摘要
移动边缘计算通过将计算任务卸载到无线网络边缘,可有效减少任务延迟与终端能耗.对于偏远地区分布的大量物联设备(如风电、光伏等电力物联终端),现有地面网络无法为其提供有效的网络服务.因此,本文重点研究空地一体化异构网络模型,通过联合设计无人机轨迹、任务卸载与计算资源分配,以最大限度地减少物联设备任务执行延迟与能耗.针对目标函数的非凸性和网络动态造成的信息不确定性,本文将问题建模为马尔可夫(Markov)决策过程,并提出一种基于MATD3的UAV轨迹与网络资源协同优化算法.实验结果表明,与基准算法相比,本文提出的方案在系统计算能耗和时延方面性能更优.
Mobile edge computing effectively reduces service latency and terminal energy consumption by offloading tasks to the edge of the wireless networks.For tremendous IoT devices distributed in remote areas(e.g.,wind power,photovoltaic,and other power IoT terminals),existing terrestrial communication networks are not able to provide effective computing services.Therefore,in this paper,we comprehensively consider an airground integrated heterogeneous network model to minimize the sum of execution delay and energy consumption of IoT devices by jointly designing UAV trajectory,task offloading,and computing resource allocation.Regarding the non-convexity of the objective function and the information uncertainty caused by network dynamics,the problem is modeled as a Markov decision process,and we propose a joint UAV trajectory and network resource optimization algorithm based on MATD3.Experimental results show that compared with other baselines,the proposed scheme has superior performance in system computing energy consumption and delay.
作者
秦鹏
王硕
付民
赵雄文
Peng QIN;Shuo WANG;Min FU;Xiongwen ZHAO(School of Electrical and Electronic Engineering,North China Electric Power University,Beijing 102206,China;Hebei Telecom Co.,Ltd.,Shijiazhuang 050036,China)
出处
《中国科学：信息科学》
CSCD
北大核心
2024年第6期1474-1486,共13页
Scientia Sinica(Informationis)
基金
国家自然科学基金(批准号:62201212)
河北省自然科学基金(批准号:F2022502017)
中央高校基本科研业务费专项资金(批准号:2023JC003)资助项目。
关键词
空地一体化异构网络
卸载决策
资源分配
UAV
轨迹优化
多智能体深度强化学习
air-ground integrated heterogeneous network
offloading decision-making
resource allocation
UAV trajectory optimization
multi-agent deep reinforcement learning(MADRL)
