Research on the Construction of Virtual-Physical Learning Environments Driven by Multimodal Brain-Computer Interfaces

Abstract

Against the backdrop of the global digital transformation of education and the deep integration of artificial intelligence technologies, achieving precise and personalized intelligent education has emerged as a pivotal scientific issue. This study employs a multimodal brain-computer interface as the experimental equipment, integrates extended reality with educational multi-agent technologies, and constructs a new paradigm of virtual-real learning environments encompassing neural perception, intelligent decision-making, and scenario adaptation. Specifically, multimodal electroencephalographic data are collected, followed by the application of Unity dynamic scene generation technology. Finally, the Coze platform is utilized to access the DeepSeek-R1 large model, thereby developing a dual-mode educational agent with both coaching and strategic capabilities. Immersive virtual teaching experiments were conducted in pilot institutions, and empirical results demonstrate that this research can effectively enhance students' path-planning abilities and learning efficiency.