"Equilibrium and learning in traffic network games"
ABSTRACT: Congestion is a common characteristic in modern telecommunication networks as well as in transportation systems of large urban areas. In this talk we review some recent developments in modeling traffic equilibrium in congested networks. Starting from the classical models of Wardrop Equilibrium and Stochastic User Equilibrium, we will present in more detail the notion of Markovian Traffic Equilibrium as well as numerical methods for computing it. We will show how all these equilibrium models admit a unified reformulation in terms of a strictly convex mathematical program. From these static equilibrium notions we move on to present a recent attempt to model the dynamic behavior of travelers by using a stochastic adaptive-learning process, and describe its asymptotic convergence towards equilibrium.