IoT-Enabled Smart Ventilation System with Edge-Based Oxygen Regulation and Embedded Clinical Training
M Omprakash, S Saisukirtha, S Shane Steffy, Dilip Kumar S · 2025
An IoT-enabled, edge-intelligent ventilation system is being designed to address the shortcomings of manual Ambu bags and expensive ventilators in the decentralized and resource-scarce settings. The system envisages the use of an ESP32 microcontroller, and a modular sensor unit namely MAX30102 for oxygen saturation and pulse rate, DHT11 for temperature and humidity, MPX5010DP for differential airway pressure, MQ-135 for CO2, SF06 for oxygen flow and AWM720P1 for real-time airflow monitoring. Sensor information is transmitted via HTTP to a cloud-based dashboard where it can be continuously visualized, trended, and alerted on. TinyML-based decision tree regressor implemented based on TensorFlow Lite for Microcontrollers predicts dynamically that oxygen is needed between 1 to 10 liters/min and controls motor-driven compression and valves. A web-based simulation interface transmits artificial ventilation characteristics and identifies respiratory condition such as normoxia, hypoxia, and hyperventilation with 94.6% accuracy. For anomaly detection, a lightweight Isolation Forest model is employed and it accurately distinguishes 91.2% abnormal physiological behaviors and sensor malfunctions. Ambient light and sound sensors enable context-aware diagnostics and a capacitive stretch sensor monitors chest expansion to evaluate breathing effort. Built-in battery management and redundancy modules provide continuous operation. The system has been tested in mock clinical settings, and is being transformed into a product-grade MVPP for the intended scalable deployment during an emergency response and clinical training scenario.