How many of the apples we eat are harvested by machines? You may be surprised to learn the number is close to zero!

AIFS faculty and UC Davis professor Dr. Stavros Vougioukas shared the latest advances in robotic fruit harvesting at the AI Speaker Series.

Dr. Vougioukas explained why we’re seeing a labor shortage in agriculture, not only in the United States but around the world. As our global population grows and the workforce available to growers decreases, we face an urgent need to increase agricultural productivity with automation.

We can google “fruit harvesting robot” and see hundreds of pictures online – even videos of automated harvesters in action. But what most people don’t realize is that a farmer can’t just order an automated harvester for their apple orchard or strawberry field. Most harvesting robots for fresh fruits cannot yet compete with manual labor. Humans are good at finding ripe fruit to pick, and they’re really fast at picking it.

Dr. Vougioukas shared research conducted to identify how humans pick and what makes them so efficient.

The two most important performance metrics for fruit harvesting are:

• Efficiency: The percentage of fruits that are picked by the machine without damage
• Speed: The number of fruits picked by the machine per unit of time

Researchers have been developing robotic harvesters for over 40 years, but most of them are still not as efficient or as fast as a human picker. Actually, to justify their cost, robotic harvesters should be able to replace the work of multiple humans.

Efficiency can increase by improving perception to identify the fruits to be picked. This is complicated by foliage, branches and other objects that hide fruits from the robot’s cameras. Once a fruit is detected, it must be picked very quickly without dropping or damaging it. Then, the fruit must be conveyed to a bin as quickly as possible, with minimal damage.

How do we achieve these capabilities? Do we use different types of cameras? Do we build robots with multiple arms? Do we use algorithms, such as numerical optimization or AI, to control and coordinate the arms?

The talk went on to describe strategies that have been attempted and the limitations faced. Dr. Vougioukas concluded by sharing where research needs to focus to solve these problems, including work being done in his lab at the University of California, Davis.

We wrapped with an engaging discussion of ideas from the audience and Dr. Vougioukas’ thoughts on the feasibility of each.

What is the timeline for developing a fast, efficient fruit-harvesting robot that’s ready for real-world adoption? That depends on who you ask. A startup might give you a different answer than a researcher. Dr. Vougioukas opted not to over-promise, recognizing that these developments can take years.

FEATURED SPEAKER: Stavros Vougioukas, Ph.D., Professor and Vice Chair, Department of Biological and Agricultural Engineering at the University of California, Davis