Train position and speed estimation using wheel velocity measurements

In order to improve safety and efficiency in the management of modern railways, several systems for monitoring and control of traffic are being developed. Automatic train protection (ATP) systems command an emergency braking procedure if dangerous situations occur, such as insufficient braking distance to one of the next target positions and target velocities. A novel ATP system named SCMT, to be installed on trains running on Italian railways, is currently being designed. One of the components of SCMT is a module for estimating train speed and positions between fixed balises, which communicate to the on-board system the distance to next targets and the velocity requirements at targets. In this paper algorithms are described for distance to target and velocity estimation, capable of compensating for poor wheel-rail adhesion conditions where conventional odometry algorithms may fail. The algorithms were derived using a variety of methods including neural networks, fuzzy logic and crisp logic. The system was designed and trained using a wide set of experimental data, obtained from test runs carried out with different types of vehicles and conditions (in particular, degraded adhesion conditions were investigated).