Research on Incentive Mechanisms and Resource Allocation of Edge Intelligence Networks for Multiple Federated Learning Services

The integration of artificial intelligence and edge computing has given rise to edge intelligence. Federated Learning (FL), as a key technology of edge intelligence, utilizes data and computation power on edge devices to achieve decentralized edge intelligence. The problem of incentive mechanism design and resource allocation in FL has been two key issues that hinder the smooth progress of the learning process. We study an edge intelligence network that supports multiple simultaneous FL services, and there are multiple task publishers seeking FL services and mobile device alliances providing services for them. We address these two key issues. First, how task publishers can incentivize device alliances to contribute resources to facilitate the execution of FL services. Second, how the device alliances allocate resources across multiple FL services that are in progress to enable bounded rational decision-making. In order to solve the former, we transform the incentive design under information asymmetry into a contract game problem. For the latter, we propose an evolutionary game model to adjust their resource allocation strategies, while considering the influence of bounded rationality and external factors on the decision-making of device alliances. The simulation results show that our incentive mechanism shows excellent performance in the case of information asymmetry. Additionally, We experimentally simulated the dynamic evolution of device alliance strategies.