Modeling And Characteristic Analysis Of Millimeter Wave Channels In High-speed Rail Scenarios

With the rapid development of high-speed rail networks,railway communication needs to develop from only processing critical signal applications to various high data rate applications.The current railway communication system performance is difficult to meet the demand for high-reliability broadband transmission.In order to achieve this demand,it is necessary to obtain sufficient system transmission bandwidth,and to explore and study the currently not fully utilized millimeter wave frequency bands,which is essential for railway wireless communication.Wireless channel modeling is the basis for the design and verification of wireless communication systems.Because propagation of signal in the millimeter wave frequency band shows different characteristics from the sub-6GHz frequency band,the traditional channel model cannot be well adapted to the millimeter wave channel modeling.At the same time,the high-speed movement of the train causes a large Doppler shift,causing the wireless channel to exhibit non-stationary characteristics.Secondly,the diversification of typical high-speed railway scenes also brings challenges to the establishment of a common channel model.This dissertation focuses on the channel modeling and channel characteristic parameter analysis in the millimeter wave frequency band in the typical high-speed rail scene.The content and research results are mainly divided into the following points:(1)Aiming at the current difficulty in obtaining channel data in millimeter-wave band at high-speed railway scene,the channel model based on ray tracing method is studied.Constructing a three-dimensional model of a typical high-speed railway propagation scene,including open fields,cuttings,and tunnels.Using ray-tracing tool to establish a deterministic channel model for the 28 GHz millimeter wave frequency band to collect comprehensive channel data under high-speed typical scenarios.The quasi-stationary interval of the channel under different scenarios is analyzed,which lays the foundation for the subsequent analysis of channel characteristic parameters.(2)Using the channel data obtained by the ray tracing method to study the channel characteristic parameters reflecting the large-scale fading characteristics and small-scale fading characteristics of the channel under different propagation scenarios,including Path loss,Shadow fading,Rice factor,Root mean square delay expansion,Coherent bandwidth,Root mean square Doppler expansion,Angle expansion.Fitting through a specific probability distribution function and giving probability distribution parameters.Provide reference for the design and evaluation of high-speed rail mobile communication system.Using the extracted channel feature information,the correlation between channel feature parameters is analyzed.We found that in different simulation scenarios,the correlation between channel characteristic parameters is consistent.It shows that the channel characteristic parameters influence each other.(3)It is proposed to establish a path loss prediction model by combining the neural network and the two-path model.The environment features are defined and extracted,and the propagation environment is described only by considering limited environment types instead of complex three-dimensional environment modeling.The low-dimensional environment features are generated through dimensionality reduction by the autoencoder,and the low-dimensional environment features are used as the input of the neural network,which reduces the complexity of the network and improves the training efficiency.By comparing the prediction models trained under the condition of no environmental features,it shows the effectiveness of the extracted environmental features to improve the model’s predictive ability.Comparing the obtained mixed prediction model with the CI model and the A-B model,it is found that the minimum root mean square error index of the mixed model is reduced by about 6.7,and the standard deviation index is reduced by about 1.5,which verifies the accuracy and stability of the mixed prediction model.There are 44 pictures,28 tables,and 56 references.