A Stochastic Performance Model For Dense Vehicular Ad-Hoc Networks

Network level modeling of vehicular networks usually takes one of two paths. Either a mobility simulator is used to generate vehicular movement traces, combined with a network simulator to simulate packet transmissions. Or, simple stochastic assumptions, such as Poisson Point Processes and Manhattan Grids are imposed to allow analytical modeling. In this paper, we use the combination of mobility and network simulations to derive more accurate analytical models for vehicular ad-hoc networks in dense urban scenarios. Our results show that cars tend to group in clusters with approximately exponential geometric densities. Furthermore, we demonstrate that the process of interference in a dense network can be accurately modeled based on a linear function of the numbers of neighbors, as well as a Gamma distributed random process.