NDSU research aimed at changing weed management

Weed management is a top priority for farmers, and finding ways to eliminate weeds without using herbicides, while also potentially saving money, is a significant challenge. Researchers at NDSU involved with the NDSU Food, Energy and Water Security initiative are developing a system that can perform both tasks.

Two faculty involved in FEWS research, Sulaymon Eshkabilov, NDSU assistant professor of agricultural machinery engineering, and Paulo Flores, NDSU assistant professor of agriculture and biosystems, have collaborated on a site-specific weed control system.

The project is called “Developing a smart soil cultivator for in-season weed control” to support site-specific weed control, and its goal is to utilize a drone imagery-based weed control prescription to automatically and individually actuate shanks on a soil cultivator so it will make soil contact only in areas of the fields where weeds are present.

“Our proposed mechanical SSWC system can offer farmers a reliable and efficient solution for weed management, particularly when targeting herbicide-resistant weeds that skipped herbicide applications, but also can be a good fit for organic farmers,” Eshkabilov said. “The system offers energy (fuel) saving, limited soil disturbance (less soil moisture loss), and reduces wear of operational parts of the tillage equipment and tractor.”

At its current configuration, the mechanical SSWC system comprises a main frame, rows of shanks equipped with sweeping blades and an electronic control unit that instructs the blades when to move. Each shank can rotate on its mounting point, and the blades are lifted or lowered by small motorized actuators. The system utilizes a prescription “weed map” loaded into the robot’s computer to identify the location of weeds. By default, the c sweeps stay in the up position, without touching the soil. The only time the sweeping blades would make contact with the soil is when they pass over an area with weeds, as indicated by the Rx map. The actuation movement of the shanks resembles fingers pressing keys on a piano.

“Sulaymon and Paulo are demonstrating how cutting-edge robotics, advanced imaging and artificial intelligence can tackle the everyday challenges producers face,” said. Frank Casey, associate director of the North Dakota Agricultural Experiment Station and leader of NDSU’s Food, Energy, and Water Security initiative, which funded the project. “Their novel solution helps manage herbicide-resistant weeds and gives organic farmers a much-needed non-chemical option. This is the kind of practical innovation that keeps North Dakota producers competitive while reducing inputs and environmental impacts, and it is exactly the type of collaboration between agronomy and engineering that FEWS is meant to accelerate for agricultural technologies that truly benefit our producers.”

Eshkabilov’s research background includes work with robotics and automation, wireless sensing systems and data-driven agriculture with AI applications used for decision support systems, crop growth monitoring and control systems.

“This site-specific mechanical weeding system is well aligned with my robotics and automation-related research domain,” Eshkabilov said.

Flores has been working to integrate drone imagery and commercial-sized sprayers to implement SSWC for several years. One day in 2023, a local farmer approached him to see if it would be possible to implement a similar approach to sugar beets, but using site-specific tillage (SST) instead of spraying. A grower asked Flores to create a prescription map that could be uploaded to a tractor’s display, so the soil cultivator is only dropped in areas where weeds are present. The grower stated that there were just certain areas of the fields populated with weeds; by cultivating these spots, he would be able to cover each field faster, save fuel, and keep moisture in the ground by reducing tillage.

“When the grower approached me to collaborate on this project, I knew that my team could tackle the weed mapping, Rx map creation and implementing the SST using the whole width of the farmer’s soil cultivator. As I visited the farm and saw that the soil cultivator had only one shank per row, I began to think about how we could develop a more efficient system than cultivating the entire width of the soil cultivator all at once. Applying what we know from sprayers, where we can control individual nozzles across the boom, I thought that we can control the individual shanks across the soil cultivator and drop them on the ground only we have a weed, using a Rx map created from the drone images.” Flores said. “The only problem with that idea was that I did not know how to implement that from an engineering perspective.”

The need for an engineering perspective led him to approach Eshkabilov to collaborate on the project.

Eshkabilov proposed designing a mechanical tillage system equipped with intelligent, piano-key-like actuators that could raise and lower individual shanks based on the presence of weeds. A weed prescription map would guide the site-specific tillage system.

In March 2025, NDSU’s Agrimechatronics Lab purchased a robot, AMIGA, and built a prototype of a three-row, site-specific weeding system that utilized linear actuators. A U.S. Patent for the system was filed shortly thereafter.

“We tested the prototype in both laboratory and field conditions, identified several issues and then built a second prototype,” Eshkabilov said. “After further testing and refinement, we now have the third version of our three-row mechanical site-specific weeding system.”

The first step to implement the approach in the field is to use a drone to collect hundreds/thousands of images from the field, which are then stitched together into a single, high-resolution picture map. The resulting image quality is sufficient to reveal even small weeds throughout the area. The drone also records highly accurate GPS coordinates — often with sub-inch precision — for each picture, which allows one to determine the precise location of every point on the map.

To improve the efficiency of the Rx map, the Flores lab is working on training a machine learning model with a large set of images. The system can learn to identify and locate weeds across the field. These weed locations are converted into a prescription map and uploaded to the tractor’s cab computer. The tractor, typically equipped with an RTK GPS receiver, provides positioning accuracy like the drone imagery — again, usually sub-inch.

As the tractor drives across the field with the field cultivator raised, the cultivator’s sweeps automatically lower to engage the soil only at the precise locations where weeds have been detected. This targeted action occurs continuously as the tractor moves through the field.

“My vision for the system is to develop a retrofit kit that can be used to convert regular soil cultivators into smart soil cultivators using the technology that we are working on,” Flores said.

The third version of the site-specific control weed system saw some adjustments and improvements.

“The third version of the system has a new capability of adjusting soil engagement depth,” Eshkabilov said. “What we learned from the first two versions of the system is that due to uneven terrain conditions in the fields and the light weight of the robot, shanks were not engaging with the soil sufficiently well in some areas. Therefore, we came up with a new version of the system. In addition, we added a load cell to the system to sense soil conditions. Moreover, in parallel, we have been working on implementing alternative, faster actuation and smart control units for individual shanks of the mechanical SSWC system to enhance its accuracy and effective field capacity.”

Eshkabilov and Flores presented the system and its performance characteristics at the Autonomous Nation Conference in 2025 and ASABE in 2025.

“The future of the system is very promising for organic and conventional farmers,” Eshkabilov said.