A Stackelberg Game and Federated Learning Assisted Spectrum Sharing Framework for IoV

Abstract

With the rapid development of Internet of Vehicles (IoV), an increasing number of vehicular users (VUEs) will connect to the Internet via 5G and beyond 5G(B5G) networks, which makes the spectrum resource becoming extremely scarce. However, the traditional spectrum allocation method cannot well adapt to the explosive growth of IoV traffic, and an efficient spectrum management for the IoV in B5G networks is an urgent challenge. In this paper, we propose a Stackelberg game and federated learning assisted spectrum sharing framework for the IoV. First, we develop a power control strategy for the region nodes considering its revenue and energy consumption, while VUEs can dynamically change their spectrum resource request strategy to maximize their revenue. We find the Stackelberg equilibrium using the alternating direction method of multiplier (ADMM) algorithm. To maximize the global revenue of region nodes, we leverage federated learning to realize the interaction between the region nodes and the central node. In specific, each region node utilizes deep learning to fit the relationship between the allocated power and its revenue. Then, they upload the network parameters to the central node. After collecting the network parameters from all region nodes, the central node can make the global decision about the spectrum allocation to the region nodes. Numerous simulation results verify the effectiveness of the proposed framework compared with the benchmark methods.