Modelling distributed real-time route guidance strategies in a traffic network that exhibits the Braess paradox

This paper examines the operational characteristics of a simple traffic network which produces a dynamic form of the Braess paradox when the in-vehicle link data base, that is used to guide vehicles in real-time through the network, is expanded to include a low capacity short cut link which was previously unknown to the drivers. The analysis focuses on the ways in which a route guidance system with distributed route selection logic, such as the TravTek system in Orlando, can be made to avoid the inefficiencies associated with the paradox without a priori knowledge of the existence of the paradox and while allowing each vehicle's RGS computer to independently make routing decisions without the explicit knowledge of the concurrent routing decisions of other drivers. It is shown that this objective can be achieved by implementing a distributed form of a system optimum routing algorithm coupled with the use of a PROBIT type of error term to reduce some of the instabilities that are intrinsic in the travel time information feedback loop.