Estimation of train speed via neuro-fuzzy techniques

Abstract

The paper describes and compares some applications of neurofuzzy (NF) systems to estimate the speed of a train from the measurement of the velocity of two axles in any wheel/rail adhesion conditions. All the presented NF approaches outperforms the firstly designed crisp algorithm in terms of computational burden and some of them achieve also a significative performance improvement, by demonstrating their capability of learning from rough data. © Springer-Verlag Berlin Heidelberg 2003.