Multilevel modelling of complex systems of traffic interactions

Paper presented at the 9th international conference on computers in urban management and urban planning, 29 June - 1 July 2005 Vehicular traffic flow is a typical complex system. Such a traffic system usually involves interactions between drivers, vehicles and the road infrastructure. The behaviour of the drivers in terms of their decisions and reactions cannot be easily understood or predicted from the characteristics of individuals or of the system. To model these complex road traffic interactions, requires a number and/or combination of techniques, for example, microscopic, macroscopic, mesoscopic and agent-based techniques. Using these techniques offers the opportunity to improve behavioural realism, whilst producing credible models that can be robustly validated. The paper examines various traffic modelling approaches. It then describes a microscopic agent based technique that we have implemented in a synthetic driving simulation framework. The technique is employed to represent complexity and uncertainty by explicitly modelling the characteristics of the traffic entities such as the vehicle/driver behaviour on specific road elements of interest.