Abstract

Urban metro rail systems are subject to high and growing demand as the populations of major cities increase. A point may be reached where improving system management using advanced control is more attractive than expanding the network. Control schemes for strengthening system performance and therefore user satisfaction typically involve measuring certain system state variables such as the numbers of passengers aboard trains and waiting at stations. Given the high cost of installing the necessary sensors, an alternative methodology is proposed for online estimation of the two variables using a particle filter. Experiments performed on a dynamic simulator show that the variable values can be inferred by measuring only train dwell-times and passengers entering stations, data on which are generally accessible without major investment. The level of accuracy of the estimates generated by the methodology is high enough to enable a model-based controller implemented in a real metro system to achieve significant performance improvements.