Traffic games with incomplete travel information

Abstract

In this paper, we consider a traffic game where a group of self-interested agents tries to optimize their utility by choosing the route with the least travel time, and propose an  -logit learning model that converges to an -Nash equilibrium (or a logit equilibrium) in the traffic game. The model consists of a N-person repeated game where the players know their strategic space and their realized payoffs, but are unaware of the information about the other players. The traffic game is essentially stochastic and described by stochastic approximation equations. An analysis of the convergence properties of the  -logit learning rule is presented. Finally, with using a single origin-destination network connected by some overlapping paths, the validity of the proposed algorithms is tested.