Research On Optimization Technology Of High-speed Railway Wireless Communication Handover Based On Machine Learning

In recent years,with the strategic implementation of the "the Belt and Road" and "Made in China 2025",China's high-speed rail technology is rapidly developing towards a safe,fast,green,and intelligent trend,and has become an important driving force for China's social and economic development.With the development of high-speed rail technology and various communication services,the existing GSM-R system has been unable to support the demand for high-speed rail communication services.LTE-R,The next-generation high-speed rail communication system will provide an efficient,high-stability and ultra-capacity wireless communication service.With the rapid improvement of computer performance,artificial intelligence technology has been widely used in various fields.Intelligentization is an inevitable trend in the development of high-speed rail technology and communication technology.High-speed rail wireless communication systems can achieve higher operational efficiency and more reasonable resource allocation with the excellent perception and learning capabilities of machine learning technology.In order to optimize the service quality of wireless communication between vehicles and the ground during the handover process of the high-speed rail wireless communication system,and to improve the effectiveness and reliability of the handover technology,the following innovative work has been done.(1)In this paper,after analyzing the mathematical characteristics of the network traffic sequence of the train control and service network,a LSTM-based train control and service network traffic prediction model is proposed to predict the wireless communication data transmission rate requirements of high-speed trains so that the communication system can allocate resources in advance.This paper proposes a beamforming power allocation algorithm based on rate requirements and geographic location information in a single-cell multi-train wireless communication scenario.It reasonably allocates more power to the directional beams pointing to the edge of the cell to improve beamforming gain.The proposed algorithm optimizes the wireless communication service quality between the train and the base station in the handover area.(2)In this paper,a K-Trend-LSTM algorithm-based high-speed rail wireless communication handover decision parameter prediction model is proposed.The handover decision parameter time series samples are aggregated into different clusters according to the similarity of their changes.The sample sequence set in various clusters is used as a data set to establish an independent LSTM neural network model for parameter prediction.This algorithm improves the prediction accuracy and reduces the workload of neural network training.Based on the parameter prediction model,this paper proposes a high-speed rail wireless communication dual-link switching scheme combined with parameter prediction under the C / U decoupling architecture.This scheme reduces the communication interruption probability of handover between macro cells by activating the dualbroadcast mechanism in the handover preparation process,and improves the system scheduling efficiency through the parameter prediction process.The proposed scheme effectively improves the effectiveness and reliability of the handover technology.This paper is aimed at the next generation high-speed rail wireless communication system,and focuses on combining machine learning technology and handover related technologies.It proposes some intelligent high-speed rail wireless communication technology Exploratory scenarios by taking advantage of the recurrent neural network excellent processing ability for time series.