3-channel representation of highway and random vehicular networks scenario for DNN-based lookahead next-hop routing

This work presents a Deep Neural Network (DNN)-based routing approach for Vehicular Ad-hoc Networks (VANETs), addressing challenges in high-mobility environments. The method introduces a grid-based representation of the vehicle environment to support local, learning-based routing decisions without relying on global topology. A U-Net-style DNN is trained to minimize per-hop transmission delay and improve real-time routing efficiency. Our dataset includes spatial-temporal snapshots of vehicular movements and communication metrics, used to train and evaluate the model.