Key Technologies Of Integrated Train Positioning Based On BeiDou And LTE-R

China’s high-speed rail lines are distributed in tunnels,vast plains,and complex environments with trees and mountains.Bei Dou Navigation Satellite System(BDS)can realize train positioning in open plains.However,its positioning accuracy is affected in complex environments with many trees and mountains by which Bei Dou satellites are partially blocked.To solve this problem,we can increase the number of satellites available and upgrade their geometric distribution by the data from the Global Positioning System(GPS)to improve the availability and positioning accuracy of the BDS;on the other hand,the BDS and the Long Term Evolution-Railway(LTE-R)can be integrated to realize train positioning in a complex environment.Specifically,the BDS/LTE-R integrated positioning data enhancement algorithm can realize continuous and high-precision train positioning.When the BDS is unavailable in a tunnel,the LTE-R wireless train location fingerprints can be used to achieve train positioning,with three methods used to optimize the accuracy and real-time performance of train location fingerprint positioning.Following the requirements of independent train positioning proposed with the new-generation train operation control system and considering the existing problems of train positioning,this dissertation deeply and systematically studies the BDS high-precision single-point train positioning,the BDS/GPS dual-mode train positioning,the BDS/LTE-R integrated train positioning,and fingerprint positioning.This dissertation starts from the theories of BDS positioning and LTE-R wireless communication positioning,aiming to fully use the equipment and the system of China’s high-speed railway.It focuses on continuous independent positioning of all rail lines to meet the requirements of real-time and high-precision positioning for high-speed trains.The innovations of this dissertation are as follows.(1)Many differential stations need to be built to realize high-accuracy train positioning using the real-time difference decomposition method.It is also challenging to fix the real-time ambiguity of the whole cycle with the medium and long baselines.To solve the problems,this dissertation adopts the carrier phase smoothing pseudo-range method based on the BDS to calculate train positions.Firstly,the optimal smoothing weight factor is introduced to reduce the smoothing noise and improve the smoothing accuracy.Secondly,the tri-frequency broadcasting of the BDS makes it possible that the linear combination of multi-frequency observation data effectively improves the period slip detection resolution and ensures positioning accuracy.Experiments show that the method not only provides real-time positioning by avoiding calculating integer period ambiguity but also realizes train positioning in a submeter range because it can effectively reduce the pseudo-range noise and eliminates the influence of the multipath effect on positioning accuracy.(2)The positioning accuracy of BDS/GPS dual-mode train positioning needs to be improved,and this positioning method is sometimes unavailable.This dissertation adopts the chaotic immune particle swarm optimization(CIPSO)algorithm to optimize and extend the Kalman filter for improving the BDS/GPS dual-mode train positioning.The simulation results show that this method can effectively reduce the influence of the observed value oscillation on the positioning.The eastward and northward positioning errors can be maintained at about4 m,ensuring higher positioning accuracy.Then this dissertation proposes to use fuzzy C-means(FCM)algorithm optimized by CIPSO to complete RAIM monitoring of multiple failed satellites.The CIPSO algorithm can optimize the initial cluster center of FCM to improve the accuracy of clustering results.The experimental results show that when the gross error is 25 m,the recognition rate of the proposed method for three faulty satellites can reach99.8%,ensuring the positioning data’s availability and reliability during train operation.(3)The BDS/LTE-R integrated train positioning method has low data fusion accuracy,and the number of visible satellites needs to be increased,making it impossible to complete RAIM monitoring and ensure the availability of positioning results.This dissertation introduces deep learning into the data fusion enhancement solution to the BDS/LTE-R integrated train positioning for the first time.The 7L-CNN positioning data fusion enhanced model is used with the second-order autocorrelation matrix of positioning data of each positioning system as input.The experimental results show that the average eastward and northward position errors can be reduced to 5 m,improving the positioning accuracy.Furthermore,This dissertation constructs virtual satellite observation information based on the LTE-R positioning results and extends the observation equation of integrated positioning to complete RAIM monitoring of integrated train positioning.This method can realize the RAIM monitoring of integrated training positioning under only two visible satellites,ensuring the availability and reliability of the positioning results.(4)The train location fingerprint positioning has low positioning accuracy and poor real-time performance.It brings a high workload of fingerprint collection.This dissertation uses the chaos particle swarm optimization＿K weighted the nearest neighbor(CPSO＿WKNN)algorithm,fuzzy least squares support vector machine(FL＿SVM),and deep convolution neural networks to improve the method of fingerprint positioning.Experiments show that the CPSO＿WKNN fingerprint positioning algorithm can achieve an 87.8% probability that the location error is less than 10 m when the fingerprint spacing is 25 m.The FLS＿SVM fingerprint positioning algorithm can shorten the calculation time for positioning to about 1min when the fingerprint spacing is 50 m,improving the real-time positioning performance.The AT-CNN location fingerprint positioning method can extend the fingerprint distance to100 m,reducing fingerprint collection workload and ensuring positioning accuracy within10 m.At the same time,the online calculation time is shortened to 10 seconds,significantly improving train positioning accuracy and realizing real-time.