CNH self-driving tractors to dodge deer, fallen trees with AI

CNH Industrial first unveiled this concept autonomous tractor in 2016 and is now refining AI to help it avoid deer and fallen branches. (CNH)

If you’re not from a farming community, it might not seem like a big deal that tractors of the future will be autonomous. However, with self-driving tractors, farmers can be freed up from time-consuming tasks of planting, spraying and harvesting. 

The transition to self-driving tractors has been underway for a long time.  CNH Industrial released an 8-minute video in 2016 showing a self-driving tractor concept that even four years later seems revolutionary.  The driverless concept tractor in the video makes it way down rows in fields on a working farm in Kentucky and “is truly independent and driverless,” according to the narrator.

Such technology opens up the possibility of 24-hour farming.  It also could help solve the problem of finding enough farm labor.  Farmers everywhere are trying to solve the problem of making more food on the same, or diminishing, amount of land even as the global population and the demand for more food expands.

Fierce AI Week

Register today for Fierce AI Week - a free virtual event | August 10-12

Advances in AI and Machine Learning are adding an unprecedented level of intelligence to everything through capabilities such as speech processing and image & facial recognition. An essential event for design engineers and AI professionals, Engineering AI sessions during Fierce AI Week explore some of the most innovative real-world applications today, the technological advances that are accelerating adoption of AI and Machine Learning, and what the future holds for this game-changing technology.

The task of using artificial intelligence and sensors technology to drive an autonomous tractor isn’t that different from building a self-driving car, but there are major differences.  Farms are not regulated the same way highways are, and the potential for liability from a crash is not as great. 

Even so, there are trees, crop boundaries, buildings, fence posts and animals that a tractor must dodge.  “We are programming for obstacles both known and found during operation,” said Steve Caudill, director of the agriculture sector, Farm/Fleet platform manager at CNH.

Caudill is based in the Chicago area, but CNH is large Dutch-domiciled multinational corporation with $28 billion in revenues and 64,000 employees in 2019.   Its top competitors are John Deere, Caterpillar, Jungheinrich, Zap and AGCO, according to analysts.

RELATED: How John Deere got good at AI

The company’s concept vehicles like the one introduced in 2016 have been in operation for some time, but Caudill said “it will be a while” before the company produces a commercially-available model.  CNH doesn’t make its own semiconductors but has done some circuit board design and uses third parties for manufacturing and works with partners or buys off-the-shelf components. 

One trend emerging in development of AI chips is to provide inferencing capabilities so that an edge device like a tractor, car or industrial robot won’t be sending enormous amounts of sensor data back to a distant server in the cloud, data center or even the kitchen of a farmhouse. Cutting out such network trips can improve split-second decisions, which might matter if a tractor on a programmed pathway suddenly comes upon a fallen tree.

“You’ll notice in the video that we are building in the vehicle AI to evaluate if an obstacle has moved--such as a deer, another vehicle or a person—in order to reduce the incidents of a vehicle being stuck when there’s no network connectivity,” Caudill said. “It’s these kinds of practical conditions that we need to address before we have a commercially-ready product.”

Caudill said the CNH approach is not reliant on real-time cloud interaction but if there’s not a live connection and the autonomous vehicle cannot execute its instructions, “you are at an impasse until it has connectivity or you have to drive to the vehicle.  For example, if the vehicle encounters an obstacle and this obstacle is small fallen branch and you are spraying crops, you might be OK going over the branch, as you’ll do more crop damage by going around. But if the branch is large enough, that isn’t wise.  This is where real-time connections to the cameras make things more practical.”

Autonomous tractors could be designed to work in groups on a farm, perhaps with one manned tractor that interacts with several unmanned farm vehicles.  CNH foresees connecting a communication hub to weather data that help farmers direct autonomous tractors to areas away from storms. 

In some areas of the Ukraine, farms are so large that tractor can stay on a single row for an entire day before reaching the end of the row.  That kind of operation makes tractor autonomy seem ideal.

Steven Caudill will join a panel discussion on Friday  at Noon ET to discuss "How diverse companies are achieving IoT success" as part of Sensors Innovation Week, a free online event. For more information and to register, check online. 

Suggested Articles

Hydrogen refueling stations are limited in the U.S., restricting interest in use of fuel cell electric cars

Silicon Labs is providing the BT module needed for detecting proximity with another Maggy device

Test automation won't fix everything, but can help, according to an automation engineer. Here are five problems to avoi to improve chances of success