联合改进最小均方误差与深度学习的V2X信道估计方案  

Joint Channel Estimation Scheme Using Improved Least Mean Square Error and Deep Learning for V2X Communications

作　　者：吕超跞 骆忠强 LYU Chaoluo;LUO Zhongqiang(School of Automation and Information Engineering,Sichuan University of Science&Engineering,Yibin 644000,China;Artificial Intelligence Key Laboratory of Sichuan Province,Sichuan University of Science&Engineering,Yibin 644000,China)

机构地区：[1]四川轻化工大学自动化与信息工程学院,四川宜宾644000 [2]四川轻化工大学人工智能四川省重点实验室,四川宜宾644000

出　　处：《无线电通信技术》2025年第2期374-382,共9页Radio Communications Technology

基　　金：国家自然科学基金(61801319);四川省科技厅项目(2020JDJQ0061,2021YFG0099);四川轻化工大学研究生创新基金(Y2023285)。

摘　　要：针对IEEE 802.11p标准中导频数量有限,难以准确追踪车联万物(Vehicle-to-Everything,V2X)通信中时变信道的问题,学者们研究了数据导频辅助(Data Pilot Aided,DPA)信道估计方案。然而,这些经典DPA方案不能在完整的信噪比(Signal to Noise Ratio,SNR)范围内给出令人满意的效果,并且其估计结果的可靠性易受误差传播的影响。研究了一种新的信道估计方案,基于使用虚拟子载波的最小均方误差(Minimum Mean Square Error Using Virtual Pilots,MMSE-VP)方案,提出一种带有时间平均操作的改进MMSE(Improved MMSE,IMMSE)方案。IMMSE方案通过利用相邻正交频分复用(Orthogonal Frequency Division Multiplexing,OFDM)符号间信道的相关性来提高MMSE-VP方案在低SNR区域的性能,达到在整个SNR区域有良好表现的目的。联合深度学习技术,采用全连接神经网络(Fully Connected Neural Network,FCNN)作为IMMSE方案的非线性后处理模块,减少误差并获得更好的估计性能。在不同实验条件下的仿真结果表明,提出的信道估计方案可以适应调制方式和车辆速度的变化,能有效应对V2X通信中的信道估计问题。To address the problem of the limited number of pilots in the IEEE 802.11p standard,which makes it difficult to accurately track time-varying channels in Vehicle-to-Everything(V2X)communications,scholars have investigated the Data Pilot Aided(DPA)channel estimation schemes.However,these classic DPA schemes cannot deliver satisfactory results across the complete Signal to Noise Ratio(SNR)range,and their estimation accuracy is prone to being compromised by error propagation.For this purpose,this paper investigates a new channel estimation scheme.Firstly,based on the Minimum Mean Square Error Using Virtual Pilots(MMSE-VP)scheme,an Improved MMSE(IMMSE)scheme that incorporates temporal averaging operations is proposed.This IMMSE scheme is designed to utilize the correlation between channels of adjacent Orthogonal Frequency Division Multiplexing(OFDM)symbols to enhance the MMSE-VP scheme's performance in low SNR regions,thereby achieving commendable performance across the entire SNR range.Subsequently,combining deep learning technology,a Fully Connected Neural Network(FCNN)is employed as the nonlinear post-processing module for the IMMSE scheme to reduce errors and achieve better estimation performance.Simulation results under various experimental conditions demonstrate that the proposed channel estimation scheme can be adapted to changes in modulation mode and vehicle velocity,proving its capability to effectively solve the channel estimation problem in V2X communications.

关 键 词：车联万物通信 信道估计 IEEE 802.11p 正交频分复用 深度学习 

分 类 号：TN911.23[电子电信—通信与信息系统]