Incentive Based Federated Learning Data Dissemination for Vehicular Edge Computing Networks

The advancements in vehicular networks and applications brings about a huge demand for vehicles to process computation-intensive tasks such as machine learning (ML) algorithms. There is a need for an efficient solution to meet up these challenges. Federated learning (FL), which is distributed and does not require the transmission of dataset to the server, is more suitable to the resource constraint vehicular edge network. However, not all vehicles will participate in FL training without any reward. This paper proposes an incentive-based federated learning scheme for vehicular networks. Vehicles participating in federated learning are selected using an auction algorithm based on maximized social welfare. Simulations are conducted to evaluate the performance of the proposed scheme.