摘要
在工业物联网中,设备的计算能力通常有限,但任务往往需要及时执行,存在延时情况。为了解决该问题,提出了一种基于移动边缘计算的分层机器学习任务分配框架。该方法通过根据每个设备不同需求,决定要卸载的任务部分,以最小化处理延迟。并结合机器学习模型复杂度和推理错误率、数据质量、设备和服务器的计算能力以及通信带宽的影响,提出了一个联合优化问题,以使总时延最小,该问题的解即为最优卸载策略。在实验部分,分析了该研究的算法有效性,并与现有方法进行了对比。结果表明,提出的算法具有较高的性能。
In the Industrial Internet of Things(IIoTs),the computing power of equipment is usually limited,but tasks often need to be executed in time,where delay situation is exist.In order to solve this problem,a hierarchical machine learning task allocation framework is proposed based on mobile edge computing.This method determines the part of the task to be unloaded according to the different needs of each device to minimize processing delay.Combined with the influence of machine learning model complexity and reasoning error rate,data quality,computing power of equipment and server,and communication bandwidth,a joint optimization problem is proposed to minimize the total delay.The solution to this problem is the optimal unloading strategy.In the experimental part,the effectiveness of the algorithm of this research is analyzed and compared with the existing methods.The results show that the proposed algorithm has higher performance.
作者
李世强
刘皓若
费海平
LI Shi-qiang;LIU Hao-ruo;FEI Hai-ping(Industrial Internet Innovation Center(Shanghai)Co.,Ltd.,Shanghai 201306,China;Shanghai Huafeng chuangxiang Internet Technology Co.,Ltd.,Shanghai 201315,China)
出处
《信息技术》
2021年第7期65-70,共6页
Information Technology
基金
国家自然科学基金(61502061)。
关键词
工业物联网
机器学习
移动边缘计算
凸优化问题
机器视觉
IIoTs
machine learning
mobile edge computing
convex optimization problem
machine vision
