HOA和SPEA2结合的边缘云计算任务卸载优化  

Optimization of Edge Cloud Computing Task Offloading Combining HOA and SPEA2

作　　者：郑文祥 刘静 陈家辉 ZHENG Wen-xiang;LIU Jing;CHEN Jia-hui(School of Computer Science&Technology,Wuhan University of Science and Technology,Wuhan 430065,China)

机构地区：[1]武汉科技大学计算机科学与技术学院,湖北武汉430065

出　　处：《计算机技术与发展》2025年第4期7-14,共8页Computer Technology and Development

基　　金：湖北省教育科学研究计划重点项目(D20201102)。

摘　　要：多接入边缘计算(MEC)已经成为一种重要的范式,通过使用MEC技术,终端设备可以将其计算任务卸载到邻近的边缘服务器或远程云服务器上执行。但对于庞大的任务量而言,它们的计算能力、内存和能源仍然受限。如何进行有效的任务分流至关重要。该文允许任务在三个位置(本地设备、边缘端、云端)处理,有效避免了传统的二元规划机制缺乏灵活性,不考虑任务依赖等问题。并提出了一种河马优化算法(HOA)与改进型强度Pareto进化算法(SPEA2)相结合的多目标河马优化算法(MHOA)。MHOA算法充分利用HOA的全局搜索能力和多样性维护等特点,以及SPEA2中Pareto前沿优化和局部开发能力,并采用特定的初始化算法,在考虑任务依赖关系的同时,解决边缘云计算环境下多用户依赖任务的卸载问题,同时最小化执行时延、能耗、资源使用成本三个约束。将MHOA与其他优化算法进行了比较,结果表明MHOA在适应度方面比第二优的优化器优化了3.5%。Multi-access edge computing(MEC)has become an important paradigm,allowing devices to offload their computations to nearby edge servers or remote cloud servers.However,given the vast amount of tasks,their computing power,memory,and energy are still limited.Effective task offloading is crucial.We allow tasks to be processed in three locations(local device,edge end,cloud),effectively avoiding the traditional binary planning mechanism lack of flexibility,does not consider task dependence and other problems.And we propose a Multi-objective Hippo Optimization Algorithm(MHOA)that combines the Hippo Optimization Algorithm(HOA)with the Strength Pareto Evolutionary Algorithm(SPEA2).This approach leverages HOA's global search capability and diversity maintenance,as well as SPEA2's Pareto frontier optimization and local development capabilities,and incorporates a specific initialization algorithm.While considering task dependencies,it addresses the problem of multi-user dependent task offloading in edge cloud computing environments,aiming to minimize execution delay,energy consumption,and resource usage cost.MHOA is compared with other optimization algorithms,and the results show that MHOA outperforms the second-best comparator by 3.5%in terms of fitness.

关 键 词：边缘计算 多目标计算卸载 任务依赖 多用户场景 多目标河马优化算法 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]