An analysis has been carried out in order to study the distribution of travel times between on-off ramps and the speed distribution along road sections. The first relevant result is that a travel time distribution really exists which is very similar to some statistical distributions. The second result is that the parameters of the distribution depend on some independent variables such as: type of vehicle, features of the way, day, hour, length of road section, flow. In order to define the most reliable distribution of travel time to consider in forecasting models, a lot of studies have been carried out about these distributions by using statistical methods and filtering techniques. Then it is possible to predict the off-ramps flow according to the features of traffic and to the in-ramps flow which can be measured or predicted itself. A model has been developed in order to predict traffic flow in a point of the network based on these concepts. This model has been tested in the case of the Italian freeway A4 (Venezia-Trieste). This approach allows to consider a lot of partially unknown random variables by using the speed distribution obtained from historical and actual data, and to create a much more realistic forecasting model. (A)
Abstract