A multi-criteria routing model for incident management

Incident management requires rerouting motorists around incidents and dispatching service vehicle to the site of the incident, often in a complex transportation network. This paper proposes a multicriteria optimization model and solution algorithm. Travel distance, expected value and variance of travel time, as well as risk constitute four important criteria in routing. Generalized dynamic programming is used to find nondominated alternate paths for motorists. At the same time, it allows describing the original network in terms of such paths. In so doing, we reduce the dimensionality of network representation. It is on this reduced network that a vehicle routing model with serving priorities is proposed. Such a model finds nondominant tours for service vehicles, dispatching them to the site of the incidents and route them back to the service depot. With a limited fleet of service vehicles, the inclusion of serving priority assures that multiple incidents happening at the same time can be served in accordance with their levels of severity. A preliminary study in Central Arkansas freeways verifies the serviceability of such a model.