A new approach to wildfire prevention - replacing labor-intensive methods with automated, accurate, and low-cost tech in partnership with AI2.
By Justin Horne
Wildfires are increasing in both frequency and intensity across the United States, creating an urgent need for better prevention strategies. While most funding and innovation have historically focused on fighting wildfires, a student team at the University of Washington’s Global Innovation Exchange (GIX) is working with experts from the Allen Institute for Artificial Intelligence (AI2) on predicting and preventing them before they start.
Accurate prediction relies on high-quality data, which has historically required time-consuming and expensive manual data collection. But Master of Science in Technology Innovation (MSTI) students from GIX, Ghea Suyono, Rishi Mullur, and Abhishek Singh, believe they can help. Their solution: a low-cost, automated sensor package that can quickly and accurately track fuel sources on the forest floor—improving efficiency and reducing time-intensive manual efforts.
The Problem: Outdated, Labor-Intensive Processes
The team’s work is supported by AI2, a Seattle-based non-profit founded in 2014 by Microsoft cofounder Paul Allen. AI2 works to develop foundational artificial intelligence with real-world impacts. AI2 is using artificial intelligence to predict wildfires, but those models rely on high-quality inputs, such as the volume and type of combustible materials on the forest floor. Currently, foresters manually record that data. “It’s an incredibly laborious task,” Singh explained. “Sometimes they go on a 3- or 4-day trek, build out a plot, and count and categorize the dry sticks in a given area.”
Collecting data is crucial for fire risk models, but the process is time-consuming, inconsistent, and often subjective. “Different foresters, with different heights and perspectives, might measure things differently. Data models need constant data to ensure objectivity and consistency,” he said. “They’re also dealing with a huge labor shortage. It’s very hard work, and that limits how many samples you can take per year.”
The Solution: Low-Cost Hardware with High Accuracy
The team’s research led them to focus on a few key factors: low cost, low human effort, high accuracy, and high consistency. Other researchers and companies have attempted to improve data-gathering in the past, using drones, LiDAR, and consumer-grade cameras. Often, they either didn’t work reliably, required skilled operators, or were prohibitively expensive.
“We studied everything, because it was an open question,” said Singh. “Do we want a zip line? Do we want a drone? Do we want some crawler that moves on the ground like a snake,” asked Singh. “Drones just don’t work well in forests. They require a skilled operator and they’re expensive."
“Our system is basically a robot that goes from one tree to another on a wire, and looks at the ground capturing the data,” said Mullur. “Similar things have been done with drones, but they were very expensive,” said Singh. “It’s about $50,000 minimum for airborne LiDAR—we've tried to achieve desired outcomes for this particular problem for under $500.”
The solution consists of two key components:
T-Rex: A compact device that captures environmental data, including humidity, temperature, and forest fuel distribution using high-quality cameras and sensors.
Raptor: A drivetrain system that moves the T-Rex across the forest via a wire strung between trees.
From Concept to Reality: Prototyping
“We’ve gone through over 30 prototypes,” said Singh. “It’s a lot of code, a lot of systems, and a lot of problems. But I don’t think you can get it right unless you go through this process.”
They faced unexpected challenges along the way. “A lot of the sensors weren’t up to the level of quality that AI2 required; we blew a couple up,” said Singh. “The placement and integration of the sensors was really difficult,” added Suyono. “Combining everything together was the real challenge.”
Field Testing and Industry Response
The team engaged directly with forestry professionals to refine their approach. “We learned very quickly that every forest has different needs,” said Suyono. “The other thing we learned is that people don’t know what they don’t know. It was crucial for us to build our prototype so that we could test it and get real feedback.”
One insight from these conversations was the need for flexibility. “We have to have a system that can easily move across a forest, that you can move from tree to tree, and that works in big areas quickly,” said Singh.
Collaboration with AI2 and the Future
The team has been working closely with their sponsor for nearly a year, iterating on their design and refining their technology. What started as a Hardware/Software Lab project at GIX program continued into the capstone phase of the program due to both AI2 and the team expressing a desire to continue working on the project.
“The experience has been incredibly educative,” said Mullur. “‘Support’ is an understatement in terms of how much they’ve done. Every week they check in to make sure we’re on track, or if they have something to teach us, or just to brainstorm.”
Ready to work on projects like these? Learn more about the MS in Technology Innovation or apply to be a student.
Interested in sponsoring a project? GIX capstone projects offer corporate sponsors up to six months of access to international and interdisciplinary teams of graduate students in the University of Washington’s project-based Master of Science in Technology Innovation. Learn more here.
Back to News
