Please wait a minute...
浙江大学学报(工学版)
Journal of ZheJiang University (Engineering Science)  2025, Vol. 59 Issue (5): 964-972    DOI: 10.3785/j.issn.1008-973X.2025.05.010
    
Semantic communication system in deterministic networking for5G-advanced vehicular network
Hongzhi LUAN1(),Shucai LI1,Yi LI2,Chuanjie QIAN2,*(),Qiong MEI2
1. Comprehensive Smart Energy Department., Shandong Electric Power Engineering Consulting Institute Limited Company, Jinan 250011, China
2. Strategic Project Department, Jiangsu Future Network Group, Nanjing 211111, China
Download: HTML     PDF(1055KB) HTML
Export: BibTeX | EndNote (RIS)      

Abstract  

A deadline-aware semantic communication (DDA-SC) system specifically tailored for deterministic network in 5G-Advanced intelligent vehicle-to-everything (V2X) environments was proposed in order to overcome the challenge of prolonged encoding time in current semantic communication (SC) systems, which impedes their direct deployment in deterministic networking (DN) for 5G-Advanced vehicle-to-everything (V2X) networks—particularly critical for intelligent transportation systems requiring high-throughput and low-latency data processing. A deadline-aware semantic encoder (DDA-SE) was employed, which adjusted the depth of semantic encoding to meet data deadlines, ensuring timely encoding and transmission. A signal-to-noise ratio (SNR) aware network based on encoding depth was introduced in order to enhance transmission reliability across different encoding deadlines. SNR was mapped to relevant signal-to-noise ratio (R-SNR) based on encoding depth during training, using an attention mechanism to sense R-SNR and better resist channel noise under various encoding deadlines. The experimental results show that this system achieve an accuracy improvement of over 80.31% in the CARS-196 vehicle identification task compared with systems that do not consider data deadlines when encoding deadlines are shorter than 5 microseconds. Results demonstrate that DDA-SC can accomplish tasks under extremely low encoding deadlines that are unachievable by conventional semantic communication systems, thus verifying the effectiveness of the method.

Key words： vehicle-to-everything (V2X)      semantic communication      5G-Advanced      deterministic networking      joint source and channel coding     
Received: 12 July 2024      Published: 25 April 2025
CLC:  TN 914  
Fund:  北京市自然科学基金资助项目 (L222043).
Corresponding Authors: Chuanjie QIAN     E-mail: Luanhongzhi@sdepci.com;qianchuanjie@fngroup.cn
Service
E-mail this article
Add to my bookshelf
Add to citation manager
E-mail Alert
RSS
Articles by authors
Hongzhi LUAN
Shucai LI
Yi LI
Chuanjie QIAN
Qiong MEI
Cite this article:

Hongzhi LUAN,Shucai LI,Yi LI,Chuanjie QIAN,Qiong MEI. Semantic communication system in deterministic networking for5G-advanced vehicular network. Journal of ZheJiang University (Engineering Science), 2025, 59(5): 964-972.

URL:

https://www.zjujournals.com/eng/10.3785/j.issn.1008-973X.2025.05.010     OR     https://www.zjujournals.com/eng/Y2025/V59/I5/964


5G-A车联网确定性网络中的语义通信系统

针对现有语义通信（SC）系统编码时间较长，无法直接应用于5G-Advanced技术框架下，需要处理大量高实时性、低延迟数据的智能车联网系统（V2X）中确定性网络（DN）的问题，提出针对5G-Advanced智能车联网中确定性网络设计的截止时间敏感的语义通信系统（DDA-SC）. 该系统采用截止时间敏感的语义编码器，通过控制编码深度适应数据截止时间，确保在有效时间内完成编码与传输. 为了提高各编码截止时间下的传输可靠性，提出基于编码深度的信噪比感知网络. 该网络在训练时将SNR映射为编码深度相关的相关信噪比（R-SNR），利用注意力机制进行R-SNR感知，以在不同编码截止时间下更好地抵御信道噪声. 实验结果表明，当编码截止时间< 5 μs时，该系统在CARS-196汽车识别任务上的准确度较不考虑截止时间的系统提高80.31%以上. 结果表明，DDA-SC可以在极低编码截止时间下完成一般语义通信系统无法完成的任务，验证了该方法的有效性.


