Scholarlink

<p>Connected and automated vehicles (CAVs) play a vital role in improving the efficiency of intelligent transportation systems (ITS). Anomalous sensor data from cyberattacks can cause a loss of vehicle control. Furthermore, existing methods typically perform fixed-scale feature extraction on time-domain data, leading to unstable detection in complex CAV environments. To mitigate redundant information interference caused by fixed-scale feature extraction, a novel anomaly detection method for ITS, i.e., dual attention multi-scale network (DAMSN), is proposed. In DAMSN, multi-scale adaptive strategies are designed to assign weights at different scales to sensor data of autonomous vehicles. Specifically, the adaptive multi-scale network (AMSN) is used to adaptively select the optimal scale to segment various sensor data of autonomous vehicles, so that the global features and local details could be better extracted. Experimental results show that the proposed DAMSN has good generalization performance for anomaly detection in CAVs, which has better performance than the existing results.</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

Dual attention multi-scale network for anomaly detection of intelligent vehicles

  • 期刊论文

  • 发表日期：

    2026年4月9日

查看全文
  • Wenxuan Xu,
  • Zhitao He,
  • Yongyi Chen,
  • Dan Zhang

摘要

Connected and automated vehicles (CAVs) play a vital role in improving the efficiency of intelligent transportation systems (ITS). Anomalous sensor data from cyberattacks can cause a loss of vehicle control. Furthermore, existing methods typically perform fixed-scale feature extraction on time-domain data, leading to unstable detection in complex CAV environments. To mitigate redundant information interference caused by fixed-scale feature extraction, a novel anomaly detection method for ITS, i.e., dual attention multi-scale network (DAMSN), is proposed. In DAMSN, multi-scale adaptive strategies are designed to assign weights at different scales to sensor data of autonomous vehicles. Specifically, the adaptive multi-scale network (AMSN) is used to adaptively select the optimal scale to segment various sensor data of autonomous vehicles, so that the global features and local details could be better extracted. Experimental results show that the proposed DAMSN has good generalization performance for anomaly detection in CAVs, which has better performance than the existing results.