If researchers achieve the goals they have set for themselves, apple and grape growers should see the day when their dormant pruning will be done by robots instead of people. Pruning is labor-intensive and accounts for about 20 percent of the cost of both grape and apple production, second only to harvesting.
Concern about future labor availability, rather than cost, is a key driver behind the $6 million, four-year project called Automation of Dormant Pruning of Specialty Crops. The project was approved earlier this year for Specialty Crop Research Initiative funding.
Leader of the multi-institution project is horticulturist Dr. Peter Hirst at Indiana’s Purdue University. Hirst has plenty of credentials for the job, being well known for his work in pruning, tree canopy structure, tree physiology, and fruit quality. But he passes on credit to Drs. Johnny Park and Avinash Kak, electrical engineers in the school of electrical and computer engineering at Purdue.
“Purdue has a reputation for excellence in electrical engineering,” Hirst said. “That’s a key component of this project.”
Park has worked on creating accurate three-dimensional reconstructions of real-world objects using laser data, and Kak has developed robotic vision systems and wireless camera networks, Hirst said.
The project’s goals are:
—To develop an imaging and decision system for creating 3-D reconstructions of leafless canopies and calculating optimal pruning points
—To develop a production-level robotic pruning system for grapevines and a decision support system and graphical user interface for apples
—To educate growers on how to use the system when it is developed
Another key partner in the project is Vision Robotics, a San Diego, California, company led by president Bret Wallach and vice president of engineering Tony Koselka. That company has already developed an advanced prototype of a robotic pruner for grapes that makes decisions about cuts and makes the cuts—but too slowly.
Hirst says that apple tree canopies are more three-dimensional than grapes, and the pruning decisions are more difficult to make. The project will work with the tall spindle apple system, which already uses a simplified renewal pruning system based on removal of the largest two or three limbs.
As Hirst envisions the project, the first step is to develop the imaging system—sensors and cameras that provide a Three-D picture of the location of all the limbs on a tree.
That visioning system was worked on in other projects that aimed to create a robot that would know the location of individual fruits on a tree and be able to pick them. But creating a mechanical hand that can gently pick fruit is much more challenging than working with pruning shears, Hirst said.
The robot must be “taught” pruning rules so it can make decisions. Those rules must be developed, and should be as useful in training human workers as in training robots, he said.
The full team
The three people from Purdue and two from Vision Robotics make up about half the project team. Also involved are horticulturist Dr. Jim Schupp, extension educator Dr. Tara Baugher, and sociologists Dr. Leland Glenna and Dr. Anouk Patel-Campillo, all from Pennsylvania State University; Clark Seavert, economist at Oregon State University; and Dr. Julie Tarara, research horticulturist with the U.S. Department of Agriculture in Prosser, Washington.
Hirst considers it a balanced team, with horticulturists, engineers, extension educators, an economist, a sociologist, and industrial developers. The horticulturists need to develop the rules that define optimal pruning.
The engineers need to develop the imaging and robot control systems. The economist and sociologists will examine the social and economic impacts of the autonomous pruning system. The extension people will work to help growers implement the new technologies.
A vitally important role is played by industrial developers who develop ideas and turn them into useful machines. Vision Robotics began field testing a prototype grapevine pruner in the 2008-2009 season and prepared a third-generation prototype for field tests in Australia this year. The company has worked on robotics and machine vision systems for apples and citrus.
The company’s experience in pruning grapes should provide a basis for work in apples as well as advance the grape pruning system toward successful commercialization.
“The current (grape) prototype scans the vines with front- and rear-mounted cameras and analyzes the images to distinguish canes, spur heads, and cordons,” Hirst said in the project proposal. “A simple set of pruning rules determine where to cut each cane. Because the scanning of each vine is complete before the cutting system reaches the vine, the pruning rules are holistically applied in order to optimize pruning quality.
“Hydraulic shears mounted on a robotic arm then cut the canes at the precise points calculated by the system.
“The pruning performance of the current prototype, however, does not yet equal that of manual pruning.”
There are five main limitations, he said.
—The vision modeling system is limited to only young, simply structured grapevines.
—The speed at which the prototype operates is slow and needs significant improvement to achieve cost requirements.
—The cutting system (shears, robot arms, system control) has limited dexterity that does not yet enable the pruner to effectively reach all cutting locations.
—The technical requirements make the overall system prohibitively expensive.
—Only a single set of pruning rules has been developed.
A project goal is to fix these problems and extend robotic pruning concepts into apples.