Scholarlink

<p>With the rapid advancement of Internet of Things (IoT) technologies and the exponential proliferation of smart mobile devices, traditional cloud-centric computing paradigms are increasingly challenged by issues such as high data transmission delay and network congestion, which substantially degrade users’ Quality of Experience (QoE). Mobile Edge Computing (MEC), with its low-delay and localized computing capabilities, has emerged as a promising solution to address these limitations. In this work, we first overcome the shortcomings of the widely adopted two-tier offloading architecture by constructing a cloud–edge–device three-layer task offloading model on the EdgeCloudSim platform, thereby effectively combining the low-delay advantages of mobile devices and edge nodes with the high computational power of cloud data centers. This design allows the system to efficiently manage large-scale, dynamic workloads. Second, we formulate a more realistic multi-objective offloading optimization problem that jointly considers device mobility, delay, energy consumption, and economic cost. Finally, we apply a Deep Double Q-Network (DDQN) algorithm integrated with an action mask to this three-tier architecture for multi-objective task offloading optimization. Experimental results demonstrate that compared to various baseline strategies, the proposed approach achieves strong performance across latency, energy consumption, cost, QoE, and task offloading success rate.</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-based adaptive task offloading for mobile edge computing

  • 期刊论文

  • 发表日期：

    2026年4月13日

查看全文
  • Jian Sun,
  • Wei Zhang,
  • Mengyue Han,
  • Muqi Shen,
  • Xue Wu

摘要

With the rapid advancement of Internet of Things (IoT) technologies and the exponential proliferation of smart mobile devices, traditional cloud-centric computing paradigms are increasingly challenged by issues such as high data transmission delay and network congestion, which substantially degrade users’ Quality of Experience (QoE). Mobile Edge Computing (MEC), with its low-delay and localized computing capabilities, has emerged as a promising solution to address these limitations. In this work, we first overcome the shortcomings of the widely adopted two-tier offloading architecture by constructing a cloud–edge–device three-layer task offloading model on the EdgeCloudSim platform, thereby effectively combining the low-delay advantages of mobile devices and edge nodes with the high computational power of cloud data centers. This design allows the system to efficiently manage large-scale, dynamic workloads. Second, we formulate a more realistic multi-objective offloading optimization problem that jointly considers device mobility, delay, energy consumption, and economic cost. Finally, we apply a Deep Double Q-Network (DDQN) algorithm integrated with an action mask to this three-tier architecture for multi-objective task offloading optimization. Experimental results demonstrate that compared to various baseline strategies, the proposed approach achieves strong performance across latency, energy consumption, cost, QoE, and task offloading success rate.