Edge-Intelligent Wearable IoT for Real-Time Stress Monitoring and Indoor Localization: A TinyML-Enabled Adaptive RPL Approach
TL;DR
An edge-intelligent wearable IoT system that integrates photoplethysmography-based sensing, edge machine learning (TinyML), adaptive networking based on the Routing Protocol for Low-Power and Lossy Networks (RPL), and zone-aware indoor tracking is proposed, representing a validated proof-of-concept toward preventive healthcare and next-generation Internet of Medical Things applications.
Abstract
The growing demand for digital health solutions, continuous monitoring, and indoor location awareness has accelerated the growth of wearable IoT systems. However, real-time analytics and low-latency communication operations are still challenging in resource-constrained environments, particularly due to limitations in processing power and network connectivity. The present paper proposes an edge-intelligent wearable IoT system that integrates photoplethysmography-based sensing, edge machine learning (TinyML), adaptive networking based on the Routing Protocol for Low-Power and Lossy Networks (RPL), and zone-aware indoor tracking. For on-device stress classification, a lightweight TinyCNN model with 32,546 parameters is deployed for embedded inference, achieving a subject-independent test accuracy of 80.11% and an AUC of 0.8814 on a large-scale wearable dataset. The system ensures real-time health analytics and low-latency alert generation, with a packet delivery rate of 98% and average latency of 51 ms. Indoor positioning is achieved without GPS or dedicated beacon infrastructure through the monitoring of gateway router associations. The proposed system enables privacy-preserving analytics and supports scalable smart healthcare systems, representing a validated proof-of-concept toward preventive healthcare and next-generation Internet of Medical Things applications. The proposed system is intended as an engineering research prototype and is not designed for clinical diagnosis.