30,000 wildfires have already burned across the U.S. in the first half of 2026, according to the National Interagency Fire Center. The standard method for assessing how dangerous those conditions are takes two days and involves carrying plant samples back to a lab, baking them in an oven, and doing math on the moisture numbers. Two graduate students at Northeastern University just built a robot-mounted sensor system that does the same thing in real time, continuously, from a moving vehicle.
The Problem With How Wildfire Risk Is Measured Today
Before a wildfire spreads, the material that feeds it, dead leaves, dry needles, twigs, discarded trash, has to be dry enough to ignite. Wildfire managers call this "fuel." Measuring how much usable fuel is present at any given location is called fuel moisture sampling, and it is one of the core inputs into fire risk modeling.
The process as it exists today is manual. Field teams walk into forests, collect samples of foliage, carry them back, weigh them, bake them in a drying oven, weigh them again, and compute moisture content by dividing the numbers and multiplying by 100. According to the U.S. Department of Agriculture, the full process takes around two days. In high-risk areas, this is done year-round, with frequency increasing during fire season.
Two days is a long time when conditions change fast. And the sampling is inherently sparse. You cannot send field teams into every corner of every fire-prone forest every week.
What the Northeastern System Actually Does
James Tukpah and Ben Cometto, graduate students at Northeastern University's Robotics and Intelligent Vehicles Research Laboratory (RIVeR Lab), built a sensor array that mounts on top of an autonomous vehicle or mobile robot and scans vegetation while the vehicle moves through a forest.
The system uses two technologies working in parallel. First, a hyperspectral camera captures light across continuous wavelengths, not the three channels a standard camera uses, which allows it to classify the moisture content of vegetation based on how plant material reflects different parts of the spectrum. Second, a lidar sensor emits pulses of light to build a 3D point cloud of the scanned area. The system then maps moisture data onto that point cloud and displays it as a gradient map, giving operators a spatial, color-coded picture of fuel conditions across a forest section.
The result, as Tukpah described it, is continuous real-time measurement instead of periodic sampling. Rather than pulling samples every few months, the robot can scan continuously as it moves, building an up-to-date picture of fuel moisture across a much larger area than a field team could cover on foot.
The system won first place at the SICK $10K Challenge, an annual collegiate robotics competition hosted by SICK Sensor Intelligence, a sensor company serving the automation industry. The Northeastern team competed against student teams from across the U.S. and Canada. The sensor array they submitted is called the SICK-IgniScan.
Why the Angle Matters: Prevention, Not Detection
Most robotic and AI systems in the wildfire space focus on detection: finding a fire once it has started and tracking its spread. Drones with thermal imaging, ground cameras with computer vision models, satellite monitoring networks. These are useful tools. They are also reactive by definition.
Tukpah and Cometto explicitly framed their work as targeting a different point in the timeline. "We figured that a lot of solutions focus on fire detection and fire prevention once the fire has already started," Tukpah said. "This gives us the opportunity to tackle wildfires from a different preventive measure."
The practical implication is that a system like this, at scale, could give fire managers continuous data on where high-fuel-moisture conditions are shifting toward high-risk levels before ignition occurs, rather than responding to an active fire after the fact. That is a different kind of tool than anything currently deployed at scale in the field.
This connects to a broader shift in how robotics is being applied to real-world infrastructure problems. Research platforms for humanoid robots have focused heavily on manipulation and mobility in controlled settings. Systems like the SICK-IgniScan point toward a different use case: autonomous sensing in unstructured outdoor environments where the environment itself is the problem to be mapped.
What Is Still Unfinished
The system is currently in field testing on Northeastern's off-road vehicle, which Tukpah has spent several years working to make autonomous. It can build a point cloud of a forest area and assign moisture data points to locations. Classifying specific types of vegetation, distinguishing a dead pine needle from a dry shrub from a piece of trash, is a capability Cometto said would likely come later.
No deployment timeline has been announced. No cost figures were provided. No field agency partnerships were named in the source. The system exists as a working prototype that won a student competition. What happens between that and actual deployment in a fire management operation is an open question the current reporting does not answer.
That gap between a working lab prototype and operational deployment is not trivial. Robots operating in dense, uneven forest terrain face conditions far more variable than a university test environment. Weather, soil type, dense canopy cover, and terrain slope all affect sensor performance in ways that field testing will have to resolve. Whether the point cloud and moisture mapping hold up under those conditions is still being tested.
For context on where autonomous ground vehicles currently are in real-world outdoor operations, even purpose-built robots deployed in structured environments like postal sorting hubs still trail human performance by significant margins in real-world conditions. Unstructured forest terrain is considerably harder.
My Take
The technology is not new. Lidar and hyperspectral imaging have both been used in remote sensing for years. What is new is the combination, mounted on an autonomous ground vehicle, aimed specifically at the fuel moisture problem. That framing matters. The wildfire robotics conversation has been dominated by detection and suppression systems. Putting a robot in the forest before the fire starts, measuring conditions that predict ignition risk, is a different intervention point.
Whether this scales past a university prototype depends on factors the current source does not address: sensor accuracy in real field conditions, cost per unit, operational durability, and whether fire management agencies have the budget and interest to deploy something like this. Those are the questions worth watching, not the competition win.
- Northeastern graduate students James Tukpah and Ben Cometto won first place in the SICK $10K Challenge for a sensor system called the SICK-IgniScan.
- The system combines lidar and a hyperspectral camera to map wildfire fuel moisture in real time from a moving autonomous vehicle.
- It replaces a 2-day manual sampling process with continuous, real-time data collection.
- The angle is preventive, not reactive: assessing fuel risk before a fire starts rather than tracking one after it does.
- The system is still in field testing. Specific vegetation classification, deployment timelines, and agency partnerships are not yet confirmed.
FAQ
What is the SICK-IgniScan?
The SICK-IgniScan is a sensor array developed by Northeastern University graduate students James Tukpah and Ben Cometto at the RIVeR Lab. It combines a lidar sensor and a hyperspectral camera to measure the moisture content of forest vegetation from an autonomous vehicle in real time. It won first place in the SICK $10K Challenge in 2026.
How does hyperspectral imaging detect wildfire fuel moisture?
A hyperspectral camera captures light across many continuous wavelengths, far more than the three channels in a standard RGB camera. Plant material reflects different wavelengths depending on its water content, so the camera can classify moisture levels in vegetation by analyzing those spectral signatures without physically collecting samples.
How is this different from existing wildfire detection systems?
Most existing systems detect or track a fire after it has started, using thermal imaging, cameras, or satellite data. The Northeastern system measures fuel conditions before ignition, giving fire managers data on where risk is building rather than where a fire is already burning.
Is this system available for deployment by fire agencies?
Not yet. As of June 2026, the system is in field testing on Northeastern's off-road vehicle. No deployment timeline, commercial availability, or agency partnership has been announced. The team has stated the goal of eventual real-world field use, but that remains unconfirmed.
What is the SICK $10K Challenge?
An annual collegiate robotics competition run by SICK Sensor Intelligence, a global sensor company focused on industrial automation. Student teams from U.S. and Canadian universities compete. The Northeastern team won first place in 2026 for the SICK-IgniScan system.
Conclusion
The SICK-IgniScan is a prototype, not a deployed product. But the problem it is aimed at is real, the approach is technically coherent, and the gap it is targeting, pre-fire fuel monitoring at scale, is genuinely underserved by current tools. What happens next depends entirely on whether the system can hold up in actual field conditions and whether the people who manage wildfires can find a path to funding and deploying something like it.
With 30,000 wildfires already counted this year, the incentive to figure that out is not abstract.
0 Comments