Scholarlink

<p>Ultra-wideband (UWB) technology offers considerable advantages for indoor positioning. However, its accuracy significantly decreases in non-line-of-sight environments, particularly in dynamic scenarios with frequent human movements. To address this challenge, this study proposed a temporal convolutional network with a channel attention module (TCN-CAM) to enhance positioning performance. The TCN architecture, employing causal and dilated convolutions, effectively mitigates the vanishing gradient problem commonly encountered in neural networks and improves the model’s capacity to capture long-range dependencies in time-series data. Concurrently, CAM enhances model adaptability by emphasizing salient features under complex conditions. Simulation and field experiments demonstrated that the TCN-CAM algorithm achieved high positioning accuracy and stability with a mean error of only 3.32&#xa0;cm. Compared with LSTM-AM, CNN-CAM, and conventional TCN algorithms, the proposed method improved positioning accuracy by 76.12%, 25.06%, and 19.42%, respectively, thereby significantly enhancing the robustness and performance of UWB-based positioning systems.</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

Application of a temporal convolutional network algorithm fused with channel attention module for UWB indoor positioning

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年1月27日

查看全文
  • Liuhui He,
  • Zengzeng Lian,
  • M. Amparo Núñez-Andrés,
  • Hao Chen,
  • Chenrui Zhao,
  • Yalin Tian

摘要

Ultra-wideband (UWB) technology offers considerable advantages for indoor positioning. However, its accuracy significantly decreases in non-line-of-sight environments, particularly in dynamic scenarios with frequent human movements. To address this challenge, this study proposed a temporal convolutional network with a channel attention module (TCN-CAM) to enhance positioning performance. The TCN architecture, employing causal and dilated convolutions, effectively mitigates the vanishing gradient problem commonly encountered in neural networks and improves the model’s capacity to capture long-range dependencies in time-series data. Concurrently, CAM enhances model adaptability by emphasizing salient features under complex conditions. Simulation and field experiments demonstrated that the TCN-CAM algorithm achieved high positioning accuracy and stability with a mean error of only 3.32 cm. Compared with LSTM-AM, CNN-CAM, and conventional TCN algorithms, the proposed method improved positioning accuracy by 76.12%, 25.06%, and 19.42%, respectively, thereby significantly enhancing the robustness and performance of UWB-based positioning systems.