Cognitive Robotics


In traditional robot applications the robot is separated from the environment while working on a recurring task in a clearly specified activity space. In contrast, modern robotic applications require the reliable execution of varied tasks in a significantly less pre-specified, open working environment. Such functionality cannot be achieved sufficiently with classical control approaches in automation technology. The scientists of the research group “Cognitive Robotics therefore focus on application-oriented Artificial Intelligence methods for controlling modern robots. These technologies, which are essential for innovative and in particular future robotic systems, supplement classical control systems with superior decision-based control functionality and are thus intended to enable robot systems to operate autonomously, robustly and especially safely.

The field of activity of the research group in particular comprises the following topics:

  • Collaborative human-robot interaction
    New opportunities in numerous areas are created by combining the strengths of humans and robots in the joint accomplishment of tasks and lifting physical barriers between them. Challenges to be solved towards this vision include assuring safety and minimizing various risks, optimizing shared work processes and establishing mutual understanding about common and each other’s goals, tasks and skills. The research group elaborates methodic approaches and application-oriented realizations for intelligent robotic systems, incorporating solutions from many related fields and allowing for humans and robots to jointly reach achievements on a new level.
  • Safety-oriented perception
    Robots operating in shared workspaces with humans raise the need for comprehensive sensory capturing of the environment combined with the corresponding appropriate situational awareness. In contrast to traditional safety systems, which act as complementary add-ons, pronounced sensory technology assures not only a future robotic system’s function but also its safety in an integrative manner. This dual purpose requires new methodic approaches for the execution of the sensory infrastructure as well as the applied data fusion and a central safety-oriented cognitive functionality.
  • Machine task planning and execution
    A robotic system acting autonomously requires the essential ability to solve complex tasks by finding a plan composed of the single steps that it needs to be perform. Hence, it is not sufficient to stubbornly and uniformly execute this plan. During the plan execution, an adequate feedback loop must be able to constantly observe the environment for relevant deviations from expected conditions, which then may require a change of the plan. Safety and robust reliability are aspects that also need to be considered in an integrative manner for this area.
  • Machine learning
    Future robot systems need to be able to deal with questions that go beyond the capabilities of conventional algorithms, be it for safety issues or the general fulfilling their tasks. For instance, they may need to evaluate or classify situations, their own or external actions, or elements of the environment, to reason about possible future events, or to judge the reliability of sensory inputs or their own conclusions. Such problems require the application and the advancement of modern approaches in machine learning.

Lakeside B13b
9020 Klagenfurt am Wörthersee
Tel.: +43 316 876-2000
Fax: +43 316 876-2010