Identification of New Patterns in Urban Traffic Flows

Abstract

Traffic flow pattern identification, as well as anomaly detection, is known to be an important component for traffic operations and control. Alongside classical applications, mainly, to improve the safety and the comfort of drivers, more recently there is a growing interest in gathering personalised route information to provide customised services. With this latter application in mind, in this paper we investigate the ability of simple macroscopic information (i.e., time varying junction turning probabilities) to identify changes in nominal urban traffic flows, most likely due to the occurrence of external events (e.g., road works or traffic congestions). Some preliminary results obtained with the use of a realistic mobility simulator are also illustrated and discussed, and some candidate applications are briefly outlined.