面向新一代无线通信的技术革新，无人机(Unmanned Aerial Vehicle，UAV)在未来通信系统中的应用越来越不可忽视。考虑传统波束管理(Beam Management，BM)方法在高动态、高频段空地链路中波束对准的高额开销，设计了一种基于人工智能(Artificial Intelligence，Al)技术的BM方案。方案基于长短时记忆(Long-Short Term Memory，LSTM)网络模型，利用深度学习(Deep Learning，DL)方法实现基站(Base Station，BS)-UAV通信过程中的BM。以参考信号接收功率(Reference Signal Receive Power，RSRP)为性能指标对BM方案进行评估，基于BM历 史经验数据将UAV终端轨迹划分为数个区域，训练特定的区域模型以更好地适应各区域中的信道传播环境特征。在模型部署阶段，根据区域划分结果按区域切换模型，实现基于模型切换的区域化AI-BM(Model Switching based Area-Specific AI-BM，MSAS AI-BM)。仿真结果表明，所提的MSAS AI-BM方案相比传统的简单穷举BM方案能够极大降低系统开销，拥有良好的RSRP保持性能。
Area-specific Intelligent Beam Management for UAV Communication
Innovative technology for the next generation of wireless communication makes the application of Unmanned Aerial Ve-hicle(UAV)in future communication systems increasingly indispensable.Considering the high overhead of beam alignment in high-dy-namic,high-frequency air to ground links with traditional Beam Management(BM)methods,an Artificial Intelligence(Al)based BM scheme is designed.This scheme is based on Long-Short Term Memory(LSTM)network models and utilizes Deep Learning(DL)methods to achieve BM in the Base Station(BS)-UAV communication process.The BM scheme is evaluated using Reference Signal Re-ceive Power(RSRP)as the performance metric.Based on historical experience data,the UAV terminal trajectory is divided into several zones,and specific zone models are trained to better adapt to environmental features in each zone.In the model deployment stage,models are switched according to the zone division results,achieving Model Switching based Area-Specific AI-BM(MSAS AI-BM).Sim-ulation results demonstrate that the proposed MSAS AI-BM scheme significantly reduces system overhead compared to traditional BM schemes with simple enumeration and exhibits excellent RSRP retention performance.
国家科技期刊平台
NSTL
万方数据
