Scholarlink

<p>In high-mobility environments, wireless communication systems experience rapid fluctuation in network quality due to vehicle motion, resulting in outdated Channel Quality Indicator (CQI) feedback and thus challenging effective network management. In this regard, we propose a position-aided CQI prediction scheme that learns both global and local dependencies from critical user equipment (UE) features, such as position, speed, and path loss by jointly using Fully Convolutional Network (FCN) layers and a self-attention-based Long Short-Term Memory (LSTM) architecture. The model utilizes positional data of both the UE and Base Station (BS) precisely, including longitude and latitude, to compute the Euclidean distance. The path loss is calculated based on this distance, while UE speed is treated as an independent feature to capture dynamic mobility. These combined features such as position, speed, and path loss are fed to the model for better CQI prediction accuracy on the baseline’s models. Compared with several state-of-the-art CQI prediction models, including Bayesian Ridge (BR), Gated Recurrent Unit, and standard LSTM models, shows that the proposed FCN Attention-LSTM model reduces mean absolute error and root mean square error by at least 34.44% in high-speed scenarios. Furthermore, the validation on the recent 5G non-standalone dataset confirms the model’s strong generalization capability to achieve high predictive accuracy while maintaining a low inference latency. Such improved accuracy can enhance communication reliability in dynamic environments and provide essential proactive CQI for intelligent network and system management in V2X systems, making the proposed model well-suited for such demanding high-speed wireless communication applications.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Position-Aided Channel Quality Indicator Prediction in V2X Scenarios using Attention-LSTM with Fully Convolutional Networks

  • 期刊论文

  • 发表日期：

    2026年2月18日

查看全文
  • Harun Ur Rashid,
  • Abate Selamawit Chane,
  • Loret Abiy Awoke,
  • Seong Ho Jeong

摘要

In high-mobility environments, wireless communication systems experience rapid fluctuation in network quality due to vehicle motion, resulting in outdated Channel Quality Indicator (CQI) feedback and thus challenging effective network management. In this regard, we propose a position-aided CQI prediction scheme that learns both global and local dependencies from critical user equipment (UE) features, such as position, speed, and path loss by jointly using Fully Convolutional Network (FCN) layers and a self-attention-based Long Short-Term Memory (LSTM) architecture. The model utilizes positional data of both the UE and Base Station (BS) precisely, including longitude and latitude, to compute the Euclidean distance. The path loss is calculated based on this distance, while UE speed is treated as an independent feature to capture dynamic mobility. These combined features such as position, speed, and path loss are fed to the model for better CQI prediction accuracy on the baseline’s models. Compared with several state-of-the-art CQI prediction models, including Bayesian Ridge (BR), Gated Recurrent Unit, and standard LSTM models, shows that the proposed FCN Attention-LSTM model reduces mean absolute error and root mean square error by at least 34.44% in high-speed scenarios. Furthermore, the validation on the recent 5G non-standalone dataset confirms the model’s strong generalization capability to achieve high predictive accuracy while maintaining a low inference latency. Such improved accuracy can enhance communication reliability in dynamic environments and provide essential proactive CQI for intelligent network and system management in V2X systems, making the proposed model well-suited for such demanding high-speed wireless communication applications.