Synergistic mechanism of virtual reality and reinforcement learning in educational adaptive learning systems

Abstract

This paper proposes a VR (Virtual Reality) and RL (Reinforcement Learning) collaborative optimization model based on the TD3 (Twin Delayed Deep Deterministic Policy Gradient) algorithm to address the ineffective integration of immersive interactive data and intelligent decision-making in educational adaptive learning systems. The model captures real-time multimodal student data via a VR headset, processes it into graph-structured state features using a lightweight GCN (Graph Convolutional Network), and drives strategy optimization with a dynamic reward function that perceives cognitive load based on predicted knowledge mastery. The TD3 algorithm, utilizing a dual-Q network and adaptive thresholds, generates adaptive instructional parameters. Experiments demonstrate the model achieves 85.6% knowledge mastery prediction accuracy at the comprehensive problem level and maintains a 278ms strategy response latency at the basic concept level, validating its effectiveness in collaborative data fusion and dynamic decision-making.