北京邮电大学学报

  • EI核心期刊

北京邮电大学学报 ›› 2019, Vol. 42 ›› Issue (2): 25-30.doi: 10.13190/j.jbupt.2018-078

• 论文 • 上一篇    下一篇

基于机器学习的MEC随机任务迁移算法

孟浩1, 霍如1, 郭倩影1, 黄韬1,2, 刘韵洁1,2   

  1. 1. 北京工业大学 北京未来网络科技高精尖创新中心, 北京 100124;
    2. 北京邮电大学 网络与交换技术国家重点实验室, 北京 100876
  • 收稿日期:2018-04-30 出版日期:2019-04-28 发布日期:2019-04-09
  • 通讯作者: 霍如(1988-),女,讲师,E-mail:huoru@bjut.edu.cn. E-mail:huoru@bjut.edu.cn
  • 作者简介:孟浩(1991-),男,硕士生.
  • 基金资助:
    北京市科技新星计划项目(Z151100000315078);国家科技重大专项项目(2018ZX03001019-003);国家高技术研究发展计划(863计划)项目(2015AA015702)

Machine Learning-Based Stochastic Task Offloading Algorithm in Mobile-Edge Computing

MENG Hao1, HUO Ru1, GUO Qian-ying1, HUANG Tao1,2, LIU Yun-jie1,2   

  1. 1. Beijing Advanced Innovation Center for Future Internet Technology, Beijing University of Technology, Beijing 100124, China;
    2. State Key Laboratory of Networking and Switching Technology, Beijing University of Posts and Telecommunications, Beijing 100876, China
  • Received:2018-04-30 Online:2019-04-28 Published:2019-04-09

摘要: 针对移动边缘计算(MEC),提出了一种基于机器学习的随机任务迁移算法,通过将任务划分为可迁移组件和不可迁移组件,结合改进的Q学习和深度学习算法生成随机任务最优迁移策略,以最小化移动设备能耗与时延的加权和.仿真结果表明,该算法的时延与能耗加权和与移动设备本地执行算法相比节约了38.1%.

关键词: 移动边缘计算, 随机任务迁移, 机器学习, 时延, 移动设备能耗

Abstract: For mobile-edge computing (MEC), a machine learning-based stochastic task offloading algorithm was proposed. By dividing the task into offloadable components and unoffloadable components, the improved Q learning and deep learning algorithm were used to generate the optimal offloading strategy of stochastic task, which minimized the weighted sum of energy consumption and time delay of the mobile devices. The simulation results show that the proposed algorithm saves the weighted sum of energy consumption and time delay by 38.1%, compared to the local execution algorithm.

Key words: mobile-edge computing, stochastic task offloading, machine learning, delay, mobile device's energy consumption

中图分类号: