A federated hybrid GNN-transformer framework for efficient traffic management in the Internet of Vehicles

Emergency healthcare transportation in the Internet of Vehicles (IoV) requires routing decisions that jointly consider traffic dynamics, wireless link reliability, medical urgency, and data privacy. Existing routing and learning-based IoV methods often address only part of this problem, which can lead to delayed ambulance movement, unstable relay selection, and exposure of sensitive mobility or healthcare information. This paper presents the hybrid graph neural network (GNN)-transformer federated opportunistic routing protocol (HGTF-ORP), a privacy-aware traffic management framework for healthcare IoV. The IoV environment is represented as a time-varying vehicle and infrastructure graph, and each candidate relay is evaluated through a healthcare-aware scoring function that combines Transformer-predicted congestion, GNN-based graph reachability, emergency priority, link reliability, buffer availability, and forwarding delay. Federated learning (FL) enables vehicles and roadside units (RSUs) to train local models without sharing raw healthcare or mobility records, while the opportunistic routing layer filters admissible neighbors and performs urgency-controlled route switching during congestion, link degradation, or RSU failure. Trace-driven simulations using Hyderabad traffic scenarios compare HGTF-ORP with baseline models using latency, packet delivery ratio (PDR), bandwidth usage, end-to-end delay, route optimality, a study-specific privacy score, robustness, and real-time scalability. The proposed protocol achieves 61.7 ms latency, 96.2% PDR, 111 ms end-to-end delay, 91.3% route optimality, a 9.3/10 study-specific privacy score, and 13.2 Mbps bandwidth usage, outperforming the strongest baseline across the main evaluation metrics. Ablation, statistical, deployment-oriented, security-threat, and scalability analyses further indicate that HGTF-ORP provides reliable and real-time routing support for emergency healthcare IoV environments.