Enhancing Safety and Performance in Intelligent Vehicular Networks Using Edge-Based Explainable AI Models

The automated cars must possess real time, efficient and understandable decisions that will ensure that there is road safety and efficiency. In the case, the researchers propose to use an edge-enabled model that is utilized to include explainable artificial intelligence (XAI) architectures, which are CNN, RNN, and GNN to provide a better safety prediction rate, low latency, and throughput when driving in dense traffic. As the analysis in the performance indicates, the deployment of the edges is significantly quicker in both the latency and the throughput like the cloud-based computation provided that the real time is performance is stable, always, with reference to the condition of peak traffic. The implementation of XAI models will cause minor trade-offs in raw accuracy and throughput but result in a notable distinction in the interpretability of the model, with stakeholders in a position to trust and audit model predictions. XAI may also yield better quality of collision prediction since as a safety measure the XAI is only effective within 2-5% change of the accuracy of the collision prediction. The findings are indicative that edge enabled XAI framework is an effective solution to intelligent transportation systems of the next generation because it is effective, safe, and transparent. Future research can be done in the field of multi-modal data integration and adaptive hybrid edge-cloud architecture to make it even more performance-driven and interpretable.