Typical speed profiles on motorways from floating vehicle data

Advanced traffic management systems (ATMS) and advanced traveller information systems (ATIS) are key applications in the field of intelligent transport system (ITS). Such applications rely on road traffic monitoring systems which represent a basic function in the design and management of any control system involving drivers and vehicles. In recent years, monitoring technologies based on probe vehicles have been emerging both as self-working systems and in cooperation with infrastructure-based systems. Floating car data (FCD) systems are based on a number of probe vehicles that are equipped with satellite positioning devices and a mobile wireless connection – usually GPRS - to send periodically their position-speed data to a central processing unit. FCD systems do not require the installation of a fixed monitoring infrastructure along the road and base their appeal on the spontaneous choice of users, who may decide to use the on-board device for various travel services, such as navigation or fleet management functions. The reliability of data collection is evidently limited by the penetration rate, but this rate has been considerably increasing in recent years. Nowadays, the wide use of positioning and communication devices are making FCD systems a promising tool for monitoring traffic flows. The proposed paper is focused on the analysis of daily speed patterns for the construction of an efficient, reliable and accurate historical database. The method aims at classifying the various road segments on the base of their typical speed profiles. In order to obtain a reliable segment classification, a preliminary analysis of homogeneous subsets of data has been carried out by recognizing the vehicle types from their observed maximum velocity and dividing the dataset in homogeneous classes associated with different vehicle types. A hierarchical technique based on the Ward’s method has been applied for the clustering of road segments providing a clear identification in the dataset of homogeneous classes of speed trends. Experiments have been carried out on a floating truck