In our daily lives, we tend to forget about the crucial role of agriculture. However, providing food in sufficient quantity, at an acceptable level of quality, and then to distribute it efficiently is far from simple. Several factors are threatening the fulfilment of these objectives in the near future. The most important of which is the demographic trend.  This includes not only rapid population growth –current projections forecast a population of 9.7 billion people in 2050 – but also the urbanization trend that is decreasing the available workforce in the agricultural sector – by 2050, 66% of the population is expected to be urban versus 54% in 2014.

Therefore, the demographic trend raises issues, namely increasing the demand and decreasing the supply of agricultural products; the latter by decreasing the relative amounts of two essential inputs: labour and land. Indeed, the reduction in agricultural land is another important factor threatening the sustainability of food production. This is partly due to the population increase, but also to the growing competition for land with biofuel producers and to the accelerated phenomena of desertification and rising sea levels. Further to labour and land, there is one last crucial input that is becoming even scarcer: water. This can also be linked to the increasing population and the competition over this resource between agriculture and other sectors.

Possible answers to these concerns can be found in Artificial Intelligence. For the scope of this article, we will define AI as any program able to process and analyse data, including Big Data, and take decisions based on these. It also includes machine learning and deep learning (itself embedded in machine learning) which can basically be described as self-improving algorithms. Although these new technologies show great promise in providing solutions to many of the arising challenges, they also have an important disruptive potential on the agricultural industry. It will be worth examining their possible impact.

Some targeted AI solutions — Food Quality & Quantity

Let us look first at the growing concern about food quality. This includes food safety but also pesticide levels and the nutrition claims of food items. One way to improve food quality is through the automated process of sorting products called optical sorting. It relies on hyperspectral cameras to analyse the material at the chemical level, which allows the system to detect hidden defects. Optical sorting systems have been increasingly used, particularly in the potato industry but also for different vegetables, fruits and nuts. These systems increase both yield and food safety as they improve the accuracy with which infected items are detected. Regarding the use of pesticide, robots can mostly improve the current situation by targeting more precisely the needed area of use, i.e. to improve spot spraying. Some automated thinning, weeding and spot spraying machines for lettuce have been developed with encouraging results.

The major innovation when it comes to increasing food supply is the UAV (Unmanned Aerial Vehicle). They can offer solutions to many of the issues of large farms by scanning vast areas of land in real-time for soil analysis, plant population count, crop anomalies, disease detection, etc. They can also be equipped to efficiently spray pesticide on crops. The main advantages of UAVs are their ability to take off and land vertically and their ability to fly at low altitudes, which makes them suitable for most kinds of terrain and crops. A second range of solutions to increase, or at least secure, food supply comes with sensor networks. An autonomous early warning system for fruit fly outbreaks has been developed. Using wireless sensor networks and GSM networks, coupled with self-organizing maps and support vector machines (two machine-learning techniques), it is able to automatically issue warning text messages to farmers and government officials when the population of fruit flies rises above a certain threshold. This system also improves food quality by reducing the need for pesticide. Another application of sensor networks lies in greenhouse climate controllers. Sensors have been part of greenhouse climate controllers for a while, but the addition of AI-based techniques like neural networks or genetic algorithms to the traditional systems have improved the yield and reduced the use of water and energy.

Some targeted AI solutions — Lack of Human Labour 

Most of the technologies already mentioned help reduce the need for human labour in the agricultural sector. Optical sorting systems, automated spot spraying machines and autonomous UAVs efficiently replace humans, but other innovations have been designed solely for this purpose. This is the case for automated harvesting machines, driverless tractors and planting robots, for example. With regard to the latter, if I can be allowed to be a little patriotic, I will take the Belgian example of Robovision. This company, based in Ghent, develops software and deep learning solutions for various industries including agriculture. Their planting robot uses deep neural networks and is able to learn within a few minutes to plant a new plant type. On the other side of the Atlantic, several institutions joined their forces to set up an efficient automated harvesting machine called Demeter. It uses two complementary navigation systems; one GPS-based and the other camera-based. As the researchers state in their paper: “Demeter is capable of planning harvesting operations for an entire field, and then executing its plan by cutting crop rows, turning to cut successive rows, repositioning itself in the field, and detecting unexpected obstacles.”

