Impact of Different Network Parameters on Federated Learning Performance in Vehicular Networks

This paper presents an in-depth examination of Federated Learning (FL) effectiveness in Vehicular Ad-Hoc Networks (VANETs) through systematic experimentation of road simulations. The paper methodically investigates the influence of various factors, such as communication range, client selection, vehicle density, and speed on FL performance within VANET systems. The findings emphasize that an increased communication range not only reduces communication overhead but also greatly enhances FL performance. Moreover, strategic client selection and optimal vehicular densities emerge as critical determinants for ensuring high-quality data and system efficiency. Remarkably, variations in vehicle speed are shown to have a minimal impact on FL performance, suggesting that vehicular speed has relatively limited effects on communication - efficiency. This paper offers a detailed understanding of the factors that affect FL performance in VANET systems, providing practical recommendations to optimize vehicular network systems.