ERI: Airborne Visual-Olfactory Remote Sensing System for Early Wildfire Detection

This Engineering Research Initiation (ERI) award supports research to develop an early wildfire detection system using autonomous unmanned aircraft systems (UASs) to provide a low-cost, flexible, and timely solution for identifying wildfires at their earliest stages. As wildfires become increasingly frequent and destructive across the United States, there is a critical need for detection methods that can identify fires before they grow out of control. Traditional approaches, such as satellite monitoring, manned aircraft surveys, and remote camera surveillance, often suffer from limited resolution, delayed response times, and high operational costs. This project introduces a dual-sensing strategy by equipping UASs with both visual and olfactory sensors. Similar to how humans can smell smoke before seeing flames, the proposed UAS would detect smoke particles and chemical signatures before visual evidence becomes apparent. The ability to trace smoke back to its source, even in low-visibility conditions such as heavy smoke or nighttime, would enable faster and more accurate fire identification. The successful outcome of this research will impact broader applications beyond wildfire detection, such as chemical leak detection and other environmental monitoring tasks. Additionally, the project will foster interdisciplinary collaboration between engineering, environmental science, and forestry, while also providing educational opportunities for students through hands-on research and field testing. This research proposes a two-phase wildfire early detection method that includes machine learning (ML) and robotics. In the first phase, an ML model will be developed to predict high-risk wildfire areas based on historical fire records, environmental conditions, and human activity data. The model will incorporate spatial and temporal analysis, along with causal reasoning techniques, to improve both prediction accuracy and interpretability. This risk map will guide where early detection efforts should be concentrated. In the second phase, UASs will be deployed to these high-risk zones and guided by a novel visual-olfactory navigation algorithm. The system will include a computer vision model to detect smoke and heat hotspots from visual imagery, as well as a biologically inspired algorithm that allows the vehicle to follow smoke plumes when visual cues are unavailable. By integrating predictive modeling with autonomous exploration, this project seeks to develop a practical, scalable system that enhances the nation’s capability for early wildfire detection and supports timely intervention before fires escalate. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.