Some targeted AI solutions — Water & Land Management

Many universities and research institutions have been working on creating more efficient irrigation systems, and AI techniques have shown to be promising in this area. Most of these innovative systems rely on sensors and parameter modelling. It is worth noting the Enorasis project, a European Union funded project that developed a smart irrigation system that combines a meteorological analysis tool and a wireless sensor network to provide optimised and sustainable water management for farmers and water management organizations. Another impressive achievement in that regard was accomplished by the company ConserWater. It developed an algorithm for the purpose of machine-learning soil moisture. It therefore works without ground sensors, using only satellite data, historical weather data and a variety of other factors. It has been shown to be highly effective and is already used by more than a hundred farmers in several countries. Its success can also be explained by the handy way it is distributed to customers: a smartphone app available on Android and iOS.

Increasing yield could be a way to reduce the need for agricultural land. This can be done through many channels, some of them already mentioned in this article (improving the sorting systems of fruits and vegetables, using sensor networks in fields and greenhouses, and so on). Another channel is plant breeding and this is being explored by FarmView, a research team at the Carnegie Mellon School of Computer Science (USA) dedicated specifically to increasing crop yield. Using sensor networks and multispectral cameras (which are similar to the hyperspectral cameras mentioned in the optical sorting systems, but with a lower spectral resolution), they try to develop a variety of grain sorghum characterised by high yield, disease resistance and drought tolerance. Machine learning helps them to determine what plant phenotypes (for example the orientation of the leaves) are linked with these desirable characteristics.

The possible impact of the AI (r)evolution

All these innovations could boost agriculture’s productivity to unprecedented levels, which is necessary to overcome the many issues that the sector is facing. But it should be emphasised that they mostly target the needs of large farms and that, even when they could be useful to smaller producers, they are currently prohibitively expensive. We should keep in mind that agriculture is highly heterogeneous in Europe. Consequently, the development of new innovative tools could enlarge the already growing gap between two types of agriculture, by increasing the concentration of production and forcing small farmers to survive in food market niches. The tendency towards concentration has been observed for more than a decade: in the period of 2005-2013, the number of small farms fell dramatically while the biggest farms saw their number increase slightly.


It is not something inherently bad, but the transition should be carefully supervised and managed. That which could happen within countries could also happen between countries. Indeed, AI technologies represent costly investments not only for farmers in terms of purchase price but also for public institutions in terms of R&D. Also, it appears that most of the R&D regarding AI is taking place in rich countries. Even assuming that developing countries will import the technologies created elsewhere, these might not be suitable for their particular agricultural sector. Therefore, it seems reasonable to believe that developing countries will import food instead of technology. Beyond the problem of different farming structures, developing countries might fail to implement AI technologies in their agricultural sector because of a lack of properly trained workers to handle these cutting-edge tools. In fact, providing enough training to farmers should be a major focus for any country willing to modernise their agricultural industry, as it is necessary in order to maximise the potential increase in productivity and the investments profitability.

The European Union seems to be aware of the coming evolution of agriculture and, more importantly, of the consequences it might have. This situation requires subtle responses in agricultural policy, but also in education and general trade policies. However, to this day, it is still unclear what framework will be provided by the post 2020 Common Agricultural Policy for innovation.

In conclusion…

The objective of this article was to give a general overview of the ongoing innovations in the field of AI and robotics targeting the agricultural sector, but also to discuss their potential to help the adaptation of this age-old and tremendously important human activity, as well as the consequences that could follow. However, despite being a fantastic means of overcoming the challenges raised by a changing situation, AI is not the only solution. Other ways towards sustainability exist, like permaculture that could drastically increase the yield of food production while improving quality and reducing the need for inputs like water and energy. Urban farming also provides a sustainable solution to the problem of available workforce due to urbanization, by relocating part of the food production closer to the people, who are both workers and consumers. It also helps boosting the image of agricultural work by underlining the environmental and community building role of growing food. Last but not least, it optimises the use of land. So agriculture has a bright future ahead, but traditional farmers might not.