Functional Data Analysis-Guided Prompt Design for RFID Sensing and Localization Using LLMs
Conference
Sun, Y, Wang, X, Cao, G et al. (2025). Functional Data Analysis-Guided Prompt Design for RFID Sensing and Localization Using LLMs
. 2015 IEEE GLOBAL COMMUNICATIONS CONFERENCE (GLOBECOM), 841-846. 10.1109/GLOBECOM59602.2025.11431647
Sun, Y, Wang, X, Cao, G et al. (2025). Functional Data Analysis-Guided Prompt Design for RFID Sensing and Localization Using LLMs
. 2015 IEEE GLOBAL COMMUNICATIONS CONFERENCE (GLOBECOM), 841-846. 10.1109/GLOBECOM59602.2025.11431647
Radio Frequency Identification (RFID) based sensing has emerged as a crucial technique in pervasive computing, making applications such as human activity recognition and indoor localization more accessible. However, traditional neural models often face challenges when dealing with the high dimensionality and temporal or spatial dependencies in RFID signal data. To address these issues, we propose a novel framework that enhances the predictive capabilities of large language models (LLMs) by constructing structured prompts tailored to RFID data. Our method leverages functional data analysis (FDA) to preprocess and smooth raw RFID signals, extracting informative features that preserve both global trends and local variations. These features are embedded into task-specific prompts, enabling LLMs to perform accurate predictions in few-shot settings. We evaluate our approach on two representative real-world datasets: XRF55 for activity classification and AdaRF for localization. Experimental results demonstrate that our framework significantly outperforms a ResNet baseline, achieving up to 15.74% higher classification accuracy and reducing root mean squared error (RMSE) and mean absolute error (MAE) by 12.85% and 16.67%, respectively, in the localization task.
