A shiny, metallic one-armed robot wheels along the rows of red pepper plants, its video camera bobbing up and down to spot the ripened fruit. When it zeros-in on one, the arm, equipped with a gripper distantly resembling a human hand, stretches out and grasps the pepper, while a quick swish of a knife cuts it off the vine. The bot may not look like the elegant C-3PO from Star Wars, but don’t shrug it off as yet another dummy automation. This pepper-picker is a new breakthrough in modern engineering: a sophisticated AI-equipped machine, built by the leading Israeli and European roboticists, that may one day pick fruit for your snacks and veggies for your dinner — a task way more difficult and important than it appears.
In Europe, Israel, and elsewhere, finding agricultural workers is getting harder, explains Yael Edan, professor at Agricultural, Biological and Cognitive (ABC) Robotics at Ben-Gurion University in Israel.
Bringing robots into the field or greenhouses can help deal with many problems associated with farming today.
“It’s a hard job, done under harsh conditions, high temperatures, and humidity, whether in the fields on in greenhouses,” Edan told Digital Trends. Plus, it’s seasonal and only serves as temporary employment, “so nobody wants to do it.”
And because fruit tends to ripen all at once, while humans can only toil so many hours a day, a good chunk of every harvest simply falls off the trees and rots, releasing the infamous greenhouse gasses instead of gracing our tables.
Which is why experts, like Edan, are turning to robotics, artificial intelligence, machine learning, and other emerging technologies as a way to combat the demands of farming in the future. High-tech tools are being tested on everything from simple harvesting tasks to sophisticated wine production and pollination. While there will always be a need for human laborers — at least, in the near future — technology could help fill the voids where farmers cannot.
Bringing robots into the field or greenhouses can help deal with many problems associated with farming today. Machines can work around the clock, including even at night with sufficient lighting. With some tinkering, they can be adjusted to tolerate heat and humidity without overheating or taking breaks. As a result, picky foodies would receive high-quality peppers, tomatoes, cucumbers, and other produce, harvested at their peak rather than too hard or too mushy. More fruit and vegetables will actually end up on our plates, because less would go to waste. The more efficient and economical process may also help balance the ever-increasing food prices too.
It’s easy to automate an assembly line, but Mother Nature doesn’t work like that.
But robots can do more than harvesting. BGU researchers are working on other “agribots,” such as drones that can pollinate flowers instead of bees, and smart sprayers that can calculate exactly how much pesticide should be spewed onto grapevines to prevent disease. Both projects could impact the future of food safety and security: As more crops move from fields into greenhouses, whether because of climate change or other reasons, pollinating can be a challenge. Plus, in the past decade bee populations around the world have been declining, because of the infamous colony collapse disorder, which poses a real threat to food supply.
The United Stated Department of Agriculture estimates that bees pollinate 80 percent of flowering plants and about 75 percent of the nuts, fruits, and vegetables Americans eat. Having robotic bee alternatives may one day come really handy if we want an uninterrupted vegetable supply. And with surging health concerns over pesticide overuse, smart sprayers can reduce the number of harmful chemicals dumped not only on grapevines, but any produce we eat and the soil it grows in.
It seems that with so many benefits, we should’ve automated husbandry decades ago. But that’s trickier than it seems. Unlike other industries, agriculture remains the unconquered frontier for robots, and for good reasons. Humans can learn to spot the ripened fruit so easily, but for robots it’s a complex mathematical and spatial task, requiring machine learning and artificial intelligence, and still hard to computerize.
It’s easy to automate an iPhone assembly line or a car conveyor belt—because sizes, lengths, locations, and parts always stay the same for a given model. The holy grail of mass-produced goods is being able to repeat the same action thousands and millions of times. A screwdriver-wielding robot at a Toyota factory is programmed to place its instrument exactly at the same position on every car of the same make. And every time it raises the screwdriver, the screw is there, waiting to be screwed in.
