Research On Nonlinear Filter-based Information Fusion Algorithm For Integrated Train Positioning

The precise train location information is a significant foundation of train operation control to guarantee the safety of train operation,shorten the time cost,and improve the traffic efficiency.In recent years,the integrated positioning algorithm based on Global Navigation Satellite System and Dead Reckoning(GNSS/DR)has experiencing a rapidly development.Accordingly,the introduction of GNSS/DR-based integrated positioning technology in train positioning systems is of great necessity,which can provide assistance to the existing train positioning systems.The core of a GNSS/DR-based integrated positioning system is the information fusion solution.However,the probable challenges under complicated operation environments along the railway tracks encourage us to find better nonlinear fusion algorithms to meet the requirement in railway applications,which can contribute to the precision and reliability of the obtained train location determination results.In this paper,the information fusion model of GNSS/DR integrated positioning system is established based on the operation environment and train’s kinematic features.A distributed information fusion algorithm is designed with respect to two levels,including the local GNSS navigation calculation level and the global GNSS/DR integration level.With this solution,performance of train positioning can be enhanced through effective data fusion.The main research of this paper can be summarized as follows.(1)Based on the kinematical characteristics of train,e.g.the slow acceleration and deceleration,large turning radius and stable climbing or descending,the "current statistical model" is involved in the state transition model.The observation model is established by considering the sensors observation characteristics.(2)Under the Bayesian estimation frame,several nonlinear Bayesian filters are analyzed and compared,including the Extended Kalman Filter(EKF),Particle Filter(PF)and Extended Kalman Particle Filter(EKPF).Taking the spatial constraint of train trajectory into consideration,an improved particle filtering method with track constraint is proposed.(3)Considering the unknown error statistics due to the probable multipath effects under complicated operation environment,raw measurement model of GNSS receivers is built.The Dirichlet process mixture model,which is a typical Bayesian nonparametric strategy,is involved in the nonlinear filtering for GNSS navigation calculation,which is capable of estimating the train position and the motion state accurately.In this paper,in order to validate the performance of proposed nonlinear estimation method,an experimental verification platform is set up.Based on the results of field experiments in Qinghai-Tibet Railway and simulations,the feasibility and performance of the proposed method are compared and analyzed with several nonlinear filters.The results illustrate that the proposed nonlinear filter-based information fusion algorithm of integrated train positioning can achieve the desired performance in complicated and uncertain train operation environments,which could meet specific requirements of train positioning in train control systems or other location-based railway applications.