Artificial Intelligence-Aided 5G Millimeter Wave Beam Management

With the rapid development of the mobile Internet and the continuous growth of mobile terminals,new application scenarios and emerging industries continue to emerge.The fifthgeneration mobile communication technology has become a necessary solution for high-speed communication,massive device connections,and ultra-low latency.Multi-antenna and millimeter wave communication are the two core technologies of 5G.Millimeter wave communication can use a large amount of idle bandwidth to improve the transmission rate and throughput of the system.Multi-antenna technology can effectively compensate for the high transmission loss faced by millimeter wave communications through beamforming.The combination of the two can well solve the problems of high-speed transmission and ultra-low delay.While bringing significant performance improvements to the system,beam-based communication also faces many challenges.Therefore,the 5G standard provides a beam management process for maintaining millimeter-wave beam communication,and solves the beam connection in the initial access and continuous tracking stages.problem.This thesis is oriented to 5G millimeter wave communication systems,using artificial intelligence technology to conduct research on millimeter wave beam management,and using a prototype verification platform to test the performance of the algorithm in actual scenarios.The specific research content is as follows:First,this thesis introduces the basic knowledge of millimeter wave beam management.Based on the introduction of beamforming technology,including the array signal propagation model and existing beamforming classification,the practical significance of the current use of analog beamforming technology based on discrete codebooks is analyzed,and the millimeter wave is analyzed from the perspective of the channel.The characteristics of transmission and the principle of beamforming to solve the problem of millimeter wave high transmission loss.Since this thesis studies beamforming based on discrete codebooks,the characteristics and precautions of different codebook designs are analyzed later,and 5G standards are summarized.The beam management process mentioned in.Then,the beam tracking problem of the mobile user terminal in millimeter wave communication is studied.Based on the modeling of the user-side mobile scene,the problems faced by beam tracking are explained,and the relevant provisions for beam measurement in the 5G standard are introduced.The design is designed according to the user’s mobile attributes and the actual air interface data collected Multi-layer LSTM network to track the optimal beam of the user end in real time.In the later test,the sensitivity of the design network to different multipath environments was discovered,which added a modularized switching network.Air interface data tests show that the proposed algorithm has a prediction accuracy rate of up to 40% higher than other traditional algorithms in mobile scenarios,and the beam tracking recovery time is less than 20%,which can effectively achieve beam tracking for mobile users.Then,it introduces the realization of the artificial intelligence-assisted receiver beam tracking algorithm on the millimeter wave transmission verification platform.Based on the overall hardware structure of the verification platform,it briefly describes the processing flow of the system,and focuses on several of the devices.The software architecture design of the system is given,including the USRP-RIO design and general purpose of the baseband part.Server-side software design,the baseband part design includes the introduction of each functional module,the neural network part design includes the function introduction of each processing thread,showing the demonstration effect of the artificial intelligence-assisted receiver beam tracking algorithm in the actual scene,including Real-time display of video transmission function and various indicators.Finally,the initial access problem of dual-end beams in millimeter wave communication is studied.On the basis of modeling the problem of double-ended beam access and reviewing several algorithms,the hierarchical codebook algorithm was studied in depth,including the analysis of beam phase and classification principles,and different multipaths and different search boxes were studied.On the impact of the hierarchical codebook algorithm,a neural network algorithm based on the hierarchical codebook is proposed according to the algorithm idea.After analyzing the characteristics of the codebook,the appropriate search box size and grading strategy are selected,and the early termination technique is used to further improve The accuracy of the network.The simulation results show that the time complexity of the neural network algorithm based on the hierarchical codebook is only 1/3 of the hierarchical codebook,and the prediction accuracy is only reduced by about 2%,which effectively reduces the initial beam access delay.