Information fusion of stationary, mobile, and wearable consumer-grade sensors to confidently estimate bedroom ventilation rates

This study adds to the body of research investigating ventilation rates measured in residential dwellings. Using open-source Consumer-Grade Sensors (CGS), we measure Carbon Dioxide (CO2) to estimate ventilation rates under one of three conditions: CO2 build-up when bedrooms were initially occupied, steady-state CO2 during the latter half of participants' sleep episodes, and CO2 decay after participants wake and vacate their bedroom. We do not ask participants to record their bedroom occupancy, but use information fusion to identify nightly occupied periods by cross-referencing home addresses to GPS measurements provided by their smartphones during sleep episodes identified by Fitbit wearable fitness trackers. We estimate 106 ventilation rates during build-up conditions which ranged from 0.04-1.36 h-1, 242 ventilation rates under steady-state conditions spanning 0.17-2.49 h-1, and 39 ventilation rates from decay conditions over a range of 0.06-1.05 h-1. In addition, we used occupied periods alongside times when GPS confirmed participants were away from home to label Indoor Air Quality (IAQ) data, and to train a Multi-Layer Perceptron (MLP) to classify occupancy based on CO2 and Total Volatile Organic Compound (TVOC) measurements. Using periods classified as occupied with at least 90% confidence, we estimated 116 additional steady-state ventilation rates spanning 0.25-2.49 h-1. Our results are consistent with more traditional and costly sensing approaches, and we conclude that information fusion of CGS is effective in measuring IAQ, detecting occupancy, and estimating ventilation rates.