Route guidance as a just-in-time multiagent task

This article discusses a prototype of a route guidance system able to cope with both long- and short-term navigation problems. The system plans the route using a classical shortest path algorithm working on an ''on-board'' road map knowledge base, and uses information acquired through a radio channel to activate route replanning. In this case the new plan is computed not only on the basis of the car and world state at the start of the replanning, but also on replanning time duration. An estimate of car and world variations during the replanning is used to build up a route that the driver can safely perform. The system builds a local world description by integrating the information computed by a boundary detection algorithm with the information stored in the road map knowledge base. This model is used to generate a car trajectory and to supply the driver with information. Although the time spent to compute the trajectory does not allow the car to be driven at a reasonable speed, the boundary detection algorithm gives information that integrates road map information and a car model to support the driver with road descriptions and guidance suggestions.