深度学习辅助的通感一体化V2I网络波束预测  

Beamforming Prediction for Integrated Sensing and Communication in Vehicle-to-Infrastructure Networks Using Deep Learning

作　　者：徐飞杭 李旻[1] 李荣鹏[1] 赵明敏[1] XU Feihang;LI Min;LI Rongpeng;ZHAO Mingmin(College of Information Science and Electronic Engineering,Zhejiang University,Hangzhou 310013,China)

机构地区：[1]浙江大学信息与电子工程学院,浙江杭州310013

出　　处：《移动通信》2023年第9期57-63,共7页Mobile Communications

基　　金：浙江省杰出青年基金项目“面向移动分布式智能业务的任务中心网络”(LR23F010005);国家自然科学基金面上项目“基于业务感知的内生智能通信网络研究”(62071425);中央高校基础基金(226-2022-00195)。

摘　　要：在V2I网络中,为了提高多车辆通信的总可达率和公平性,研究通感一体化与深度学习辅助V2I通信性能优化问题。采用卷积神经网络与长短期记忆网络相结合的方式用于波束形成预测。该网络包含两个子网络,分别用于获取车辆的状态信息和功率分配方案,然后计算波束形成向量。所设计的网络综合了监督学习和无监督学习的特点,在保证用户公平性的同时,最大限度地提高总可达率。仿真结果表明,与现有的基准方案对比,所提算法能显著提高系统的总可达率和公平性。In vehicle-to-infrastructure networks,the performance optimization problem of vehicle networking is investigated to improve the overall achievable rate and fairness of multiple vehicles.Deep learning techniques,specifically the combination of convolutional neural networks and long short-term memory networks,are employed for beamforming prediction.The proposed network comprises two sub-networks:one for acquiring vehicle state information and the other for power allocation,subsequently calculating the beamforming vectors.The designed network leverages the strengths of both supervised and unsupervised learning,aiming to maximize the overall achievable rate while ensuring vehicle fairness.Simulation results demonstrate that the proposed algorithm significantly improves the system's overall achievable rate and fairness of vehicles compared to existing benchmark solutions.

关 键 词：通感一体化 波束跟踪 波束形成预测 深度学习 功率分配 

分 类 号：TN92[电子电信—通信与信息系统]