Modeling And Evaluating Data Traffic Of Vehicular Networks

As an important supporter of intelligent transportation, vehicular network has attracted great attention of the world auto industry and research in the academic field.Drivers can get the information out of his sight through the communication of vehicle to vehicle, vehicle to roadside units, or vehicle to infrastructure, such as the status of other vehicles and the real time traffic information. Meanwhile, the passengers have the same mobility with the vehicles, they get access to network services via personal mobile devices.With the rapid development of the vehicular network and its applications,large scale of vehicular terminals and passengers' devices get access to the network and produce a large amount of data traffic. The huge amount of data traffic not only bring great pressure on the network infrastructure, but also has negative effect on the quality of the service provided by the network. Therefore, analyzing and modeling the data traffic of vehicular network is of great significance.At present, there is very few research on vehicular network traffic, and the traffic model has many limitations. In the definition of vehicular network traffic, only vehicles' or passengers' traffic is considered. In analyzing of vehicular network traffic, the mobility of the vehicles has been ignored. In the modeling of vehicular network traffic, there are very few models specifically used to describe the data traffic. Therefore, there are many challenges in the modeling vehicular network traffic.This paper is based on the collected data traffic of vehicular, which contains communication scenarios of V2V and V2R, including traffic generated by both vehicle applications and user applications. Then we proposed a new traffic modeling method, i.e. TL-HHMM (Two Level Hierarchical Hidden Markov Model) method to analyze the vehicular network traffic. It uses the method of time series segmentation and K-Means clustering to process the raw data, and uses the Baum-Welch method to learn the traffic, to establish a vehicular network traffic model.The modeling method takes into consideration the periodicity and stage from the large-scale point of view, and the fluctuation and hierarchy from the small-scale point of view. The experimental results show that the proposed model can effectively capture the characteristics of large-scale and small-scale vehicular network traffic, and can also accurately describe the self-similarity in traffic.Our results can be used as foundation to assess the impact of vehicular network traffic on the network infrastructure and the H2H flow. The results can also be used for improving the performance of network infrastructure,vehicular network QoS, and the bandwidth utilization.