When the citrus industry began to pursue the idea of a robotic harvesting system several years ago, it had plenty of labor and was looking at robotics to serve as a back-up if there was a shortage of labor, Ted Batkin, president of the California Citrus Research Board, said during the Washington State Horticultural Association’s annual meeting in Wenatchee.

But the harvester had to be economically viable, too. The Research Board calculated that the cost of picking fruit by hand was $18 per bin. In comparison, the cost of buying a robotic harvester from Vision Robotics Corporation in California and servicing the debt over five years, plus the cost of operation and down time for maintenance, was estimated at $17 a bin.

"It kind of knocked our socks off," Batkin said. "We had been looking at the costs of machines from Europe being $800,000 to $1 million per machine."

Thanks to advances in technology and changes in the concept, the cost of the Vision Robotics harvesting system, which includes a scout and a picking machine, is estimated at only $350,000 to $400,000.

In the past two years, since the Citrus Research Board embarked on the project, labor costs have jumped from $18 to $24 a bin. This year, the cost might be closer to $26, Batkin said. Meanwhile, due to inflation in the cost of developing the machine, the estimated cost of robotic picking has risen from $17 to just $18 a bin.

Cost saving

"We’re now looking at this as a labor-saving and cost-saving device as well as having a labor solution," he said. "We’re in the position where we not only have a back-up, but we have a system that could potentially save us a considerable amount of money and make the fresh-market industry competitive with their foreign competitors.

"These are the kind of technologies that we, as growers, think will enhance our profitability and keep us competitive," he said. "There will always be the person who’s afraid to try something, and that’s fine. Let everybody else make the mistakes and correct them. What’s important is we have enough innovators in our industry to move this forward. It takes that aggressive, innovative nature to be the first out there and not be afraid to fail. That’s the –mentality we have applied to this whole situation."

The Washington Tree Fruit Research Commission is also funding Vision Robotics to develop a robotic harvesting system for tree fruits. Derek Morikawa, chief executive officer of Vision Robotics, said tree fruits are more suited to robotic harvesting because the trees are smaller and the fruit within easier reach. Citrus trees are typically 24 feet high and picked by people on 18-foot ladders.

"As this technology develops, whether it be for apples, pears, peaches, plums, or cherries, we will probably be reshaping our crops to a fruiting wall," Batkin said. "There will be an adaptability curve as we move forward to make them the most efficient systems possible."

The Vision Robotics system consists of a robotic scout that maps the location of the fruit and a separate picking robot that picks the fruit according to the map.

Morikawa said it’s already been proven that it’s possible to pick fruit with a robot-and-camera system. "But in order to save money for the grower and have a real payback, you need to come up with a clever system," he said. "This idea of scouting first and knowing where everything is so you can pick efficiently and economically will be a breakthrough."

Crop estimate

Although the scout and harvester will work in tandem at harvest, the scout can be used independently for other purposes.

Batkin said the scout is a valuable part of the project because of its ability to assess the crop and estimate crop load and fruit size. "We spend a lot of money on crop –estimates. With the scouting system, you can run it through the grove 30 to 40 days before harvest and know exactly what you’ll have at harvest time."

It also allows the crop to be selectively picked based on size. Batkin said it’s possible that technology for nondestructively measuring the soluble solids content and the color of the fruit could be incorporated into the scout. The scout might also be used to look for pests, so that only infested parts of the orchard would need to be treated.

"The sky’s the limit," he said.

Morikawa said as his company developed the system, he talked to growers to find out their priorities. "We’re not interested in building equipment that stays under a tarp. It has to be something you find to be vitally necessary to use and addresses some significant economic problem for you," he told growers.

One of the greatest engineering challenges in developing the system is designing a commercially viable automated picking hand for the machine, and a group of senior students at Olin College of Engineering in Massachusetts has taken on the project.

Robotics professor Dr. David Barrett said Olin College operates like a trade school, and has a $2.5-million machine shop. By the time students reach their senior year, they are working at almost a professional level.

The students designing the picking hand began by picking apples in local orchards to understand the problems and to gain a cultural awareness of the problem they are trying to solve. It has to be something that people would actually use and not be too expensive, too complicated, or violate the way that people think picking should be done, Barrett said.


Initially, the scout and harvester will be towed by a tractor driven by a person, Morikawa said, but as growers become comfortable with it, it could be made into an unmanned system.

"The cost of paying the driver to sit on the tractor and keep it going in a straight line is a significant cost of operating the machinery," he said. "Eventually, there will certainly be a drive to make the machines more and more autonomous if and when we’re comfortable with the safety systems for other humans and creatures in the field."

The harvester would move through the orchard at around one mile per hour. Asked to comment on the safety of autonomous machines, Barrett said he thought a robotic tractor going through an orchard at one mile per hour would be a lot safer than a human traveling at one mile per hour.

Asked whether such systems could work on hillsides, Morikawa said even flat ground is not perfectly smooth, so assemblies for articulation and leveling are built onto robotic systems already. For slopes, those assemblies would simply need to be larger and longer, though that would add somewhat to the cost and complexity.