It is accepted that parallel computing is a promising approach to developing faster-than-real time traffic models for supporting advanced traffic management systems. In this paper a new technique for modelling the flow of motor vehicle traffic in congested urban networks using parallel processing is presented. MOSTFLOW (MOdelling Saturated Traffic Flow), a microscopic and parallel process-oriented traffic network model is being developed to meet the above needs. The model is considered to be especially flexible, due to its modular construction, which permits the modelling of arbitrarily complex traffic networks, and its use of decision tables, which may be easily modified for calibration purposes, calibration purposes, to represent individual vehicle and driver behaviours. MOSTFLOW operates by splitting the catch lane up into slots of 5.75 metres. A simulated vehicle moves from one slot to the next, suitably delayed, corresponding to the speed of the vehicle. If a slot is full then another vehicle cannot enter that slot and similarly once a slot becomes empty a following vehicle can enter the slot. In this way queues are managed automatically. This level of control is provided by the underlying parallel processing system. Models are built of every junction and feature all that makes up the network. Thus MOSTFLOW is able to capture the behaviour of signal and priority controlled junctions, roundabouts, all types of pedestrian crossing and other aspects such as bus stops. This means that a complete model of an urban network can be built which can be used to evaluate the effect of traffic assignment plans developed for isolated junctions using other modelling packages.
Abstract