Machine Learning Applications in V2X Networks: a systematic review

Abstract

Transport systems are evolving to be increasingly intelligent and connected, making it necessary to connect vehicles to pedestrians and network and traffic infrastructure elements. In this context, vehicle-to-everything (V2X) networks are of great importance, providing adequate connectivity requirements and quality of service parameters. Machine learning algorithms have great potential for solving problems in V2X and several works have been done in this area. This article is a systematic review of works in the area of Machine Learning applied to problem solving in V2X networks. Using the applied research method, 64 works published in journals were reviewed. Five research questions (RQ) were evaluated in the proposed methodology: research problem, datasets used, machine learning techniques adopted, hyperparameter tuning methods and computational tools for simulation. For future work, it is necessary to do more work in computational simulations, making the simulated context increasingly realistic, enabling the creation of datasets that can be applied to train ML models that will solve the problems of V2X networks in the real world. From this review, ML shows great potential for solving problems in V2X networks, being essential for development of intelligent transport systems.