边缘环境下基于移动群智感知计算卸载的数据汇聚  

Data aggregation based on mobile crowd sensing and computation offloading in edge environment

作　　者：杨桂松[1] 桑健 Yang Guisong;Sang Jian(School of Optical-Electrical&Computer Engineering,University of Shanghai for Science&Technology,Shanghai 200093,China)

机构地区：[1]上海理工大学光电信息与计算机工程学院,上海200093

出　　处：《计算机应用研究》2024年第9期2705-2711,共7页Application Research of Computers

基　　金：国家自然科学基金资助项目(61602305,61802257);上海市自然科学基金资助项目(18ZR1426000,19ZR1477600)。

摘　　要：当前“云-端”式移动群智感知(mobile crowd sensing,MCS)系统面临负载过重的问题,导致数据汇聚过程中时延和能耗显著增加,从而降低了数据汇聚的效率。针对该问题,提出了一种基于AP-DQN的“云-边-端”MCS计算卸载算法。首先,考虑时延和能耗的均衡优化建立效用函数,以最大化系统效用作为优化目标。其次,优化P-DQN算法,提出一种联合资源分配的计算卸载算法AP-DQN,结合MCS优势,将空闲用户作为卸载设备之一。最后,使用该方法求解问题。实验结果显示,与已有算法相比,该方法能有效提高数据汇聚效率,并具有很好的稳定性。The conventional“cloud-end”MCS system currently faces problems of excessive load,leading to a significant increase in delay and energy consumption during the data aggregation process,inevitably causing a decrease in data aggregation efficiency.To tackle this issue,this paper proposed a“cloud-edge-end”MCS computation offloading algorithm based on AP-DQN.Firstly,it established a utility function considering the balanced optimization of delay and energy consumption,with the maximization of system utility as an optimized goal.Secondly,improving the P-DQN algorithm,it proposed a computational offloading algorithm AP-DQN for combining resource allocation.This algorithm,leveraging the advantages of MCS,designated idle users as one of the offloading devices.Finally,the problem was solved using the proposed method.Experimental results show that,compared to existing algorithms,the proposed method significantly improves data aggregation efficiency and maintains excellent system stability.

关 键 词：移动群智感知 边缘计算 数据汇聚 计算卸载 资源分配 深度强化学习 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]