Scholarlink

<p>In Mobile Edge Computing (MEC), Smart Mobile Devices (SMDs) can offload tasks to nearby servers, creating a fundamental trade-off among latency, energy, and cost. This paper studies the trade-off among Average Completion Time (ACT), Average Energy Consumption (AEC), and Average Price Cost (APC) in MEC, formulating it as a Multi-objective Computation Offloading Problem (MCOP) with task priority as a new constraint. To address it, an enhanced Multi-objective Evolutionary Algorithm based on Decomposition (MOEA/D), called MOEA/D-CO, is proposed, which incorporates two performance improvement schemes. First, a population initialization scheme is employed that utilizes a delay-aware execution location (EL) method to determine the initial offloading decision (local SMD or MEC server) for each task. Then, an energy-saving scheme is introduced that leverages Dynamic Voltage and Frequency Scaling (DVFS) to reduce energy consumption without extending the application’s makespan. Simulation results confirm that our algorithm yields superior offloading solutions, achieving lower completion time, energy consumption, and cost.</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

Latency-energy-cost optimization with task priority constraints for edge computing offloading

  • 期刊论文

  • 发表日期：

    2026年6月1日

查看全文
  • Zi-xin Chai,
  • Guo-qi Feng,
  • Xi-Jin

摘要

In Mobile Edge Computing (MEC), Smart Mobile Devices (SMDs) can offload tasks to nearby servers, creating a fundamental trade-off among latency, energy, and cost. This paper studies the trade-off among Average Completion Time (ACT), Average Energy Consumption (AEC), and Average Price Cost (APC) in MEC, formulating it as a Multi-objective Computation Offloading Problem (MCOP) with task priority as a new constraint. To address it, an enhanced Multi-objective Evolutionary Algorithm based on Decomposition (MOEA/D), called MOEA/D-CO, is proposed, which incorporates two performance improvement schemes. First, a population initialization scheme is employed that utilizes a delay-aware execution location (EL) method to determine the initial offloading decision (local SMD or MEC server) for each task. Then, an energy-saving scheme is introduced that leverages Dynamic Voltage and Frequency Scaling (DVFS) to reduce energy consumption without extending the application’s makespan. Simulation results confirm that our algorithm yields superior offloading solutions, achieving lower completion time, energy consumption, and cost.