Jul 3, 2008

Industry invests dollars to allay labor fears

Over the last three years, the Washington Tree Fruit Research Commission has invested $670,000 to back the research efforts of one company: Vision Robotics of San Diego.

That’s a lot of eggs to place in one basket – more spending this year will boost the number to $900,000 – but it shows commitment to a principle. Vision Robotics wants to mechanize orchard and vineyard operations, including harvest, using robots.

“It supports the premise of our National Tree Fruit Technology Roadmap,” said Jim McFerson, the commission’s manager. “We have not wavered on that. If we don’t create a more attractive workplace, we’re truly doomed.

“We are attempting to improve the workplace by making it safer and more attractive to workers. We will always need people to work in our industry, and we want them to be able to earn a secure living year round. In essence, we are not eliminating workers. We are transforming the workplace and creating new job opportunities while improving the quality and efficiency of our production and processing activities.”

As McFerson sees it, the U.S. tree fruit industry has a choice: It can adopt technology or it can act like a third-world country. In the United States, there are huge problems that come with relying on unreliable, foreign, seasonal labor. The money being invested in Vision Robotics rejects this “third-world” choice.

While much attention has been focused on the need for a better way to harvest fruit, what McFerson likes about Vision Robotics’ approach is it addresses more than just harvest labor problems.

“Virtually every orchard operation requires hand labor,” he said. “Vision Robotics offers a broader portfolio than just harvest. It covers a whole range of other technologies.”

Thinning, pruning, precision spraying – all could be handled by robots. Moreover, robotic technology is not specific or restricted to one or two kinds of fruit, so it could be broadly applied and affect the entire fruit and vegetable industry and the way growers use workers. Once you accept the idea of robots, the tasks robots can do are endless.

In a way, that’s how Derek Morikawa, CEO of Vision Robotics, got involved with fruit harvesting.

Vision Robotics started in 1999, focused on developing consumer tools like robotic vacuum cleaners. Credit goes to Tad Batkin, president of the California Citrus Commission, for coming to Vision Robotics in 2004 and asking the question, “Can you make a machine that will harvest oranges?” Since then, the California Citrus Commission has invested more than $1 million with Vision Robotics to develop a way to do that.

Sometimes, solving a problem requires properly identifying it first. Morikawa identifies the key problem in fruit harvest as knowing where the fruit is.

More than 25 years ago, he said, experimental robots were developed that could pick fruit. A robot equipped with a camera could locate a piece of fruit and pick it. But this one-at-a-time, find-and-pick approach was very inefficient. The robot was unable to make good choices because it was seeing one fruit at a time. Dashing from one to another with no overall plan, it would pick 70 percent of the fruit and never finish a tree, Morikawa said.

Technology now in use in Europe has demonstrated that human pickers can be made more efficient if a human brain makes decisions and uses tools to carry them out. Putting pickers on a moving platform and equipping them with vacuum hoses to deliver apples gently to a bin improves worker efficiency and shows that apples can be picked without bruising them. Fruit can be thinned and trees pruned using similar harvest aids.

Eliminating the need to position and climb ladders takes a large step in the right direction, but the 11 software programmers and mechanical engineers that work for Vision Robotics have developed a different, more totally mechanized, solution. It is a two-robot solution.

The first robot, called the “scout,” exists in prototype form. It is a tow-behind trailer equipped with many cameras that combine their images to give what Morikawa calls “stereoscopic vision,” a 3-D picture of an orchard and the location of the fruit in it. Computers can handle huge quantities of data, so it’s “not a problem” to have a map of an orchard with every fruit on every tree located.

This photography operation can be done any time of the year and the data map used in a number of ways.

Robots equipped to carry out specialized operations would operate using data gathered by the first robot. If you don’t want a fruit where one exists, you could remove it as a blossom or small green fruit. You could examine trees for pests and spray where needed. You could use the data to select branches to prune.

When it comes time to pick fruit, you create the map to locate each one and know its size or color or defects.

The scout is the data-collecting tool. The map it creates can be used in all kinds of further operations. The concept was tested in Washington orchards last season and a prototype version of the scout will be deployed again this summer.

The next step after creating the map is developing tools to do specific tasks – and developing “rules” governing how they’re used. A robotic arm that wields pruning shears is half the equation; the other half is the software that tells the robot which limbs to cut. The same applies to fruit thinning and fruit harvest. The robot needs to know where the fruit is and apply “rules” that tell what size or color it is supposed to pick.

The robot also uses the location data to make decisions. Given where the fruit is on a tree, how can the robot pick it most efficiently? Like a worker deciding where to put the ladder to gather the most fruit, the robot uses the data to decide in what order it will pick the fruit.

“It’s just math,” Morikawa said. “Once you know where all the fruit is, it is easy to calculate the most efficient way to pick it.”

The harvesting robot he envisions for oranges is a multi-armed creature that uses all its arms swiftly and at once, without tangling them up. The arms would reach into the canopy and snip the stems off flush, just as humans do now.

Apple trees, on the other hand, can be grown in many shapes, and Morikawa thinks the fruiting wall concept will make finding and picking fruit much easier. Apple trees can be grown to meet the needs of the best robotic concept.

Olin College of Engineering in Boston is working on a mechanical hand that can pick an apple without bruising it, which is just about worked out, Morikawa said. Several projects are underway, evaluating bin-filling devices that lay apples into bins without damaging them. There are any number of concepts for belts and conveyors that gently move fruit from one point to another.

Vision Robotics is developing another machine for pruning grapes. Like the system for other fruit, it works by creating a three-dimensional map showing cane location and applies pruning principles that tell where to make the cuts.

Current machines work without that finer distinction. Usually, he said, machines are pre-pruners that cut canes back to a certain length, leaving 8 to 12 inches on which a crew of workers can make more precise cuts.

“We want to develop a machine that will precision prune as well as would a well-trained hand pruner,” Morikawa said.

Another interesting feature of the work at Vision Robotics is that it is entirely funded by industry associations. McFerson said funding from the Washington Tree Fruit Research Commission does “not commonly” go to companies and that land-grant universities and USDA have been the traditional providers of new technology.

The U.S. government has not been as actively involved in finding engineering solutions as it was during the 1960s and 1970s. Funding for that kind of research has not been there to support either the Agricultural Research Service or land-grant university engineers.

McFerson, looking at language in the newly passed Farm Bill, said there might be money for things they have asked for. He gives credit to USDA for working with specialty crop organizations to create the Specialty Crop Research Initiative – and many of its provisions were incorporated into the Farm Bill.

The USDA Cooperative State Research, Education and Extension Service has announced its intention to issue, in July, a “request for proposals” that will pave the way for $27 million in federal grants in five focus areas that were legislatively mandated in the new Farm Bill.

Among these five areas are “new innovations and technology, including improved mechanization and technologies that delay or inhibit ripening.” These grants must be allocated by the end of September.

Automation, long delayed, is coming to U.S. agriculture. The new Farm Bill provides hope for that, and both the Washington Tree Fruit Research Commission and the California Citrus Commission have been betting on it for the last few years.