Young Minds Develop a New Technology to Prevent Roadkill

• Colorado is home to four aspiring computer science engineers who last year began working on a device that detects a deer in a vehicle’s path and sends out a warning to the driver.

• The students won their state’s Samsung Solve for Tomorrow award, which earned their school $12,000 to help pay for a working prototype of their Project Deer device as they continue the project this year.

• Automakers and suppliers are working toward a 2029 NHTSA mandate requiring all new light vehicles to detect pedestrians (and ostensibly wildlife) to deploy automatic emergency braking. The Project Deer team claims a unique approach using artificial intelligence to model thermal images of deer.

Colorado is a big, beautiful state, with ski resorts and lots of forest, and with the Rocky Mountains bifurcating east and west. No surprise then, that on its myriad highways, motor vehicles have experienced unfortunate contact with wildlife.

The state’s Department of Transportation reports nearly 4,000 vehicle crashes involving wildlife every year, resulting in injuries to humans and fatalities to wildlife, and an estimated $80 million in property damage. The collisions involve small mammals, elk, bear, and moose, though the vast majority are deer, the DOT says.

No surprise then, that Colorado is home to four aspiring computer science engineers who last year began working on a device that detects a deer in a vehicle’s path and sends out a warning to the driver.

Surprise: The four aspiring computer science engineers are high school students. High school girls in fact, enrolled at STEM School Highlands Ranch, a science, technology, engineering, and mathematics public charter school located in an unincorporated part of Douglas County and within the Denver-Aurora-Lakewood metropolitan statistical area.

“When we did research about, what problems are there? Especially in our community, we wanted to target where we live,” said Siddhi Singh, who selected the three other female students for the project after computer science teacher Tylor Chacon told her he would sponsor the project.

“We found out that wildlife collisions are huge burdens to the state, and there’s millions of dollars poured into making these fences and stationary devices,” Singh said.

The other students are Dhriti Sinha and Bri Scoville, also juniors this year, and Robyn “RJ” Ballheim, a sophomore. STEM School Highlands Ranch has a bit more than 1,400 students, of whom about 39% are female.

Last year, as three sophomores and a freshman, they won their state’s Samsung Solve for Tomorrow award, which earned their school $12,000 to help pay for a working prototype of their Project Deer device as they continue the project this year.

Like players on the varsity football team who prove themselves stars by their second year in high school and thus have a couple of years to make it to the state championship game, team Project Deer has this school year and the next to make it to the Samsung Solve for Tomorrow national finals, which awards up to $100,000 in Samsung tech products and classroom materials.

Automakers and their suppliers have been working for a couple of decades on night vision and infrared nighttime detection of pedestrians and wildlife. The National Highway Traffic Safety Administration is requiring all new light vehicles by 2029 to have the necessary equipment for automatic emergency braking as standard, but the 317-page final rule talks only about pedestrians, not wildlife.

One supplier, Magna International, already has a thermal imaging technology for pedestrians and wildlife, expected to be ready for market by the time the mandate takes effect. But the Project Deer team claims a unique approach with an emerging technology: Siddhi Singh worked with University of Colorado-Boulder PhD students during an internship to train an artificial intelligence model using thermal images of deer.

“We have two components,” Singh said. “One is the main processing and one is the camera. First, the camera’s going to take eight pictures every second and store it.

“We have a machine-learning model on the processor, which through all these images will run through a processor and will output deer in this image, deer in (another) image, send a signal to an LED light. Run all thermal images through this model, which will output deer and trigger a response.”

The team also hopes to outsource various images for the AI to process, like elk, bear, moose, and small mammals.

Estimated cost, including a special, $500 chip built for AI models, is $1,000 to $2,000, said Chacon, the teacher.

As an aftermarket device, it can be added to any passenger vehicle and still cost less than the luxury cars that are usually first to get in-dash pedestrian detection technology, and far less than the millions of dollars Colorado’s DOT has spent on stationary warning devices scattered along the state’s roads and highways.

Next phase of Project Deer prototyping is to attach the camera to a windshield arm, clamp down the microprocessor to the dash with small suction cups and test at a nature preserve with deer near the road, Chacon said.

“We’re aiming for 60 mph, highway speeds. Our chip is fast enough and our camera is fast enough—that’s why it’s more of a reality now,” because of the AI chip, he said.

Because this is an enthusiasts’ publication, you’ll want to know that early test vehicles will be Singh’s parents’ Honda CR-V and Chacon’s 2014 Dodge Dart, but probably his 2003 Chevrolet Trailblazer first, because it has a big windshield and better outward visibility. Ultimately, the Project Deer team hopes to test it in a variety of passenger cars and trucks.

“We don’t have unlimited funding,” Chacon said, “but if anyone wants to support us, they can send us a check or send us the parts or contact us at stepk12.org and consult our communications department… because it would be nice to have an in-classroom model and one for the car. Right now, it’s just one.”

Whether or not Project Deer receives any such outside support, the high school girls out to save deer from becoming roadkill have come a long way in two worlds otherwise heavily dominated by men: automotive and computer tech.

“It feels really surreal to me, because even when I started this project, I had no idea where it was going to go,” Singh said. “We were just a team of girls, working on this project after school, with, like, these $5 sensors. We didn’t know how far it could really go.

“So, I would say, passion goes a long way. If you’re passionate about something, you can make it happen. I feel like we were able to get this far because we all are passionate about this.

“This is definitely a project we want to continue into the future and keep making it better, keep refining it. This is a problem that’s really important to us, and it’s something we really care about.”

Have you collided with a deer or other wildlife on the road? If so, how did you, the car, and the animal come through? Please comment below.