John Deere’s Path to Autonomy
John Deere has unveiled the next stage in its autonomous machinery development, expanding beyond traditional row crops into functions for tillage, orchard spraying, landscaping, and construction. By prioritising ease of use and high-speed precision, the company is aiming to integrate driverless technology into the daily rhythm of the modern farm.
A Three-Step Approach to Autonomous Tillage
John Deere is focused on bringing its autonomous solution to tillage applications first because the task is relatively simple. The company has developed its autonomous system to be as straightforward for the farmer as possible, with three steps to get started:
1. Planning and Data Integration
Users first need to set up and plan in the John Deere Operations Centre. Here, they make work plans for the autonomous tractor, guided by autonomy boundaries and agronomic data collected during harvest. Farmers then send this work plan to the tractor, where John Deere’s tech stack from the last 20 years - such as AutoTrac Turn Automation, AutoPath, and Operations Centre Mobile - kicks in.
2. Field Activation and Remote Monitoring
Once the plan is sent, a farmer needs to bring their tractor to the field and enable autonomy mode on their G5 display. From there, all they need to do is launch the Operations Centre mobile on their phone and swipe on the activation screen to start tillage.
Farmers can then leave the field and move on to another task. They will be able to remotely monitor the tractor in operation, allowing them to stop and start the machine as needed. If the tractor encounters an obstacle, it isn’t able to verify; the farmer can view a live camera feed to determine the next action. When work is complete, the farmer can retrieve their tractor and move it to the next field.
3. Post-Work Analysis
Finally, farmers can return to the Operations Centre to analyse the generated data. Farmers can identify which parts of the field were productive and diagnose areas that weren’t, using that analysis to inform future autonomous plans.
Redesigning Perception: The 16-Camera Array
The next-generation perception system has been redesigned to expand its capabilities and enable quick retrofits onto farmers’ existing tractors. Autonomous machinery uses stereoscopic cameras mounted at various points on the tractor to create a single image and spatially process the environment while it works.
To get a full view of the field, Deere implemented a perception system with 16 cameras mounted atop the tractor cab, evenly distributed across all four sides. Rather than the standard two-camera stereoscopic vision overlap, Deere has implemented triple overlap. These cameras can correct themselves - without any moving parts - to create a steady image, even as the tractor works in unstable conditions.
High-Speed Efficiency and Rugged Processing
The advanced vision system enables the tractors to work at 20kph, 40% faster than with the previous system. This speed increase significantly impacts the bottom line; in testing, it shortened post-harvest tillage work for one farmer by two weeks. The 20kph working speed also allows the system to react to any potential obstructions in a timely manner. To ensure safety, Deere has evaluated its perception system in tens of thousands of scenarios.
Rather than needing a constant data connection, the visual processing is done locally by graphics processing units (GPU), allowing for near-instant calculations. These GPUs are ruggedised and capable of operating in extreme conditions, from the searing heat of summer to the cold of winter.
