针对终端用户产生计算任务大小动态变化以及在工业物联网场景下业务的低时延、低能耗需求,提出了一种基于用户意愿度的D2D(device to device)协助的工业物联网资源分配模型.首先在用户层,每隔时隙t,由概率分布函数更新用户成为资源给予端的意愿度,在移动边缘计算(MEC)服务器层,使 MEC具有决策功能,能对终端上传任务做出判断,寻找出合适的 MEC处理;其次基于K-means聚类算法,将终端产生的任务匹配到对应的层进行处理;最后在资源分配阶段,为解决Q-learning里Q表难以实时更新的问题,提出N-DQN算法,使用双层神经网络相互拟合.仿真表明所提策略较传统方法,系统能耗降低约10%,系统时延降低约12%.
Resource allocation of industrial internet of things based on user willingness D2D assistance
In view of the dynamic changes in the size of computing tasks generated by end users and the low-latency and low-energy consumption requirements of services in industrial IoT scenarios,a D2D-assisted industrial IoT resource allocation model based on user willingness is proposed.First,at the user layer,every time slot t,the probability distribution function is used to update the user's willingness to become a resource giver.At the mobile edge computing(MEC)server layer,the MEC is given a decision-making function that can make judgments on terminal upload tasks and find the appropriate solution.MEC processing;secondly,based on the K-means clustering algorithm,the tasks generated by the terminal are matched to the corresponding layer for processing;finally,in the resource allocation stage,in order to solve the problem that the Q table in Q-learning is difficult to update in real time,N-DQN is proposed algorithms,fit each other using two-layer neural networks.Simulation shows that the proposed strategy reduces system energy consumption by 10%and system delay by 12%compared with traditional methods.
willingnessindustrial internet of thingsedge computingK-meansresource allocation
NETL
NSTL
万方数据
