AI-controlled robots: Platform shows potential and limits

Robots that learn independently and work closely with humans - the combination of robotics with artificial intelligence (AI) methods has great potential for society and the economy. Robots capable of learning can relieve the burden on employees, alleviate the shortage of skilled workers and support people with disabilities. However, there is still a long way to go before AI-controlled helpers become part of our everyday lives. In its new web special, Plattform Lernende Systeme provides an overview of adaptive robotics: a dashboard shows areas of application and developments, while experts from the platform shed light on opportunities and hurdles from various perspectives.

Germany is the largest robotics market in Europe and ranks third worldwide with 415 robots per 10,000 employees. According to forecasts, the German robotics and automation industry is heading for a record value of 16.8 million euros. For many years, automated robots have been taking over - predominantly rigid - work processes in structured and shielded environments, particularly in industrial production and logistics. The integration of robotics with AI methods such as machine learning is now bringing a new quality: robots are becoming capable of learning. They can take on increasingly complex tasks in changing environments and work in close cooperation with humans.

"The possible tasks that robots will be able to perform in everyday life in the future are diverse - as are the preferences of their users as to how they want to be supported by a robot. This makes pure pre-programming of future robots almost impossible. The ability to learn new tasks in interaction with humans is therefore becoming a key component for the development of intelligent robotic systems," says Dorothea Koert, researcher at the Technical University of Darmstadt.

How robots learn

Robots can learn through demonstrations by having a human show them their task, or they can improve what they have already learned through human feedback. Thanks to generative AI, robots can also be easily operated using natural language instead of code. This makes controlling robots much more intuitive for humans and no longer requires any special programming knowledge.

The adaptive robotic systems can be used in more and more areas of application beyond industry. Whether as delivery vehicles, in the event of a disaster or in the home - robots will provide support in many areas and relieve people of strenuous or dangerous tasks. One promising area of application is care and rehabilitation: "The advantage of robotic systems that can learn is that their ability to learn allows them to adapt better to the user, to the caregiver or the patient, for example," explains Elsa Kirchner, Professor of Medical Technology Systems at the University of Duisburg-Essen and head of the Learning Robotic Systems working group of Plattform Lernende Systeme. "A patient who has suffered a stroke may be paralyzed on the right side of their body, for example. The aim of an adaptive exoskeleton system is to get the patient moving again, so that the brain can relearn how to coordinate arm movements, for example."

A long but rewarding journey

However, adaptive robotic systems still have to overcome many challenges on their way from research to practice. "Robots have achieved unprecedented leaps in performance over the last ten years. However, there are still seemingly insurmountable hurdles in robot learning, such as the fact that machine learning is not able to utilize the understanding of our physical reality," says Sami Haddadin, Executive Director of the Munich Institute of Robotics and Machine Intelligence at the Technical University of Munich.

"There is still a long - but worthwhile - way to go before robots can function reliably in an "open world" and at the same time meet social expectations, for example in terms of empathic behavior," sums up Andreas Angerer, Head of Research & Innovation at XITASO.

The new website on AI-controlled, adaptive robotic systems can be accessed at https://www.plattform-lernende-systeme.de/robotik.html. All statements and contributions by the experts have been released for editorial use.