Team Crossfire Continues to Build Autonomous Wildfire Suppression Systems Despite Semifinals Loss

Team “Crossfire,” a University of Maryland multidisciplinary effort to contend in the XPRIZE Competition to End Destructive Wildfires, will continue to work towards its mission of developing autonomous response systems for wildfires despite not qualifying for the finals.

The team, co-led by the Department of Fire Protection Engineering and xFoundry@UMD, had presented a wildfire response system consisting of drone-based detection and suppression technologies to a panel of judges last October. The team, which included students and faculty in the Departments of Aerospace Engineering, Mechanical Engineering, Maryland Robotics Center, UAS Research and Operations Center (UROC), and the Maryland Fire and Rescue Institute (MFRI) had been competing in the $5M Autonomous Wildfire Response Track.

“Crossfire put on a great demonstration showcasing what a dedicated team of students and faculty can achieve with a fraction of the budget and personnel of some of our competitors,” said Fernando Raffan-Montoya, an assistant professor in fire protection engineering and a technical lead in the project.

In their suppression system, a Freefly Alta X drone, hovered about 14 meters above a wood pallet fire and dropped a smart liquid suppressant payload designed to burst at a predetermined height above the flames, ensuring water dispersal extinguished the fire in its entirety. This payload was designed, built and tested over the course of several semesters by FPE and ME students, several of which were onsite to see their work in action. A second smaller drone (Chimera D, custom built by UROC) surveyed the fire scene from a higher vantage point to instruct the Alta X drone operator where to position the aerial vehicle for precise payload deployment.

Earlier, identical pallet fires were detected and localized with sub-meter precision using a custom data processing pipeline using visible and thermal camera footage and data captured by a DJI MAtrice 4TD. The processing algorithms used a deep learning model that recognized fire in the images and further confirmed an active fire using data from thermal cameras. To train the model, students fed 40,000 images to it so that it automatically identified high-risk and decoy fires, which was a requirement for the competition. Judges observed the detection and localization in realtime on a monitor showcasing the detection steps and adding localization dots to a map of the test site.

“We have developed novel pieces of technology that we intend to mature further as sponsored research projects and commercialization. I’m proud of what we achieved,” said Raffan-Montoya.

Now, finalists in the competition’s autonomous response track are developing systems that would detect and suppress a high-risk fire in 10 minutes or less over a large, environmentally complex area that is roughly the size of San Francisco, Oakland, and San Jose combined. Fire protection engineering experts think that this timeframe may not be fast enough, depending on wind conditions. Despite that, it is estimated that the solutions developed by the teams can still be put into use in more developed areas where fires spread more slowly, while firefighters evacuate civilians.

Despite not advancing to the finals, Crossfire is considering plans to commercialize their developing solution. Derek Paley, Willis H. Young Jr. Professor of Aerospace Engineering Education and a technical lead in the project, told the IEEE Spectrum in October that he has started conversations with potential customers in fire departments and government agencies for some of the technology developed by his research group, including fire detection and localization.

“The University of Maryland is well-positioned to seek public safety applications of our drone technology, thanks to our strong participation in team competitions like XPRIZE, which seek to spur research on topics with high societal impact,” said Paley.

In the News