Research on hybrid precoding for millimeter-wave massive MIMO by integrating Gumbel-Softmax and CNN

Liu Qingli; Zhang Zhaoqing

doi:10.11959/j.issn.1000-0801.2026019

您当前的位置：

首页 >

文章列表页 >

Research on hybrid precoding for millimeter-wave massive MIMO by integrating Gumbel-Softmax and CNN

Research and Development | 更新时间：2026-03-05

- Research on hybrid precoding for millimeter-wave massive MIMO by integrating Gumbel-Softmax and CNN
- Telecommunications Science Vol. 42, Issue 2, Pages: 75-90(2026)
- 作者机构：
  
  1.大连大学信息工程学院，辽宁大连 116622
  2.大连大学通信与网络重点实验室，辽宁大连 116622
- 作者简介：
- 基金信息：
  
  The National Natural Science Foundation of China(61931004)
- DOI：10.11959/j.issn.1000-0801.2026019
  CLC： TN929.5
- Received：24 September 2025，
  
  Revised：2025-09-15，
  
  Accepted：10 October 2025，
  
  Published：20 February 2026
- 稿件说明：
移动端阅览
刘庆利,张兆庆.融合Gumbel-Softmax与CNN的毫米波大规模MIMO混合预编码研究[J].电信科学,2026,42(02):75-90.

Liu Qingli,Zhang Zhaoqing.Research on hybrid precoding for millimeter-wave massive MIMO by integrating Gumbel-Softmax and CNN[J].Telecommunications Science,2026,42(02):75-90.
刘庆利,张兆庆.融合Gumbel-Softmax与CNN的毫米波大规模MIMO混合预编码研究[J].电信科学,2026,42(02):75-90. DOI： 10.11959/j.issn.1000-0801.2026019.

Liu Qingli,Zhang Zhaoqing.Research on hybrid precoding for millimeter-wave massive MIMO by integrating Gumbel-Softmax and CNN[J].Telecommunications Science,2026,42(02):75-90. DOI： 10.11959/j.issn.1000-0801.2026019.

摘要

在毫米波大规模多输入多输出（multiple-input multiple-output，MIMO）系统中，自适应全连接结构存在二值约束、恒模约束以及信道信息利用不充分问题，导致频谱效率和能量效率性能受限。为此，提出一种融合Gumbel-Softmax与卷积神经网络（convolutional neural network，CNN）的混合预编码方法。该方法设计了两个卷积神经网络子网——发送端开关网络（transmit switching network，TsNet）和发送端相移网络（transmit phase-shift network，TpsNet），分别用于优化开关预编码矩阵和相移预编码矩阵。在TsNet中，创新性地引入Gumbel-Softmax方法，将离散二值约束嵌入网络；TpsNet则通过相位层将输出限制在移相器有效相位区间，并借助C2层满足恒模约束。TsNet和TpsNet以并联方式构建预编码联合网络（precoding coordinated network，PCNet），通过残差网络提取毫米波信道特征。两子网并行训练，共享残差网络参数以增强特征一致性，使生成的预编码矩阵接近最优。仿真结果表明，PCNet相较于其他对比算法，频谱效率和能量效率均有显著提升。

Abstract

In millimeter-wave massive multiple-input multiple-output (MIMO) systems

adaptive fully connected architectures suffered from binary constraints

constant modulus constraints

and insufficient utilization of channel information

resulting in limited spectral and energy efficiency. To address this issue

a hybrid precoding method integrating Gumbel-Softmax and convolutional neural network (CNN) was proposed. Two CNN subnetworks

TsNet and TpsNet

were designed to optimize the switch precoding matrix and the phase shift precoding matrix

respectively. In TsNet

the Gumbel-Softmax method was innovatively introduced to embed discrete binary constraints. TpsNet used a phase layer to constrain the output to the effective phase range of the phase shifter and utilized the C2 layer to satisfy the constant modulus constraint. TsNet and TpsNet were combined in parallel to form a joint network

PCNet

which extracted millimeter-wave channel features using a residual network. The two subnetworks were trained in parallel

sharing residual network parameters to enhance feature consistency

resulting in a near-optimal precoding matrix. Simulation results show that PCNet achieves improved spectral and energy efficiency compared to other competing algorithms. This method significantly enhances system spectral and energy efficiency.

