Track-constrained GNSS/odometer-based train localization using a particle filter

The accurate and reliable localization of the trains is one decisive factor for a lot of specific location-based railway applications. Considering the cost-efficiency of construction and maintenance, the Global Navigation Satellite System (GNSS) is an effective approach for train localization systems which aim to replace the track-side Balises with on-board sensors. Thus, the accumulative error of the odometer is calibrated by the GNSS receivers and the autonomy of the on-board equipment is surely improved. In order to cope with the uncertainties in raw sensor measurements, the Bayesian filtering frame is adopted to obtain an accurate estimation of the train's state. Based on that, an enhanced particle filter solution is presented to realize iterative estimation. In this method, the cubature Kalman filter (CKF) is involved to generate the proposal distribution by using the track constraint, which indicates a modified kinematical model and an extended measurement model. The coupling of track constraint is designed to generate the importance proposal distribution for the update stage of the sequential importance sampling. Results from simulation with field data demonstrate the capability of the track-constrained particle filter for train localization using GNSS and odometer, which is with great potential for enabling the next generation GNSS-based railway systems.