The Science and Technology Directorate (S&T), the University of California San Diego (UC San Diego), and other partners recently collaborated with local fire departments in California to evaluate WIFIRE’s edge computing and artificial intelligence (AI) capabilities.
UC San Diego’s WIFIRE Lab creates science-driven technologies by combining data, hardware, software, and AI to support the fire management community. It is focused primarily on providing fire personnel with enhanced situational awareness about wildfires using next-generation data collection and processing tools.
“This is achieved via real-time simulations that are quickly and easily accessible, even by firefighters on the trench lines. We are working with UC San Diego to develop the most accurate fire simulation modeling possible by combining existing technologies and methodologies with new AI, hyper-localized data collection, and edge computing,” said S&T Program Manager Norman Speicher.
WIFIRE’s use of edge computing means that when the deployed sensors send their data to be crunched into a simulation, it is not sent out to the cloud or a distant datacenter, but instead the processing is done within its own local network.
Last summer, the team traveled to Kern County, California, about 130 miles north of Los Angeles, to conduct a prescribed burn scenario. This would be a true “live fire” exercise, where the Kern County Fire Department would use the WIFIRE Edge platform for a planned brush fire.
“The concept that we were setting out to test is, if we can capture hyper-localized data, including dynamic updates on the constantly changing conditions in and around a fire, supplement it with other existing data sources, and do all the processing at the edge… would it improve the fire models? If so, we can save more people’s lives and protect more people’s property from destruction,” said Speicher.
“WIFIRE Edge provides high-resolution information on the fire environment. What that really means is data from the fire front. It includes better and more accurate weather data, and better and more accurate location information,” added WIFIRE Lab Founding Director and Principal Investigator Ilkay Altintas.
This incredibly detailed data is collected by a series of sophisticated sensors literally attached to the firefighters and their equipment, as shown above. The sensors capture different environmental measurements, including GPS location, temperature, humidity, barometric pressure, air quality, CO2 levels, 2D and 3D wind speed, and more.
“The integration of advanced sensor technology with edge computing enables us to collect data about the fire environment from the personnel at fire’s edge, and then compute it right there on the remote site network. We developed two realistic scenarios to test and evaluate the impact of edge computing,” said WIFIRE Lab Director of Product Management Shweta Purawat.
The first part of the prescribed burn test began before the fires were even ignited. WIFIRE BurnPro3D, a prescribed burn planning and management tool developed by the WIFIRE Lab, created simulations that predicted fire behavior based on weather forecast data. Using WIFIRE BurnPro3D, Burn Bosses (the individuals responsible for the planning and safety of the burn) gained a data-driven understanding of the risks and trade-offs associated with the prescribed burn. This insight helped them determine if environmental conditions were suitable for the operation. If conditions were too dry or too windy, the risk of a ‘slop-over,’ where the fire spreads outside of the intended area, increases and could make the burn too risky. In such cases, the Burn Boss makes a Go/No-Go decision to conduct or postpone the exercise.
Once it was determined that the conditions were favorable, a series of small fires were lit in a prescribed pattern.
During the burn, the WIFIRE Edge continuously collected sensor data and generated time-series graphs. This allowed firefighters to monitor localized weather data, air quality parameters, and real-time trends. Using real-time measurements, fuel moisture was also continuously calculated at the edge. Additionally, live ignitor locations were transmitted through WIFIRE Edge to personnel’s mobile devices, enabling tracking of ignition patterns.
During the exercise, WIFIRE Edge successfully provided enhanced situational awareness, while the firefighters executed their mission of culling the dangerous tinder.
“Any analysis that we can do at the edge, that doesn't have to go through an outside network, will help us to understand the fire environment better and respond and react to things a lot faster. Then you combine that with information coming from other ground-based sensors like cameras or from satellites or from drones. The idea is to bring together all this multi-modal data and turn it into useful knowledge that can inform and support better fire management decisions,” said Altintas.
The second ‘Initial Attack’ fire response scenario simulated an unplanned fire. This hypothetical scenario was conducted about 45 miles east of Los Angeles in Corona, California, in November. It began with a mock call for a vegetation fire being reported to a 911 call center. From there, firefighters with body-worn and truck-mounted sensors responded to the fictional fire location. Once on scene, their sensors gathered environmental data, fed it to the WIFIRE platform at the edge, and it was then combined with additional data streams from traditional sensing apparatus. Then, dynamic simulations were created in real-time.
Had this been a real fire, potentially hours of waiting for a simulation to be built and shared could have been reduced to only minutes—and since the WIFIRE simulation is not just a static snapshot in time, the model would be constantly updating.
Doing the processing at the edge saves significant amounts of time, and in a fire, time is everything.
Based upon the situational awareness that was provided to the fire management team by WIFIRE, firefighters were dispatched to specific areas to dig trenches and douse the would-be flames. To assist with developing a common operating picture, the firefighters were also equipped with the S&T-funded Team Awareness Kit (TAK).
The feedback from the firefighters using the WIFIRE platform was impressive.
“Imagine a video game where you're looking at a detailed map and you can see where you are in relation to a wildfire. And you can see other emergency personnel, and you can see fire engines and police cars. And you can see where the structures are and where the public is. And you can see where the fire is heading. And it is all pretty much in real-time,” said Captain Andreas Johansson of the Corona Fire Department. “At that point, you know what you have to do to get the public out of harm’s way. And that’s what this whole thing is all about. It’s about saving lives.”
“The most at-risk areas for wildfires in the U.S. are in the Wildland Urban Interface (WUI), where internet connection can be weak or non-existent, and firefighting resources may be stretched thin. The WIFIRE Edge with its predictive AI, real-time edge computing, and ability to seamlessly connect with other S&T-developed technologies like smoke/fire sensors and TAK is a force-multiplier. These tests were a huge success, and I can see this making a big difference in protecting the public, as well as those in the fire service,” said Speicher.
Since these two field tests, the WIFIRE Edge team has continued to improve the platform with enhanced sensor capabilities and back-end system upgrades. Additional field testing and deployment at fire response incidents are planned for summer and fall of 2024.
For related media inquiries, contact STMedia@hq.dhs.gov. Visit S&T’s First Responder Capability page for additional information about our ongoing research and development efforts for the fire service.