Neural network modeling of vehicle discharge headway at signalized intersection: model descriptions and results

Vehicle discharge headway at signalized intersections is of great importanc e in junction analysis. However, it is very difficult to simulate the disch arge headway of individual queued vehicle because of the great variations i n the driver behaviors, vehicle characteristics and traffic environment. Th e current study proposes a neural network (NN) approach to simulate the que ued vehicle discharge headway. A computer-based three-layered (NN) model wa s developed for the estimation of discharge headway. The widely used backpr opagation algorithm has been utilized in training the NN model. The NN mode l was trained, validated with field data and then compared with other headw ay models. It was found that the NN model performed better. Model sensitivi ty analysis was conducted to further validate the applicability of the mode l. Results showed that the NN model could produce reasonable discharge head way estimates for individual vehicles. (C) 2001 Elsevier Science Ltd. All r ights reserved.