High-density orchards that are precisely planted and had good initial pruning for uniform growth will be most suitable for automated pruning.
Photos by Sally Colby

No one would argue the fact that orchards are labor-intensive. Three of the most critical aspects of orchard operation – thinning, pruning and harvesting – are extremely labor-intensive, and maintaining a skilled orchard labor force is challenging. Although developments in mechanically assisted thinning and harvesting have helped growers significantly, automated pruning hasn’t been addressed until recently.

Dr. Jim Schupp, associate professor of pomology at Penn State University, is part of a team working on the development of automated pruning. Team members include horticultural and robotics experts from Penn State, Oregon State University, Purdue University, the U.S. Department of Agriculture (USDA) and Vision Robotics Corp.

During a recent field day at the Penn State Fruit Research & Extension Center, Schupp introduced a project known as “Automation of Dormant Pruning of Specialty Crops.” He started with a review of proper pruning techniques for high-density apples, and used a simple diagram of a lateral fruiting branch to demonstrate where pruning cuts should be made.

“Historically, horticulturists would say that there are two kinds of cuts. There are heading cuts, where you remove a portion of the branch, and there are thinning cuts, where you remove a branch at its point of origin in its entirety. Now we’re making a third cut, a hybrid cut, and it has characteristics of both the heading cut and thinning cut,” explained Schupp. This hybrid cut creates a window for air and light movement throughout the tree, and the small duckbill stub that remains after the cut allows for renewed lateral branches. Schupp said that current hand-pruning techniques should help establish “rules” that will be necessary to develop algorithms for automated pruning.

“This new cut is what we’re going to ask the engineers to make,” said Schupp, who described the four basic orders of pruning cuts based on branch size. “First, if a branch is too large, cut it off. The second order involves branch density. One of the functions of pruning is to balance vegetative and reproductive growth. We need to remove some of that vegetation to get a good balance, so we remove large branches and also remove branches to get down to a certain number of branches for the size of the tree. The third order is branch angle, removing branches because they’re either too upright or too pendant. The fourth order involves branching complexity; as we move into high-density systems, we want a low order of complexity, with primary and secondary branches on a leader.”

Schupp said the team must be able to provide information about pruning cuts so the engineering team can program the computer that will operate the automated pruner. “They’ll have a vision system that ‘sees’ the canopy, a robotic arm, and a computer that transfers that information and says, ‘That branch violates a rule. If it’s too big, cut it off,'” he said. “Unlike harvest assist, we aren’t trying to gently cradle a delicate, bruisable piece of fruit from the tree to the bin in an efficient manner. It’s a dormant branch; all we need to do is figure out what we want to remove, then cut it off and let it fall to the ground.”

Although the project is still in the early stages of development, Schupp and his colleagues had to determine and define several key factors that could be transferred to computer technology. One was pruning severity: How many branches should be removed? “That’s a question a computer has to have an answer for,” stated Schupp. “We used the premise that any given tree needs a certain number of branches and a certain number of branches taken off. Somewhere along the way, you’ve heard that pruning is an art and a science. In the old days, with large 3-D canopies and multiple layers of branches and leaves, that was true. On those trees, it was very hard to say whether one cut was better than another. It wasn’t an exact science.”

At a field day held recently at the Penn State Fruit Research & Extension Center in Biglerville, Pa., Ph.D. candidate Tom Kon holds a diagram as Dr. Jim Schupp explains the pruning cuts that will provide a basis for computerized automated pruning.

In an automated pruning system, there’s no room for the “art” aspect of pruning. Cuts must be precisely defined so the computer can transfer accurate information to the pruner. Schupp said there are correct and incorrect pruning cuts, as well as levels of severity for those cuts. The initial phase of the project involves establishing pruning rules: What do the engineers need to know to make machines do this? “I think we can do that with this kind of tree,” said Schupp, referring to high-density plantings. “Along the way, it will make you a better pruner.”

As research continues, the team will consider crop loads, fruit size and distribution, and fruit quality. “Our hope is that we’ll have some tight numbers for the engineers to let them know what to measure, including trunk and limb diameter,” said Schupp. “The computer can do the calculations very quickly, and the camera can capture information, and based on that can prune to the correct level.”

The automated pruning system involves removing only primary branches from the leader. There are no other cuts being made, and no criteria for removing a limb other than size. “Removing those branches from the central leader might be all we have to tell [the engineers] to do for this kind of tree,” said Schupp, adding that grower awareness of simplified pruning techniques will eventually lead to a better automated system. “They might not need any more explanation. We can give the engineers a very simple target.”

Dr. Johnny Park, research assistant professor in the School of Electrical and Computer Engineering at Purdue University, is working on developing the computer aspect of the project. “Because canopies are 3-D, we need a sensing system that can capture 3-D information about trees,” said Park, describing the early work. “We’re working on very novel sensing technologies that will allow us to capture 3-D information, and then based on that, compute all of the attributes that are relevant to pruning rules. Then we decide which are the best pruning points, and the robotic system can prune based on the optimal points.”

Park said a working prototype, developed by Vision Robotics in San Diego, is currently being used to prune grapevines. Vision Robotics is also working with the team to develop the automated pruning system for orchards.

The author is a frequent contributor and freelance writer who farms and raises Great Pyrenees in south-central Pennsylvania. Comment or question? Visit http://www.farmingforumsite.com and join in the discussions.