Max Planck Institute for Intelligent Systems, Tübingen site

Max Planck Institute for Intelligent Systems, Tübingen site

Intelligent systems can optimise their structure and properties in order to successfully function within a complex, partially changing environment. Three sub-areas – perception, learning and action – can be differentiated here. The scientists at the Max Planck Institute for Intelligent Systems are carrying out basic research and development of intelligent systems in all three sub-areas. Research expertise in the areas of computer science, material science and biology is brought together in one Institute, at two different sites. Machine learning, image recognition, robotics and biological systems will be investigated in Tübingen, while so-called learning material systems, micro- and nanorobitics, as well as self-organisation will be explored in Stuttgart. Although the focus is on basic research, the Institute has a high potential for practical applications in, among other areas, robotics, medical technology, and innovative technologies based on new materials.

Contact

Max-Planck-Ring 4
72076 Tübingen
Phone: +49 7071 601-1700

PhD opportunities

This institute has an International Max Planck Research School (IMPRS):

IMPRS for Intelligent Systems

In addition, there is the possibility of individual doctoral research. Please contact the directors or research group leaders at the Institute.

Attack on autopilots

How fast the development from assisted to fully automated vehicles will progress is uncertain. One crucial factor here is the reliability with which a vehicle can navigate in its surroundings and react to unforeseeable incidents. Our group at the Max Planck Institute for Intelligent Systems showed that methods for motion analysis based on deep neural networks – likely components in future autonomous vehicles – can be confused by small patterns designed to “attack” these networks.

more
Max Planck ETH Center for Learning Systems extended by five years

Max Planck Institute for Intelligent Systems and ETH Zurich strengthen cooperation

more
Due to technical problems unfortunately the CoroNotes app is currently not available.

The CoroNotes app uses anonymous health data for a better understanding of the novel coronavirus

more

Machine learning methods which calculate movements are susceptible to interference

more
Körber Prize 2019 for Bernhard Schölkopf

This year, the scientific distinction with the highest prize money in Germany goes to the pioneer of artificial intelligence

more

In the future, it will be more and more common for computers to make decisions about human beings – whether they are granting loans or assessing applicants. However, it happens occasionally that the automated systems that are already in use discriminate against certain groups of people. Niki Kilbertus and Bernhard Schölkopf, researchers at the Max Planck Institute for Intelligent Systems in Tuebingen, want to change this by developing fair algorithms.

As domestic help, healthcare assistants or emergency response units: robots are suitable for these jobs only if they are capable of learning and acting independently, at least to a certain extent. Stefan Schaal and the members of his Autonomous Motion Department at the Max Planck Institute for Intelligent Systems in Tübingen are teaching machines to become flexible and autonomous.

A time may yet come when everyone has their own chauffeur-driven car – if robots take the wheel, that is. In order for autonomous vehicles to become a reality without huge technical outlay, however, computers will have to be able to assess complex traffic situations at least as well as drivers do. Andreas Geiger and his team at the Max Planck Institute for Intelligent Systems in Tübingen are working to develop the necessary software.

The life of an avatar is dependent on technology, including even the very act of its birth. For the virtual figure to look true to life and move realistically in its computer world, its creators need to have detailed information about the body of the real-life model, as well as about its movement. This is precisely the data that the first four-dimensional full-body scanner provides. This device was developed by Michael J. Black, Director at the Max Planck Institute for Intelligent Systems in Tübingen, together with American company 3dMD. With 22 stereo cameras and 22 color cameras taking 60 images per second, the scanner captures a person in various positions and activities that Javier Romero, a scientist at the institute, demonstrates here. For the scan, red and blue squares are printed on Nick Schill, a professional model, and then illuminated with a quickly pulsating spot pattern. The two patterns help the researchers reconstruct the three-dimensional surface of the body and the skin naturally. Not only can this method be used to create true-to-life figures for computer games and films, but it also offers interesting perspectives for research in psychology and medicine. In this way, it will soon be possible to use the realistic avatars in conducting perception experiments on body awareness– for instance to prevent eating disorders.

A Way Out of the Inner Prison

Material & Technology

The paralysis starts gradually, but in the course of time, it affects the entire body. People with amyotrophic lateral sclerosis eventually reach a point where they are no longer able to move any muscles or communicate with the outside world. A research group headed by Moritz Grosse-Wentrup at the Max Planck Institute for Intelligent Systems in Tübingen is looking for ways to help ALS patients break out of their isolation by teaching computers to read their minds.

No job offers available

Colour patch could throw self-driving vehicles off track

2019 Anurag Ranjan, Joel Janai, Andreas Geiger, Michael J. Black

Computer Science

In our team of researchers at the Max-Planck-Institute for Intelligent Systems in Tübingen we show that optical flow systems based on deep neural networks – a likely component of future autonomous cars – are vulnerable to adversarial attacks. The computer vision experts are shaking up the automotive industry by warning car manufacturers around the globe that it could take a simple color pattern to put the brakes on computer vision systems in autonomous cars.

more

Robots with their own thirst for action

2017 Georg Martius

Computer Science Material Sciences

Robots as helpers in everyday life can make our lives better in the future. However, there is much research needed to get there. One problem is the hardware. It needs to withstand everyday usage without being bulky or dangerous. The bigger problem, however, is to develop the right "brain”. To come somewhere close to human skills, a robot has to learn a lot by itself. The researchers of the Autonomous Learning Group at MPI for Intelligent Systems are working on artificial curiosity and the associated learning methods so that artificial systems can improve themselves in the future.

more

Computing with Uncertainty

2016 Hennig, Philipp

Computer Science

Machine learning requires computer hardware to reliable and efficiently compute estimations for ever more complex and fundamentally incomputable quantities. A research team at MPI for Intelligent Systems in Tübingen develops new algorithms which purposely lower the precision of computations and return an explicit measure of uncertainty over the correct result alongside the estimate. Doing so allows for more flexible management of resources, and increases the reliability of intelligent systems.

more

Robots learn how to see

2015 Geiger, Andreas

Computer Science

Autonomous vehicles and intelligent service robots could soon contribute to making our lives more pleasant and secure. However, for autonomous operation such systems first need to learn the perception process itself. This involves measuring distances and motions, detecting objects and interpreting the threedimensional world as a whole. While humans perceive their environment with seemingly little efforts, computers first need to be trained for these tasks. Our research is concerned with developing mathematical models which allow computers to robustly perceive their environment.

more

Learning robots

2014 Trimpe, Sebastian

Computer Science

An exploded power plant, collapsed buildings after an earthquake, a burning vehicle loaded with hazardous goods – all of these are dangerous situations for human emergency responders. What if we could send robots instead of humans? Researchers at the Autonomous Motion Department work on fundamental principles required to build intelligent robots which one day can help us in dangerous situations. A key requirement for making this happen is that robots must be enabled to learn.

more
Go to Editor View