Vacuum cleaner principle and AI: new ideas for a changing working world

From automotive companies to master carpenters, from engineers to tradespeople: more and more people are encountering robotics and using technologies in their day-to-day work. How can working with cobots be made as easy and intuitive as possible? Researchers in the OPERA research project are developing innovative ideas to address this issue.

Teresa Scheunert

2025年3月19日

Imagine

読了時間： 5分間

Robots are useful helpers. They take on tasks that are monotonous or strenuous and help to increase efficiency. However, there are obstacles to the use of robotics, especially for small and medium-sized companies: they often cannot afford to employ their own programmers, the integration of a robot into an overall concept of controls and software is often complex, and on top of that, they may have a production process with small batch sizes and a constantly changing environment. So how do you set up a robot cell quickly and cost-effectively?

Cobot operation with new ideas

New technologies offer a solution: intuitive operating concepts and the use of artificial intelligence (AI) are making robotics more accessible. This means that sensitive cobots can be “taught” their tasks by the user performing the desired motion sequences by hand with them. This hand-guided programming makes deployment easier and faster, especially for medium-sized companies without a great deal of experience.

“We built on this basic idea and developed new features,” says Dr. Niels Dehio, OPERA project manager at KUKA and an expert in AI. Together with partners from the DLR in Oberpfaffenhofen, he worked on the next stage of AI-based automation in the Bavarian research project, which ran until the end of 2024. “The movements programmed by humans in the classical way are often rather rigid and do not always follow the optimal path for a robot. In addition, modern KUKA robots not only perform fixed pre-programmed tasks monotonously but should also adapt to changing conditions – in other words, they must demonstrate a certain degree of flexibility and intelligence.”

Modeling according to the vacuum cleaner principle

One crucial point here is to execute robot movements safely and without collisions. To do this, the robot's environment can be virtually modeled with a high degree of precision so that the robot is aware of the environment in which it is moving and takes it into account during its movements to avoid breaking anything. However, many smaller companies have a flexible production environment that is constantly changing. It is precisely for such cases that the researchers at OPERA have developed concepts for modeling robot cells quickly, easily and cost-effectively.

Test setup: the cobot with obstacles in its work environment

“The cobot is moved by hand for a few minutes, and the space around the robot is virtually swept like with a hand-held vacuum cleaner,” explains Dehio. This is then used to create a precise obstacle model – so the robot knows exactly where it can move without problems and where obstacles such as pillars or tables are located. An algorithm then optimizes the cobot's movements, fully automatically. Thanks to the new obstacle model, this ensures that the robot doesn't bump into anything – without the need for expensive sensors and in a way that is intuitive for new users as well.

Learning by imitation – and AI

In addition, the team working on the OPERA project has been using AI to teach the cobot how to handle process variance. To do this, the robot uses a camera to capture the scene and, again with the help of a human guide, is shown the desired movement multiple times with slight modifications. On this basis, the developers created an AI model so that the robot can adapt independently to slightly modified situations. Environmental modeling based on the vacuum cleaner principle serves as an additional safety measure to prevent collisions.

The cobot captures the scene with a camera and is shown the desired movement with variations.

New ideas for a changing production world

Such innovative concepts can change the way robots are used and make the use of robotics easier and cheaper – especially in a production environment that is constantly changing. Even after the three-year project ended, an international team of doctoral students at KUKA continued to work on this topic. “Artificial intelligence in combination with vision systems opens up great potential in robotics,” says Dr. Dehio. And to exploit this potential for more intuitive robotics, Niels Dehio, his colleagues and other KUKA teams have their sights firmly set on the AI topics of the future.