Abstract In order to enhance the realism and diversity of traffic flow modelling, this chapter presents a data-driven traffic behavior model based on neural networks in a real-virtual interaction traffic simulation system. First, we extract individual personalized real trajectories from each vehicle, then use neural networks to develop specific traffic models from the trajectories of each vehicle. In contrast to traditional, manually-defined traffic models, we aim to develop a data-driven model to describe the relationship between the traffic states faced by a driver and the driver’s resultant actions. In this model, a driver’s behavior is influenced by the current traffic states of the leading vehicle and the following vehicle. This is a regression problem for which the inputs of the model are the traffic states of the leading and following vehicles. The output is the action of the current vehicle. Finally, this chapter presents a real-virtual interaction system. In detail, real trajectories are introduced directly into the simulation process to maximize the characteristics of real traffic flow. In comparison to existing simulation methods, traffic flows simulated by this method can depict irregular vehicle driving behavior.
Vehicle Behaviors Simulation Technology Based on Neural Network
2018-10-11
18 pages
Aufsatz/Kapitel (Buch)
Elektronische Ressource
Englisch
Artificial neural network-based system for pre-judging behaviors of surrounding vehicles
Europäisches Patentamt | 2015
|The Intelligent Vehicle Control System Based on the Fuzzy Neural Network Technology
Tema Archiv | 2012
|