Modeling and identification of nonlinear dynamics for freeway traffic by using information from a mobile cellular network

The high coverage of the territory by cellular networks and the widespread diffusion of mobile terminals aboard vehicles allow one to collect information on the traffic behavior. The problem of selecting a dynamic model to describe the freeway traffic by using the information available from a wireless cellular network is addressed by assuming the distribution of mobile terminals aboard vehicles to be uniform along the carriageway. Two different nonlinear parametrized models of freeway traffic are investigated: the first is an extension to a well-established macroscopic model, while the second is based on a black-box approach and consists in using a neural network to approximate the traffic dynamics. The parameters of such models are identified off line by a least-squares technique. Traffic measurements obtained from a cellular network are employed to identify and validate the proposed models, as shown by means of simulations.