A hierarchical federated learning incentive mechanism in UAV-assisted edge computing environment

Federated learning can utilize the local resources of user devices for model training while protecting their private security. Combining edge computing with federation learning can migrate computational tasks to the network edge, reducing communication latency and overhead. However, traditional edge stations are fixed in location and costly to deploy, making it difficult to cope with traffic surges, infrastructure failures, and device migration. Therefore, we introduce flexible and low-cost UAVs into federated learning, which can establish line-of-sight transmission paths with user devices and receive data for local aggregation. In federated learning, user devices need to contribute their own resources, and without sufficient rewards they may not be willing to participate, so we proposes a hierarchical federated learning incentive mechanism, which is designed based on contract theory considering data volume, data quality and cost input in an information asymmetry scenario. The experimental results compared with other benchmark schemes verify that the contractual design of this paper satisfies incentive compatibility and maximizes the utility of model owners.