Federated Learning for Vehicular Networks
share
Machine learning (ML) has already been adopted in vehicular networks for such applications as autonomous driving, road safety prediction and vehicular object detection, due to its model-free characteristic, allowing adaptive fast response. However, the training of the ML model brings significant complexity for the data transmission between the learning model in a cloud server and the edge devices in the vehicles. Federated learning (FL) framework has been recently introduced as an efficient tool with the goal of reducing this transmission overhead while also achieving privacy through the transmission of only the gradients of the learnable parameters rather than the whole dataset. In this article, we provide a comprehensive analysis of the usage of FL over ML in vehicular network applications to develop intelligent transportation systems. Based on the real image and lidar data collected from the vehicles, we illustrate the superior performance of FL over ML in terms of data transmission complexity for vehicular object detection application. Finally, we highlight major research issues and identify future research directions on system heterogeneity, data heterogeneity, efficient model training and reducing transmission complexity in FL based vehicular networks.
READ FULL TEXT
