Agricultural equipment maker John Deere has announced its latest piece of autonomous farming kit: a package of hardware and software that combines machine learning with the company’s GPS-powered auto-steer features to create a “fully autonomous tractor.”
The technology to support autonomous farming has been developing rapidly in recent years, but John Deere claims this is a significant step forward. With this technology, farmers will not only be able to take their hands off the wheel of their tractor or leave the cab — they’ll be able to leave the field altogether, letting the equipment do the work without them while monitoring things remotely using their smartphone.
“This is not a demo. It’s not a concept machine. It’s something we’ve had in the field with farmers for years and will be taking to production in fall,” Deanna Kovar, vice president of production and precision ag production systems at John Deere, told The Verge.
This may seem like an unexpected breakthrough, but the farming world has arguably made more consistent progress with autonomous driving than automakers or tech startups, mostly due to the simplicity of the task at hand. Although plowing or seeding a field is certainly a difficult job — requiring farmers to navigate the contours of their land while operating complicated equipment — the driving component is relatively straightforward: operators follow set lines without having to worry about pedestrians or other road users.
Because of this, companies like John Deere have been able to automate many aspects of farm driving over the past decades. Mostly, they offer auto-steer systems which use GPS to locate and guide tractors. Farmers first map the boundaries of their fields, often using beacons or by driving around the perimeter, and the software then plots a route. The driver — sitting in the cab of their tractor — can then oversee this path and correct it if necessary.
“We’re not going from no tech all the way up to an autonomous machine,” says Kovar. “John Deere’s AutoTrac solution has taken the job of steering in the field out of the operators’ hands for almost 20 years now.” Today’s announcement, she says, builds on these solutions.
“We’re not going from no tech all the way up to an autonomous machine.”
The big difference with this new technology is that drivers will now be able to set-and-forget some aspects of their self-driving tractors. The company’s autonomy kit includes six pairs of stereo cameras that capture a 360-degree view around the tractor. This input is then analyzed by machine vision algorithms, which spot unexpected obstacles.
“All [farmers] need to do is transport [their tractor] to the field, get it set, get out the cab, and use their mobile phone to ‘swipe to farm,’” says Kovar. “And every eight hours, they return to give it fuel and move it from field to field.”
Although John Deere is presenting this as an autonomous system, it’s worth noting that there are humans in the loop, and not just farmers. When the company’s algorithms spot something unexpected, images from the cameras will be sent to “tele-operators” — essentially a call center of third-party contractors who will manually check if the obstacle is a false positive or if the problem has resolved itself. If it’s a real issue, they’ll escalate things to the farmer via an alert on their mobile app. The farmer can then view the images themselves and decide if they want to plot a new course or check the situation in person.
“We’ve trained the algorithm to know that those are birds flying, you don’t have to stop for birds. But if you have, say, a dog in the field, then we’ll stop,” says Kovar. “We don’t want to always alert the farmer because this could be two in the morning. Part of the value of autonomy is allowing farmers to focus on other tasks.”
This system won’t be able to handle all aspects of tractor work, though. Right now, John Deere is focusing on the job of tillage — preparing soil for cultivation, either by turning over the earth, removing crop residue, or plowing this material back into the field to return nutrients to the soil. This is a “competing priority” job that’s usually done around harvest time, says Kofar, meaning farmers may set it aside in favor of more pressing tasks. That makes it a perfect target for automation.
Of course, the proof of the pudding is always in the eating, and despite years of testing, there will no doubt be teething problems when it comes to using this technology in farms. Driving a tractor isn’t just about steering round obstacles, and farmers also have to check that their equipment is working and adjust it to environmental changes. Kovar says the company’s software can monitor some of these variables, like checking that individual shanks on tillage tools are still operational, but there are bound to be other issues.
The company will be selling its new autonomy package as equipment to be retrofitted onto a number of its more recent tractors. But it has not released pricing — either upfront costs or annual subscriptions (which it charges for its autosteer products). The underlying equipment, though, is already extremely pricy. A John Deere 8R tractor and chisel plow used for tillage will set farmers back hundreds of thousands of dollars. There’s also the contentious issue of right to repair. Deere has been criticized strongly for locking farmers out of their own machines, and adding more computation will only speed this trend.
As Kovar says, though, this is just another step in John Deere’s journey towards ever-greater automation in agriculture. “This is a huge fusion of all of the technologies that agriculture has been leveraging for a couple of decades now,” she says. “There’s tons of opportunity for autonomy to stretch all the way through the production cycle, and at John Deere, we’re committed to that.”