电信科学 ›› 2021, Vol. 37 ›› Issue (6): 14-22.doi: 10.11959/j.issn.1000-0801.2021135

• 专题:5G+6G • 上一篇    下一篇

基于生成对抗网络的MIMO信道估计方法

华郁秀1, 李荣鹏1, 赵志峰1,2, 吴建军3, 张宏纲1   

  1. 1 浙江大学信息与电子工程学院,浙江 杭州 310027
    2 之江实验室,浙江 杭州 311121
    3 华为技术有限公司,上海 200120
  • 修回日期:2021-06-10 出版日期:2021-06-20 发布日期:2021-06-01
  • 作者简介:华郁秀(1993- ),男,浙江大学信息与电子工程学院博士生,主要研究方向为无线资源管理、网络切片、深度学习、强化学习
    李荣鹏(1989- ),男,浙江大学信息与电子工程学院副教授,主要研究方向为智能通信网络、网络智能、网络切片
    赵志峰(1975- ),男,浙江大学信息与电子工程学院教授,之江实验室科研发展部主任,主要研究方向为认知无线电、无线mesh 网络和SDN 在无线通信中的应用
    吴建军(1977- ),男,华为技术有限公司未来网络架构实验室主任,主要研究方向为6G网络架构、5G端到端切片、下一代移动通信标准及产业发展
    张宏纲(1967- ),男,浙江大学电子与信息工程学院教授,主要研究方向为认知无线电、绿色通信和下一代异构蜂窝网络架构
  • 基金资助:
    国家重点研发计划项目(2020YFB1804804SCI);国家自然科学基金资助项目(61731002);国家自然科学基金资助项目(62071425);浙江省重点研发计划项目(2019C01002);浙江省重点研发计划项目(2019C03131);华为合作项目:之江实验室资助项目(2019LC0AB01);浙江省自然科学基金资助项目(LY20F010016)

GAN-based channel estimation for massive MIMO system

Yuxiu HUA1, Rongpeng LI1, Zhifeng ZHAO1,2, Jianjun WU3, Honggang ZHANG1   

  1. 1 College of Information Science and Electronic Engineering, Zhejiang University, Hangzhou 310027, China
    2 Zhejiang Lab, Hangzhou 311121, China
    3 Huawei Technologies Co., Ltd., Shanghai 200120, China
  • Revised:2021-06-10 Online:2021-06-20 Published:2021-06-01
  • Supported by:
    The National Key Research and Development Program of China(2020YFB1804804SCI);The National Natural Science Foundation of China(61731002);The National Natural Science Foundation of China(62071425);Zhejiang Key Research and Development Plan(2019C01002);Zhejiang Key Research and Development Plan(2019C03131);Huawei Cooperation Project, The Project sponsored by Zhejiang Lab(2019LC0AB01);The Natural Science Foundation of Zhejiang Provincial of China(LY20F010016)

摘要:

作为 5G 的一项关键技术,大规模多输入多输出(MIMO)系统通过在基站配备大量天线可以显著提高频谱效率和能效。然而,在大规模MIMO 系统中,精确的信道估计面临严峻挑战。为了在导频序列长度小于发射天线数量以及信道噪声强烈的情况下进行精确的估计信道,提出了一种基于生成对抗网络的大规模MIMO 信道估计方法N2N-GAN。N2N-GAN首先对接收端的导频信号进行去噪,然后使用条件生成对抗网络根据去噪后的导频信号估计信道矩阵。仿真实验证明,与传统的信道估计算法和基于深度学习的算法相比, N2N-GAN 对环境噪声具有更高的鲁棒性,而且可以适应更少导频符号和更多天线数量的场景。

关键词: 信道估计, 大规模MIMO, 信号降噪, 生成对抗网络

Abstract:

As a key technology of 5G, massive MIMO system can significantly improve spectrum efficiency and energy efficiency by equipping a large number of antennas in base stations.However, in massive MIMO system, accurate channel estimation faces severe challenges.In order to estimate the channel accurately when the pilot sequence length is smaller than the number of transmitting antennas and the channel noise is strong, the estimation method N2N-GAN was proposed.N2N-GAN firstly denoised the pilot channel at the receiving end, and then used the conditional generative adversarial network to estimate the channel matrix according to the denoised pilot signal.Simulation experiments show that N2N-GAN achieves better robustness against noise compared with traditional channel estimation algorithms and deep learning-based methods.Meanwhile, it can adapt to scenarios with fewer pilot symbols and more antennas.

Key words: channel estimation, massive MIMO, signaldenoising, generative adversarial network

中图分类号: 

No Suggested Reading articles found!