| 摘要: |
| 针对车联网场景下多入多出-正交时频空（Multiple-Input Multiple-Output-Orthogonal Time Frequency Space,MIMO-OTFS）系统的信道状态信息（Channel State Information,CSI）反馈问题，提出了一种面向时延-多普勒（Delay-Dopler,DD）域CSI反馈的时间差分架构Transformer网络（Time-differencing Architecture Delay-Doppler Transformer Network,TA-DD-TransNet）,引入分时反馈机制，将残差信息建模与压缩反馈相结合。网络结构融合Transformer的全局建模能力与卷积神经网络的局部特征提取优势，在保持CSI重构精度的同时显著降低了反馈比特数与计算复杂度。在不同车速、信噪比及非完美信道估计条件下的仿真实验结果表明，所提方法在归一化均方误差（Normalized Mean Squared Error,NMSE）和余弦相似度指标上均优于CsiNet、CsiNet+和BCsiNet。在60 km/h、30 dB信噪比、1/4压缩率下，TA-DD-TransNet的NMSE约-27 dB，余弦相似度达0.96。复杂度分析显示，TA-DD-TransNet在1/4压缩率下的编码器和解码器浮点运算次数分别为1.809×107和2.281×107，参数量均为8.4×106左右，显著低于CsiNet+。所提方法能满足车联网中对高可靠低时延通信的实际需求。 |
| 关键词: 车联网(IoV) MIMO OTFS CSI反馈 时延-多普勒域 |
| DOI：10.20079/j.issn.1001-893x.250327001 |
|
| 基金项目:重庆市自然科学基金面上项目（CSTB2023NSCQ-MSX0025） |
|
| TA-DD-TransNet:a CSI Feedback Method for Delay-Doppler Domain |
| LIAO Yong,LUO Yu,LIAO Yang,YE Yanshao |
| (1.School of Microelectronics and Communication Engineering,Chongqing University,Chongqing 400044,China;2.Three Gorges High-tech Information Technology Co.,Ltd.,Yichang 443002,China;3.College of Physics,Chongqing University,Chongqing 401331,China) |
| Abstract: |
| To address the channel state information(CSI) feedback challenge in multiple-input multiple-output-orthogonal time frequency space(MIMO-OTFS) systems for Internet of Vehicles(IoV),a time-differencing architecture delay-Doppler Transformer network(TA-DD-TransNet) tailored for CSI feedback in the delay-Doppler(DD) domain is proposed,the time-sharing feedback mechanism is introduced and residual information modeling is combined with compressive feedback.Because the network structure integrates the global modeling ability of Transformer and the local feature extraction advantages of convolutional neural network,it significantly reduces the number of feedback bits and the computational complexity while maintaining the reconstruction accuracy of CSI. The simulation experiments under different vehicle speeds,signal-to-noise ratios(SNRs) and imperfect channel estimation conditions show that the proposed method is superior to CsiNet,CsiNet+ and BCsiNet in terms of normalized mean squared error(NMSE) and cosine similarity indicator.Under the condition of 60 km/h speed,30 dB SNR,and 1/4 compression ratio,the NMSE of TA-DD-TransNet is approximately -27 dB,and the cosine similarity is 0.96. Complexity analysis shows that the number of floating-point operations of the encoder and decoder of TA-DD-TransNet at the 1/4 compression ratio is 1.809×107 and 2.281×107 respectively,and the number of parameters is about 8.4×106 for both,which is significantly lower than that of CsiNet+.The proposed mathod can meet the actual demand for highly reliable and low-latency communication in IoV. |
| Key words: Internet of Vehicles(IoV) MIMO OTFS CSI feedback delay-Doppler domain |
