A multimodal thermal comfort perception system with edge-cloud collaboration and augmented reality

Thermal comfort perception is a fundamental challenge in intelligent environmental systems. The adjustment of the smart system could enhance the human well-being and keep resilient buildings. However, traditional methods struggle with limited environmental context awareness and a lack of personalized adaptation. This paper proposes a multimodal thermal comfort perception system integrating edge computing, cloud-based Retrieval-Augmented Generation (RAG), and Augmented Reality (AR) technologies. The system achieves low-latency, high-precision personalized thermal environment regulation through heterogeneous sensor data acquisition, lightweight model inference, and knowledge-driven decision-making. Experiments demonstrate that compared to traditional methods, the system improves thermal comfort prediction accuracy while maintaining an average response latency in the acceptable level. AR interaction significantly enhances user satisfaction (NPS 90). The study validates the theoretical innovation and application potential of the edge-cloud hierarchical architecture and AR visualization in privacy preservation, real-time performance, and human-machine collaboration.