摘要
移动边缘计算(Mobile Edge Computing,MEC)技术的发展有效缓解了移动设备计算和存储需求与系统资源受限之间的矛盾。通过将计算任务卸载至移动边缘计算服务器,可以显著提升移动设备的服务质量。然而,在实际网络环境中,任务卸载存在潜在的失败风险,导致任务仍然需要在本地计算,从而产生延迟等额外开销。因此,如何制定合理的卸载决策并优化资源分配,以适应不同任务类型,仍然是MEC系统面临的关键挑战。针对上述问题,基于异步优势动作—评价(Asynchronous Advantage Actor-Critic,A3C)框架,提出一种混合任务卸载与资源分配算法,该算法结合前景理论,以平衡任务卸载的风险与收益,确保更合理的决策。此外,根据任务的截止时间确定其优先级,并据此分配MEC计算资源,保障高优先级任务获得所需的计算资源。算法考虑了卸载失败的风险,并利用深度强化学习自适应优化卸载决策,以适应动态网络环境。仿真结果表明,与基线算法相比,所提算法在获得更优卸载方案的同时,有效降低了任务延迟和能耗。
The development of mobile edge computing(MEC)technology has effectively alleviated the contradiction between the growing computational and storage demands of mobile devices and their limited system resources.By offloading computational tasks to MEC servers,the service quality of mobile devices can be significantly enhanced.However,in real-world network environments,task offloading is subject to potential failure risks,leading to local execution and additional latency overhead.Therefore,making reasonable offloading decisions and optimizing resource allocation to accommodate various task types remain critical challenges in MEC systems.To address these challenges,this paper proposed a hybrid task offloading and resource allocation algorithm based on the asynchronous advantage actor-critic(A3C)framework.The algorithm integrated prospect theory to balance the risks and rewards associated with task offloading,ensuring more rational decision-making.Additionally,task priorities were determined based on their deadlines,and MEC computing resources were allocated accordingly to guarantee that high-priority tasks received the necessary computational resources.The proposed algorithm considered the risk of offloading failure and leveraged deep reinforcement learning to adaptively optimize offloading decisions in dynamic network environments.Simulation results demonstrated that,compared to baseline algorithms,the proposed approach achieved superior offloading strategies while effectively reducing task delay and energy consumption.
作者
唐山林
吴涛
周启钊
Tang Shanlin;Wu Tao;Zhou Qizhao(School of Computer Science,Chengdu University of Information Technology,Chengdu 610225,China)
出处
《西南大学学报(自然科学版)》
北大核心
2026年第1期229-239,共11页
Journal of Southwest University(Natural Science Edition)
基金
四川省科技研究开发项目(2024ZHCG0195,2023jdzh0034)
成都信息科技大学人才引进研究启动项目(kytz202269)。
关键词
移动边缘计算
任务卸载
资源分配
mobile edge computing
task offloading
resource allocation
