Filtering approach to online estimate the position of high-speed train

It's of great significance to achieve high-accuracy positioning for the safe operation and running efficiency of high-speed trains. Positioning error will accrete as train is moving away from balise. Not until the train passes through next balise group will the accumulated error be corrected. In this paper, integrated filtering approach is presented to online estimate the position of high-speed train. We firstly proceed the statistical analysis of relative range error defined by the ratio of correcting value to the link distance of balise. It indicates that the relative error conforms to the normal distribution. Then an approximate model is proposed for range acquisition, which can be offline and online identified by least squares support vector regression. Basing on the noise statistics and model of range acquisition, a linear model for updating relative position of high-speed train is proposed. And we use linear Kalman filter to estimate the position states, according to linear model. Thus obtaining the absolute position of train referring to the location of balise coordinate. Simulations are conducted with field data from type tests of trains. Results show that the procedure presented in this paper can bring at least 35.34% higher precision of positioning as regards Mean Absolute Percentage Error.