Cognitive Robotics Department

Robotics is closely linked to the topic of artificial intelligence (AI). After all, robots are built to do something – and in order for them to know what to do and how, they need a certain intelligence. Every robot the institute develops should solve its tasks as efficiently and purposefully as possible. To this end, the department of Cognitive Robotics works on various core AI topics. These topics are relevant for many application domains – from space assistance to robots in care and factories of the future.

Design of shared control skills for assistive robotics
DownloadDownload

Achieving goals requires a robot to know what it is doing and why it is doing it. The first focus of the department is therefore to represent such knowledge explicitly, and make it accessible to the robot. This involves knowledge of the environment, as well as the goals and skills of the robot itself. It is important that this AI ​​is explainable: the robot should be able to explain to the humans whom it cooperates with what it is doing and what goal it is pursuing. The research on this topic is pursued in the group Transferable Explainable Knowledge.

Investigation of human sensorimotor performance in weightlessness
Credit:
Roskosmos (links), Grafik DLR CCBY 3.0 (rechts)
DownloadDownload

There are almost always several paths leading to a goal. The more skills a robot has, the more (diverse) paths it can take. And it also needs a diverse set of skills to be able to generate a path to the goal, even if errors occur and unforeseeable situations arise. The second focus of the department is therefore to enable robots to quickly acquire, learn and improve new skills through intuitive programming, demonstration and reinforcement learning. The research on this topic is pursued in the group Interactive Skill Learning.

A third focus is on the processing and interpretation of biological datastreams – such as an electromyogram and electrodermal reactions. The aim is to recognize and even anticipate people's intentions, so that they can intuitively operate various assistance devices, such as prostheses or wheelchairs. This research on this topic is conducted in the team Reenabling Systems.

InFuse consortium group picture
Image: 1/13, Credit:
Cooperation between humans and robots
On board the ISS, NASA astronaut Frank Rubio is in control of the robotics team on "Earthly Mars" at DLR in Oberpfaffenhofen. As part of the first surface avatar experiments, humans and robots had the task of retrieving a scientific instrument - here a seismograph - from the lander and placing it in the mission area.
Image: 2/13, Credit:
DownloadDownload
Design of shared control skills for assistive robotics
Using assistive robotic systems, people with severe motor impairments can interact with their environment again. However, the direct control of such systems is cumbersome, especially for activities that require high precision or coordinated motion. In order to facilitate the usability of assistive robots, we develop methods for semi-autonomous support. To this end, we define robot skills as finite state machines, which constrain the robot's motion in such a way that the task is successfully executed. In combination with the control interface, the user can determine which task is executed and always has control over the robot system also during task execution. With this technology, we aim to make the use of assistive robots in performing various activities of daily living easy and intuitive. (exemplary scenario with a non-disabled person).
Image: 13/13, Credit:
DownloadDownload

Contact

Dr. rer. nat. Freek Stulp

Head of Department
Institute of Robotics and Mechatronics
Cognitive Robotics
Münchener Straße 20, 82234 Oberpfaffenhofen-Weßling