Major agricultural machinery manufacturers today produce tractors and equipment with such technological solutions and equipment that they are considered the “beginning of robotics” in agriculture.
Lorenzo Benvenuti
The technological content of agricultural machinery is growing rapidly. Until a few years ago, it was mainly due to the implementation of electronic and mechatronic devices aimed at improving the efficiency of specific subsystems and adopting the methods of precision farming. Today, among the innovations, systems that favour machine connectivity seem to prevail: the fleet of machines becomes a coordinated fleet that can be managed with a smartphone and whose entire activity can be recorded. However, in this proliferation of electromechanical and IT innovations, the trend, which has been underway for several years now, towards the production of increasingly autonomous agricultural machinery is evident. Thus agricultural machines is increasingly equipped with advanced sensors to detect environmental parameters and crop status, with IoT connectivity (i.e. the so-called Internet of Things, which allows smart machines to be created) to exploit management platforms and have real-time analyses, and, in the most advanced cases, equipped with artificial intelligence (AI) and machine learning software to make autonomous decisions and improve performance.
The historic manufacturers of agricultural machinery now produce tractors and equipment with such technological solutions and equipment that we can already consider them the antechamber of robotics. In short, they are going around it, testing the waters, and above all experimenting. Many have already created prototypes of autonomous tractors, even without a cab, sometimes with an electric motor (easier to robotise, but more difficult to power) and aim to enter the market shortly with real robotic tractors. For example, Deana Kovar, president of John Deere’s Worlwide Agriculture & Turf Division for Europe Asia and Africa, recently stated that by 2030 she expects to launch a line of fully autonomous arable farming machines on the market, with tractors equipped with hybrid technology for medium to large power units, and electric for the others. Karl, the self-driving tractor by Kuhn, is potentially already operational. Other manufacturers are working to enter the market with fully electric tractor lines, such as Fendt. However, only the promulgation of the new European regulations on agricultural machinery, robots, AI, civil liability, etc. will be able to boost this market segment.
Building a tractor that operates without a driver in the cultivation of arable land is a great challenge and an excellent training ground for future technological developments, but it cannot be a point of arrival for agriculture and will have little impact on the agricultural system. With today’s means and the most advanced cultivation techniques involving minimum surface tillage, high-speed sowing, irrigation with pivot or rainger, and by entrusting harvesting to mechanisation companies, a specialised operator can dominate areas close to 1,000 hectares with 3 or 4 crops in rotation, so as to defer sowing times. The real business of robotics is another.
Start-ups and similar
Start-ups and hyper-technological companies have understood this well and have entered the market a straight leg, by overturning the canons of conventional agricultural machinery by offering original machines in terms of architecture, size and power; real robots capable of carrying out some operations in full autonomy (or almost). A burst of innovation that has stimulated private research and created an extremely interesting satellite industry. Although it is difficult to generalise, these robots are generally intended to carry out mainly those agricultural operations that are considered problematic.
For example, in weed control, robots are proposed that allow for physical weeding (mechanical, thermal, laser, etc.) or chemical weeding but targeted at individual weeds, and which already represent a valid alternative to spraying over the entire field. In some horticultural crops, intelligent weed control operated by robots is an alternative to the manual work still needed on densely sown horticultural crops that do not allow the use of mulching films.
The issue of labour is of course central, but the robot can be fully successful when it is proposed as a replacement for unskilled seasonal labour (technologically speaking, because any agricultural manual labour requires skill and expertise). In this context, the great challenge is the harvesting of the fruit of a plant that may be herbaceous, shrubby or arboreal. In fact, this is almost always a burdensome operation for the operator, expensive for the strictly seasonal entrepreneur, and creates managerial difficulties. This is why freeing fruit picking from human labour by entrusting it to autonomous machines is, especially for large companies, one of the objectives that will emerge strongly in the short term.
