(ITA). IEEE: 1-8.
ALRABEIAH M, ALKHATEEB A, 2020. Deep learning for mm wave beam and blockage prediction using sub-6 GHz channels[J]. IEEE Trans Commun, 68(9): 5504-5518.
CHEN Z, MA X, ZHANG B,et al, 2019. A survey on terahertz communications[J]. China Commun, 16(2): 1-35.
CHOI J, 2015. Beam selection in mm-wave multiuser MIMO systems using compressive sensing[J]. IEEE Trans Commun, 63(8): 2936-2947.
ECHIGO H, CAO Y, BOUAZIZI M,et al, 2021. A deep learning-based low overhead beam selection in mm wave communications[J]. IEEE Trans Veh Technol, 70(1): 682-691.
GAO F, LIN B,BIN C,et al, 2021. FusionNet: Enhanced beam prediction for mm wave communications using sub-6GHz channel and a few pilots[J]. IEEE Trans Commun, 69(12): 8488-8500.
HASHEMI M, KOKSAL C E, SHROFF N B, 2017. Out-of-band millimeter wave beamforming and communications to achieve low latency and high energy efficiency in 5G systems[J]. IEEE Trans Commun, 66(2): 875-888.
HEATH R W, GONZALEZ-PRELCIC N,RANGAN S, et al, 2016. An overview of signal processing techniques for millimeter wave MIMO systems[J]. IEEE J Sel Top Signal Process, 10(3): 436-453.
HUANG Z, WANG Z, CHEN S, 2024. Sub-6GHz assisted mm wave hybrid beamforming with heterogeneous graph neural network[J]. IEEE Trans Commun, 99: 1.
JEONG C,PARK J,YU H,2015.Random access in millimeter-wave beamforming cellular networks: Issues and approaches[J]. IEEE Commun Mag, 53(1): 180-185.
KAYA A Ö, VISWANATHAN H, 2021. Deep learning-based predictive beam management for 5G mm wave systems[C]//IEEE Wireless Communications and Networking Conference(WCNC). Nanjing,China:IEEE: 1-7.
KIM J, MOLISCH A F, 2014. Fast millimeter-wave beam training with receive beamforming[J]. J Commun Netw, 16(5): 512-522.
KUTTY S, SEN D, 2015. Beamforming for millimeter wave communications: An inclusive survey[J]. IEEE Commun Surv Tutor, 18(2): 949-973.
LIM S H, KIM S, SHIM B, et al, 2021. Deep learning-based beam tracking for millimeter-wave communications under mobility[J]. IEEE Trans Commun, 69(11): 7458-7469.
LIU L, OESTGES C, POUTANEN J, et al, 2012. The COST 2100 MIMO channel model[J]. IEEE Wirel Commun, 19(6): 92-99.
LONG Y, CHEN Z, FANG J, et al, 2018. Data-driven-based analog beam selection for hybrid beamforming under mm-Wave channels[J]. IEEE J Sel Top Signal Process,12(2): 340-352.
LUO X, LIU W, WANG Z, 2019. Calibrated beam training for millimeter-wave massive MIMO systems[C]//IEEE 90th Vehicular Technology Conference(VTC2019-Fall). Honolulu, Hawaii, USA:IEEE: 1-5.
MA K, HE D,SUN H, et al, 2021a. Deep learning assisted calibrated beam training for millimeter-wave communication systems[J]. IEEE Trans Commun, 69(10): 6706-6721.
MA K, HE D,SUN H, et al, 2021b. Deep learning assisted mm wave beam prediction with prior low-frequency information[C]//IEEE International Conference on Communications (ICC). Montreal, QC, Canada:IEEE: 1-6.
MA K, ZOU H, SUN C, et al, 2022. Deep learning assisted adaptive mm wave beam tracking: A sum-probability oriented methodology[C]// IEEE Global Communications Conference (GLOBECOM). Rio de Janeiro, Brazil: IEEE: 573-578.
MA K, ZHANG F,TIAN W,et al, 2023a. Continuous-time mm wave beam prediction with ODE-LSTM learning architecture[J]. IEEE Wirel Commun Lett, 12(1): 187-191.
MA K, DU S,ZOU H,et al, 2023b. Deep learning assisted mm wave beam prediction for heterogeneous networks: A dual-band fusion approach[J].IEEE Trans Commun, 71(1): 115-130.
MA W, QI C, LI G Y, et al, 2020. Machine learning for beam alignment in millimeter wave massive MIMO[J]. IEEE Wirel Commun Lett, 9(6): 875-878.
RAPPAPORT T S, SUN S, MAYZUS Ret al, 2013. Millimeter wave mobile communications for 5G cellular: It will work![J]. IEEE Access, 1: 335-349.
ROH W, SEOL J Y, PARK J, et al, 2014. Millimeter-wave beamforming as an enabling technology for 5G cellular communications: Theoretical feasibility and prototype results[J]. IEEE Commun Mag, 52(2): 106-113.
UWAECHIA A N,MAHYUDDIN N M,2020.A comprehensive survey on millimeter wave communications for fifth-generation wireless networks: Feasibility and challenges[J]. IEEE Access, 8: 62367-62414.
WEI L, LI Q, WU G, 2017. Exhaustive, iterative and hybrid initial access techniques in mm wave communications[C]//IEEE Wireless Communications and Networking Conference (WCNC). San Francisco, CA, USA: IEEE: 1-6.
XIAO M, MUMTAZ S, HUANG Y, et al, 2017. Millimeter wave communications for future mobile networks[J]. IEEE J Sel Area Commun, 35(9): 1909-1935.
XIE T, DAI L, NG D W K, et al, 2019. On the power leakage problem in millimeter-wave massive MIMO with lens antenna arrays[J]. IEEE Trans Signal Process, 67(18): 4730-4744.
ZHAO Y, ZHANG X, GAO X, et al, 2024. LSTM-based predictive mm wave beam tracking via sub-6GHz channels for V2I communications[J]. IEEE Trans Commun, 99: 1.
ZHANG J, HUANG Y, WANG J, et al, 2020. Intelligent interactive beam training for millimeter wave communications[J]. IEEE Trans Wirel Commun, 20(3): 2034-2048.
0
浏览量
0
下载量
0
CSCD
关联资源
相关文章
相关作者
相关机构