Positioning and Perception: The Complexity of Producing Safe Autonomous Machines in Agriculture

The global demand for increased and more sustainable production in agriculture is driving the need for automation and autonomy in the field, making the development of this technology more urgent than ever. However, the complexity of achieving the full autonomy of tractors and farm machinery safely, is very high. 

Two of the most critical components of safe autonomous machines are machine positioning and perception. Knowing the accurate position of an unmanned piece of agricultural equipment provides the foundation for all operations, operating area and precision agriculture practices. GNSS (Global Navigation Satellite System) is one of the safest, most precise and universally and globally available methods of machine positioning when used with quality and reliable solutions.

GNSS is made up of satellite constellations that provide signals from space and transmit position and timing data so receivers can determine their location. On its own, it can provide a geographic accuracy of around one metre, however, different types of errors can lead to errors of over 7 metres. With a high-quality receiver and a signal correction solution, the machine's accuracy can be pinpointed to just a few centimetres. The challenge and risk, comes when solutions that cannot handle the accuracy, reliability and resiliency needs of autonomous equipment are used in production.

Have you ever requested a car or had something delivered through an app and watched as the vehicle rotates on the map or suddenly jumps in position? Think of the danger this lack of precision could cause when used with driverless agricultural machines. While the technological premise is the same, the positioning solution for agriculture autonomy must have the highest level of confidence possible when the machine is both moving and static. 

The presence of trees or electrical wiring can cause problems for autonomous agricultural machinery by blocking satellite visibility for short periods of time. That's why Hexagon has developed a proprietary technology through its NovAtel brand called SPAN, which provides position smoothing during outages, in which the machine can operate reasonably until it regains satellite visibility, allowing for greater machine uptime and less frustration for both farmers and manufacturers. 

However, these are not the only issues to be resolved in the development process of autonomous farm machines. Safety in the field is an essential concern that must be treated seriously by companies that set out to develop agriculture robotics. On a farm, there is always the possibility of people, animals and other objects moving near machines. Therefore, the perception or “vision” of equipment is critical. 

Vision-based perception uses cameras to detect objects in the field. When combined with a reliable object classification database, a machine can detect and identify an obstacle, allowing developers to create algorithms that provide the machine with information on what to do when faced with certain objects. However, the data collection and machine learning required to build a complete library of identified objects in agriculture scenarios is extensive. 

When you think of all the different types of scenarios an autonomous machine may be in, and what it may encounter, the possibilities are almost endless. Just to allow for the accurate identification of people, the cameras need to be trained with images that include different variations of humans, standing, squatting, lying down – imagine someone fainting in the heat of the countryside near a self-driving machine – at every possible angle. In the variability a farm can provide, the scope of objects that can be identified is limitless, from crops to animals, machines, etc.

The complexity of these foundational pieces of agriculture autonomy is part of the reason why autonomy in the field is more difficult than many developers expected. The responsibility put on developers to create reliable and safe solutions to support the needed production growth from farmers is immense. Strong partnerships between automation and autonomy technology developers and agricultural machinery manufacturers will be the basis for creating smarter and safer autonomous equipment. Thus, supporting farmers globally to operate more sustainably while increasing production to meet our future food demands.