关键词： 车联网(V2X),  语义通信,  5G-Advanced,  确定性网络,  联合信源信道编码 
Fig.1 Joint source channel coding system
Fig.2 Deadline-aware semantic communication System
算法1　根据截止时间进行语义编码
输入：信源消息${\boldsymbol{x}}$.
需要：接收端反馈的信噪比$s$、编码器的参数集$({\boldsymbol{\theta}} ,{\boldsymbol{\varTheta}} ) = \{ ({{\boldsymbol{\theta}} _1},{{\boldsymbol{\varTheta}} _1}),({{\boldsymbol{\theta}} _2},{{\boldsymbol{\varTheta}} _2}), \cdots ,({{\boldsymbol{\theta}} _n},{{\boldsymbol{\varTheta}} _n})\} $、编码截止时间${T_{\text{s}}}$、编码开始时间${t_{{\text{start}}}}$、当前时间$ {t_{{\text{curr}}}} $、平均各编码层所需要的编码时间${t_{{\text{ave}}}}$、当前编码层编号$i$.
输出：传输符号${{\boldsymbol{T}}_{\boldsymbol{x}}}$.
1 初始化：${t_{{\text{start}}}} \leftarrow 0$, ${t_{{\text{curr}}}} \leftarrow 0$, $k \leftarrow 0$;
2 While ($ {t_{{\text{curr}}}} - {t_{{\text{start}}}}+{t_{{\text{ave}}}} \leqslant {T_{\text{s}}} $ AND $ k \leqslant n $) Do
3${\boldsymbol{x}} \leftarrow {f_{{{\boldsymbol{\varTheta}} _i}}}({f_{{{\boldsymbol{\theta}} _i}}}({\boldsymbol{x}}),s)$;
$ {{\boldsymbol{T}}_{\boldsymbol{x}}} \leftarrow {\boldsymbol{x}} $;
$i \leftarrow i+1$;
${t_{{\text{curr}}}} \leftarrow $当前时间;
4
5
6
7 Return ${{\boldsymbol{T}}_{\boldsymbol{x}}}$
 
Fig.3 Coding-depth based signal-noise ratio aware network
Fig.4 Training process of deadline-deadline aware Semantic Communication System
算法2　截止时间敏感的语义通信系统的训练算法
输入：训练集${{D}_{\text{I}}} = \{ ({{\boldsymbol{x}}^{(k)}},{\boldsymbol{y}}_{{\text{truth}}}^{(k)}),k = 1,2, \cdots ,N\} $.
需要：初始化的补偿因子$\gamma = \{ {\gamma _1},{\gamma _2}, \cdots ,{\gamma _n}\} $、信道模拟函数${\mathrm{CH}}({{\boldsymbol{T}}_{\boldsymbol{x}}},s)$、编码器的参数集$({\boldsymbol{\theta}} ,{\boldsymbol{\varTheta}} ) = \{ ({{\boldsymbol{\theta}} _1}, {{\boldsymbol{\varTheta}} _1}), ({{\boldsymbol{\theta}} _2},{{\boldsymbol{\varTheta}} _2}), \cdots ,({{\boldsymbol{\theta}} _n},{{\boldsymbol{\varTheta}} _n})\} $、解码器参数${\boldsymbol{\phi}} $、训练轮次T、编码器参数的学习率$\alpha $、解码器参数的学习率$\beta $、当前迭代轮次$j$、总误差$L$.
输出：${\boldsymbol{\theta}} $、${\boldsymbol{\phi}} $.
1 初始化：$j \leftarrow 0$,初始化${\boldsymbol{\theta}} $、${\boldsymbol{\phi}} $;
2 对训练集${{D}_{\text{I}}}$中的样本随机排序;
3 While ($j < $T) Do
4 从0 dB到20 dB中随机采样一个SNR，记为$s$
5 For $k = 1,2, \cdots, N$ Do
6　　$L \leftarrow 0$, ${\boldsymbol{z}}_0^{(k)} \leftarrow {{\boldsymbol{x}}^{(k)}}$;
7　　For $i = 1 ,\cdots ,n$ Do
8　　　　$ {\boldsymbol{z}}_i^{(k)} \leftarrow {f_{{{\boldsymbol{\varTheta}} _i}}}({f_{{{\boldsymbol{\theta}} _i}}}({\boldsymbol{z}}_{i - 1}^{(k)}),{\gamma _i} s)$;
9　　　　$\hat {\boldsymbol{z}}_i^{(k)} \leftarrow {\mathrm{CH}}({\boldsymbol{z}}_i^{(k)},{\gamma _i} s)$;
10　　　 $ {\boldsymbol{y}}_i^{(k)} \leftarrow {f_{\boldsymbol{\phi}} }(\hat {\boldsymbol{z}}_i^{(k)}) $;
11　　　 $L \leftarrow L+L\;({\boldsymbol{y}}_i^{(k)},{\boldsymbol{y}}_{{\text{turth}}}^{(k)})$;
12　 ${\boldsymbol{\theta}} \leftarrow {\boldsymbol{\theta}} - \alpha {\nabla _{\boldsymbol{\theta}} }L$; ${\boldsymbol{\varTheta}} \leftarrow {\boldsymbol{\varTheta}} - \alpha {\nabla _{\boldsymbol{\varTheta}} }L $; $\gamma \leftarrow \gamma - \alpha {\nabla _\gamma }L$;
　 ${\boldsymbol{\phi}} \leftarrow {\boldsymbol{\phi}} - \beta {\nabla _{\boldsymbol{\phi}} }L$; $j \leftarrow j+1$;
13 Return ${\boldsymbol{\theta}} $,${\boldsymbol{\varTheta}} $,$\gamma $,${\boldsymbol{\phi}} $
 
