Node Cluster Stability in Vehicular Ad hoc Networks

In recent years, efforts have been made to deploy communication capabilities in vehicles and the transport infrastructure, leading to a potential of vehicular ad hoc networks (VANETs). In the envisioned VANET, communications among vehicles will enhance the intelligent transportation systems (ITS) and support not only public-safety applications, but also a wide range of infotainment applications. Urban roads and highways are highly susceptible to a large number of vehicles and traffic jams. Therefore, the networking protocols for VANETs should be scalable to support such large sized networks. Node clustering (i.e., organizing the network into smaller groups of nodes) is a potential approach to improve the scalability of networking protocols for VANETs. However, high relative vehicle mobility and frequent network topology changes inflict new challenges on maintaining stable clusters. The communication links between network nodes play an essential role in determining the VANET topology. This thesis presents a stochastic microscopic vehicle mobility model to capture the time variations of the distance between two consecutive vehicles on a highway. The proposed mobility model is used to characterize the length and the duration of a communication link connecting two nodes in the network for different vehicular traffic flow conditions. Vehicle trajectory data from real and simulated highways are used for performance evaluation. In a highly dynamic VANET, vehicles join and leave clusters along their travel route, resulting in changes in cluster structure. This thesis investigates the impact of vehicle mobility on node cluster stability. A lumped stochastic model is proposed to describe the temporal variations of a system of intervehicle distances, where each intervehicle distance is represented by the proposed microscopic mobility model. Two metrics are used to measure cluster stability: the time period of invariant cluster-overlap state between two neighboring clusters as a measure of external cluster stability, and the time period of invariant cluster-membership as a measure of internal cluster stability. Using the proposed lumped stochastic model, the two cluster stability metrics are probabilistically characterized for different vehicular traffic flow conditions. Additionally, the limiting behavior of a system of two neighboring clusters is modeled, and the steady-state number of common/unclustered nodes between two clusters is approximately derived. To the best of our knowledge, this is the first mathematical characterization of node cluster stability which takes account of the effect of microscopic vehicle mobility. In addition to the impact of vehicle mobility on node cluster stability, the notion of cluster stability is also related to the network protocol requirements. This thesis explores