GNN-Based Proportional Fair Dynamic Bandwidth Allocation in Wireless Vehicular Networks

In wireless vehicular networks, dynamic bandwidth allocation (DBA) faces two main challenges: mobility and heterogeneity. Rapidly changing vehicle locations make it difficult to predict and allocate bandwidth efficiently, and the diverse communication capabilities and requirements of vehicles can make fair allocation a challenge. The solution proposed in this paper involves two algorithms. The first algorithm uses Graph Neural Networks (GNNs) to predict the connection topology of the vehicular network based on historical data. This topology can be used to prioritize vehicles for bandwidth allocation based on their Quality of Service (QoS) requirements and proximity to other vehicles. The second algorithm dynamically allocates available network resources based on demand, ensuring flexible, efficient, and reliable communication services to all vehicles. These algorithms work together to address the mobility and heterogeneity challenges while providing a fair distribution of network resources to all users. A novel approach is used to optimize resource allocation in a volatile network, ensuring that users with higher QoS requirements receive a larger share of bandwidth while maintaining fair distribution.