Poster Session II
Project Type
Poster
Project Funding and Affiliations
NSF EPSCoR SMART FIRES' Seed Funding Award Program
Faculty Mentor’s Full Name
Anh Nguyen
Faculty Mentor’s Department
Computer Science
Abstract / Artist's Statement
Prescribed fires are deliberately conducted to reduce hazardous fuel loads, mitigate wildfire risk, and sustain ecosystem function. Their operational effectiveness, however, is limited by sparse real-time environmental sensing, inadequate situational awareness, and latency in decision support, factors that impede accurate forecasting of fire spread, smoke emissions, and associated safety hazards. This project proposes developing a prototype Internet-of-Things platform, FireFly, that addresses these limitations through distributed, multimodal sensing and on-device machine-learning inference. The system will deploy low-power sensor nodes measuring wind direction and speed, soil moisture, air quality, temperature, and humidity during controlled burns. Each node integrates an edge-computing unit capable of local classification without reliance on continuous internet connectivity, enabling reliable operation at remote burn sites.
Using the collected environmental data, FireFly will classify burn states: firelines, hotspots, and smoke dispersion patterns, in real time to improve operational responsiveness and support adaptive coordination of aerial and ground robotic assets. The prototype will be evaluated during controlled burn exercises (e.g., campus burn laboratory or smokejumper training facility), emphasizing sensing fidelity, communication reliability, and inference robustness. The resulting dataset and models will advance our IoT field of study applied in wildfire research by characterizing fire behavior under known environmental conditions and quantifying environmental dynamics throughout prescribed burn operations.
Category
Physical Sciences
FireFly
UC South Ballroom
Prescribed fires are deliberately conducted to reduce hazardous fuel loads, mitigate wildfire risk, and sustain ecosystem function. Their operational effectiveness, however, is limited by sparse real-time environmental sensing, inadequate situational awareness, and latency in decision support, factors that impede accurate forecasting of fire spread, smoke emissions, and associated safety hazards. This project proposes developing a prototype Internet-of-Things platform, FireFly, that addresses these limitations through distributed, multimodal sensing and on-device machine-learning inference. The system will deploy low-power sensor nodes measuring wind direction and speed, soil moisture, air quality, temperature, and humidity during controlled burns. Each node integrates an edge-computing unit capable of local classification without reliance on continuous internet connectivity, enabling reliable operation at remote burn sites.
Using the collected environmental data, FireFly will classify burn states: firelines, hotspots, and smoke dispersion patterns, in real time to improve operational responsiveness and support adaptive coordination of aerial and ground robotic assets. The prototype will be evaluated during controlled burn exercises (e.g., campus burn laboratory or smokejumper training facility), emphasizing sensing fidelity, communication reliability, and inference robustness. The resulting dataset and models will advance our IoT field of study applied in wildfire research by characterizing fire behavior under known environmental conditions and quantifying environmental dynamics throughout prescribed burn operations.