Interference Analysis And Network Quality Evaluation For Rail Transit CBTC System Based On Machine Learning

In recent years,urban rail transit has become one of the main travel modes for people due to its features of rapidity,timeliness,and ultra-high passenger capacity.For high-speed rail trains,the effectiveness and reliability of the communication between the train and the ground should be focused.Communication based train control(CBTC)system is the most widely employed automatic control system for subway trains.However,the open space transmission of wireless communication makes CBTC wireless communication network vulnerable to interference,which affects the reliability of wireless transmission and even the safty of rail transit.In order to improve the reliability of wireless transmission,this thesis studies the wireless interference and the network quality assessment in CBTC system.The main work of the thesis is as follows:Firstly,this thesis studies the wireless interference in CBTC system based on LTE-M.A feature extraction scheme adopting a sliding window of signal spectrum is proposed and employed to study the recognition of interference signal based on machine learning.CBTC system interference signal data set is constructed by both software simulation and air interface acquisition.Experiments are performed on this data set adopting feature extraction scheme,and thus effectiveness of feature extraction method in identifying different interference types is verified.Secondly,this thesis also applies feature extraction method to the analysis of the physical characteristics of the CBTC interference signal to achieve a more accurate detection of the interference signal.A regression experiment is carried out on the aforementioned interference signal data set by using neural network model.Meanwhile,the levels of the interference signals and their central frequency are detected,and then the detection accuracy is evaluated and analyzed.Finally,this thesis proposes a machine learning-based network quality assessment strategy of wireless communication system in CBTC which adopts the WLAN standard.With the analysis of the wireless signal strength and handover related parameters during train operation,the sparse autoencoder neural network is employed to extract relevant features,and the corresponding supervised learning model is trained based on the XGBoost algorithm,so as to realize the network quality assessment in CBTC system.The thesis employs the actual measurement data from one subway line in China to conduct experiments,and thus the effectiveness of the algorithm is verified.