基于深度强化学习的IRS辅助NOMA-MEC通信资源分配优化  

IRSassisted NOMAMEC Communication Resource Allocation Optimization Based on Deep Reinforcement Learning

作　　者：方娟 刘珍珍 陈思琪 李硕朋 FANG Juan;LIU Zhenzhen;CHEN Siqi;LI Shuopeng(Faculty of Information Technology,Beijing University of Technology,Beijing 100124,China)

机构地区：[1]北京工业大学信息学部,北京100124

出　　处：《北京工业大学学报》2024年第8期930-938,共9页Journal of Beijing University of Technology

基　　金：国家自然科学基金资助项目(61202076);北京市自然科学基金资助项目(4192007)。

摘　　要：为了解决无法与边缘服务器建立直连通信链路的盲区边缘用户卸载任务的问题,设计了一个基于深度强化学习(deep reinforcement learning, DRL)的智能反射面(intelligent reflecting surface, IRS)辅助非正交多址(non-orthogonal multiple access, NOMA)通信的资源分配优化算法,以获得由系统和速率和能源效率(energy efficiency, EE)加权的最大系统收益,从而实现绿色高效通信。通过深度确定性策略梯度(deep deterministic policy gradient, DDPG)算法联合优化传输功率分配和IRS的反射相移矩阵。仿真结果表明,使用DDPG算法处理移动边缘计算(mobile edge computing, MEC)的通信资源分配优于其他几种对比实验算法。To solve the problem of blind spot edge user offloading tasks where direct communication links cannot be established with edge servers,an intelligent reflecting surface(IRS)-assisted non-orthogonal multiple access(NOMA)communication resource allocation optimization algorithm based on deep reinforcement learning(DRL)was designed.The algorithm aimed to obtain the maximum system benefit weighted by system sum rate and energy efficiency(EE)for green and efficient communication.The deep deterministic policy gradient(DDPG)algorithm was adopted to jointly optimize the power allocation and phase-shift matrix.The simulation results show that DDPG algorithm is superior to other comparative experimental algorithms in dealing with the communication resource allocation of mobile edge computing(MEC).

关 键 词：非正交多址(non-orthogonal multiple access NOMA) 智能反射面(intelligent reflecting surface IRS) 深度确定性策略梯度(deep deterministic policy gradient DDPG)算法 移动边缘计算(mobile edge computing MEC) 能源效率(energy efficiency EE) 系统收益 

分 类 号：TN929.5[电子电信—通信与信息系统]