Research On High Speed Train Positioning System Based On Interactive Multi-Model Algorithm

Precise location information is the basis for safe and efficient operation of high-speed trains.The classical positioning methods represented by orbital circuit method,query/responder positioning,and electronic axle counter positioning can no longer meet the requirements for real-time and accurate positioning of high-speed trains.Although the satellite positioning method represented by GPS has high precision and good real-time performance,it completely relies on the characteristics of the external signal transmission to make it have a great shortage of safety and reliability.The above method and the LTE-R positioning system based on the TDOA positioning principle are briefly analyzed,focusing on the influence of the filtering algorithm on the positioning accuracy.As an important part of the positioning system,the filtering algorithm directly affects the positioning accuracy of the positioning system.In the existing related research,the research focus is usually focused on the use or improvement of the classic algorithms such as Newton iteration method,least square method,Kalman filter algorithm,etc.In addition,the high-speed train motion model used in simulation verification is mostly a single model.It does not take into account the rapidity and the diversity of the motion modes of high-speed trains in actual motion.Therefore,on the basis of classical filtering algorithms such as EKF and UKF,an interactive multi-model algorithm that is widely used in aerospace and other high maneuvering target tracking/positioning fields is used to achieve real-time and accurate positioning of high-speed trains("positioning" are synonymous with "tracking" in this thesis,both of them obtain the position information of target).The main research work is as follows:(1)Firstly,combining the basic theory of maneuvering target tracking algorithm and the characteristics of high-speed trains,three kinds of motion models,CV,CA,and CT are established.They are the basis for improving the filtering accuracy.Then introduced two commonly used filtering algorithms,EKF and UKF,and compared them in the set simulation environment.The simulation results show that when the high-speed train has three motion modes: CV,CA,and CT,the two positioning algorithms have large errors,which provides a theoretical basis for the improved algorithm.(2)Analyzed the filtering performance of IMM-EKF and IMM-UKF algorithms in high-speed train positioning system.In order to adapt to the diversity of high-speed trains' motion patterns,based on the analysis of the principle of the IMM algorithm,the IMM-EKF and IMM-UKF algorithms were designed by combining them with EKF and UKF algorithms respectively,and they were simulated and analyzed.The results show that compared with the single-model EKF and UKF algorithms,they reduce the filtering error and improve the positioning accuracy.The accuracy of the IMM-UKF algorithm is higher than that of the IMM-EKF algorithm.However,these algorithms have the problem of increasing the error before and after the model switching time.(3)Through the improvement of the algorithm,the error before and after the model switching time is reduced.First,the reasons for the increase of the error between the IMM-EKF and the IMM-UKF algorithm before and after the model switching time are analyzed in detail.Then based on this,an AMP-IMM algorithm using the posterior information to correct each element in the Markov probability transfer matrix is proposed.The algorithm overcomes the disadvantages of the constant value of elements in the Markov matrix.Finally,the comparative analysis of simulation experiments shows that while inheriting the advantages of the IMM-UKF algorithm,it reduces the filtering error before and after the model switching time,and then improves the positioning accuracy within the corresponding time period.The above research content has proved the superiority of the IMM algorithm as a filtering algorithm in the high-speed train positioning system through the combination of theory and simulation experiments,and provides a new method for realizing high-speed train real-time and accurate positioning.