Advanced OR and AI Methods in Transportation A REINFORCEMENT LEARNING MODEL FOR SIMULATING ROUTE CHOICE BEHAVIOURS IN TRANSPORT NETWORK

This paper proposes a new algorithm for finding disaggregate user equilibrium on a congested network, in which a driver is assumed to be an agent who learns from driving experiences to get maximal payoffs under the condition of incomplete travel information. Day-to-day route choice behaviours of each driver are formulated as a kind of repeated game with learning, and a simple adoptive procedure that lead to Nash equilibrium is proposed. The model presented here can cover a wide range of network equilibrium concepts from deterministic to stochastic user equilibriums.