Improving Neighbor Localization in Vehicular Ad Hoc Networks to Avoid Overhead from Periodic Messages

Vehicular Ad Hoc Networks (VANETs) have been widely studied, and the deployment of such networks is likely to happen soon as the required technology and commercial opportunities are already available. Most of applications proposed for these networks require a localization mechanism with reasonable accuracy. In addition to applications, most protocols rely on the availability of a system that determines vehicles´ positions. As many vehicles are (or are in the process of being) equipped with GPS, the accurate localization of vehicles in VANETs is achievable. However, one issue that has been ignored is the localization of neighboring vehicles. In VANETs, nodes usually move at high speeds; the information about the position of a neighbour vehicle is therefore fast outdated. The naive approach for handling this issue is to increase the frequency by which periodic messages containing a node´s position are exchanged. This solution might lead an unfeasible overhead significantly dropping the bandwidth available for the exchange of services´ data. In this paper, we study this problem further, and propose a solution where vehicles predict the position of a neighbor for the near future. Through extensive experiments, we show that a prediction model of low complexity was able to considerably increase the accuracy of neighbor localization. Our mechanism has achieved an accuracy of 50 centimeters with a frequency of exchange of beacons 75% smaller than the naive approach.