A DIGITAL TWIN-ASSISTED MOBILITY PREDICTION MODEL FOR SEAMLESS TASK HANDOVER IN IOV-ENABLED FOG COMPUTING

Abstract

The Internet of Vehicles (IoV) in 6G ecosystems generates ultra-low-latency, mobility-driven workloads from connected and autonomous vehicles (CAVs), where seamless task handover across fog-edge nodes is critical to avoid service disruption during high-speed movement (up to 120 km/h) and frequent handovers (every 5–15 seconds in dense urban scenarios). Traditional static or reactive offloading strategies suffer from prediction errors, excessive migration overhead, and increased latency under dynamic trajectories. This review presents a digital twin-assisted mobility prediction model that creates real-time virtual replicas of vehicles, road segments, network topology, and fog nodes to enable proactive, anticipatory task migration and resource orchestration. The framework integrates: (i) multi-modal trajectory forecasting using graph attention networks (GAT), spatio-temporal transformers, and LSTM variants fused with LiDAR/radar/HD-map data; (ii) digital twin-driven simulation of vehicle mobility, signal quality (RSRP/RSRQ), and computational load; (iii) predictive handover decision engine based on deep reinforcement learning (DRL) or model predictive control (MPC) that optimizes task splitting, migration timing, and target node selection; and (iv) blockchain-secured or federated learning-enhanced state synchronization for privacy-preserving twin updates across distributed fog domains. Evaluations on realistic vehicular traces and NS-3/OMNeT++ simulations demonstrate 40–75% reduction in handover latency, 25–60% lower service interruption probability, and 15–45% improved task success rate compared to baseline mobility-unaware or reactive schemes. The approach supports URLLC requirements (latency <1 ms, reliability >99.999%) for safety-critical applications (platooning, collision avoidance, remote driving) while mitigating signaling storm and resource fragmentation in dense IoV deployments. Challenges twin synchronization overhead, model fidelity under uncertainty, and scalability in massive CAV fleets are addressed through lightweight twin updates, edge-based inference, and hierarchical twin architectures. Digital twin-assisted mobility prediction emerges as a cornerstone enabler for resilient, anticipatory computing in future 6G-enabled intelligent transportation systems.