Scholarlink

<p>Internet of Underwater Things (IoUT) networks face unique challenges, including limited bandwidth, high delays, severe energy constraints, and dynamic topology changes due to underwater currents. Traditional routing protocols often fail to address these challenges effectively, resulting in a shortened network lifetime. The proposed comprehensive Stochastic Network Calculus (SNC)-based Deep Reinforcement Q-Learning (DRQL) routing model combines the adaptive learning capabilities of deep reinforcement learning (DRL) with SNC to develop energy-efficient routing strategies with probabilistic bounds. The DRQL model employs deep Q-learning with prioritized experience replay to learn opportunistic routing while handling void areas, balancing energy efficiency and performance. The SNC component provides a rigorous mathematical framework for deriving probabilistic bounds on key performance metrics, which are integrated into the DRQL reward function to ensure QoS guarantees. The DRQL model incorporates underwater acoustic channel characteristics to reflect realistic communication conditions. The DRQL model attains 7% to 9% diminution in energy utilization, an 18% improvement in network lifetime, a 12% increase in Packet Delivery Ratio (PDR), and an 8% reduction in end-to-end delay. The integration of SNC with DRL provides performance guarantees that are validated through simulation results, with the derived bounds being both tight and reliable. This SNC based DRQL approach represents a significant improvement in IoUT energy-efficient routing, with performance guarantees in underwater environments.</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

Deep Reinforcement Learning Model for Energy-Efficient Routing in Internet of Underwater Things using Stochastic Network Calculus

  • 期刊论文

  • 发表日期：

    2026年1月21日

查看全文
  • Saravanan M,
  • Kalaivani Chellappan

摘要

Internet of Underwater Things (IoUT) networks face unique challenges, including limited bandwidth, high delays, severe energy constraints, and dynamic topology changes due to underwater currents. Traditional routing protocols often fail to address these challenges effectively, resulting in a shortened network lifetime. The proposed comprehensive Stochastic Network Calculus (SNC)-based Deep Reinforcement Q-Learning (DRQL) routing model combines the adaptive learning capabilities of deep reinforcement learning (DRL) with SNC to develop energy-efficient routing strategies with probabilistic bounds. The DRQL model employs deep Q-learning with prioritized experience replay to learn opportunistic routing while handling void areas, balancing energy efficiency and performance. The SNC component provides a rigorous mathematical framework for deriving probabilistic bounds on key performance metrics, which are integrated into the DRQL reward function to ensure QoS guarantees. The DRQL model incorporates underwater acoustic channel characteristics to reflect realistic communication conditions. The DRQL model attains 7% to 9% diminution in energy utilization, an 18% improvement in network lifetime, a 12% increase in Packet Delivery Ratio (PDR), and an 8% reduction in end-to-end delay. The integration of SNC with DRL provides performance guarantees that are validated through simulation results, with the derived bounds being both tight and reliable. This SNC based DRQL approach represents a significant improvement in IoUT energy-efficient routing, with performance guarantees in underwater environments.