组件名称结构名称结构尺寸
截止时间敏感的语义编码器图像重整维度224×224×3
图像块维度32×32×3
嵌入向量尺寸384
总编码层数10
多头/单头单头注意力
基于编码深度的信噪比感知网络自适应平均池化384→96
全连接层196×24
全连接层224×1
全连接层351×50
联合信源信道解码器线性层1图像块数量×1
线性层2嵌入向量尺寸×类别数量
Tab.1 Model parameter for deadline-aware semantic communication system
超参数数值
学习率$\alpha = \beta = 0.001$
批量大小1 000
优化器Adam
损失函数交叉熵
训练轮次200
模型选择策略在验证集上性能最优早停
Tab.2 Setting of training hyperparameter
Fig.5 Impact of coding depth on model performance
Fig.6 Model performance with different coding deadline
Fig.7 Model performance at different signal-to-noise ratio
Fig.8 Ablation experiment in coding-depth based signal-noise ratio aware network
[2]   ZHAI Shuo, QIAN Benhua, WANG Rui, et al. Overview of the development and application of vehicle networking technology [C]// Proceedings of the 24th China Conference on System Simulation Technology and Applications . Hefei: Hefei University of Technology Press, 2023.
[3]   曹志义 汽车智能制造中的工业物联网技术应用[J]. 汽车与新动力, 2023, 6 (5): 1- 4
CAO Zhiyi Application of industrial internet of things technology in intelligent manufacturing of automobiles[J]. Automobiles and New Power, 2023, 6 (5): 1- 4
[4]   张应辉, 李国腾, 韩刚, 等 5G车联网中安全高效的组播服务认证与密钥协商方案[J]. 电子与信息学报, 2024, 46 (7): 3026- 3035
ZHANG Yinghui, LI Guoteng, HAN Gang, et al A secure and efficient multicast service authentication and key agreement scheme in 5G vehicular networks[J]. Journal of Electronics and Information Technology, 2024, 46 (7): 3026- 3035
doi: 10.11999/JEIT231118
[5]   黄晓舸, 吴雨航, 尹宏博, 等 车联网中基于有向无环图区块链的个性化联邦互蒸馏学习方法[J]. 电子与信息学报, 2024, 46 (7): 2821- 2830
HUANG Xiaoge, WU Yuhang, YIN Hongbo, et al A personalized federated mutual distillation learning method based on directed acyclic graph blockchain in vehicular networks[J]. Journal of Electronics and Information Technology, 2024, 46 (7): 2821- 2830
doi: 10.11999/JEIT230976
[6]   王毅成 第六代WiFi技术探析及其与5G关系的探究[J]. 信息通信, 2020, 29 (5): 1- 3
WANG Yicheng Analysis of sixth generation WiFi technology and its relationship with 5G[J]. Information and Communications, 2020, 29 (5): 1- 3
[7]   肖扬 5G-Advanced扩展现实增强技术研究[J]. 电信科学, 2024, 40 (4): 160- 169
XIAO Yang Research on 5G-advanced extended reality enhancement technology[J]. Telecommunications Science, 2024, 40 (4): 160- 169
[8]   YANG Y, GUO C, LIU F, et al Semantic communications with artificial intelligence tasks: reducing bandwidth requirements and improving artificial intelligence task performance[J]. IEEE Industrial Electronics Magazine, 2023, 17 (3): 4- 13
