Neural network based vehicle-following model for mixed traffic conditions

Car-following behaviour is well studied and analyzed in the last fifty years for homogeneous traffic. However in the mixed traffic, following behaviour is found to vary based on type of lead and following vehicles. In this study, a neural network based model is proposed to predict the following behaviour for different lead and following vehicle-type combinations. Performance of the model is studied using data collected for six vehicle-type combinations. A multi-layer feed-forward back propagation network is used to predict vehicle-type dependent following behaviour by incorporating the vehicle- type as input into the model. The neural network model is then integrated into a simulation program to study the macroscopic behaviour of the model. Performance of the proposed neural network model is compared with the conventional Gipps? model at microscopic and macroscopic level. This study prompts the need for considering vehicle-type dependent following behaviour and ability of neural networks to model this behaviour in mixed traffic conditions.