Artificial Intelligence of Things (AIoT) Enabled Floor Monitoring System for Smart Home Applications

TLDR

Floor monitoring systems using triboelectric sensors can capture daily activity data without cameras, yet their high humidity sensitivity and long‑term instability hinder reliable deployment. This work aims to create a robust, smart floor monitoring system by integrating highly reliable triboelectric coding mats with deep‑learning‑assisted data analytics. The system employs quaternary coding electrodes normalized against a reference electrode, a universal mask‑based screen‑printing design for cost‑effective fabrication, distinct wiring for each mat to enable parallel arrays, and deep‑learning analytics for position, identity, and control tasks. The resulting low‑cost, large‑area, reliable floor monitoring platform demonstrates promising advancement for smart home applications.

Abstract

To enable smart homes and relative applications, the floor monitoring system with embedded triboelectric sensors has been proven as an effective paradigm to capture the ample sensory information from our daily activities, without the camera-associated privacy concerns. Yet the inherent limitations of triboelectric sensors such as high susceptibility to humidity and long-term stability remain a great challenge to develop a reliable floor monitoring system. Here we develop a robust and smart floor monitoring system through the synergistic integration of highly reliable triboelectric coding mats and deep-learning-assisted data analytics. Two quaternary coding electrodes are configured, and their outputs are normalized with respect to a reference electrode, leading to highly stable detection that is not affected by the ambient parameters and operation manners. Besides, due to the universal electrode pattern design, all the floor mats can be screen-printed with only one mask, rendering higher facileness and cost-effectiveness. Then a distinctive coding can be implemented to each floor mat through external wiring, which permits the parallel-array connection to minimize the output terminals and system complexity. Further integrating with deep-learning-assisted data analytics, a smart floor monitoring system is realized for various smart home monitoring and interactions, including position/trajectory tracking, identity recognition, and automatic controls. Hence, the developed low-cost, large-area, reliable, and smart floor monitoring system shows a promising advancement of floor sensing technology in smart home applications.