关键词

Keywords

references

Venkateswaran V , Van Der Veen A J . Analog beamforming in MIMO communications with phase shift networks and online channel estimation [J ] . IEEE Transactions on Signal Processing , 2010 , 58 ( 8 ): 4131 - 4143 .

Yu X H , Shen J C , Zhang J , et al . Alternating minimization algorithms for hybrid precoding in millimeter wave MIMO systems [J ] . IEEE Journal of Selected Topics in Signal Processing , 2016 , 10 ( 3 ): 485 - 500 .

Lin T , Cong J Q , Zhu Y , et al . Hybrid beamforming for millimeter wave systems using the MMSE criterion [J ] . IEEE Transactions on Communications , 2019 , 67 ( 5 ): 3693 - 3708 .

Yuan M H , Wang H , Yin H , et al . Alternating optimization based hybrid transceiver designs for wideband millimeter-wave massive multiuser MIMO-OFDM systems [J ] . IEEE Transactions on Wireless Communications , 2023 , 22 ( 12 ): 9201 - 9217 .

Shi B Z , Liu F L , Du R Y . A MCUR-TS hybrid precoding and combining algorithm for MIMO-OFDM systems [J ] . IEEE Transactions on Wireless Communications , 2024 , 23 ( 6 ): 5488 - 5502 .

LóPez-Valcarce R , GonzáLez-Prelcic N . Hybrid beamforming designs for frequency-selective mmWave MIMO systems with per-RF chain or per-antenna power constraints [J ] . IEEE Transactions on Wireless Communications , 2022 , 21 ( 8 ): 5770 - 5784 .

Zhang P , Pan L R , Laohapensaeng T , et al . Hybrid beamforming based on an unsupervised deep learning network for downlink channels with imperfect CSI [J ] . IEEE Wireless Communications Letters , 2022 , 11 ( 7 ): 1543 - 1547 .

Liu Z Y , Yang Y W , Gao F F , et al . Deep unsupervised learning for joint antenna selection and hybrid beamforming [J ] . IEEE Transactions on Communications , 2022 , 70 ( 3 ): 1697 - 1710 .

Jin W J , Zhang J , Wen C K , et al . Model-driven deep learning for hybrid precoding in millimeter wave MU-MIMO system [J ] . IEEE Transactions on Communications , 2023 , 71 ( 10 ): 5862 - 5876 .

Liu F L , Li X Y , Yang X H , et al . Deep learning based joint hybrid precoding and combining design for mmWave MIMO systems [J ] . IEEE Systems Journal , 2024 , 18 ( 1 ): 560 - 567 .

Huang H J , Song Y W , Yang J , et al . Deep-learning-based millimeter-wave massive MIMO for hybrid precoding [J ] . IEEE Transactions on Vehicular Technology , 2019 , 68 ( 3 ): 3027 - 3032 .

Peken T , Adiga S , Tandon R , et al . Deep learning for SVD and hybrid beamforming [J ] . IEEE Transactions on Wireless Communications , 2020 , 19 ( 10 ): 6621 - 6642 .

Salh A , Alhartomi M A , Ali Hussain G , et al . Deep reinforcement learning-driven hybrid precoding for efficient mm-wave multi-user MIMO systems [J ] . Journal of Sensor and Actuator Networks , 2025 , 14 ( 1 ): 20 .

Gao J B , Zhong C J , Li G Y , et al . Deep learning-based channel estimation for wideband hybrid MmWave massive MIMO [J ] . IEEE Transactions on Communications , 2023 , 71 ( 6 ): 3679 - 3693 .

Park S , Alkhateeb A , Heath R W . Dynamic subarrays for hybrid precoding in wideband mmWave MIMO systems [J ] . IEEE Transactions on Wireless Communications , 2017 , 16 ( 5 ): 2907 - 2920 .

