DTMR

Abstract

The emerging application of machine learning (ML) in different areas and the good results obtained have motivated its inclusion in the intelligent transport system (ITS) with smart cities and also in vehicular ad hoc networks (VANETs). In this sense, the main contribution of this work is the proposal of a decision tree-based multimetric routing protocol to make more intelligent forwarding decisions in the selection of the best next-hop neighbour node to transmit packets to the destination. To the best of our knowledge, most of the available datasets regarding vehicular networks are related to mobility patterns. Thus, we have collected our targeted dataset from several simulations runs over different urban vanet scenarios. Besides, we have included the evaluation of the importance of each routing metric by applying regularization. The goal here is to include the more relevant metrics to support the ML in the routing decisions. The performance evaluation shows significant improvements in terms of packet losses and end-to-end delay.