John Deere announced three more autonomous machines on the eve of CES 2025 on Monday, expanding beyond its trademark 8R self-driving tractor announced three years ago. It described a giant articulated dump truck (ADT) for quarries as well as a commercial mower and tractor for pulling a sprayer into orchards.
The traditional 8R tractor will be able to use a second-generation autonomy kit with 16 cameras for a 360-degree view, while the ADT is Deere’s first autonomous piece of construction equipment. At 34 feet in length and 12 feet high, it handles 92,000 pounds of material.
The autonomous commercial mower has 4 pairs of stereo camera and can operate for up to 10 hours on battery electric power with 21 kWh on board.
Meanwhile, Deere announced an autonomous diesel tractor for pulling a sprayer in orchards that uses nine cameras, lidar and other sensors. For almond trees, repeat sprayings are needed for 10 hours a day for up to eight times in a single season. The company will also introduce an electric version of the orchard-focused tractor, officials announced on Monday.
Deere officials showed off the four machines (including the 8R tractor with a new autonomy kit) to reporters and analysts at an outdoor November event in Gilroy, Calif., where the commercial mower made repeated passes on a grassy plot. Deere employees pointed out the value of the quiet machine for use in frequent commercial mowing on campuses and office parks.
The tractor pulling the orchard sprayer will eliminate the need for a driver, often a hazardous job.
In traditional orchard spraying, drivers often wear hazardous materials suits to reduce the impact of chemical sprays.
Despite its intentions of showing reporters the autonomous ADT moving about a field in Gilroy, the giant machine sat at rest with a video playing nearby to show how it might be used around rugged terrain in a quarry. Deere officials said the ADT wasn’t moving during the demonstration because it had shut down as expected when it performed its autonomy sequence, but they didn’t elaborate any further.
Noting the enormous amount of material needed for building an interstate highway, officials said it would take 1,900 ADT loads for a single mile of interstate or three truckloads of sand, gravel and stone for an average house. Using an autonomous ADT would save time for quarry workers, they said. Contractors are expected to need to build 96,000 new homes per day by 2030.
Deere has been a champion of the value of autonomous tractors amid a global shortage of farm workers and a growing need for more food as the global population increases to 10 billion people by 2050, up from 8 billion today. By 2033, ag worker employment is expected to decline byu 2%.
Deere CTO Jahmy Hindman told reporters the autonomous tractor is “the answer to labor availability” with every farmer “challenged to find labor do the work when needed.”
Deere is engaged in offering “autonomy in action across all verticals.” Autonomous Deere 8R tractors are deployed mainly in the upper Midwest, Hindman noted. He said the cost of a $500,000 -$600,000 tractor is being addressed by relying on a licensing model for the various types of jobs a tractor will perform
Blue River Technology CEO Willy Pell told reporters that the training data for Deere’s autonomous machines has been gathered over years with existing Deere pilots and other sources. He said the portion of synthetic data used in the data gathering process is “in the single digits.”
"Mechanical tolerances and vibration and temperature swings effectively limit how far you can sense at some distance accurately, so to solve this, we created what we're calling camera arrays. Instead of two cameras overlapping, we have many cameras overlapping, so each camera can correct the position and orientation of every camera, every single frame….This real-time calibration allows us to run the wider baselines and have more accurate depth and accurate range,” Pell said.
Matt Potter, director of robotics and mobility technology, explained: "So the way this works is when an object is viewed by two cameras, there’s a slight angular offset at each of the other spaces. So with knowledge of the system geometry, we can compute the range to that object using triangulation. One common technique for doing this is called rigid baselines stereo camera. In a rigid baseline stereo, the two cameras are separated and held positionally and orientation-wise from each other mechanically. To increase the sensing range of the system like this, we have to widen that baseline so widening that baseline makes it harder and harder to preserve those angular orientations between the cameras.”