LEO星座边缘计算网络中的动态计算卸载策略
Dynamic computing offloading strategy in LEO constellation edge computing network
高玉芳 1姬智 2赵康僆 1李文峰 1胡佩聪3
作者信息
- 1. 南京大学电子科学与工程学院,江苏 南京 210046
- 2. 陆军工程大学通信工程学院,江苏 南京 210007
- 3. 中国人民解放军63893部队,河南 洛阳 471003
- 折叠
摘要
在卫星边缘计算网络中,当过多用户通过同一信道接入卫星时,产生的同信道干扰会导致边缘计算性能下降.为了解决该难题,在动态环境低地球轨道(LEO)星座边缘计算网络的系统模型下,提出了一种基于随机博弈的多用户计算卸载策略.在考虑用户的自私性、星地信道的随机特性和地面用户接入的动态特性的前提下,从博弈论的角度,将动态环境下地面用户的卸载决策过程表述为随机博弈,证明了所制定的随机博弈等价于具有至少一个纳什均衡(NE)的加权势博弈,并且NE最小化系统开销.为了在动态环境下以分布式方式达到NE,基于随机学习设计了一种智能随机学习算法,以高效达到所提随机博弈的NE.仿真结果表明,与基准算法相比,所提算法能够显著降低同信道干扰和系统开销,并达到接近最优的性能.
Abstract
In satellite edge computing networks,when too many ground users access the satellite through the same chan-nel,the resulting co-channel interference will lead to edge computing performance degradation.To address this problem,a multi-user computing offloading strategy based on stochastic game was proposed under the system model of dynamic environment low earth orbit constellation edge computing network.On the premise of considering the selfishness of us-ers,the stochastic characteristics of the satellite-ground channel and the dynamic nature of ground user access,from the perspective of game theory,the offloading decision-making process of ground users in the dynamic environment was for-mulated as a stochastic game.It was proved that the formulated stochastic game was equivalent to a weighted potential game with at least one Nash Equilibrium(NE),and the NE minimized the system overhead.In order to achieve NE in a distributed manner under dynamic environment,an intelligent stochastic learning algorithm based on the stochastic learn-ing was designed to efficiently achieve NE for the proposed stochastic game.Simulation results show that compared to the benchmark algorithm,the proposed algorithm can significantly reduce the co-channel interference and the system overhead,and achieve near-optimal performance.
关键词
空间物联网/边缘计算/计算卸载/动态环境/势博弈Key words
Internet of space things/edge computing/computing offloading/dynamic environment/potential game引用本文复制引用
基金项目
国家自然科学基金资助项目(62131012)
中央高校基本科研业务费专项资金资助项目(021014380187)
出版年
2024
国家科技期刊平台
NSTL
万方数据