Navigation coopérative de véhicules autonomes basée sur la communication V2X dans un réseau de 5ème génération

Abstract

In today’s world, road transport is essential to our daily routines and business activities. However, the exponential growth in the number of vehicles has led to problems such as traffic congestion and road accidents. Vehicular communication presents an innovative solution, envisaging a future where vehicles communicate with each other, the road infrastructure, and even the road itself, sharing real-time data to optimize traffic flow and enhance safety. This thesis focuses on 5G and Beyond 5G (B5G) technologies, which promise to revolutionize Vehicle-to-Everything (V2X) communication. With the emergence of millimeter-wave (mmWave) communication, high-speed, low-latency data transmission is essential for vehicular networks. However, mmWave communication faces problems with signal attenuation and interference. Our research focuses on solving these problems using a deep learning-based approach. Three significant contributions are proposed. First, we introduce a classical optimization technique, the simulated annealing algorithm, to improve beam alignment in 5G vehicular networks. This reduces latency and improves data transmission between millimeter-wave base stations and vehicles. Our second contribution is a new approach involving a hybrid deep-learning model that predicts optimal beam angles. Combining a 1D CNN and a BiLSTM improves th accuracy of the prediction and reduces errors. This approach eliminates time-consuming computations and iterations critical to the success of B5G vehicular networks. The third contribution introduces a BiLSTM-based model to select the optimal beam pair angles at the mmWave base station (mmBS) and the moving vehicle side. This approach improves the reliability of data transmission while minimizing the error probabilities and overheads during beam search. This research contributes to advancing vehicular communications, offering innovative solutions for 5G and B5G networks. We aim to enhance the efficiency, reduce the latency, and improve the reliability of communications for connected vehicles. This thesis explores beam alignment through classical and deep learning techniques and presents solutions for the challenges of millimeter-wave vehicular networks. Our research provides the foundation for the next generation of vehicular communication and its vital role in making road transport safer and more efficient.