移动通信2024，Vol.48Issue(7)：73-79,7.DOI:10.3969/j.issn.1006-1010.20240701-0001
复杂约束条件下的无线资源优化:AI视角下的方法和洞察
Wireless Resource Optimization under Complex Constraints:Methods and Insights from an AI Perspective
摘要
Abstract
This paper summarizes various deep learning-based methods for tackling the complex constraints commonly encountered in wireless resource optimization problems.Although existing deep learning-based methods have achieved great success in various power allocation and beamforming design problems,most of the considered problems are restricted to simple constraints(such as power budget constraints),which can be satisfied by a simple projection operation.However,it is still challenging to tackle the more complex constraints,such as non-convex quality-of-service constraints,where the optimization variables and wireless channels are highly coupled.Aiming at the wireless resource optimization problem under complex constraints,this paper first introduces the applicability and the deficiencies of existing deep learning-based methods,which are classified into three categories:supervised learning methods,penalty learning methods,and Lagrangian duality methods.Then,a penalty-dual learning framework based on the augmented Lagrangian approach is proposed,which alternately trains two independent neural networks to infer the original variables and the corresponding Lagrange multipliers,respectively.In addition,by applying the proposed penalty-dual learning framework to two typical wireless resource optimization problems,we show through simulation experiments that the proposed penalty-dual learning framework outperforms state-of-the-art deep learning-based methods and traditional optimization algorithms in terms of constraint violation and computational time.关键词
无线资源优化/学习优化/非凸优化/惩罚对偶/复杂约束Key words
wireless resource optimization/learning to optimize/non-convex optimization/penalty-dual learning framework/complex constraints分类
信息技术与安全科学引用本文复制引用
李洋,徐凡,张纵辉,刘亚锋..复杂约束条件下的无线资源优化:AI视角下的方法和洞察[J].移动通信,2024,48(7):73-79,7.基金项目
国家重点研发计划项目"学习优化理论与方法及其在5G网络中的应用"(2022YFA1003900) （2022YFA1003900）
国家自然科学基金项目"面向大规模无线资源管理的AI辅助优化方法研究"(62101349),"面向天基信息实时服务的星地一体化传输关键技术"(U23B2005),"数据与模型双驱动的大规模MIMO高速传输关键技术研究"(62071409),"混合整数规划的人工智能方法"(11991021) （62101349）
鹏城实验室重大攻关任务"面向6G智能通信的数理基础与核心算法"(PCL2023AS1-2) （PCL2023AS1-2）
深圳市科技创新委员会优秀科技创新人才培养(杰出青年基础研究)"面向超大规模天线通信系统的分布式信号处理技术与基础理论"(RCJC20210609104448114) （杰出青年基础研究）
广东省大数据计算基础理论与方法重点实验室 （）
