Learning robots: How humans and machines work together

The integration of robotics with Artificial Intelligence (AI) methods takes robotics to a new level. In close cooperation with humans and controlled by voice, gestures or interactive learning, robots can adapt flexibly to different tasks, people and environments. This opens up new fields of application, even in complex environments - for example in care, medical technology or the trades. This is associated with far-reaching economic and social potential. In a new white paper, Plattform Lernende Systeme uses use cases to illustrate specific application scenarios for robots with learning capabilities, describes technological developments and identifies design options for practical use.

Germany is well positioned in many respects to utilise the potential of interactive, adaptive robotics, but faces strong competition in research and development as well as in application. At the same time, remarkable progress is being made in machine learning, while the costs of robots and components are falling. More powerful computing architectures combined with AI methods provide solutions to complex problems in real time thanks to powerful hardware. Learning through interaction in robotics has become a strategically important technology for maintaining and expanding competitiveness and technological sovereignty in Germany and Europe.

Potential for the economy and society

Robotics is benefiting from developments in data-supported machine learning - through the improved perceptual capabilities of robots and the ability to control them by voice with the help of large language models. They learn by demonstration when a human shows them their task, or they improve what they have already learnt through human feedback.

Robots that require little expertise on the part of the user and at the same time save on programming costs open up far-reaching possibilities. The tasks that robots will be able to take on in the future go beyond their use in industry alone. They can contribute to overcoming many of the current challenges of our time, be it maintaining and expanding competitiveness, building a circular economy or tackling the shortage of skilled labour. The white paper shows possible use cases: In recycling, for example, robots can be used to precisely separate recyclable materials and sort out specific objects. In the care sector, it should be possible to use robotics for simple tasks in future so that carers have more time for their core work and interaction with people.

Utilising synergies and creating safety

Analysing the various use cases reveals similarities and differences. The utilisation of synergies can play a particularly important role in practical implementation. When developing robotic systems, attention should therefore be paid to modularity and cooperation between different application areas.

‘It is crucial to develop complete systems that are as modular and universally applicable as possible for different areas,’ says Elsa Kirchner, Professor at the University of Duisburg-Essen and Head of the “Intelligent Healthcare Systems” department at the German Research Centre for Artificial Intelligence (DFKI) and co-author of the new white paper of Plattform Lernende Systeme. For example, we have developed exoskeleton technology alternately for space and medicine. We have therefore developed different components and transferred them between the application areas. Such universal solutions are possible in many areas and reduce costs, both in research and in development and application.’

The authors of the white paper also recommend pushing ahead with technical integration. The aim should be to transfer individual available technologies, such as interactive learning, cloud or edge computing and deep learning, into overall systems. Interdisciplinary research is still required for the development of future secure and human-centred learning algorithms. The quality and quantity of available data plays a decisive role here. Open source databases, data cooperation or federated learning for particularly sensitive data could address the data shortage.

The use of robots in close proximity to humans requires trust in the application and acceptance of the technology in a complex social environment. This is primarily achieved through safe, reliable and traceable technologies, continuous further development of safety concepts and early, close involvement of all stakeholders, including the users.

About the white paper

The white paper "KI in der Robotik. Flexible und anpassbare Systeme durch interaktives Lernen" was written by members of Learning Robotic Systems working group of Plattform Lernende Systeme. Members of the working group Future of Work and Human-Machine Interaction and the working group Health Care, Medical Technology, Care were involved in the creation of use cases, among other things. The white paper is available to download free of charge.

An interview with Jürgen Beyerer, co-author of the white paper and member of Plattform Lernende Systeme is available for editorial use.