深度强化学习的空天地架构移动边缘计算卸载策略
Mobile Edge Computing Offloading Strategy based on Deep Reinforcement Learning for Space-Air-Ground Integrated Networks
徐飞 1王泽轩 1宁辛1
作者信息
- 1. 西安工业大学计算机科学与工程学院,西安 710021
- 折叠
摘要
针对传统无人机边缘计算卸载的网络时延高、所需能耗大、计算资源有限等问题,文中提出一种LEO-UAV辅助任务卸载的集成空天地网络架构,为地面设备提供更多可用的计算资源和网络需求.为了最大限度地降低卸载任务所产生的延迟和消耗的能量,将该问题转化为马尔可夫决策模型,进一步提出利用多智能体深度确定性策略梯度(MADDPG)算法来解决.实验结果表明,与基线算法相比,MADDPG算法能够有效缩短44.45％的系统卸载任务时延,节省61.35％的能耗,证实了 MADDPG算法在处理移动边缘计算卸载方面的可靠性.
Abstract
To address the problem of high network latency,high energy consumption,and limited computational resources in traditional drone-based edge computing offloading,the paper presents an integrated air-space-ground network architecture with Low Earth Orbit Unmanned Aerial Vehicle(LEO-UAV)assisted task offloading,which can provide more available computational resources and network services for ground devices.To minimize the delay and energy consumption produced in offloading tasks,the problem is formulated as a Markov Decision Process(MDP)and further solved by the Multi-Agent Deep Deterministic Policy Gradient algorithm.Experimental results demonstrate that the MADDPG algorithm can effectively reduce the system's task processing delay and energy consumption by 44.45％and 61.35％,respectively,which verifies the reliability of the MADDPG algorithm in handling mobile edge computing offloading tasks.
关键词
卫星网络/移动边缘计算/计算卸载/深度强化学习Key words
satellite networks/Mobile Edge Computing(MEC)/computational offloading/deep reinforcement learning引用本文复制引用
出版年
2024
NETL
NSTL
万方数据