Robotic platform boosts harvest efficiency in tests

A robotic platform for fruit picking has performed well in tests, increasing harvesting speeds and impressing employees who worked on the platform, according to a University of California agricultural engineer.

Stavros Vougioukas, a professor in the Department of Biological and Agricultural Engineering at UC Davis, has been researching the robotic platform in a project that started in 2016 and ended last year. The work received funding from the National Robotics Initiative of the U.S. Department of Agriculture.

Existing orchard platforms typically allow four to six people to harvest, but there can be variances between the amount of fruit coming in and the fruit-picking capacities of the employees, Vougioukas said.

For instance, one employee may be allocated a higher zone to harvest, while another is allocated a lower zone. Most platforms remain static, with elevations preset for the employees, but fruit in the canopy is not uniformly distributed.

"Actually, you don't even know the distribution. It changes," Vougioukas said. "In one row, the fruit could be high, low or just random."

So, how fast each worker picks really depends on several factors: his or her ability and energy level, and the speed of the platform. If the platform moves too slowly, some employees may be idle. If it goes too fast, they can't keep up and fruit will be left on the tree. The end result, Vougioukas said, is an imbalance in labor supply and demand.

To address the inefficiency, his team modified a standard platform, the Bandit Express, retrofitting it with hydraulic cylinders to change the elevation of the pickers.

"Carnegie Mellon University provided a vision system that would find the visible fruits in front of the platform, and then we also developed some smart picking bags," Vougioukas said.

The bags can measure, in real time, how much weight is inside, to track how fast the employees pick.

"Now, we had all the pieces of information we needed," Vougioukas said. "We had the distribution of the fruits, we had the individual picker speeds and we had a mechanical system that could take the pickers up and down at will, by using hydraulics."

That information went into a computer on the platform, which controls three elements in real time: the elevation of the front picker, the elevation of the rear picker and the speed of the platform.

With this information, the platform could vary the height of the pickers, so the person picking faster stays in a zone with more fruit, and the person picking slower remains in a zone with less fruit, Vougioukas said.

"Because this changes all the time, the computer may vary that in real time, and that way you could bring back balance to the supply and demand problem," he said. "It's controlled by the computer, because there's too much information to digest, and the picker's job is to pick."

The robotic platform collaborates with humans, he said, with the humans actually standing on the robot.

"They're working on top of the robot, because now the platform is a robotic machine," Vougioukas said. "It has sensors, it has a camera, it has actuators, it has a computer that's running advanced control software, and you can see the system as a machine and people working together."

Experiments were run at the university and then in commercial orchards. Vougioukas said his team used an apple orchard, comparing the robotic platform to a standard platform.

Overall, the automated platform saw harvesting speed gains of about 25%, he said.

"This means that by intelligently controlling the speed of the platform and the heights, we could get a significant increase in performance," Vougioukas said, noting that if only elevation were controlled, efficiency still improved 10-11%.

There was also positive feedback from the employees using the robotic platform. With the platform moving up and down, it was ergonomically friendly, Vougioukas said, meaning people were bent over less often and didn't have to reach for the fruit, being able to pick what was right in front of them.

Nothing has been commercialized at this point, Vougioukas said, as the project only concluded at the end of 2019.

"We will be talking to the manufacturer and see if there is an interest on their side to adopt the technology, especially the variable speed," he said.

The robotic platform isn't restricted to a specific crop, but rather to a specific tree architecture: Trees need to be trellised and flat, and specific varieties work best, Vougioukas said.

"The trees must be relatively uniform, the canopy must be a uniform surface and not very wide, because otherwise you can't really reach out and pick the fruit," he said.

Growers could hedge trees in established orchards, but the trees could become less productive as a result.

"It's not simple," Vougioukas said. "I don't think it's possible to just go into any orchard and hedge it so that it's nice and geometrical, and then you go and pick fruit. I think it needs to be specific cultivars and also training systems. You need to plant with automation in mind."

The ideal, he said, would be to establish new orchards so they can be picked from a platform. Having visibility and reachability means the orchard has geometrical canopy and could reap the benefits of either existing platforms or robots in the future, Vougioukas said.

There are also economics to consider: the cost to structure the orchard for automation, increased management and denser planting.

"So, it's complicated. It depends also on the yield of those trees," Vougioukas said. "You need to look into the yields coming out of those trees, the cost of establishing and maintaining the trees, and the benefit of automation."

Vougioukas said he has just started a new project with the platform at UC Davis, with funding from USDA under the National Robotics Initiative, to build a multi-arm, fruit-harvesting robot.

(Kathy Coatney is a reporter in Bend, Oregon. She may be contacted at kacoatney@gmail.com.)