Scholarlink

The Blind Adaptive Rate (B-ADR) algorithm, designed by Semtech for LoRaWAN networks, is tailored for mobile applications to allocate the spreading factor before each uplink transmission. This enhances reliability even in areas with fluctuating signal quality, a common occurrence in mobile applications. However, B-ADR predominantly focuses on spreading factor (SF) adjustment and overlooks dynamic transmission power control, leading to increased energy consumption. Consequently, this article proposes a novel and efficient B-ADR methodology tailored for mobile applications, leveraging Tiny Machine Learning (TinyML). In this approach, TinyML is deployed on each End Device (ED), enabling intelligent selection of optimal transmission power alongside SF before each uplink transmission. The proposed framework entails offline model training followed by real-time prediction execution on deployed EDs. Simulation results demonstrate superior performance of the proposed approach compared to typical B-ADR implemented on EDs. Thus, the proposed methodology proves to be well-suited for dense mobile applications within LoRaWAN networks.

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

TinyML-Based Approach for Dynamic Transmission Power in LoRaWAN Network

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Muhammad Ali Lodhi,
  • Lei Wang,
  • Khalid Ibrahim Qureshi,
  • Khalid Mahmood

摘要

The Blind Adaptive Rate (B-ADR) algorithm, designed by Semtech for LoRaWAN networks, is tailored for mobile applications to allocate the spreading factor before each uplink transmission. This enhances reliability even in areas with fluctuating signal quality, a common occurrence in mobile applications. However, B-ADR predominantly focuses on spreading factor (SF) adjustment and overlooks dynamic transmission power control, leading to increased energy consumption. Consequently, this article proposes a novel and efficient B-ADR methodology tailored for mobile applications, leveraging Tiny Machine Learning (TinyML). In this approach, TinyML is deployed on each End Device (ED), enabling intelligent selection of optimal transmission power alongside SF before each uplink transmission. The proposed framework entails offline model training followed by real-time prediction execution on deployed EDs. Simulation results demonstrate superior performance of the proposed approach compared to typical B-ADR implemented on EDs. Thus, the proposed methodology proves to be well-suited for dense mobile applications within LoRaWAN networks.