The solutions are very different, some decidedly fanciful, but all try to solve the problem of safeguarding the integrity of the fruit which, let us remember, is intended for fresh consumption, its identification within the foliage, the evaluation of its degree of ripeness, and the methods of detachment. Just think how complex it can be not to crush a strawberry or blueberry during picking, the difficulty of detaching a pepper or courgette from the plant, of reaching the fruit on large plants such as apple and pear trees. Here come robotic hands and grippers, suction cups and other solutions capable of handling the grip with sufficient skill and guarantee of integrity.
Recognition uses colour, when it can, or shape, or increasingly in AI-equipped machines it uses a complex set of parameters chosen directly by the machine through machine learning processes. The robot, when dedicated to the harvesting of gradually ripening fruits, can also be equipped with devices that analyse the quality of the harvested product in real time, providing the robot with feedback for verification. These are robots that operate according to the Japanese philosophy of Kaizen, i.e. continuous self-development.
Another sector with a heavy use of seasonal and unskilled labour is transplantation. In this case, however, robotic or semi-robotic equipment has already been on the market for several years. These are machines in which the mechanical aspect predominates, and electronic systems should basically coordinate the different components of the transplanter. However, industrial research is attempting to apply systems that recognise the quality of the plant (essentially dimensional) and its physiological state, so that only healthy and vigorous seedlings can be transplanted. Again, this involves applying AI systems equipped with machine learning.
The machine-plant relationship is constantly evolving. In fact, it is taken for granted that recognition and then analysis take place using optical devices or alternatively electromagnetic waves outside the visible range, such as infrared (think of NIRS) or violet and ultraviolet. In reality, this is not always the case. For example, some researchers have developed a sensor that identifies different species of plants at various stages of growth by “touching” their leaves with an electrode inspired by human skin. The robot is able to measure properties such as surface texture and water content that cannot be determined with current visual approaches. Such a method could make it possible to monitor crop health and growth and support decisions on the amount of water and fertiliser to be given to plants and on the control of pests and diseases.
Shortcuts
There are devices on the market that can be applied to normal tractors of different brands and power ratings, transforming them into autonomous machines. A retrofit kit and a software system are tipically provided that transform tractors into autonomous vehicles capable of towing, transporting and operating a range of agricultural equipment, without the technical need for ongoing supervision (suveillance, however, may be mandatory under regulatory requirements).
Installation is quick and can also take place on existing tractors, allowing them to perform a range of field operations autonomously. Of course, the kits include advanced connection systems that allow the activity to be recorded, remotely controlled and managed in the same way as Sabanto’s Steward system. It is also necessary to detect the field boundary and any obstacles by manually guiding the tractor. The optimal path is then defined by the software according to the geometry of the field, the real-time position of the tractor within the field, the degree of overlap, and the working mode at the headland. Throughout the activity trajectories and consumption are transmitted to the cloud, where they can be monitored by the farmer or, if necessary, by our service team.
There are many similar solutions such as the one proposed by Same Deutz-Fahr (SDF) which transforms its TTV 5115 into a Smart Vineyard Tractor. This tractor gains the ability to operate autonomously thanks to front, rear and side cameras, GPS antenna and safety bumpers. The autonomous technology of TTV 5115 comes from VitiBot, the French manufacturer that developed the Bakus robot, designed for mowing and spraying vineyards. The Deutz-Fahr TTV 5115 narrow tractor is expected to hit the market no earlier then two years.
Sensors also help to integrate nature and agricultural production. For agricultural areas bordering natural areas, which are common in the Alps and Apennines, Pöttinger developed a wildlife scanner years ago, but now improved in its effectiveness. The new Sensosafe wildlife scanner can be mounted on the front of mowers, combine harvesters and other harvesters. When it detects birds, deer or other wildlife hidden by the crop, it alerts the driver to stop the machine. The scanner uses infrared light that interacts with wild animals, which seems more effective than thermal imaging cameras.
