Digital Twin–Enabled Resource Scheduling Optimization for Business English Mobile Interactive Systems

Abstract

Heterogeneous real-time tasks within business English mobile interactive systems impose stringent demands on edge resource scheduling. Conventional reactive scheduling paradigms are ill-suited to time-varying user behaviors and differentiated quality-of-experience requirements, often resulting in degraded service quality and suboptimal resource utilization. To address these challenges, a digital twin-enabled dual-timescale resource scheduling optimization framework was proposed. A lightweight learner-specific digital twin model was constructed, integrating multidimensional features from terminal devices, network conditions, and user behaviors. Gated recurrent units (GRUs) were embedded to enable proactive resource demand forecasting, while an event-driven synchronization mechanism was employed to balance model fidelity and system overhead. A hierarchical scheduling scheme driven by quality of experience was developed, comprising a large-timescale component that optimizes resource reservation policies online via an upper confidence bound algorithm and a small-timescale component that performs real-time resource allocation using task-specific quality-of-experience sensitivity heuristics. A bidirectional closed-loop feedback channel was established to facilitate continuous iterative refinement of both the model and the policy. Simulation results demonstrated that the proposed framework substantially outperformed mainstream baseline algorithms in terms of average quality of experience, latency satisfaction ratio, and edge resource utilization, with performance gains being particularly pronounced under high-concurrency scenarios. This study provides theoretical and technical foundations for efficient edge resource allocation in mobile learning environments.