Towards Anticipatory Traffic Forecast – Modelling of Route Choice Behaviour

This paper discusses a framework which considers the behaviour of drivers as a mean to achieve an anticipatory traffic forecast. We introduce a two-layer agent-based architecture for the simulation of traffic flow and consider a simple scenario in which agents (drivers) receive dynamic information about traffic conditions. It is shown that in such scenarios, with information being a key commodity, a layered architecture has advantages over the traditional single-layer simulators, which deal only with the reactive tasks associated with driving. Introducing dynamic information into the system can destabilise traffic patterns in an undesirable way.