The most useful type of information to a driver faced with travel choices is reliable predictive information. Provision of such information is, however, extremely difficult because it must be based on projected traffic conditions, which are themselves dependent on the ways in which drivers respond to the information. In this paper, for the case of non-recurrent traffic congestion, the theoretical possibility of providing drivers with travel time forecast is discussed in the context of drivers' route choice. A concept of recursive forecasting is proposed which can help to construct predictive messages consistent with drivers' future experience. An illustrative example shows that providing the recursive forecast makes traffic smoother than it would have been without information.
Abstract