Real-time estimation of incident delay in dynamic and stochastic networks

The ability to predict the link travel times is a necessary requirement for most intelligent transportation systems (ITS) applications such as route guidance systems. In an urban traffic environment, these travel times are dynamic and stochastic and should be modeled as such, especially during incident conditions. In contrast to traditional deterministic incident delay models, the model presented explicitly considers the stochastic attributes of incident duration. This new model can be used for predicting the delay that a vehicle would experience as it travels through nonrecurring congestion brought about by an incident. The model is operational in the sense that it does not require significant data and computational abilities beyond that which is traditionally used and can be used within traffic models or within actual ITS implementations. A mixed discrete and continuous vehicle-delay model is first derived and estimators of the mean and variance of vehicle delay are identified. A sensitivity analysis subsequently is performed, and a method for updating the estimated delay as new information becomes available is provided.