Robovision develops and hires out artificial ‘brains’. These make it very easy for companies to automate complex production systems and machinery, such as a combine harvester. We spoke to Jonathan Berte, the (human!) brain behind the Belgian company Robovision.
The new generation of smart software, with neural networks and algorithms that are able to partially imitate the human brain, is creating an agricultural revolution. Jonathan Berte, founder of Robovision says: “Smart machines with artificial intelligence (AI) will supersede dumb implements.”
This is already a reality, in his view. “Farmers with smarter, more efficient machines have a competitive advantage, but these smart implements require large investments. The size of a combine makes no difference to an artificial brain. Or rather, you will see a return on it sooner with a very large combine than with a small one. This technology will therefore become profitable sooner on large farms, which will continue to drive scaling up in agriculture.”
Text continues underneath image
Jonathan Berte, founder of Robovision: “Smart machines with artificial intelligence (AI) will supersede dumb implements.” - Photo: Peter Roek
Ideal in agriculture
Berte graduated from Ghent University as a engineering physicist and spent a year studying applied neurology at the University of Zurich. The insights he developed there became the blueprint for the artificial mini-brains that Robovision is now road testing.
Jonathan explains why AI and machine learning is such a major advance. “Take autonomous driving: a computer has to simultaneously process a huge amount of information and make decisions in countless different situations. A programmer would then have to write a new algorithm for each situation, and that’s impossible. The imitation brain is able to create and use its own algorithms.”
Automation of combines
Robovision, which is based near Ghent in Belgium, was established in 2008 and it operates in three markets: agriculture, industry (detecting production faults) and media & security (automatic human detection). A major manufacturer of tractors and implements was recently added to its customer base for the automation of combines. “Modern combines already have their own kind of nervous system (electronic control unit) with senses (sensors), so it’s a logical step to connect an artificial brain,” says Berte.
Text continues underneath image
Illustration: Future Farming
How a combine with an artificial brain works
A combine is like a factory on wheels. Its performance is affected by its design, threshing conditions, and crop characteristics. The combine has numerous variables that affect threshing quality and capacity, including travel speed, reel speed (1) and height, cutting height (2), the speed of the threshing drum (3) and its distance from the thresher cover (4), the wind speed (5) through the sieve (6) and the sieve openings. In modern combines, the driver can change all of these settings from a terminal in the cabin.
The optimum settings are a complex interplay of all threshing functions and the changing threshing conditions. If the latter, in particular, are changeable during threshing, for example the moisture content of the straw, adjusting the settings is a considerable challenge, even for experienced drivers.
With the advent of artificial intelligence and computer vision, all of these threshing functions can now be controlled automatically and in real time. Using cameras (7) in the elevator to the grain tank and above the straw walkers (8), deep neural networks (9) are able to continuously monitor the quality of the grain and analyse losses, and subsequently optimise all of the combine’s settings.
The concept of an AI-optimised combine allows the driver to download an optimally trained mini-brain for any crop from a cloud-based platform (10). These mini-brains are continuously updated with experience data from across the world. This means that a cereal farmer in Canada can benefit from the experience gained during threshing in Australia.
Hiring out artificial brains
Robovision hires out its artificial mini-brains to companies and assists with training and installing them. The way in which those companies pass on the cost of hiring to their customers depends on the situation. “A large tulip grower in the Netherlands, for example, pays us a licence fee per bulb for our artificial brains in the planting robot. The combine manufacturer will have to include the hire costs one way or another in the purchase price of its machines. Maybe as a subscription, similar to those offered by smartphone app providers.”
Text continues underneath video
Robovision does not wish to disclose the name due to competition considerations. One combine manufacturer is John Deere, the world’s largest manufacturer of agricultural machinery but, since it bought the Blue River Technology company in 2017, it has got AI and machine-learning expertise in house.
Jonathan recognises that independents such as Robovision risk suffering the same fate as Blue River Technology, which means that all AI technology will end up in the hands of a few large multinationals. “If you consider how powerful Amazon and various large Chinese corporations have become, that can’t be a positive development,” he muses, “also because of how much political effort it takes to curb the power of large high-tech companies.”
One of the best things about AI is that you no longer need to be a smart software engineer to automate your machinery, tractors or implements
Understandably, Berte is keen to emphasise how much agriculture will benefit from AI: “One of the best things about it is that you no longer need to be a smart software engineer to automate your machinery, tractors or implements. You just need to give the artificial mini-brain a little help at the start by training it in basic tasks, and it doesn’t matter where you do that. For example, the mini-brain we provided for a cutting planter at a Dutch chrysanthemum grower was trained by Syrian refugees who have settled in Belgium. They were able to do the job by indicating on the photos the optimum ‘grab target’ for the robot.”
AI creates diversity in agriculture
Berte also believes that intelligent machinery will contribute to greater diversity in what is currently large-scale, monoculture farming: “With AI-controlled autonomous field robots, crop farmers will soon be able to sow and cultivate a variety of crops with greater ease and efficiency, or grow a niche product with specific consumer requirements. What’s more, consumers will be able to view all the data that prove that your vegetables were grown in harmony with nature.”
What is what?
Algorithm: a process or set of rules to be followed in calculations or other operations by a computer. Artificial intelligence: the theory, and development, of computer systems able to perform tasks typically requiring human intelligence, such as visual perception, speech recognition and decision-making. Neural network: A computing system set up to mimic a biological brain and how it functions. Just like the brain, the network can make spontaneous connections and improve without coding. Deep learning: A computer making use of AI and neural networks to analyze input and discover unseen patterns that will help to achieve a specific goal. Computer vision: Essentially the science of teaching computers to ‘see’ – interpreting images captured by cameras, to understand what they show, and enable autonomous decision-making based on what has been learned.