基于无线供能和节点间合作的二进制计算卸载方案
Binary Computing Offloading Scheme Based on Wireless Power Supply and Node Cooperation
秦娥 1许方耀 1徐也淳 2池凯凯 1毛科技 1李卫锋 3何文秀4
作者信息
- 1. 浙江工业大学计算机科学与技术学院,浙江 杭州 310023
- 2. 浙江工业大学公共管理学院,浙江 杭州 310023
- 3. 浙江华通云数据科技有限公司,浙江 杭州 310030
- 4. 浙江工业大学之江学院,浙江 绍兴 312030
- 折叠
摘要
针对性研究了无线供能下的移动边缘计算,其中计算节点能够自主执行计算任务或将其卸载给空闲节点或边缘服务器,以最大化节点的总计算速率(Sum Computing Rate,SCR).首先,将SCR最大化问题建模为一个非凸问题,考虑到能量因果和任务因果等约束.接着,提出了一个基于深度强化学习的解决方案,采用深度神经网络输出近似最优的二进制卸载决策.最后,设计了高效算法来解决在给定卸载决策下的子问题.该方案具备在线学习能力,具有快速收敛和低计算复杂度的特点,实现了近似最大SCR.
Abstract
Mobile edge computing under wireless power supply is focused on,where computing nodes can independently perform computing tasks or unload them to idle nodes or edge computing server to maximize the sum computing rate(SCR)of nodes.Firstly,the SCR maximi-zation problem is modeled as a non-convex problem,taking into account the constraints such as energy causality and task causality.Next,a solution based on deep reinforcement learning is proposed,which uses deep neural networks to output approximately optimal binary off-loading decisions.Finally,an efficient algorithm is designed to solve the subproblem under a given offloading decision.This scheme has the ability of online learning,fast convergence,and low computational complexity,achieving the approximate maximum SCR.
关键词
移动边缘计算/无线供能/节点间协作/二进制卸载Key words
mobile edge computing/wireless power supply/node cooperation/binary offloading引用本文复制引用
基金项目
浙江省基础公益研究计划项目(LGG22F020014)
国家自然科学基金项目(62072410)
出版年
2024