Scholarlink

<p>The rapid growth of IoT applications has increased the demand for low-latency and energy-efficient computation at the network edge. However, existing MEC systems struggle to handle dynamic workloads because task offloading is often performed without considering real-time network conditions, load imbalance, or multi-constraint resource requirements. This study addresses the problem of minimizing reaction latency in SDN-assisted MEC environments through a hybrid dynamic offloading and resource allocation framework. The method integrates density-based task clustering (DBSCAN), a Mixed Integer Programming (MIP) driven Heap-Based Optimizer (HBO) for offloading decisions, and a Deep Q-Network (DQN) for multi-constraint resource allocation. Real-time network information is incorporated through OpenFlow to support adaptive routing and load-aware scheduling. Prototype evaluation on an SDN–MEC testbed shows that the proposed system significantly reduces response latency, energy consumption, and load imbalance compared to Greedy-Q and traditional Q-learning approaches.</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

Computation offloading and scheduling in mobile edge networks and SDN

  • 期刊论文

  • 发表日期：

    2026年4月21日

查看全文
  • S. Dinesh Kumar,
  • R. M. S. Parvathi

摘要

The rapid growth of IoT applications has increased the demand for low-latency and energy-efficient computation at the network edge. However, existing MEC systems struggle to handle dynamic workloads because task offloading is often performed without considering real-time network conditions, load imbalance, or multi-constraint resource requirements. This study addresses the problem of minimizing reaction latency in SDN-assisted MEC environments through a hybrid dynamic offloading and resource allocation framework. The method integrates density-based task clustering (DBSCAN), a Mixed Integer Programming (MIP) driven Heap-Based Optimizer (HBO) for offloading decisions, and a Deep Q-Network (DQN) for multi-constraint resource allocation. Real-time network information is incorporated through OpenFlow to support adaptive routing and load-aware scheduling. Prototype evaluation on an SDN–MEC testbed shows that the proposed system significantly reduces response latency, energy consumption, and load imbalance compared to Greedy-Q and traditional Q-learning approaches.