90636: Computer scientist, mechanical engineer, electrical engineer or similar (f/m/x) - Context-sensitive task representations for assistance robots

Area of research:

Scientific / postdoctoral posts,Sonstiges

Job description:

The Institute of Robotics and Mechatronics is working on the development of assistance robots that support people with physical disabilities to lead as self-determined life as possible. The user should always have the feeling of being able to freely determine the actions of the robot. The "Shared Control Templates" (SCTs) developed at the institute are a successful approach to this. These allow the user to dynamically control the relevant degrees of freedom of the system depending on the task, the state of the robot and its environment. This makes it possible to solve even complex tasks without overwhelming the operator.

A current limitation of the approach is that the task must be known to the system. With few interaction possibilities, this can be solved using simple heuristics. However, in real scenarios, such as taking a specific product from a well-stocked fridge, this reaches its limits. In order to solve this problem, the task representations of the SCTs are to be extended. The first step here is to merge different "tasks". If two objects are close to each other, it is difficult for the robot to deduce which object should be grasped, but the necessary movements for both objects are initially identical. By merging different SCTs into a combined, situational SCT, the decision can be postponed to a later and more relevant point in time. The second approach is to use information beyond the physical situation to infer the intended task. The challenge here is the heterogeneity of the information. It is possible to draw conclusions about the probable tasks from higher-level tasks, habits and also spontaneously occurring events. A context model is therefore to be developed which extracts the aspects relevant for task selection from this information and makes them usable.

Your area of responsibility will focus on the following:

• conception and development of methods for situational fusion of SCTs
• design and development of a context model to support task selection
• conception and development of a context-based "Task Selection"
• application and evaluation of the methods on the assistance robot "EDAN"
• collaboration in projects in the application area of assistance robotics
• publication of results in relevant journals and conferences

This research center is part of the Helmholtz Association of German Research Centers. With more than 42,000 employees and an annual budget of over € 5 billion, the Helmholtz Association is Germany's largest scientific organisation.