Motorway speed pattern identification from floating vehicle data for freight applications

• We examine a wide set of data collected from heavy vehicles on Italian motorways.
• We propose various rules for classifying speed profiles for days and road sections.
• We apply a clustering method for identifying the typical speed profiles for a road.
• A small number of profiles can model the main speed behavior for all road sections.
• Anomalies detected using the speed profiles are confirmed by actual incident data.

Nowadays, the diffusion of in-car navigators, location-enabled smartphones and various reasons for tracking vehicles – either for insurance and recovery, fleet management or for electronic tolling – are making floating car data (FCD) a leading solution for traffic monitoring. In the next years, this solution might be much more strengthened by the introduction and diffusion of black boxes, installed on commercial or private vehicles devoted to monitor or validate new safety technologies (e.g., the automatic in-vehicle emergency call service eCall in Europe).1 FCD, possibly integrated with data coming from infrastructure-based monitoring systems, represents a valuable platform for intelligent transport systems (ITS). Traffic monitoring based on FCD relies on a processing algorithm for aggregating the measured data into an accurate and complete traffic map. In this paper we present an experimental study on FCD processing based on a unique large amount of data in Italy, provided by heavy-duty vehicles used as probes over the Italian A4 motorway. A processing procedure is proposed for identifying the typical speed patterns, to be used as baseline for automatic anomaly detection, transport planning or traffic analysis applications. A first assessment based on real traffic-event information shows that the comparison of the probe data to previously identified historical speed patterns allows a clear detection of anomalous events.