doi: 10.1109/MIE.2022.3174331
[9]   徐英姿, 刘原, 时梦然, 等 语义在通信中的应用综述[J]. 电信科学, 2022, 38 (Suppl.1): 43- 59
[1]   濮赞成 第五代移动通信技术应用开发研究[J]. 现代工业经济和信息化, 2018, 8 (14): 117- 118
PU Zancheng Research on application development of fifth generation mobile communication technology[J]. Modern Industrial Economy and Informatization, 2018, 8 (14): 117- 118
[9]   XU Yingzi, LIU Yuan, SHI Mengran, et al Overview of the application of semantics in communication[J]. Telecommunications Science, 2022, 38 (Suppl.1): 43- 59
[10]   LIU C, GUO C, YANG Y, et al Adaptable semantic compression and resource allocation for task-oriented communications[J]. IEEE Transactions on Cognitive Communications and Networking, 2023, 10 (3): 769- 782
[2]   翟硕, 钱本华, 王睿, 等. 车联网技术的发展与应用综述[C]//第24届中国系统仿真技术及其应用学术年会论文集. 合肥: 合肥工业大学出版社, 2023.
[11]   乔萱, 李宗辉, 刘强, 等 无线时间敏感网络技术综述[J]. 软件学报, 2025, 36 (1): 184- 202
QIAO Xuan, LI Zonghui, LIU Qiang, et al Overview of wireless time-sensitive networking technology[J]. Journal of Software, 2025, 36 (1): 184- 202
[12]   潘强. 面向图像分割的车载语义通信技术研究[D]. 北京: 北京邮电大学, 2023.
PAN Qiang. Research on semantic communication technology for image segmentation in vehicular networks [D]. Beijing: Beijing University of Posts and Telecommunications, 2023.
[13]   陈九九. 语义驱动的车联网资源分配算法研究[D]. 北京: 北京邮电大学, 2024.
CHEN Jiujiu. Research on semantic-driven resource allocation algorithms in vehicle networking [D]. Beijing: Beijing University of Posts and Telecommunications, 2024.
[14]   FENG Y, SHEN H, SHAN Z, et al. Semantic communication for edge intelligence enabled autonomous driving system [J]. IEEE Network, 2024, 39(2): 149-157.
[15]   SUN L, YANG Y, CHEN M, et al Adaptive information bottleneck guided joint source and channel coding for image transmission[J]. IEEE Journal on Selected Areas in Communications, 2023, 41 (8): 2628- 2644
doi: 10.1109/JSAC.2023.3288238
[16]   RUMELHART D E, HINTON G E, WILLIAMS R J Learning representations by back-propagating errors[J]. Nature, 1986, 323 (1986): 533- 536
[17]   DENG L The MNIST database of handwritten digit images for machine learning research [Best of the Web][J]. IEEE Signal Processing Magazine, 2012, 29 (6): 141- 142
doi: 10.1109/MSP.2012.2211477
[1] Yong LI,Zhiqiang LIU,Maoxing TIAN,Songlin JIA. Semantic compression algorithm based on global correlated semantic importance[J]. Journal of ZheJiang University (Engineering Science), 2025, 59(4): 795-803.