Identification of freeway-traffic dynamic models: a real case study

This paper deals with the problem of black-box identification of dynamic motorway traffic models devoted to the prediction of origin/destination traffic volumes. A real case study is addressed regarding an entire motorway stretch located in the north-east part of Italy. Different kinds of prediction models are considered including standard time-variant input-output and state-space models and time-variant models with periodic parameters. Thanks to the availability of a rather large database of real traffic data, an extensive experimental identification comparison between all the above dynamic prediction models is reported showing the possible relative benefits of each model type towards the design of an accurate and reliable predictor of traffic volumes.