Hand-eye coordination from Micropsi Industries helps robots to carry out production processes which could not previously be automated.

Covid-19 has hit the world by its Achilles heel. The virus not only exposes human mortal­ity; it shows the vulnerability of the economy, which is dependent on interna­tional supply chains. For example, production stops in China brought production in the domestic automotive industry to a standstill. In order to become resilient after the pan­demic, companies must increasingly rely on local production, it is said. But how can a high-wage country like Germany keep up with the cheaper competition from Asia? And how can mass production take place when, due to distance regulations, not so many workers can be employed in factories as before? ”Now the hour of the robots has come“: with this headline "Der Tagesspiegel" recently provided an answer to these questions.

From emotion to added value

"Yes, many people are now saying this", says Ronnie Vuine, co-founder and CEO of the Berlin start-up ”Micropsi Industries“ (pronounced: Meikropsi). "We've been saying this for a while now." Ever since his student days at Humboldt University Berlin in the early 2000s, a group around Vuine has been working on the topic of artificial intelligence, more precisely on the "PSI" model of the German psychologist Dietrich Dörner. This involved the programming of emotional robot personalities.
When the introduction of deep learning expanded the technological possibilities, the time had come for the former student experiment to grow up: in 2014 Vuine founded the start-up in Berlin-Neukölln with Dominik Welland, Joscha Bach, Priska Herger and Ulrik Deichsel. The name of the student working group - Micropsi (Micro-Psi) - has remained the same over the years. However, the team has now moved away from the idea of emotional robots. Rather, it has concentrated on where the technology could be used to create value. An investor put them on the right track: robotics in automa­tion. "We Germans know how to produce and automate things," says Vuine, describing a principle of success of the domestic economy, "we have accumulated an enormous amount of know-how in this field over the last 30 years.”

^{CEO Ronnie Vuine © micropsi Industries GmbH}

Germany: country with the highest robot density

In fact, Germany is one of the countries that have automated their production most. It is surpassed only by Singapore and South Korea. According to figures from the International Federation of Robotics (IFR), there are 338 industrial robots for every 10,000 workers in Germany. The global average is 99. "Automation in Germany, how­ever, is only really high in the automotive industry," says Ronnie Vuine. In fact, 59 per cent of all robots installed in Germany are used in this industry. This is followed by the metalworking industry with 14 per cent and the plastics and chemical industry with eight per cent. Since the emergence of the trend towards robotics in the 1980s, indus­trial robots have traditionally been used for tasks that required a particularly high degree of power or precision. These include building car bodies, painting or welding vehicles. "The robots are programmed and are given commands for a single movement that is always the same," Vuine explains the principle. But this only works in those rare cases where the workpiece to be processed is always in exactly the same place. If the environment changes by just one millimetre, the machine quickly reaches for nothing. "This is all the more true when a workpiece has to be moved from A to B, but A or B is not known," says the managing director, quoting the example of a Berlin company which produces plastic housings for small computers. When they fall from an injection moulding machine onto the assembly line, they do not always end up in the same place. Until now, it was the job of a worker to take the housing off the assembly line, lay it in a nest and place the manufacturer's logo. Such activities are not uncommon: "In an astonishing number of production processes, people carry out very repetitive tasks which require tact and hand-eye coordination," Vuine explains. Here, he says, the use of human labour is more economical than having complex sensor evaluation soft­ware developed for an often highly specific process step. "It is often overlooked that not always technological feasibility was an obstacle to automation, but also the costs or the skills of those who have to set up the robots," the computer scientist adds. "Medium-sized companies in particular often cannot do this and would have to obtain very expensive help.”

Micropsi Industries recognized this niche as early as 2014, and developed a solution to meet these multiple challenges with one stroke: the team in Berlin-Neukölln, which has now grown to 28 experts, developed a system that enables robots to learn movements which are not the same with every repetition. So-called "hand-eye coordination" helps machines from ABB or Universal Robots to perform their tasks even in dynamic environ­ments. This is achieved with the help of computer vision and artificial intelli­gence. Instead of each individual step having to be programmed in a complex and expen­sive way, the robot is simply taken by the hand. This is meant quite literally: the "MIRAI" software (Japanese word for "the future") learns by the robot arm being guided through the task by a human. A camera and sensors on the arm record the movement performed. In just a few hours, Micropsi Industries' artificial intelligence learns to understand the data thus obtained. "It understands how the workpiece and the movement relate to each other, and understands the principle of the movement itself," Vuine explains. In this way, the robot learns to perform the task in real time and can also be flexibly trained at any time to perform new movement steps.