A Predictive and Multi-Objective Cloud–Edge Scheduling Framework with Task-Aware Encoding for Real-Time Educational Services

INTRODUCTION: With the rapid expansion of real-time interactive educational applications, including smart classrooms, online assessments, and learning behavior analytics, cloud–edge collaborative computing has become a critical infrastructure for ensuring low-latency and high-reliability service delivery. However, existing scheduling frameworks often rely on reactive or coarse-grained strategies that inadequately capture task-level latency sensitivity and system dynamics, resulting in inefficient resource utilization and unstable performance under high-concurrency educational workloads. OBJECTIVES: This study aims to design a latency-aware cloud–edge collaborative scheduling architecture capable of accommodating heterogeneous educational tasks, dynamically adapting to fluctuating resource and network conditions, and providing differentiated service prioritization to meet stringent real-time performance requirements. METHODS: The proposed architecture integrates three tightly coupled components: (1) a task feature encoder that models multimodal educational tasks in terms of computational load, data scale, and latency sensitivity; (2) a latency prediction model that leverages historical system states to anticipate future end-to-end latency trends; and (3) a multi-objective scheduling strategy that jointly optimizes latency, resource utilization, and migration cost through adaptive decision-making across cloud and edge nodes. RESULTS: Experimental evaluations conducted on real-world educational task datasets demonstrate that the proposed approach significantly outperforms mainstream reinforcement learning–based schedulers. Specifically, it reduces average end-to-end latency by 18.1%, improves average resource utilization by 8.0%, and achieves a task success rate of 96.8% under high-concurrency conditions, while maintaining lower latency jitter and migration overhead. CONCLUSION: The proposed latency-aware cloud–edge collaborative scheduling architecture provides an effective and scalable solution for guaranteeing low-latency performance in real-time educational services. By combining task-aware representation, predictive latency modeling, and multi-objective optimization, the framework offers strong practical value for deployment in smart classrooms and large-scale online education platforms.