This paper explores the integration of Federated Learning with autonomous driving simulation using the CARLA simulator, the YOLO object detection algorithm, and the Jetson Nano as a representative edge device. Synthetic data were extracted from CARLA simulations, with annotations converted to the YOLO format, eliminating the need for manual labeling. The proposed approach allows multiple vehicles to locally train object detection models using camera video streams and subsequently aggregate their parameters through a Federated Learning mechanism, without sharing raw data. This method enhances data privacy and scalability in distributed autonomous driving systems.