A two-tiered incentive mechanism design for federated crowd sensing

Mobile crowd sensing uses the combined effects of a large number of users to collect, process, and reuse data for different types of applications. Federated Learning enables training a global model without compromising users’ privacy. In this work, we attempt to explore a new distributed sensing and learning paradigm, called Federated Crowd Sensing (FCS). Specifically, we propose a two-tiered incentive mechanism for FCS. First, we design an incentive mechanism in the participant recruitment stage where we consider the heterogeneous network effect, where larger fraction of participants will give potential mobile users an added value, but different users perceive it differently. Second, we design another incentive mechanism in the task result collection stage where we propose a hybrid uploading strategy selected by the users after completing the FCS tasks. Using the proposed algorithm for optimal uploading mechanism, the participants can increase their own utility. The numerical results show that platform can attract more potential mobile users, gain higher utility for platform and participants, and reduce the overall cost.