Artificial Intelligence-Based Wireless Channel Characteristics Prediction And Scenario Classification For B5G High-Speed Train Communications

With the growing number of high-speed train(HST)users and the diversification of user needs,the existing HST communication system is far from being able to meet the growing demand.The new generation of HST communication system is urgently needed to provide massive access,high energy efficiency,low latency,and high reliability communication services.A channel model that accurately and effectively describes the unique channel characteristics of HST is critical to the design and network evaluation of future intelligent HST communication systems.The next generation of HST communication system will introduce some B5G key technologies.New technologies introduction makes the future HST wireless channels expand in multiple dimensions,such as time domain,frequency domain,air domain,and scenarios.This led to the increasing complexity of traditional HST channel modeling methods,and limited the rapid evaluation and application of HST wireless communication technologies.All of these pose new challenges to the traditional channel modeling methods.In addition,the rapid movement of HST through multiple scenarios makes it difficult to measure,and the expensive test equipment,which makes HST field measurement data is limited.This paper addressed the limitations of traditional HST channel measurement and modeling studies,and considered the combination of big data analysis and HST channel modeling.Artificial intelligence(AI)was used for channel characteristic prediction and partial forecasting of future channel state information.At the same time,the mapping relationship between HST wireless channel data and radio wave propagation scenario was studied through accurate identification of HST scenarios.Driven by the background of big data of HST,AI-based predictive modeling research of HST wireless channels was carried out.The main research works in this paper are as follows:1.The typical scenarios of HST were reconstructed using the ray tracing(RT)software Wireless Insite verified by real measurement data.It includes four HST scenarios:viaduct,open space,cutting,and hilly.A wireless channel database of typical HST scenarios in Sub-6 GHz and millimeter wave(mmWave)bands was established.2.Based on the acquired HST wireless channel data,a research on HST wireless channel characteristics of different scenarios in Sub-6 GHz and mmWave frequency bands was carried out.Analyzed and compared the typical HST wireless channel characteristics in different frequency bands and different operation scenarios.At the same time,a database of HST wireless channel characteristics of typical frequency bands and typical scenarios was established.3.Based on the wireless channel database of HST established in 1,a research on the prediction of typical channel characteristics of HST based on neural network was carried out.Currently,most common prediction characteristics are large scale fading,while there are fewer studies on channel characteristic prediction for small scale fading.In this paper,a channel characteristic prediction model based on Back Propagation Neural Network(BP-NN)and Radial Basis Function Neural Network(RBF-NN)was proposed.A large number of HST wireless channel datasets were used to train the network,and the prediction of received power,Rice K-factor,root mean square delay spread and angular spread was achieved.Finally,the model was evaluated by the prediction performance indexes.4.Based on the wireless channel characteristics database of HST established in 2,a research on the identification and classification of typical propagation scenarios of HST based on neural network was carried out.Considering that scenario categories are closely related to wireless channel characteristics,15 different combinations of channel characteristic datasets were used as network inputs.The HST scenario labels were numbered and the network was trained to achieve HST scenario classification.It investigated the contribution of different channel characteristics to the propagation scenario recognition and classification.Finally,the classification performance indexes of BP-NN and RBF-NN corresponding to the 15 combined datasets,including accuracy,F-scores,and so on,were calculated to evaluate the classification effects of different algorithms.