Scholarlink

Long Range Wide Area Network (LoRaWAN) has attracted much attention in research and business as a significant IoT enabler. It offers a desired choice for applications that utilize hundreds or thousands of actively linked devices to monitor a process or environment or aid in supervising a specific operation. Measurement of the Received Signal Strength Indicator (RSSI) of signals from numerous transmitters or reference sites is the foundation of Received Signal Strength (RSS)-based localization. The position of the devices or objects can be ascertained by contrasting the RSSI readings from these established locations. In this study, we investigated an indoor localization mechanism based on a Deep Learning (DL) algorithm called Gated Recurrent Unit (GRU) and by the strength of the received signal. To further improve the accuracy of the GRU, we have used the Bidirectional GRU (BiGRU), which gives the highest accuracy of 93%. The experimental results also analyzed the comparison of regression, classification, and softmax layers in the set of forward and backward layers in the BiGRU model.

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

RSS-Based Localization Using GRU and BiGRU Deep Learning Models in LoRaWAN-IoT Networks

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • R. Swathika,
  • S. M. Dilip Kumar

摘要

Long Range Wide Area Network (LoRaWAN) has attracted much attention in research and business as a significant IoT enabler. It offers a desired choice for applications that utilize hundreds or thousands of actively linked devices to monitor a process or environment or aid in supervising a specific operation. Measurement of the Received Signal Strength Indicator (RSSI) of signals from numerous transmitters or reference sites is the foundation of Received Signal Strength (RSS)-based localization. The position of the devices or objects can be ascertained by contrasting the RSSI readings from these established locations. In this study, we investigated an indoor localization mechanism based on a Deep Learning (DL) algorithm called Gated Recurrent Unit (GRU) and by the strength of the received signal. To further improve the accuracy of the GRU, we have used the Bidirectional GRU (BiGRU), which gives the highest accuracy of 93%. The experimental results also analyzed the comparison of regression, classification, and softmax layers in the set of forward and backward layers in the BiGRU model.