Researchers advance orchard automation technology with AI pruning and thinning

Researchers are testing orchard automation technology that uses AI and robotics to improve thinning and pruning in fruit crops. Oregon State University (OSU) and Washington State University scientists, along with colleagues in mechanical engineering and computer science, are also exploring how these tools can apply to other crops, including berries.

In a Prosser, Washington field trial, a manipulator mounted on a utility vehicle used an “eye-in-hand” camera configuration to identify pruning points. Electric bypass shears then autonomously cut limbs, giving researchers valuable insights into the possibility of robotic pruning.

Advancing AI pruning research

Joe Davidson, assistant professor of robotics and mechanical engineering at OSU, outlined the research at the FIRA USA robotics in agriculture conference in Salinas, California. His work is part of the AgAID Institute, a USDA-NIFA-funded collaboration also supported by the National Science Foundation. Partners include:

• University of California-Merced
• University of Virginia
• Heritage University
• Wenatchee Valley College
• Kansas State University
• IBM Research
• innov8.ag

Davidson said the institute focuses on practical field trials rather than only large-scale tests. “We went to the field early to work on system integration and do preliminary field trials,” he said. “We wanted to identify where our ideas were viable and where we should go back to the drawing board.”

Researchers used advanced AI techniques to develop human-robot collaborative systems. They’re designed to support orchard workers, improve fruit quality and yield, and boost labor productivity.

Using data to strengthen automation

Davidson and his colleagues have collected extensive data from apple and sweet cherry orchards, focusing on high-density, planar orchard systems that favor automation. Their AI models rely on RGB “truecolor” images to guide pruning tools with precision.

Software trained with generative AI in simulations has proven more adaptable to varying orchard conditions, tree architectures and weather patters. However, cloudy conditions and shifting light still pose challenges.

Because vision alone is not always enough, researchers combine visual feedback with force feedback — essentially giving machines a “sense of touch” — to help perform precise cuts on small branches without damaging nearby limbs.

Creating digital trees for training

To train deep learning algorithms, researchers labeled images of dormant trees, a process that is both time-consuming and prone to error.

Davidson noted that digital models can accelerate this work. “There are tools out there where you can create digital trees, using realistic growth models with light distribution, carbon transport and all the things horticulturists care about study,” he said. “You can actually create digital environments where you can train robot controllers.”

These digital trees allow reinforcement learning, where robots improve performance through simulated practice. By segmenting orchard images, AI can isolate the tree to be pruned from the background, making the process more accurate.

Expanding orchard automation technology

Next steps include testing the system with professional pruners to refine rules and techniques. Davidson is also developing digital training tools for seasonal workers to speed up onboarding for pruning tasks.

Researchers are broadening their studies beyond apples and cherries to include blueberries, a significant crop in Oregon and Washington. They are also adapting pruning technology for other orchard tasks, including measuring tree trunk cross-sectional area, canopy volume, color changes and vigor for use in precision fertilization projects.

Looking at labor and grower perspectives

Alongside technical work, Davidson and his team are conducting interviews and observational studies with orchard managers to understand how labor interacts with automation. As co-lead of the labor intelligence effort at the AgAID Institute, Davidson spends much of his time studying how automation can complement the workforce.

By continuing to refine orchard automation technology, researchers hope to deliver tools that improve efficiency, reduce labor burdens, and ensure the long-term sustainability of fruit production.

— By Doug Ohlemeier, assistant editor