Relay Selection Based on Deep Learning for Broadcasting in VANET

In vehicular ad hoc networks, multi-hop broadcast plays a relevant role in data dissemination between vehicles. Several applications rely on the broadcasting protocol performance to provide best QoS for VANET services users. Generally, the most efficient way for broadcasting a data packet is selecting some vehicular nodes as a relaying set. The initiator node just needs to multicast the message to these nodes and then it will be forwarded in the same way to the rest of whole network. Here, the election of the best relaying set is a key step in the broadcast procedure. In this paper we propose a supervised learning-based relay selection algorithm. Starting from linear classification of candidate nodes factors we end by sorting these candidates according to their relay quality and best ones are selected. We verify the applicability of our theoretical method with an implementation of the learning-based classification and prove its efficiency compared to other methods.