系统工程与电子技术 ›› 2025, Vol. 47 ›› Issue (9): 3093-3098.doi: 10.12305/j.issn.1001-506X.2025.09.31

• 通信与网络 • 上一篇    

TS-GRU-VTA:基于深度学习的车辆信道估计方案

季策1, 马相宇1,*(), 牟晓宇1, 赵家毅2   

  1. 1. 东北大学计算机科学与工程学院,辽宁 沈阳 110819
    2. 东北大学秦皇岛分校计算机与通信工程学院,河北 秦皇岛 066004
  • 收稿日期:2024-07-22 出版日期:2025-09-25 发布日期:2025-09-16
  • 通讯作者: 马相宇 E-mail:2505812102@qq.com
  • 作者简介:季 策(1969 —),女,副教授,博士,主要研究方向为正交频分复用关键技术研究、盲信号处理
    牟晓宇(2001—),男,硕士研究生,主要研究方向为基于深度学习的信道估计
    赵家毅(2001—),男,硕士研究生,主要研究方向为基于张量的信道估计

TS-GRU-VTA: vehicle channel estimation scheme based on deep learning

Ce JI1, Xiangyu MA1,*(), Xiaoyu MU1, Jiayi ZHAO2   

  1. 1. School of Computer Science and Engineering,Northeastern University,Shenyang 110819,China
    2. School of Computer and Communication Engineering,Northeastern University at Qinhuangdao,Qinhuangdao 066004,China
  • Received:2024-07-22 Online:2025-09-25 Published:2025-09-16
  • Contact: Xiangyu MA E-mail:2505812102@qq.com

摘要:

在IEEE 802.11p标准下,传统的数据导频辅助(data-pilot aided, DPA)估计器难以有效追踪时变信道,尽管基于深度学习的估计算法得到广泛研究,但通常面临复杂度高或性能不佳的问题。基于此,提出了一种低复杂度的信道估计方案,通过结合时域采样、门控循环单元(gated recurrent unit, GRU)和变系数时间平均方法,有效降低复杂度并提高性能。仿真结果表明,该算法在性能和复杂度上均优于对比算法。

关键词: 低复杂度信道估计, 时域采样, 深度学习, 门控循环单元网络, 变系数时间平均

Abstract:

Under the IEEE 802.11p standard, traditional data-pilot aided (DPA) estimators struggle to effectively track time-varying channels. Although deep learning-based estimation algorithms have been widely studied, they often face issues of high complexity or poor performance. Based on this, a low-complexity channel estimation scheme that combines time-domain sampling, gated recurrent units (GRU), and a variable coefficient temporal averaging method is proposed to effectively reduce complexity and enhance performance. Simulation results demonstrate that this algorithm outperforms comparative algorithms in both performance and complexity.

Key words: low-complexity channel estimation, time-domain sapling, deep learning, gated recurrent unit (GRU) network, variable coefficient temporal averaging

中图分类号: