Competence Groups

System Architecture & Embodiment

Christian Goerick
Chief Scientist
Sensory Processing & Learning

Heiko Wersing
Chief Scientist
 Intelligence is in our understanding a property of a com- plete system. It is the results of the well orchestrated inter-  We believe that learning methods are the best approach to create algorithms for transforming complex sensory  infor-

play between several elements organized in a structure defining architecture and embodied in the real world. Consequently, the research questions range from the theoretical description and modeling of large scale intelligent systems to their efficient implementation and integration guided by scientific principles. On the functional method level we research several specific elements that require a more holistic approach like task specific perception, context based situation modeling and prediction, behavior generation and learning.

mation into meaningful descriptions of the world. Our inspiration is the astonishing efficiency of biological sensory systems, enabling intelligent behavior in highly variable and dynamic environments. This motivates our main research questions: How can we employ the most recent sensor technology for sensing the environment? How can we decompose visual and auditory information into meaningful features? How can we orchestrate learning at different time scales and levels of sensory abstraction?

Complex System Optimization & Analysis

Markus Olhofer

Chief Scientist

Cognitive Systems & Representation

Julian Eggert

Chief Scientist

Major challenges in the design and the optimization of technical systems arise from their growing complexity and Key ingredients of a scalable vision system are efficient representations, situation-dependent selection mechanisms,
the increase of complex interactions in a network of systems. An additional difficulty is the necessity of a fast adaptation to changing requirements and constraints. The utilization of computational intelligence methods like population based search and advanced modelling techniques in the design methodologies are the key to cope with these challenges and to create optimal systems, which have the requested functionality and at the same time help to conserve our natural resources. fast and scalable computation of "sufficiently good" solutions, and flexibility through adaptation and learning.
Using engineering principles which closely follow the latest development in mathematical and computer science as well as strategies used by the brain, the dream of autonomous cognitive vision is getting closer to becoming reality.