Adaptive Teaching Path Modeling and Dynamic Intervention Mechanisms for Higher Education Driven by Interactive Mobile Technologies

Abstract

The proliferation of mobile intelligent terminals and ubiquitous networks has established mobile learning as a core modality within the digital–intelligent higher education ecosystem. Conventional adaptive teaching systems, which rely on centralized cloud architectures, suffer from limited contextual awareness dimensionality, delayed intervention responses, and pronounced privacy risks—deficiencies that render them ill-suited for the fragmented and dynamically fluctuating contexts characteristic of mobile learning. To address these challenges, a three-tier adaptive teaching intervention framework was constructed, integrating on-device perception, edge-side inference, and cloud-based optimization. A gated adaptive multi-modal context fusion mechanism was designed to integrate multiple contextual dimensions and cognitive states into a unified joint representation. A lightweight efficiency prediction network was trained via knowledge distillation, enabling real-time local path fine-tuning on mobile terminals through constrained optimization, while raw data remained entirely on the device. An edge-empowered end–edge–cloud collaborative intervention strategy was proposed, wherein structured pruning facilitated low-latency inference at the edge layer, and a heterogeneity-aware personalized federated learning algorithm was introduced to perform global distributed optimization. A dual-timescale closed-loop feedback mechanism was established to synchronously enable instantaneous policy iteration and long-term knowledge system refinement. This study achieves a deep integration of mobile terminal sensing capabilities with adaptive teaching logic, reconciling real-time adaptability, global optimization, and privacy preservation, thereby providing technical support for the scalable deployment of digital–intelligent learning ecosystems in higher education.