XAI-based Robust Train Communication Framework to Analyze Vehicle Path Loss for Public Safety

In the dynamic setting of train communication systems, ensuring consistent and reliable communication, especially during emergency situations, is of utmost for public safety. The framework predicts Vehicle Path Loss (VPL), a critical metric that serves as an indicator of signal propagation characteristics and communication quality. The prediction of VPL is influenced by multiple features, including ’RSRP_mean_interpolated_outdoor’, ’RSRP_mean_interpolated_indoor’, and ’BS_distance. The proposed framework leverages Artificial Neural Networks (ANN) to predict VPL within train communication systems. Furthermore, the interpretability of the predictions is provided by Explainable Artificial Intelligence (XAI) models like SHAP and LIME. The use of XAI fosters trust and transparency in decision-making processes. The results obtained showcase the effectiveness of the proposed framework, achieving a mean squared error of 0.0431 and a mean absolute error of 0.02678. The proposed framework prioritizes interpretability along with dependability in order to increase safety and efficacy in communication during emergency situations onboard trains.