The Washington Tree Fruit Research Commission has given the go-ahead to a California robotics company to begin developing a robotic apple harvester. When engineers at Vision Robotics Corporation in San Diego began developing a robotic fruit harvester for the citrus industry, they recognized that a similar harvester could work for harvesting other types of fruits.

The machine will have two parts: a scout that moves through the grove or orchard first to produce a three-dimensional map showing the position and size of all the fruit on the trees, and a multi-armed harvester that will pick the fruit according to the map. Though the hardware required to harvest big, bushy 18-foot-tall citrus trees will be different from that needed to pick fruit from high-density orchards, much of the software will be the same.

The Tree Fruit Research Commission is providing $283,000 in funding to develop an orchard scout, while the Citrus Research Board is paying a similar amount for Vision Robotics to develop a citrus scout. The synergy between the two has minimized development costs.

By next year, Vision Robotics should have a prototype of the apple scout to demonstrate in orchards in the hopes of attracting private investment to develop the complete harvesting system. The scout will be ready to go into production in 18 months, and the harvester should be ready 18 months after that, if the financing goes well, according to Vision Robotics’s chief executive officer Derek Morikawa.

His company is interested in a new processing chip named “cell” that will be used in the new Sony PlayStation. For the type of calculations that Vision Robotics must do for its vision systems, the cell chip is 20 times more powerful, though ten times more expensive, than existing microprocessors.

Confidence

Morikawa and engineer Tony Koselka visited Washington State this fall to look at the various types of environments where a tree fruit scout and harvester would need to work, and were pleasantly surprised by the consistency of the trees and the fruit in the orchards they visited.

“My overall impression is a greatly increased confidence that we will be able to handle a pretty good percentage of orchards—at least those that are not the traditional old trees,” Morikawa said. “The caveat to that is that apples, pears, and peaches clearly need to be handled more delicately, so that part needs to be solved.”
The robot will have multiple arms that reach into the trees to pick fruit on both sides of the drive row. Morikawa believes it will be able to harvest apples in an average orchard for $15 a bin. That includes both operating and capital costs and assumes that the equipment will be used 16 weeks per season for five seasons. The machine should be able to pick 50 bins of apples in an eight-hour shift, but it could work day and night.

Morikawa figures the cost of picking from trellised trees might be a third less than picking from full trees, which would make robotic picking more economical to the grower.

He envisages that a large grower-packer operation might own one and move it to various orchard locations as needed, or that fruit industry service companies might rent them to growers.

Estimate

The scout might be used separately from the harvester during the growing season to map the fruit and estimate the crop. It might just sample some of the trees, perhaps every tenth or fiftieth tree, Morikawa said. At harvest, the scout would go through the orchard just before the picker, or with it, to map the location and size of all the fruit to be picked.

The final version of the scout should be able to recognize both the red color and background color of the apples in order to pick them at the right maturity. He does not expect that the scout will have trouble distinguishing green Granny Smith apples from the surrounding foliage.

Morikawa said it will not do a perfect job of color picking because it won’t spin the fruit to see all sides, but he feels sure it will perform at least as well as human pickers do now. He believes internal quality sorting is best done at the packing house, rather than on the tree.

The robotic system would also be able to do other orchard tasks, such as selective pruning, thinning, or even spraying. The company has already developed a robotic wine grape pruner, which will enable grape growers to cut their labor costs in half for pruning, Koselka said. That machine will pay for itself in two to five years.