Dynamic Federated Learning-Based Economic Framework for Internet-of-Vehicles

Federated learning (FL) can empower Internet-of-Vehicles (IoV) networks by\nleveraging smart vehicles (SVs) to participate in the learning process with\nminimum data exchanges and privacy disclosure. The collected data and learned\nknowledge can help the vehicular service provider (VSP) improve the global\nmodel accuracy, e.g., for road safety as well as better profits for both VSP\nand participating SVs. Nonetheless, there exist major challenges when\nimplementing the FL in IoV networks, such as dynamic activities and diverse\nquality-of-information (QoI) from a large number of SVs, VSP's limited payment\nbudget, and profit competition among SVs. In this paper, we propose a novel\ndynamic FL-based economic framework for an IoV network to address these\nchallenges. Specifically, the VSP first implements an SV selection method to\ndetermine a set of the best SVs for the FL process according to the\nsignificance of their current locations and information history at each\nlearning round. Then, each selected SV can collect on-road information and\noffer a payment contract to the VSP based on its collected QoI. For that, we\ndevelop a multi-principal one-agent contract-based policy to maximize the\nprofits of the VSP and learning SVs under the VSP's limited payment budget and\nasymmetric information between the VSP and SVs. Through experimental results\nusing real-world on-road datasets, we show that our framework can converge 57%\nfaster (even with only 10% of active SVs in the network) and obtain much higher\nsocial welfare of the network (up to 27.2 times) compared with those of other\nbaseline FL methods.\n