Stackelberg Game-Base Incentive Scheme for Federated Learning in Artificial Intelligence of Things
Shaowen Qin, Jin Ye, Xin Tang, Xiaohuan Li · 2023
Artificial Intelligence of Things (AIoT) is an integrated technology that combines Artificial Intelligence with the Internet of Things. Due to the large number of nodes in AIoT, which require significant computational and communication resources and raise privacy concerns, the combination of federated learning (FL) with AIoT has become a hot research topic. However, the participation of edge nodes requires significant computational and communication overheads, so selfish edge nodes may be reluctant to participate in federated learning, and we need to incentivize nodes to contribute their resources. Therefore, it becomes a challenge to develop an incentive scheme that leads to optimal results with a limited budget. In this paper, we propose an incentive mechanism applied to federated learning to motivate edge nodes to contribute to model training, balancing overhead and model training under limited budget. Specifically, we model the utility between the server and the edge nodes and use the Stackelberg game to obtain the optimal solution of the utility model. Simulation results show that our proposed mechanism has significant performance.