Project Evergreen provides an innovative approach to weeding large tree nurseries
Each year, Professor John Shovic awards the “Crown of Destruction” to the student who wreaks the most havoc in the Center for Intelligent Industrial Robotics in Coeur d’Alene. The tongue-in-cheek honor may go to Brent Knopp this year, but the solution arising from the damage could boost reforestation efforts in Idaho and nationwide.
The U.S. Forest Service called on Shovic and his students to develop an efficient solution to eliminating weeds from its nurseries. The U of I team embarked on Project Evergreen to build a machine that can autonomously move through a nursery, locate, identify and kill destructive weeds.
As Knopp and the Project Evergreen team built their robot designed to clear weeds from tree nurseries, challenges emerged. But when it comes to innovation, sometimes things must break before a breakthrough.
This project is everything I envisioned for graduate school. I wanted to apply machine learning to robots, and the scope of work in this project is great experience.
Brent Knopp
U of I graduate student
“The high voltage on the weed-killing mechanism was coupling with the frame, and we blew up the robot two days before a demonstration for six nurseries,” Shovic said. “That’s when Brent earned the ‘Crown of Destruction.’ We give it to the student that does the most damage — whoever does the best job Vandalizing stuff.”
Nurseries can spend up to $100,000 annually on weeding and frequently struggle to find workers willing to do the monotonous task. The U of I robotics group tackled the challenge by designing an autonomous, GPS-guided robot equipped with a camera and an AI system to identify weeds and a subsystem to kill the weeds.
Shovic tapped Garrett Wells to lead the project. The computer science doctoral student welcomed the challenge.
“It was definitely daunting to have a project that has so many different components, from hardware to software,” said Wells, who earned his bachelor’s and master’s degrees at U of I. “We’re really fortunate with the people that came together on the team.”
Knopp earned his bachelor’s degree in electrical engineering before starting his master’s in computer science, so the team relied on his expertise to develop the weed-killing component. After much brainstorming and experimentation, the Navy veteran designed an electric zapper to electrocute and kill weeds.
The explosive setback days before their demonstration eventually led to solutions. Redemption came last summer when the team demonstrated its prototype robot for the Forest Service.
“It felt really great to show them our work,” Knopp said.
The next steps for the project include adding an autonomous computer that identifies weeds to run the AI program. After that, all the computer systems must be integrated. Commercial production of the weed-killing robot is likely years away, but the students continue to expand upon Project Evergreen, gaining tremendous skills in the process.
“This project is everything I envisioned for graduate school,” Knopp said. “I wanted to apply machine learning to robots, and the scope of work in this project is great experience.”