Understanding and modelling information dissemination patterns in vehicle-to-vehicle networks

Advances in wireless communication technology have enabled information exchange opportunities between moving vehicles within proximity. Potentially through such physical contacts a piece of information can diffuse to the entire network. While there has been extensive research on information diffusion in social networks, we do not know much about the spatial patterns in vehicle motion and how such patterns can support information dissemination. To this end, in this paper, we provide a systematic study of three large-scale data sets of taxi GPS traces from three big cities. The study shows the following properties universal of the three data sets: 1) the small world property, that information can be disseminated to almost the entire set of participants, within a very small number of hops; 2) certain physical contacts can be extremely effective in exchanging messages and such effectiveness shows a power law distribution; 3) the lack of hubs, no vehicle behaves as major hubs; removing top 20% nodes that have the highest number of physical contacts does not affect the effectiveness of information dissemination. 4) the information dissemination exhibits strong spatial temporal correlation. Finally, to explain the observations in particular the small world property, we develop mathematical models of the taxi movement patterns such that on graph topologies exhibiting properties of real-world road networks a number of observations can be rigorously proved.