But Mother Nature doesn’t work like that. It is utterly random. Even when produce is grown in greenhouses where many conditions like temperature, light, and humidity are controlled, exactly where each plant will decide to bloom is completely unpredictable. Unlike with a car assembly, you can’t program a shrub to flower at specific coordinates, and thus you can’t program robots to look for peppers at any exact locations.
“We don’t expect these robots to completely replace humans in the fields. We expect them to help with tasks humans can’t and don’t want to do.”
As a result, the simple action of fruit picking varies drastically, depending on the location of the fruit, its size, shape, and leaves it can hide under. The machines must be intelligent enough to recognize peppers or cucumbers by appearance, which, depending on the specific variety can differ in sizes, shapes, and color. The robots also need to learn to peek behind or under bunches of lush foliage, and they need to know how to handle their crops gently so they don’t squish them into a pepper puree or don’t pull out the entire plant along with one fruit.
To top it all, Edan says, traditionally there’s been little money for agricultural innovations, so robotic husbandry has been quite literally left out in the boondocks. But now that paradigm is changing. As part of the European Union’s SWEEPER collaboration with Dutch, Swedish, and Belgian researchers, Edan’s team is now testing the pepper-picking robot in some greenhouses in the Netherlands.
“It took hundreds of thousands of pepper pictures—huge databases of pictures to get the machine learning algorithm to recognize the peppers properly,” says Polina Kurtser, a Ph.D. student at Edan’s lab. “But now the computer can handle that.” The team is testing different harvesting strategies—cutting fruit off versus sucking it in with a vacuum, and other methods.
Other agribots are also sprouting roots. Shai Arogeti, Yael’s collaborator at BGU, is working on a pollinator drone. Still in early stages of development, the drone neither lands on flowers nor looks like a bee, but it can quite literally create enough buzz to achieve the same results.
Some plants, like tomatoes, don’t require bees to pollinate them—they can pollinate each other nicely in the wind. So Arogeti has been testing a bot that can fly between the rows of greenhouse plants, creating a gentle breeze with its spinning blades and blowing plants’ pollen around. The task deals with a similar precision challenge—the bot has to be smart enough to fly in between the lush vegetation without crashing into it. To prevent the bot mowing down the plants, it is tethered to “a base” and can only fly straight through the isles,” but the robo-pollination prove of concept works.
Another computerized helper in the works is the pesticide spraying robot—the brainchild of Ron Berenstein, formerly at BGU and now at the University of California, Berkeley. To keep fungi, insects, and other crop pests under control, farmers spray millions of gallons of chemicals all over the world. Quite often however, they only need to sprinkle a small amount of the specific affected spots rather than showering it all over the patch and soil. Berenstein’s vine-spraying robot can identify grape clusters and suggest to farmers where and how much pesticide should be sprayed. The goal is to reduce the amount of toxic chemicals sprayed over the vines and soil, which ultimately seeps into the grapes themselves and into the wine made from them. Berenstein’s design also allows operating the robo-sprayer remotely, reducing the farm workers’ exposure to pesticides.
Will robot farmers cause massive loss of jobs in agriculture? Scientists don’t think so.
“We don’t expect these robots to completely replace humans in the fields,” Edan says. “We expect them to help with tasks humans can’t and don’t want to do,” enabling us to grow the best produce possible.
Exactly when will robots join us in the field is a bit less clear. “It’s a million dollar question,” Bernstein chuckles, adding that commercializing depends on funding, investors interested in the project, and other things. “I think five years is a reasonable estimate—because the technology exists.”
So, some time next decade your perfect-looking strawberry pack may indeed say “robot-picked.” Or even, “robot-pollinated and sprayed.”
- Optical illusions could help us build the next generation of AI
- The future of A.I.: 4 big things to watch for in the next few years
- Scientists are using A.I. to create artificial human genetic code
- A.I. translation tool sheds light on the secret language of mice
- Neuro-symbolic A.I. is the future of artificial intelligence. Here’s how it works