Position Computation Model And Algorithms For High-speed Train

In recent years, the train control system puts forward a higher requirement on positioning accuracy with the running speed of high-speed railways improves. Furthermore, positioning accuracy of high-speed train is determined by odometer and balise. Positioning error is zero when the train is passing through a balise. However, positioning error between adjacent balises is caused by idling and slipping of train wheels and it is going up as the train is moving away from the previous balise. At present, methods for reducing positioning error between adjacent balises are mainly based on adding extra equipment. It is clear that this method greatly increases the cost and, the extra equipment introduced has the potential of adding complexity and unreliability for the train control system. In this paper, we focus on applying machine learning methods to increase the positioning accuracy between adjacent balises for high-speed train with balises and speed sensors only. This paper mainly research on several aspects of positioning problem as follows:Firstly, we formulate a mathematical model based on the analyzing of wireless message from the train control system to better illustrate the position computation problem, which is not affected by external environment factors. And this paper put forward that the travelling distance is only related to the speed of train at each position report point and travelling time.Secondly, it is difficult to get the accurate position for high-speed train running between adjacent balises. Therefore, we assume that the positioning error for each traveling distance is proportional to its value. Through this way, we can reasonably estimate each accurate traveling distance between two adjacent balises. LSM, SVM, and LSSVM models are employed to compute the position of high-speed train based on the mathematical model. Furthermore, we put forward parameter updating method for each position computation model.Finally, the proposed models are trained and tested by the field data collected in Wuhan-Guangzhou high-speed railway. The results show that:1) Compared with average speed method (ASM), the three models proposed are capable of reducing positioning error;2) Compared with ASM, the percentage error of LSM model is reduced by45.9%in training and53.9%in testing;3) Compared with LSM model, the percentage error of SVM model is further reduced by38.8%in training and14.3%in testing.4) Although LSSVM model performs almost the same with SVM model, LSSVM model has advantages over SVM model in terms of running time.5) With the online learning methods proposed, better positioning accuracy is obtained. Taking the positioning error and running time into consideration, we suggest that LSSVM model with parameter updating method is the best model for high-speed train position computation among the three models.