Banerjee B , Elliott R C , Krzymień W A , et al . Hybrid beamforming for mmWave massive MIMO systems using conditional generative adversarial networks [J ] . IEEE Transactions on Vehicular Technology , 2024 , 73 ( 10 ): 15803 - 15808 .

Liu F L , Kan X D , Bai X Y , et al . Two-stage hybrid precoding algorithm based on switch network for millimeter wave mimo systems [J ] . Progress in Electromagnetics Research M , 2019 , 77 : 103 - 113 .

Hei Y Q , Liu C , Li W T , et al . CNN based hybrid precoding for MmWave MIMO systems with adaptive switching module and phase modulation array [J ] . IEEE Transactions on Wireless Communications , 2022 , 21 ( 12 ): 10489 - 10501 .

Liu F L , Zhang L J , Yang X H , et al . DL-based energy-efficient hybrid precoding for mmWave massive MIMO systems [J ] . IEEE Transactions on Vehicular Technology , 2023 , 72 ( 5 ): 6103 - 6112 .

Saleh A A M , Valenzuela R . A statistical model for indoor multipath propagation [J ] . IEEE Journal on Selected Areas in Communications , 1987 , 5 ( 2 ): 128 - 137 .

Telatar E . Capacity of multi-antenna Gaussian channels [J ] . European Transactions on Telecommunications , 1999 , 10 ( 6 ): 585 - 595 .

El Ayach O , Rajagopal S , Abu-Surra S , et al . Spatially sparse precoding in millimeter wave MIMO systems [J ] . IEEE Transactions on Wireless Communications , 2014 , 13 ( 3 ): 1499 - 1513 .

Gao X Y , Dai L L , Sun Y , et al . Machine learning inspired energy-efficient hybrid precoding for mmWave massive MIMO systems [C ] // Proceedings of the 2017 IEEE International Conference on Communications (ICC) . Piscataway : IEEE Press , 2017 : 1 - 6 .

Jang E , Gu S X , Poole B . Categorical reparameterization with gumbel-softmax [PP ] . V5. arXiv ( 2017-08-05 )[ 2025-08-24 ] . arXiv:arXiv. 1611.01144.

Lin B , Gao F F , Zhang S , et al . Deep learning-based antenna selection and CSI extrapolation in massive MIMO systems [J ] . IEEE Transactions on Wireless Communications , 2021 , 20 ( 11 ): 7669 - 7681 .

Cao Y S , Lv T J , Lin Z P , et al . Complex ResNet aided DoA estimation for near-field MIMO systems [J ] . IEEE Transactions on Vehicular Technology , 2020 , 69 ( 10 ): 11139 - 11151 .

Tian Y , Pan G F , Alouini M S . Applying deep-learning-based computer vision to wireless communications: methodologies, opportunities, and challenges [J ] . IEEE Open Journal of the Communications Society , 2021 , 2 : 132 - 143 .

Gao X Y , Dai L L , Han S F , et al . Energy-efficient hybrid analog and digital precoding for MmWave MIMO systems with large antenna arrays [J ] . IEEE Journal on Selected Areas in Communications , 2016 , 34 ( 4 ): 998 - 1009 .

Yu X H , Zhang J , Letaief K B . A hardware-efficient analog network structure for hybrid precoding in millimeter wave systems [J ] . IEEE Journal of Selected Topics in Signal Processing , 2018 , 12 ( 2 ): 282 - 297 .

Views

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

Research on human action recognition based on mobile communication network signals

An OMP-based hybrid precoding algorithm with column-orthogonality constraint for cell-free networks

A neural convolutional network intrusion detection model based on autoencoder dimension reduction

Multi-user hybrid precoding design based on QR decomposition and beam gain alignment

Related Author

Wu Zhefu

Wan Yuxuan

Lu Chengling

Qian Neng

Wang Anding

Zhang Sihan

FENG Guanghui

DING Jiawei

Related Institution

Institute of Cyberspace Security, Zhejiang University of Technology

School of Electronics and Communication Engineering, Zhejiang Post and Telecommunication College

School of Information and Electronic Engineering, Zhejiang Gongshang University

School of Information Engineering, Zhengzhou University of Industrial Technology

School of Computer Science and Cyber Engineering, Guangzhou University

⁰