John Deere tractors have been equipped with some automated features for 20 years, but now the company has unveiled a fully autonomous tractor ready for large-scale production that will be on sale in the fall.
What “fully autonomous” means is that a farmer can transport the tractor to a field and configure it for autonomous operation using a mobile app to start the machine and monitor its operation as it moves up and down a field while the farmer performs other tasks. About the only time the farmer must be present is to re-fuel the tractor or respond to the scene if the machine can’t make its way around an unexpected obstacle.
While the machine is plowing or planting, a farmer can leave the field to conduct the myriad jobs on a typical farm.
The tractor will be shown at a CES 2022 booth this week. “This machine is not for demonstration but for real work in real farm fields and with real customers,” said Deanna Kovar, vice president of production and precision agriculture production systems for Deere, in an interview with Fierce Electronics.
Deere provided video of its Deere 8R tractor pulling a Truset-enabled chisel plow as it tills soil after a harvest making it ready for the next season’s crops. A GPS guidance system helps in creating a geofence to keep the tractor on course, accurate to within an inch. The tractor is also equipped with six pairs of stereo cameras for 360-degree obstacle detection and to calculate distances. Images from the cameras are routed through a deep neural network via two Nvidia Jetson GPUs that help classify each pixel in 100 milliseconds to determine if the machine should continue or stop for an obstacle.
The mobile app allows a farmer to see as the tractor sees with separate views that could help decide whether help is needed to move a fallen tree or if the tractor can be routed around it. Kovar said one advantage of the real-time video is that a farmer can judge whether the tillage operation is deep enough, depending on the moisture of the soil that can vary over an average-sized farm of 4,000 acres.
Deere provides real-time telesupport to see what’s happening if a tractor encounters an obstacle. If it is only a shadow or a deer passing across the tractor’s path, telesupport might not bother the farmer. If a billboard falls into a field, telesupport might not let the tractor clear it and would alert the farmer on the mobile app.
The machine learning behind the autonomous capability involves training on 50 million images from farms collected over five years, Kovar said. If the algorithm detects something it hasn’t seen before, the machine will stop.
A tillage operation with a chisel plow is designed to inject nutrients into the soil as stubble from a previous crop is chopped up. “It’s really important in the fall to do this at the right level or it could impact next year’s crop,” Kovar said.
Full autonomy for a typical farmer could be a godsend, since many must spend 12 to 16 hours a day in the cab of a machine during the busiest times. Finding skilled labor to help is a challenge and there is a tight window to get the work done in any season.
“Autonomy is really the answer for us,” Kovar said, to meet mushrooming demands for food production that must grow by 70% in coming years to feed a growing world.
Deere’s fully autonomous tractor has been tested on farms in the Midwest and many farmers are already accustomed to riding in tractors with assisted features. The adjustment to full autonomy could be routine.
“I would say farmers will start by spending more time watching [the autonomous tractor],” Kovar said. “Farmers we have worked with are confident. They are probably not going to run off to Hawaii, but as they watch a field [with a fully autonomous tractor], it’s noneventful. Farmers are excited about autonomy.”
Deere has approached autonomous technology with the goal of improving efficiency in farming.
“We’re pretty confident this approach can be more productive,” Kovar said. “This tractor never stops to sleep or calls in sick.”
RELATED:Imagimob agritech demo shows off TinyML’s promise for edge IoT applications