Research On Key Technology Of Millimeter Wave Beam Management Based On Machine Learning In High-speed Railway

With the development of China’s high-speed railway,it not only needs to support higher operating speeds,but also needs to support higher data transmission rates.In March 2018,the Liaoning section of the Beijing-Shenyang high-speed rail launched the "Comprehensive Experiment on Key Intelligent Technologies of High-speed Railways" to promote the construction of China’s smart high-speed rail,which is inseparable from the support of the high-speed rail wireless communication system.High-speed rail communications have changed from only supporting simple train control services to the era of high data rates that need to support data services on mobile phones,tablets,and laptops.This poses huge challenges to the capacity,bandwidth,delay,and security of high-speed rail wireless communication systems.In fact,the development of public mobile communication networks has entered the 5G era.As a dedicated wireless communication system,the high-speed rail mobile communication system also needs to evolve to the next generation of more efficient and intelligent high-speed rail mobile communication systems.The millimeter wave frequency band of 5G has abundant spectrum resources,and the signal wavelength of the millimeter wave frequency band is small,which can accommodate more antenna arrays in a limited physical space,and beamforming technology can be used to obtain a higher system data transmission rate.On the other hand,the development of artificial intelligence technology is also obvious to all.By combining the learning ability of machine learning and its ability to perceive the environment,the intelligent high-speed rail wireless communication system can obtain smaller transmission delay and more computing resources.Based on the special scenarios of high-speed rail,this paper uses millimeter wave massive MIMO technology and machine learning theory to carry out in-depth research on beamforming technology.The specific work is as follows:First,this article introduces the propagation characteristics of millimeter wave frequency band signals and analyzes the key technologies used in high-speed rail millimeter wave communications,including traditional MIMO technology,massive MIMO technology,and relevant knowledge and theories of beamforming technology.Aiming at the special scene of high-speed rail communication,a multi-beamforming scheme based on optimal uniform beam width is proposed.In this solution,a mobile relay is installed on the top of each car of the train.The millimeter wave base station uses the train’s position information and channel state information(CSI)to form a multi-beam to communicate with each mobile relay.To achieve the effect of adaptively adjusting the number of beams and beam width,it can effectively increase the directional gain and maximize the system transmission capacity.And this scheme considers the dual-base station mode,and further improves the throughput of the train at the edge of the cell through joint transmission with the multi-beams emitted by the adjacent millimeter wave base stations.Then,aiming at the low latency and reliability problems of high-speed rail millimeter-wave wireless communication systems,this paper combines machine learning related theories and proposes an adaptive multi-beam selection scheme based on the assistance of Generalized Regression Neural Network(GRNN).This scheme also takes the position information and channel state information of the train as the basis,and trains the GRNN model at every moment in the train operation process,and finally obtains the optimal adaptive beam selection result.After simulation analysis,the prediction result of the proposed scheme not only does not reduce the data accuracy,but also greatly reduces the computational complexity and communication transmission delay of the traditional adaptive multi-beam selection algorithm.Improve the stability and reliability of high-speed rail trains during high-speed operation.