Abstract

Machine learning algorithms are used in manycomputing systems due to their abilities to learn the conditionsof a computer system without being explicitly programmed. Inparticular, these algorithms have a great potential to be exploitedin Vehicular Ad-hoc Networks (VANETs) because these networkshave very dynamic topologies and also very dynamic trafficloads. In fact, since each node or vehicle in the VANET hasa process unit, we can use the existing VANET architectureto develop machine learning-based algorithms to exploit theirbenefits. In this work, we present the design of an ArtificialNeural Network (ANN) classifier that enables VANET vehiclesto determine the Communication Network Traffic Load scenarioby means of evaluating three key metrics of the MAC layerprotocols in VANETs, namely, the Channel Busy Ratio (CBR),the Normalized Broadcast Received (NBR), and the NormalizedTimes Into Back-off (NTIB). The ANN classifier will eventuallybe part of a Context-aware System designed for improving thedissemination of safety messages. Our ANN Classifier has beenimplemented and validated through simulations, whose resultsdemonstrate that it provides an accuracy of success classificationof 98% for low density scenarios and a success classification of94% for high density scenarios.