Vehicle-to-Everything (V2X) networks may experience intermittent connectivity and frequent link disruption due to high node mobility, variable density, and heterogeneous mobility patterns. When a contemporaneous end-to-end path cannot be guaranteed, Delay-Tolerant Networking (DTN) enables message delivery through store-carry-forward communication. Among DTN routing protocols, PRoPHET exploits historical and transitive encounters to estimate destination-specific delivery predictability. However, its conventional Greater Than (GRTR) forwarding rule has limited contextual awareness because relay selection is based primarily on relative delivery predictability. This paper proposes ProphetRouterRL, a two-stage routing protocol that augments PRoPHET with Q-learning. First, PRoPHET identifies eligible relay candidates satisfying ΔP > 0. A Q-learning agent then determines whether to send or hold each message using a six-dimensional state comprising relay-buffer availability, encounter history, message load, remaining time-to-live, hop count, and node type. ProphetRouterRL was evaluated against conventional PRoPHET using the ONE Simulator in a map-based urban scenario representing the UKM Bangi area in Malaysia. The seven-day simulation comprised 129 heterogeneous nodes, including pedestrians, cyclists, cars, buses, metro trains, and stationary communication points. ProphetRouterRL increased the delivery probability from 46.47% to 93.12%, while reducing routing overhead and average latency by 39.4% and 41.4%, respectively. These results demonstrate the potential of combining PRoPHET’s delivery-predictability estimation with context-aware learned forwarding decisions in opportunistic V2X environments.

A Delay Tolerant Network (DTN) is designed for intermittent connectivity and long delays environments, using a store-and-forward approach to transmit data. Machine Learning in Delay Tolerant Networks (DTNs): Prediction of link availability, routing optimization and adaptation to network dynamics. Machine learning models improve the efficiency of data transfer, thus reducing latency and enhancing decision making. This improves the robustness of the communication in the applications of remote sensing and disaster recovery. In addition, machine learning has shown promising application to design intelligent routing strategies for DTNs, which can improve communication efficiency in highly dynamic environments. The data unbalance due to different message generation rates and network connectivity has a negative impact on the performance of ML-based DTN routing algorithms. This results in biased predictions and sub-optimal routing decisions. The issue of imbalanced data in machine learning based DTN routing is examined in this study using four common protocols: Epidemic, Spray and Wait, Prophet, and MaxProp. We compare different oversampling strategies with six classifiers: Random Forest, Extra Trees, XGBoost, Bagging Classifier, Decision Tree and K-Nearest Neighbors. We discuss different data pre-processing techniques, algorithmic approaches and pertinent evaluation measures. Our analysis finds the best approaches for handling data imbalance problems. Among the methods considered, SMOTENN with Extra Trees classifier consistently obtained the highest accuracy in all four procedures. The application of these tactics resulted in an impressive improvement in model performance - accuracy increased by 13.58% (from 81% to 92%) and thus improving the robustness and fairness of ML-based DTN routing protocols.

Vehicle-to-Everything (V2X) networks may experience intermittent connectivity and frequent link disruption due to high node mobility and variable density. When an end-to-end path cannot be guaranteed, Delay-Tolerant Networking (DTN) enables message delivery through store-carry-forward communication. Among DTN routing protocols, PRoPHET exploits historical and transitive encounters to estimate destination-specific delivery predictability, enabling forwarding decisions without requiring a contemporaneous end-to-end path. However, its Greater Than (GRTR) forwarding rule has limited contextual awareness because relay selection relies on relative delivery predictability. This paper proposes ProphetRouterRL, a two-stage routing protocol that augments PRoPHET with Q-learning. PRoPHET first handles contacts, predictabilities, and direct delivery; a Q-learning agent then chooses SEND or HOLD for each message from a six-dimensional state (hop count, relay buffer space, relay node type, relay as destination, remaining time-to-live, and destination-encounter history). ProphetRouterRL was evaluated against conventional PRoPHET using the ONE Simulator in a map-based urban scenario of UKM Bangi, Malaysia, with 129 heterogeneous nodes. Across four 12-h runs with different movement seeds, ProphetRouterRL delivered 90.26% (SD 0.54%) against 46.40% (SD 0.20%) for PRoPHET, with a 36.8% lower overhead ratio and 39.5% lower latency ($p<0.001$). These results demonstrate the potential of context-aware learned forwarding decisions in opportunistic V2X environments.