The fifth (5G) and emerging sixth (6G) generations of cellular networks aim to support vehicular communications, including communication between vehicles, pedestrians, and road infrastructure. These communications often face difficult propagation conditions due to fast-varying wireless channels and must support low latency, high reliability, and high data rates, with vehicles forming dynamic topologies. Leveraging cellular networks, vehicular applications increasingly rely on machine learning services to enable assisted and autonomous driving. Sensors embedded in vehicles generate large volumes of privacy-sensitive data. While centralized machine learning can leverage this data for inference and real-time decisions (e.g., 2D and 3D object detection), its centralized nature is often unsuitable. Since data is inherently distributed across vehicles, federated learning (FL) emerges as a promising alternative, enabling local training while improving communication efficiency and privacy. FL trains a global model via a central server and a subset of devices, which use local data to optimize local objectives and iteratively exchange model updates. FL is particularly suited for scenarios requiring limited communication, data privacy, and adaptation to heterogeneous devices with varying data and compute capabilities. To realize the vision of self-driving systems, it is essential to integrate wireless communications and distributed machine learning. FL algorithms must be adapted to vehicular communication constraints, and wireless resource allocation must be tailored to FL tasks. This project investigates distributed learning for assisted driving systems that benefit from vehicular communications, with a focus on how mobility impacts learning performance. We aim to evaluate FL from both the learning and communication perspectives, analyzing communication efficiency, privacy, and suitability for vehicular tasks such as object detection, and understanding how mobility affects model transmission, energy consumption, and system performance. The research is carried out at Uppsala University by a team comprising one master’s and one doctoral student, each working on specific aspects: • Master’s student: Federated Diffusion Models for Vehicular Applications. • Doctoral student:Distributed Machine Learning over Vehicular Communications.