Prediction-based connection manager for vehicular networks

Vehicular Ad-hoc NETworks (VANETs) have unique characteristics, such as the high mobility of nodes, which cause frequent losses of connectivity, turning technology and network selection into a complex problem. Moreover, due to the instability of networks and its fast change, this network selection requires information of the networks in advance, so that it can predict the quality of the networks to connect. This paper proposes a predictive connection manager that is able to choose in advance the best network and technology a vehicle can connect to. This approach will use the information of previous connections to a station or road side unit, such as the throughput and amount of transferred traffic, and combines this information with sensing information of the quality of available networks. With both previous and current information, it will choose the network that has been able to provide the best performance in previous connections and remains with a good quality. The results of the proposed approach, tested in a real vehicular network deployed in Portugal, show that key performance metrics are improved, such as connection range, throughput and the number of handovers, with a general improvement on connection stability.