A Resource Allocation Strategy in Internet of Vehicles Based on Multi-Task Federated Learning and Incentive Mechanism

With the continuous emergence of Internet of Vehicles (IoV) applications, the demand for computational resources of many resource-intensive applications in IoV has shown an explosive growth trend, which poses a serious challenge to the limited computational resources of the vehicles themselves. This paper designs a federated learning structure with a two-layer game for vehicular networks, using intelligent roadside terminals for federated optimization. Meanwhile, this paper proposes a Federated Learning and Cloud-Edge Gaming with Incentive-Driven (FL-CEGID) algorithm for dynamic task offloading in IoV. Our proposed algorithm optimizes vehicle and computing resource allocation as well as cache updates through a hierarchical distributed approach, which has separate vehicle and edge intelligence strategies for offloading decisions and caching strategies. The experimental results show that our proposed FL-CEGID has significant improvements in transmission capacity, transmission delay, and advantages in different key tasks and times in IoV compared to other schemes.