The application of artificial neural networks to anticipate the average journey time of traffic in the vicinity of merges

摘要

A microscopic simulation model representing traffic behaviour in the vicinity of merges, especially under congested situations, was developed. The simulation model was applied to produce a set of data representing traffic patterns in the merge area, ramp metering rates, and the corresponding vehicle journey times. The data were used to develop an artificial neural network (ANN) model, which anticipates the average journey time of mainline vehicles that enter an upstream section during a 30s interval. The ANN model was validated with an independent data set. An investigation was made to ensure that the ANN model and the simulation model are capable of demonstrating the onset of flow breakdown at high combinations of the mainline and the entry ramp traffic flow. The ANN model can be applied to develop an ANN based feedback control system, which adjusts ramp metering rates to keep the average journey times of vehicles close to their desired or target value, and to reduce congestion in the vicinity of merges.