移动边缘计算中能量感知的动态软实时任务卸载算法  

Energy-aware Dynamic Soft Real-time Task Offloading Algorithm in Mobile Edge Computing

作　　者：周治华 刘静 ZHOU Zhihua;LIU Jing(School of Computer Science and Technology,Wuhan University of Science and Technology,Wuhan 430065;Hubei Province Key Laboratory of Intelligent Information Processing and Real-time Industrial System,Wuhan 430065)

机构地区：[1]武汉科技大学计算机科学与技术学院,武汉430065 [2]智能信息处理与实时工业系统湖北省重点实验室,武汉430065

出　　处：《计算机与数字工程》2024年第10期2872-2879,2931,共9页Computer & Digital Engineering

基　　金：教育厅科学研究计划重点项目(编号:D20201102)资助。

摘　　要：移动边缘计算可以将计算密集型任务进行卸载以提高设备的计算能力并降低能耗。然而,现存的算法无法在任务的软实时和并行依赖限制下有效进行计算卸载和资源分配以降低设备的执行代价(能耗和完成时间的加权和),因此论文提出了一种分布式计算卸载算法来决定任务的执行位置、CPU频率和传输功率,以降低任务执行代价。为了实现优化目标,首先将原始优化问题建模为凸优化问题,然后根据任务的截止时间预分配计算资源,最终通过分布式算法得出资源分配和计算卸载决策。实验结果表明,论文提出的算法可以有效降低任务执行代价,并且可以降低任务超时的时间。Mobile edge computing can offload computation-intensive tasks to enhance the computational capability of a device and reduce energy consumption.However,existing algorithms cannot efficiently perform computation offloading and resource alloca-tion to reduce the execution cost(weighted sum of energy consumption and completion time)of the device under the soft real-time and parallel dependency constraints of the task,so this paper proposes a distributed computation offloading algorithm to determine the execution location,CPU frequency,and transmission power of the task to reduce the execution cost.In order to achieve the opti-mization goal,firstly,the original optimization problem is modeled as a convex optimization problem,and computational resources are pre-allocated according to the task's deadline,and finally the resource allocation and computational offloading decisions are de-rived by the distributed algorithm.The experimental results show that the algorithm proposed in this paper can effectively reduce the task execution cost as well as reduce the time of task timeout.

关 键 词：边缘计算 计算卸载 软实时 资源分配 凸优化 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]