Scholarlink

LPWAN technologies like LoRaWAN connect objects to the Internet using a base station or gateway. These networks must meet several requirements: low deployment costs, low energy consumption, high capacity, and data security. They often use simple communication protocols like ALOHA to reduce traffic, which can lead to increased packet collisions as the number of connected devices grows. Recent advancements propose using reinforcement learning techniques to improve LoRaWAN network performance. For instance, decentralized learning at the device level can optimize radio parameters for each packet transmission, reducing interference and enhancing reliability. Algorithms like the Multi-Armed Bandit and Upper Confidence Bound (UCB) have shown promise in reducing collisions and extending device battery life with minimal processing and memory overhead. This article further explores reinforcement learning for ultra-dense LoRaWAN networks to mitigate radio collisions and minimize energy consumption. The study demonstrates that these techniques can effectively enhance network performance.

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

Enhancing LoRaWAN Network Performance with Reinforcement Learning: Mitigating Collisions and Reducing Energy Consumption

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Pape Abdoulaye Fam,
  • Ibrahima Faye,
  • Papa Silly Traore,
  • Mamadou Lamine Ndiaye

摘要

LPWAN technologies like LoRaWAN connect objects to the Internet using a base station or gateway. These networks must meet several requirements: low deployment costs, low energy consumption, high capacity, and data security. They often use simple communication protocols like ALOHA to reduce traffic, which can lead to increased packet collisions as the number of connected devices grows. Recent advancements propose using reinforcement learning techniques to improve LoRaWAN network performance. For instance, decentralized learning at the device level can optimize radio parameters for each packet transmission, reducing interference and enhancing reliability. Algorithms like the Multi-Armed Bandit and Upper Confidence Bound (UCB) have shown promise in reducing collisions and extending device battery life with minimal processing and memory overhead. This article further explores reinforcement learning for ultra-dense LoRaWAN networks to mitigate radio collisions and minimize energy consumption. The study demonstrates that these techniques can effectively enhance network performance.