Using Probabilistic Estimates to Guarantee Reliability in Crossroad VANETs.

We consider an intelligent crossroad where conventional traffic lights are substituted by a roadside unit (RSU), which synchronizes vehicles at the intersection, minimizing waiting time and energy consumption (by avoiding unnecessary braking and accelerating). Clearly, in this case, a reliable communication needs to be guaranteed between vehicles and the RSU, for which we investigate the design and analysis of specialized Vehicular Ad Hoc Networks (VANETs). It turns out that reliability strongly depends on the number of vehicles at the crossroad, i.e., the more vehicles, the more interference and, hence, the lesser reliability. As a result, to guarantee a desired level of reliability, we first need to estimate the worst-case number of vehicles at the crossroad. However, straightforward, deterministic approaches --- computing the maximum number of vehicles that physically fit into the crossroad's area --- lead to a great amount of pessimism and overdesign. In this paper, we propose using probabilistic estimations for the number of vehicles instead, which greatly reduces the amount of pessimism while still guaranteeing safety. Our approach is based on vehicles' statistical information and allows computing the probability of having a certain number of vehicles at the crossroad in the worst case. We incorporate this probabilistic estimate into the VANET's design and analysis to derive guarantees on reliability. Finally, we illustrate the benefits of the proposed approach by means of a detailed case study and simulations using OMNeT++.