Robots of the Future: Cooked for and Cared for by Gentle Machines

“Darmstadt University of Technology professor Georgia Chalvatzaki wants to develop robots that are partners to humans. They should be able to learn from experience and adapt their behavior to their environment. However, there is still a long way to go.

Tiago is extremely agile. He can move up and down, rotate, and move sideways, forward, or backward. And all of this simultaneously, if necessary. Right now, the white robot is supposed to open a kitchen door. In the laboratory of robotics researcher Georgia Chalvatzaki at Darmstadt University of Technology, her team has set up a complete kitchen unit – with a stove, sink, cabinets, and cooking utensils. The robot is supposed to be able to grasp and operate all of these things.

Using a remote control, a doctoral student guides Tiago's robotic arm towards the rectangular handle, which is bright pink on the kitchen door. Computers analyze every movement, which is recorded by a camera and sensors. This is how data is collected for the complex system that is required even for such a seemingly simple grasping movement.

Tiago can't yet work in the kitchen like a human. "But someday it will be like we've perhaps seen in movies," says Georgia Chalvatzaki, laughing. "Then the robot will clean or do the dishes while we read a book, have more time for family or leisure." Ideas for self-driving cars existed decades ago. Today they are a reality. Robots that interact gently with humans and learn in real time will also require years of development, but that is exactly what the Greek-born researcher, who has been a professor of interactive robot perception and learning in the Department of Computer Science at Darmstadt University of Technology since 2023, is researching. Robots to fill gaps in healthcare

The 37-year-old scientist views robots, humans, and their environment as an integrated system. She and her team want to develop robots that can learn from experience and interaction with humans and their surroundings, and continuously adapt their behavior. Chalvatzaki sees these machines not as replacements for humans, but as their partners. She envisions "home robots" that could take over household chores, assist in the care of elderly or sick people, and thus fill gaps in social life and healthcare that will become increasingly larger in the future. For her promising efforts to achieve this goal, she recently received the Alfried Krupp Prize, endowed with 1.1 million euros.

Even her doctoral thesis, which the researcher, who studied electrical and computer engineering in Athens, completed in 2019, focused on human-centered approaches in assistive robotics. For this work, she maintained contacts with clinics in Germany and Heidelberg, among others, with whom she collaborated on the development of a robotic walker. "Many older people asked me back then whether the robot could also bring them their medicine or food."

This motivated her to continue researching in this direction. As a postdoc, she came to TU Darmstadt because she absolutely wanted to work with Jan Peters. The professor is an expert in autonomous robotics and reinforcement learning based on a reward system. Chalvatzaki also applies this method in her own research.

Recognizing objects and understanding their function

In her lab, she works with 13 doctoral students and postdocs. The researchers want to feed the robot's brain with algorithms that are not based on millions of data points like other AI applications. Having a human control the robot and teach it every task beforehand is one solution, but a very complex one, as Chalvatzaki explains. Therefore, she and her team want to find data patterns that make it possible for the robot to independently grasp a cup or open a door. To achieve this, they use simulations and visual data. For example, they have introduced geometry-based learning methods that integrate the structure of three-dimensional space, including position and orientation, into the training. One goal is for the robot to be able to navigate in the 3D world, recognize objects, and understand their function.

The scientists want to enable the transfer between the simulated and real worlds. To this end, they allow the robot to try out actions and learn through experience and corrections. For example, they incorporate into their models the knowledge that opening a coffee pot requires rotating hand movements. This data is then transferred to other tasks. This enables the robot to perform precise and fluid movements.

Future capabilities will include skills such as cooking, cleaning, and caregiving tasks, as well as human-like speed and gentle interaction. "This is a major difference compared to industrial robots, which operate according to predefined procedures," says Chalvatzaki, who has received numerous awards for her work, including acceptance into the Emmy Noether Program of the German Research Foundation.

 

She is certain that robots will become part of society in the future, acting as helpful partners and complements to everyday life, providing independence and mobility not only to the elderly or sick.  For this to happen, they must be absolutely reliable – for example, when it comes to dispensing daily medication. In the near future, she plans to establish contacts with care organizations and clinics to gain them as partners and learn more about where there is a need for robots.

The professor believes that the first prototype of a home robot could be on the market in perhaps ten to fifteen years.  She doesn't think it necessarily needs to look like a human. Robots should remain intelligent machines – even in their appearance. The potential is certainly enormous. Developments are already underway in Europe in which Chalvatzaki and her team have participated. Robots are working as shelf stockers in supermarkets, in baggage handling at Fraport airports in Greece, and in English agriculture, where they are used for potting plants, as Chalvatzaki reports. "This is already working successfully."”