Link travel time prediction for decentralized route guidance architectures

A critical problem in decentralized route guidance is to communicate anticipated congestion to individual drivers in such a way that the routes chosen are likely to be consistent with the forecast. We propose a prediction technique for decentralized route guidance architectures to identify time-dependent link travel times which when communicated to drivers leads to time-dependent fastest paths consistent with this forecast. The fixed-point property of the forecast is assured by an iterative process of traffic simulations followed by dynamic route determinations until the routes and hence the resulting dynamic link times become stable. The resulting routes yield an inherently accurate forecast of congestion as well as being user-optimal by construction. A back-dating process is utilized to insure the discovery of a stable routing after a finite and usually small number of iterations, An empirical case study based on the roadway network in Troy, MI is included.