Assessing the consistency between observed and modelled route choices through GPS data

In traffic engineering, different assumptions on user behaviour are adopted in order to model the traffic flow propagation on the transport network. This paper deals with the classical hypothesis that drivers use the shortest possible path for their trip, pointing out the error related to using such approximation in practice, in particular in the context of dynamic origin-destination (OD) matrix estimation. If this problem is already well known in the literature, only few works are available, which provide quantitative and empirical analysis of the discrepancy between observed and modelled route sets and choices. This is mainly related to the complexity of collecting suitable data: to analyse route choice in a systematic way, it is necessary to have observations for a large period of time, since observing trajectories for the single user on a specific day could not be enough. Information is required for several days in order to analyse the repetitiveness and understand which elements influence this choice. In this work the use of the real shortest path for a congested network is evaluated, showing the differences between what we model and what users do. Results show that there is a systematic difference between the best possible choice and the actual choice, and that users clearly consider route travel time reliability in their choice process.