北京邮电大学学报

  • EI核心期刊

北京邮电大学学报 ›› 2020, Vol. 43 ›› Issue (6): 51-58.doi: 10.13190/j.jbupt.2020-162

• 论文 • 上一篇    下一篇

IRS辅助的边缘智能系统中基于数据重要性感知的资源分配

田辉1, 倪万里1, 王雯1, 郑景桁1, 贺硕2   

  1. 1. 北京邮电大学 网络与交换技术国家重点实验室, 北京 100876;
    2. 郑州大学 信息工程学院, 郑州 450001
  • 收稿日期:2020-09-02 出版日期:2020-12-28 发布日期:2020-11-30
  • 作者简介:田辉(1963-),女,教授,博士生导师,E-mail:tianhui@bupt.edu.cn.
  • 基金资助:
    国家重点研发计划项目(2019YFC1511400)

Data-Importance-Aware Resource Allocation in IRS-Aided Edge Intelligent System

TIAN Hui1, NI Wan-li1, WANG Wen1, ZHENG Jing-heng1, HE Shuo2   

  1. 1. State Key Laboratory of Networking and Switching Technology, Beijing University of Posts and Telecommunications, Beijing 100876, China;
    2. School of Information Engineering, Zhengzhou University, Zhengzhou 450001, China
  • Received:2020-09-02 Online:2020-12-28 Published:2020-11-30

摘要: 针对智能反射面(IRS)辅助的边缘智能系统中模型参数汇聚的问题,提出一种基于数据重要性感知的资源分配算法.利用凸优化和分支定界等方法交替优化用户的发射功率、传输次数和智能反射面的相移矩阵.仿真结果表明,所提算法能够基于本地数据的重要性差异有效汇聚分布式智能体的模型参数,并最大化加权和速率.

关键词: 智能反射面, 模型汇聚, 重要性感知, 资源分配

Abstract: In order to solve the problem of model aggregation in intelligent reflecting surface (IRS) aided edge intelligent system, a data-importance-aware resource allocation algorithm is proposed by using convex optimization and branch-and-bound methods to alternately design the user's uplink power, transmission time, and the phase shifts of IRS. Simulation results show that the proposed algorithm can effectively aggregate the model parameters of the distributed agents based on the importance difference of local data, and can maximize the uplink weighted sum rate.

Key words: intelligent reflecting surface, model aggregation, importance aware, resource allocation

中图分类号: