John Deere reveals fully autonomous ADT (VIDEO)

John Deere has taken its autonomous equipment development program to a new level with the presentation of an autonomous articulated dump truck (ADT), two tractors, and an electric mower at CES 2025. Building upon the tech stack from the first-generation autonomous tractor introduced at CES 2022, John Deere has expanded and refined the autonomous kit to produce an impressive, fully autonomous 460 P-Tier ADT nicknamed "Dusty".

Skilled labour shortage across industries inspires technology development

The driving force behind John Deere's autonomy development program is a shortage of skilled labour for key roles across the construction/aggregates, agriculture, and commercial landscaping industries. All of these industries must find solutions to efficiently address this skilled labour shortage. 

"This is the kind of challenge that we love to work on and help solve for our customers and we believe that autonomy is part of that solution," says Jahmy Hindman, John Deere's senior vice president and chief technology officer. "We're taking our tech stack, which is nearly three decades in the making, and we're extending it to more of our machines to safely run autonomously in these unique and complex environments that our customers work in every day. Just because a farm doesn't look anything like a rock quarry or an office park doesn't mean that we can't deploy the same or similar technology across those machines operating in those environments."

The ADT, the tractors, and the electric mower all operate with full autonomy. There is no one in the machine, and no one is controlling the machine remotely. Autonomous machines are managed through the John Deere Operations Centre app, where users can assign their autonomous machines to tasks on the job site, but they do not need to provide continuous oversight. Users have access to live video, images, data, and metrics, and they can adjust various factors such as speed. The app user will be notified if there is a problem with the job quality or with the machine.

John Deere develops autonomy in iterations

According to Matt Potter, director for Robotics & Mobility Technology at John Deere, limiting the scope of capability of the first-generation autonomous tractor was foundational to the success that we see with the four new machines today. Getting the project right, where full autonomy could be achieved with no operator in the cab, is essential for these machines to be useful for customers. John Deere designed the first-generation fully autonomous tractor to work on a slow, easy farming task without time sensitivity: fall tillage using a chisel cloud.

This first generation used stereo (two camera) vision, which is used in robotic systems and computer vision to add the depth measurement of objects. The two cameras "see" an object and the computer uses the angular difference between their views to compute the distance. 

To extend the range of the system, the cameras can be moved farther apart, but Willy Pell, CEO of Blue River Technology, a subsidiary of John Deere that develops intelligent machines for the agriculture industry, notes that maintaining their wider alignment to accurately make those measurements is too difficult for stereo cameras in a rugged environment.

To get around this problem, Pell says that John Deere's second-generation system adds many more cameras around the cab so that they often have triple overlap. No longer operating in stereo, this camera array can compensate for vibration as they move relative to each other. 

The NVIDIA GPU (graphics processing unit) sits in a rugged housing that John Deere calls the VPU (vision processing unit) where it calculates the position and orientation of each of the cameras on a per-frame basis, which eliminates the need for the cameras to be rigidly mounted with respect to each other. This continuous processing of all the camera views keeps the machine running safely and reliably and all of these calculations are done on the machine — not in the cloud.

This advancement opens up the opportunity to expand autonomy to faster, more complicated tasks. The system will also withstand a harsh environment, including moisture, vibration, and temperature changes.

The key is now accounting for edge cases — all of the unusual variables that the ADT may encounter during operation — and having enough information in the system to train the model to recognize those scenarios. John Deere has built a library with hundreds of thousands of images to train this second-generation model.

John Deere's 460 P-Tier autonomous articulated dump truck

ADTs require less skill to operate than other construction equipment, and valuable skilled operators are best utilized in the cabs of machines that fully take advantage of their talent. The ADT is also suited to consistent, repetitive work patterns, such as in quarries, which further makes it ideal for automation.

Built on John Deere's expertise in components for rugged environments, the autonomous 460 P-Tier ADT features high-performance electronics and a rugged component stack consisting of satellite, internet, advanced controls, and high-performance GPUs. The ADT can dynamically adjust to factors in its environment, routing around objects, or pulling aside as needed. The system is continuously making the same decisions that a human driver would face.

According to Maya Sripadam, senior product manager, Autonomy, Blue River Technology, "Our customers are managing a lot of inputs and what we want to do with autonomy is provide a safe, productive option to enhance their workflow and reduce the cognitive load that our operators face."

The autonomous machines are presently in the testing stage and the autonomous kit is not yet available for purchase. It is designed to be a retrofit kit that customers can choose to add to their machine after purchase, giving them the flexibility to modify their fleets and meet the challenging demands of the modern job site.