Plowing through code: Robotics student cultivates innovation with agricultural robots

In a brown dirt field, wet from irrigation and washed in early morning sunshine, a curious metal contraption moves like a mini tractor.

As large as a love seat, and set on caterpillar tracks, the machine hums and zaps as it rolls at a snail’s pace over rows of small pines at the U.S. Forest Service tree nursery in Coeur d’Alene.

“This machine can potentially save us a half million dollars annually in manual labor costs across our six nurseries,” said retired Forest Service Senior Research Scientist Kas Dumroese, M.S.’86, Ph.D. ’96.

The machine, a weeding robot developed by this year’s University of Idaho robotics team, is designed to kill weeds in the nursery’s seedling beds. The team is comprised of graduate and undergraduate students and is based at the North Idaho College campus in Coeur d’Alene.

The Forest Service nursery in Coeur d’Alene grows conifer trees for replanting efforts following fires or logging activities. Each year, between 4 and 5 million seedlings are shipped from the nursery. The tiny trees go to federal forests in Montana and North Idaho. Field grown nursery seedlings — called bareroot seedlings — are also earmarked for the Colville National Forest, the Okanogan- Wenatchee and parts of the Umatilla National Forest. Trees are also grown for the Bureau of Land Management, Coeur d’Alene Tribe and Nez Perce Tribe.

Like rows of tomatoes, the small seedlings must compete with weeds in their beds, and usually people are hired seasonally to keep the Coeur d’Alene nurseries, two, 5-acre bareroot plots, weed free.

This is where the robot comes in.

In a collaborative effort between the USFS and U of I, Dumroese and robotics instructor John Shovic developed a plan for an artificial intelligence weeding machine that would use rechargeable batteries and could systematically remove weeds from the plant beds day or night.

“It can basically do its job, weeding the beds 24/7 as long as the batteries are charged,” Shovic said.

Garrett Wells, a doctoral student, developed the robot along with graduate student Brent Knopp and undergraduate Kevin Wing.

The robot travels rows via a programmed route while following data to differentiate between seedlings and weeds. It uses a moveable arm and electricity to “zap” weeds, killing them with a more than 95% efficiency, Wells said.

“This three-foot wide autonomous wheeled robot uses artificial intelligence to scan, identify and precisely locate weeds within up to a half inch,” Wells said. “Its visual recognition identifies weeds and makes contact with individual weeds to deliver a five-kilovolt jolt with no soil disturbance.”

Using electricity to kill weeds is an alternative to using herbicides. It comes at a time when the agriculture industry faces herbicide resistance issues and a shift to more environmentally friendly growing processes.

“It’s more sustainable and better for the environment,” Wells said.

It’s all about sustainability, Dumroese said.

“We wanted a system we could use to kill weeds that would reduce our labor costs and reduce the use of petrol chemicals,” he said. “We wanted to maintain the impact while being more environmentally conscious and less resource driven.”

More than 20 Forest Service nursery employees attended a July demo of the robot, watching as it rolled along the aisles of tiny seedlings. Its three cameras scanning the rows, as its killer arm, complete with electrical probe, slid back and forth on the metal frame — called a gantry. Wells said his team is still working on a name for the robot, and so far, “Jethro” is the top choice.

Nursery manager Aram Eramian, MS ‘86, was pleased with Jethro’s work.

“We’re looking at this as a cost savings,” Eramian said. “You don’t have to feed it or fill it. You just power up the batteries, and it can run all day, killing weeds in these beds with 90-plus percent efficiency.”