"RoboCrop" Calculates The Odds of Success When Picking Fruit, Its Creator Says

Osaka Metropolitan University's Takuya Fujinaga is looking to ease the workload on fruit farms by teaching robots to pick delicate tomatoes — leaving humans only the more challenging examples left to grab.

"This moves beyond simply asking 'can a robot pick a tomato?' to thinking about 'how likely is a successful pick?,' which is more meaningful for real‑world farming," Fujinaga claims of the work. "This research establishes 'ease of harvesting' as a quantitatively evaluable metric, bringing us one step closer to the realization of agricultural robots that can make informed decisions and act intelligently."

The idea of having robots perform the back-breaking labor of fruit picking isn't new, but is a longstanding challenge. Fruit doesn't grow neatly or in always-predictable places, and must be handled carefully both to make sure the fruit isn't bruised and that the plant it came from isn't damaged. This is particularly difficult when it comes to soft fruit like tomatoes — which is where Fujinaga's research comes in.

Rather than trying to teach the robot to pick all possible fruits, Fujinaga's approach has it estimate how likely it is to succeed in picking a given fruit — and if the success projection is too low, skip it and move onto the next. Using this, the robot was able to successfully pick fruit 81% of the time — a big gain over current predictors. The system even showed the ability to "learn" from its mistakes, revisiting fruits that it had failed to pick front-on and tackling them from the sides instead.

"This is expected to usher in a new form of agriculture where robots and humans collaborate," Fujinaga says of the project. "Robots will automatically harvest tomatoes that are easy to pick, while humans will handle the more challenging fruit"

Fujinaga's work has been published in the journal Smart Agriculture Technology under open-access terms.