Launching an agricultural AI startup isn’t as simple as building a robotic farmer or gathering data sets for computer vision systems. It requires identifying specific use cases, as well as which fruit and vegetable growers to work with. When Burro.ai built a robot that uses autonomous driving to ferry produce between workers, it chose to initially focus on table grapes.
By contrast, other agricultural AI startups like Ceres Imaging focused on high-value orchard crops like almonds and specialty crops like wine vineyards. Security drone maker Sunflower Labs is being used for automatic deployments on the perimeter of outdoor marijuana-growing operations.
Last month, the company began delivering its first commercially available robots to grape growers near Coachella, California. This builds on earlier work with the California Table Grape Commission and the Western Growers Association.
Now in its sixth generation, Burro ferries grapes from pickers in the field to packers putting grapes into clamshells or bags before the fruit gets loaded up and shipped to grocery stores.
The startup chose grapes as an initial application for Burro robots, CEO Charles Andersen told VentureBeat in a phone interview, because of a high concentration of growers between Coachella in Southern California and the San Joaquin Valley. The limited geographic expanse helped the Burro team manage trials without becoming overextended.
Burro’s robots are being created specifically for industries reliant on human labor to pick produce, the opposite of farms that depend on a highly mechanized John Deere tractor, for example. A group of six robots can support a human crew of up to 60 people.
“Beyond table grapes, we’ve done paid trials in blueberries, blackberries, raspberries, nursery crops, persimmons, and stone fruit. So there’s a whole other gamut of different crops where having digital train tracks with a small autonomous ground vehicle running around next to people, supplementing people and potentially eventually replacing them, becomes pretty compelling,” Andersen said.
The autonomous platform on wheels is laden with 22 sensors, including six cameras on the front and on the back. Four of the six cameras are devoted to depth detection. Burro uses computer vision for getting around, following a two-step process that trains the robot to understand the path from a packing table to a grape picking row. The training process is repeated for each row of grapes.
To test and train Burro, the team completed 4,000 hours of operation on farms last year. “We have seen most of the scenarios you can imagine, whether that be a cooking fire in a row [or] seemingly six-inch deep puddles that turn out to be very deep,” Andersen said.
He added that the company is initially focused on collecting the data necessary for a range of helper robots to navigate farming environments and augment or replace human activity.
“If you imagine a world in which you have something akin to Wall-E running around doing the work that people do today, how does that product actually start? In our heads, it starts as a little autonomous ground vehicle running autonomous routes that’s cloud-connected and modularly expandable; it begins as something like Burro,” he said.
The company hopes to begin with grapes in the U.S. and eventually expand to vineyards in other parts of the world or to new crops.