Optimized Predictive Graph Learning Framework for Adaptive Routing in VANETs

Vehicular Ad-Hoc Networks (VANETs) are essential for enabling real-time vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) communication in intelligent transportation systems. However, frequent topology changes, varying traffic densities, and dynamic road conditions present critical challenges to stable and efficient data routing. This paper proposes an Adaptive VANET Routing Framework that combines Predictive Graph Learning using Graph Neural Networks (GNNs) with Swarm Intelligence inspired by Ant Colony Optimization (ACO). The proposed hybrid framework models the road network and traffic conditions as dynamic graphs, where GNNs predict congestion and assign real-time weights to road segments. These predictions are used by ACO to discover optimal communication paths that minimize delay, avoid congestion, and improve packet delivery. The framework adapts to real-time traffic variations, resulting in enhanced routing performance, reduced packet loss, and increased overall network throughput. Simulation results validate that the proposed method outperforms traditional routing protocols in terms of adaptability, energy efficiency, and communication reliability.