基于强化学习和神经网络的导频功率动态优化  

Dynamic optimization of pilot power based on reinforcement learning and neural network

作　　者：李烨[1] 肖梦巧 LI Ye;XIAO Mengqiao(School of Optical-Electrical and Computer Engineering,University of Shanghai for Science and 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：《智能计算机与应用》2022年第9期94-99,104,共7页Intelligent Computer and Applications

基　　金：华为技术有限公司合作项目(YBN2019115054)。

摘　　要：为了更好地实现用户对移动通信网络的无线接入,合理分配基站导频功率十分重要,本文研究了无线接入网中基站导频功率的动态优化问题,设计了一种结合强化学习和神经网络的导频功率优化模型,以感知无线接入网的变化。其次,利用Q学习算法来维持基站与外部环境的连接和信息交互;在Q学习算法中,利用神经网络学习Q值,避免了状态爆炸问题;最后设计了关键性能指标保护机制和回退机制,以满足工程要求。仿真结果表明,提出方案能很好地适应无线网络频繁变化,并获得显著的性能增益。In order to better realize the wireless access of users to the mobile communication network, it is very important to reasonably allocate the pilot power of the base station. This paper studies the dynamic optimization of the pilot power of the base station in the radio access network. Firstly, a pilot power optimization model combining reinforcement learning and neural network is designed to sense the changes of the radio access network. Secondly, the Q-learning algorithm is used to maintain the connection and information interaction between the base station and the external environment. In the Q-learning algorithm, the neural network to learn the Q value is used for avoiding the state explosion problem. Finally, a key performance indicator protection mechanism and a rollback mechanism are designed to meet engineering requirements. Simulation results show that the proposed scheme can well adapt to the frequent changes of wireless networks and achieve significant performance gains.

关 键 词：无线接入网 导频功率 Q学习 神经网络 关键性能指标保护机制 回退机制 

分 类 号：TN929.5[电子电信—通信与信息系统]