Is it possible to automatically build a large scale Metropolitan traffic model? Evidence from a study of connected transportation applications in Montreal

Intelligent Transport Systems (ITS) can improve transport systems while reducing their negative impacts. The deployments of ITS applications are highly dependent on telecommunications technologies, which must be tested either in field experiments or in simulation. The traffic simulation is the least expensive and most flexible solution. Yet, models require large amounts of data and expertise to develop, and the impacts of their parameters on their outputs for ITS applications have not been much investigated. Using the Simulation of Urban Mobility (SUMO) tool as an example, this article attempts to build automatically a simple large-scale traffic model for the Island of Montreal and evaluates how this model performs. This paper considers microscopic and a mesoscopic models, along with three traffic assignment methods. The outputs are compared to travel times from global navigation satellite system (GNSS) data and from Bluetooth sensors, and a sensitivity analysis is also performed for several output indicators at different scales. The results highlight the trade-offs between more detailed and accurate microscopic models, and faster, less accurate, mesoscopic models.