Masters Thesis: Information Dissemination in Vehicular Networks

A Vehicular Ad-Hoc Network (VANET) is a network where vehicles can disseminate information such as traffic conditions, availability of free parking lots or content. In order to fulfill VANET application requirements (Time To Live (TTL), information distance spread), we need to use the available network resources (e.g., radio capacity) optimally depending on the traffic conditions. While cellular-based approaches (e.g., 3rd Generation mobile telecommunications (3G)) is suited for low traffic conditions, dense traffic (where users might already use 3G) is particularly suited for relying on Vehicle-to-Vehicle (V2V) communication (e.g., ad-hoc WIFI). There is a clear need to adapt the dissemination strategy to the current traffic conditions and application requirements. The optimal solution is based on combining or alternating between cellular-based and opportunistic dissemination strategies. In this thesis, we aim at modeling epidemic dissemination in vehicular networks for prediction and optimization purposes. Contrarily to state-of-the-art works (e.g., Susceptible, Infectious (SI) models; VANET) we consider the dissemination as a function of time, but also distance. To do so, we first abstract vehicular mobility with a lattice grid and homogeneous transition probabilities from which we derive a dissemination model (with upper and lower bounds). We evaluate our model numerically and find a transition threshold also known as disease outbreak probability. In a next step, we depart from this abstract lattice model to Manhattan mobility and show by simulations that a similar behavior is observed hence validating the analytical model. We further show that our lattice model is also a good approximation of realistic vehicular mobility. Based, on these results we can illustrate how vehicles can adopt the best dissemination strategy (e.g., V2V, 3G or a combination) depending on the traffic conditions (sparse vs. dense) and application requirements (TTL, distance). Our work is a major contribution as it provides a model of spreading which accounts for distance with direct applications to vehicular networks for optiMaster of Science Thesis Tamás Zoltán Veres