Spatio-temporal Information Ranking in VANET Applications

Vehicular ad-hoc networks (VANETs) is a promising approach to the dissemination of spatiotemporal information such as the current traffic condition of a road segment or the availability of a parking space. Due to the constraint of the communication bandwidth, only a limited number of information items may be transmitted upon a vehicle-to-vehicle communication opportunity. Ranking becomes critical in this situation, by enabling the most important information to be transmitted under the bandwidth constraint. In this paper we propose a method for online learning of spatio-temporal information ranking in VANETs. In this method, mobile nodes such as vehicles judge the relevance of incoming information items and use them as training examples for Naive Bayesian learning. Additionally, a separate machine learning algorithm is used to estimate the probability of a duplicate item being transmitted. The method is used in place of commonly used heuristics, and is evaluated for travel time and parking availability dissemination applications.