Enhancing the Granularity of Spatial Sensing Using a Low-Cost IoT-Based Sensing Network

Pennsylvania State University, Department of Architecture, University Park, PA, US

^* Corresponding author: This email address is being protected from spambots. You need JavaScript enabled to view it.

Abstract

Indoor carbon dioxide (CO₂) concentration is a widely used indicator of ventilation effectiveness, occupancy-driven emissions, and indoor environmental quality, with direct implications for health, comfort, and cognitive performance. As demand increases for continuous and spatially resolved indoor CO₂ monitoring, low-cost sensing systems integrated with advanced modeling techniques offer a promising alternative to traditional, sparsely deployed reference instruments. This study presents a hybrid fixed-mobile, low-cost Internet-of-Things (IoT) sensing framework to reconstruct high-resolution spatiotemporal indoor CO₂ distributions from limited sets of sensing data collected using fixed-mobile sensing systems. After validating individual sensor nodes against reference instrumentation, a distributed sensing network was deployed in a controlled indoor environment using a combination of continuously operating stationary sensors and sequentially repositioned portable devices. A grid-based stop-and-measure protocol was adopted to ensure measurement stability while achieving full spatial coverage under sparse and asynchronous sampling conditions. To reconstruct CO₂ concentration fields from incomplete spatiotemporal observations, Gaussian Process (GP) and Random Forest (RF) models were evaluated under random, periodic, and spatial cross-validation schemes. Results show that GP consistently outperformed RF in terms of variance preservation, spatial generalization, and normalized error metrics, aligning with ASTM D5157-19 recommendations for indoor air quality model evaluation. This work demonstrates that combining low-cost hybrid sensing with uncertainty-aware spatiotemporal modeling provides a scalable and cost-effective approach for indoor CO₂ monitoring. Future work will extend the framework with autoregressive spatiotemporal formulations to better capture short-term temporal dependence and localized transient peaks under occupied conditions, varying ventilation states, and real-time digital twin integration.