Researchers at the University of Minnesota Twin Cities have developed a new system of aerial robots that use artificial intelligence to detect, track and analyze wildfire smoke plumes. The technology is expected to help create more accurate computer models, which could improve predictions of air pollution levels during wildfires and prescribed burns. The findings were published recently in the journal Science of the Total Environment.
The need for better smoke monitoring tools has grown in recent years. Between 2012 and 2021, 43 wildfires resulted from 50,000 prescribed burns, according to an Associated Press report from 2024. Understanding how smoke spreads is critical because smaller smoke particles can travel long distances and remain suspended for longer periods, affecting air quality in areas far from the fire.
The research team addressed limitations in current smoke modeling systems, which previously struggled to gather detailed real-time data during fires. To improve this, the scientists used coordinated swarms of aerial robots equipped with AI and sensors. These robots can identify smoke, fly directly into it and capture multiple angles, allowing researchers to create 3D reconstructions of smoke plumes and closely study flow patterns. Traditional drones and satellite-based tools have been less effective because they are either too costly or limited in detail.
The new approach enables high-resolution data collection across large areas at a relatively low cost, said Nikil Krishnakumar, the study’s lead author and a graduate researcher at the Minnesota Robotics Institute. Professor Jiarong Hong, senior author of the study, explained that better understanding smoke composition and dispersion will support improved hazard response planning.
The team says this technology could also be adapted for other airborne hazards such as sandstorms and volcanic ash clouds. Their next steps focus on developing practical systems for early fire detection and rapid response. Earlier work by the group involved autonomous drones that could detect and track smoke in real time. Now, they are aiming to improve particle characterization using Digital Inline Holography and coordinated multi-drone systems. They are also testing fixed-wing VTOL drones capable of taking off vertically and flying long distances, which could provide extended surveillance.
The project was supported by the National Science Foundation and carried out in collaboration with the St. Anthony Falls Laboratory. The researchers hope that as development continues, the system can be deployed widely to help communities prepare for and respond to increasing wildfire risks.