An AI-Enabled Multi-Sensor Embedded IoT Framework for Real-Time Indoor Air Quality Monitoring and Adaptive Air Purification
DOI:
https://doi.org/10.7492/91qyx134Keywords:
Indoor Air Quality, LSTM, HEPA Purification, Edge Computing, Multi-Sensor Fusion, Adaptive ControlAbstract
Indoor air quality (IAQ) poses a critical public health challenge, as individuals spend over 90% of their time in enclosed spaces where pollutant concentrations frequently
exceed WHO-recommended thresholds. This paper presents an AI-enabled, multi-sensor embedded IoT framework for real-time IAQ monitoring and adaptive air
purification. The proposed system integrates heterogeneous low-cost sensors including PM2.5/PM10, CO₂, CO, VOC, temperature, and humidity transducers interfaced
with an ESP32/Raspberry Pi edge node for continuous multi-parameter data acquisition. Raw sensor streams undergo hardware-level fusion and are processed locally
using a lightweight Long Short-Term Memory (LSTM) deep learning model optimized for edge deployment, enabling real-time pollutant classification and short-term
forecasting. A fuzzy-logic-based adaptive control engine dynamically regulates HEPA filtration, activated carbon modules, and variable-speed ventilation actuators in
response to predicted IAQ indices. Cloud connectivity via MQTT protocol supports remote dash boarding, alert generation, and over-the-air model updates.
Experimental validation across residential and office environments demonstrates an average pollutant prediction accuracy of 94.3% and a 78% reduction in PM2.5
concentration within 25 minutes of automated purification activation. The framework offers a scalable, energy-efficient, and cost-effective solution for smart building
automation, promoting healthier indoor environments with minimal human intervention.








