The worldwide increase of vehicles is demanding the deployment of an intelligent transportation system for the urban environment. Recently, cloud computing technology is utilized to make the vehicles on the roads smarter and offer better driving experience. However, the intrinsic client-server communication model in the cloud-assisted service cannot meet the increasing demands for intensive computing in vehicles. To solve this challenging issue, we investigate another form of computing service, vehicular fog computing (VFC), which is a group of nearby smart vehicles connected via peer-to-peer communication model. Though VFC can provide computing service to any task initiator, its computational capability, i.e., the ability to provide computing service to the initiator, might be severely constrained by the realistic environments including limited communication ranges, high speeds and unpredictable mobility patterns of vehicles. In this paper, we characterize the computational capability (indicated by the product of processor speed and the time length from receiving the task) of VFC in a practical scenario through studying real-world vehicular mobility traces of Beijing. Specially, we propose a time-varying graph model to access the capability of VFC in such a large-scale urban environment with different scenarios. Based on this model, we reveal the temporal and spatial characteristics of the computational capability with different number of task initiators and portray its distribution of the number of connected vehicles and the computational capability. The distribution of the computational capability is also portrayed. Based on these observations, we define two modes to depict two different models of task distribution. Furthermore, we reveal the relationship between the computational capability and system parameters of computation delay, communication radius, and the number of initiators.
展开▼
