A UAV-integrated multi-sensor platform for hazardous gas detection, threat-level classification, and environmental monitoring using machine learning. Selected as 1 of 6 nationwide teams for the ₱100,000 Green Rising Fund grant, in partnership with UNICEF Philippines and the Positive Development Youth Network (PYDN).
Uncontrolled gas leaks from industrial facilities, storage systems, and residential equipment are a persistent, often undetectable health hazard — gas-related incidents account for more than 60% of confined-space fatalities worldwide. AEROVANT is a UAV-mounted electronic-nose (e-nose) system built to detect hazardous gas mixtures in real time and classify their cumulative health risk as the target gases accumulate in the environment.
The system integrates five MQ-series metal-oxide gas sensors (butane, ethanol, LPG, hydrogen, ammonia) with a DHT11 environmental sensor, mounted on a drone platform for mobile, real-time monitoring — overcoming the coverage limits of fixed gas-detection installations.
A two-stage machine learning pipeline converts raw sensor voltage into actionable risk classifications:
Stage 1 — Calibration: Ensemble regression models (Random Forest, XGBoost) convert raw MQ-sensor signals into accurate parts-per-million (ppm) concentrations, correcting for temperature and humidity drift.
Stage 2 — Classification: A Multi-layer Perceptron (MLP) neural network classifies cumulative health risk as "Stable" or "Critical," referencing the ACGIH Threshold Limit Value (TLV) Mixture Formula and OSHA/ACGIH exposure guidelines.
My contribution focused on the machine learning pipeline — sensor calibration modeling and the health-risk classification stage — alongside the wider team's UAV integration and field-testing work.
Kein Jake Anquillano Culanggo · Keith Laspoña · Eurice Jean Ca-ay Lopez · Chris Jallaine Sua Mugot · Qylle Christian Salvo Quiño — Department of Data Science, University of Science and Technology of Southern Philippines (USTP CDO).