An incentive scheme for federated learning in the sky

Abstract

The enhanced capabilities of Unmanned Aerial Vehicles have promoted the rapid growth of the Drones-as-a-Service (DaaS) market. To enable privacy-preserving collaborative machine learning among independent DaaS providers, we propose a Federated Learning (FL) based approach. There exists a tradeoff between Service Latency (SL), i.e., the time taken for the training request to be completed, and Age of Information (Aol), i.e., the time elapsed between data aggregation to completion of the FL based training. Given that different training tasks may have varying Aol requirements, we propose a contract-theoretic task-aware incentive scheme that can be calibrated based on the weighted preferences of the model owner. Performance evaluation validates the incentive compatibility and flexibility of our contract design amid information asymmetry.