Mobile fog computing must support latency-sensitive applications under dynamic user mobility and time-varying network conditions. Existing mobility-aware scheduling approaches are largely reactive and often ignore prediction uncertainty, resulting in service disruptions and inefficient task migration. This paper proposes an uncertainty-aware digital twin-based orchestration framework for proactive mobility-aware fog computing. The framework maintains real-time synchronized digital twins of users and fog nodes and integrates a hybrid Gated Recurrent Unit-Exponentially Weighted Moving Average (GRU-EWMA) mobility prediction model with fog-load forecasting to enable joint mobility- and load-aware decision-making. An entropy-based confidence mechanism is introduced to regulate proactive handover and task migration, thereby reducing unnecessary task migrations when predictions are uncertain. The proposed framework is implemented in the MobFogSim simulator and evaluated against state-of-the-art baselines. Experimental results demonstrate that the proposed approach reduces the average task delay by up to 28.1%, decreases energy consumption by up to 9.5%, and improves the task success rate to 99.1%, while incurring only a modest digital-twin computational overhead. These results confirm that integrating uncertainty-aware mobility prediction with digital twin–driven orchestration significantly enhances reliability and efficiency in mobile fog computing environments.
Kaur et al. (Thu,) studied this question.