Research of the Knowledge-Based Systems (KBS) group focuses on methods and algorithms for controlling embedded agents to acquire, reason in, and maintain symbolic knowledge about the world to achieve purposeful and goal-directed behavior in the current context, and do so by computation on-line and on-board the agents. We call Embedded Knowledge-Based Systems (EKBSs) the resulting control systems. Autonomous robots are typical agents considered here, but pure software agents are of interest, too. At least some of the respective knowledge must be symbolic in order to be inspectable and communicable. Regarding environment sensing, the focus is on 3D sensor data, such as point clouds, and on methods and algorithms for efficient handling and semantic interpretation.
The KBS group is closely connected with the DFKI Plan-Based Robot Control Group.
Videos about our projects and current research can be found on our YouTube Channel.
In quite a few projects world-wide, 3D Mapping technology has been used for documenting historical buildings – typically involving a massive amount of time and person power for both the data capture and their processing for public access (if that is granted at all). This can be afforded for eminent buildings like the Pantheon in Rome or the Cologne Cathedral; it is prohibitive for the many thousands of listed historic buildings constituting the cultural heritage of a whole region. 3DInOs was a study for developing and exemplifying a concept how to document historical buildings in a level of detail satisfying art historians’ needs and involving much limited cost for capturing, processing, and archiving the data, owing to the automatization and standardization of many of the processes and data formats involved.
The LiSA project was funded by BMBF (Bundesministerium für Bildung und Forschung). Project duration was 1.4.2006 - 31.7.2009. The goal was to develop an autonomous mobile robot as a system assisting in a life-science research lab environment.
MACS – Multi-sensory Autonomous Cognitive Systems. The MACS project is a Specifically Targeted Research Project (STReP) for Exploring and Exploiting the Concept of Affordances for Robot Control. The project is funded by the European Commission’s 6th Framework Programme IST under contract/grant number FP6-004381. The Commission’s support is gratefully acknowledged.
Das interdisziplinäre Forschungsprojekt SOILAssist befasst sich mit dem landwirtschaftlichen Bodenschutz bei der Befahrung von Ackerflächen und verfolgt das Ziel, negative Veränderungen der Bodenstruktur und -funktionalität frühzeitig zu erkennen und zu vermeiden. Durch angepasstes Management soll ein günstiger Bodenzustand erhalten und verbessert werden. Im Rahmen des Teilprojektes ‘Adaptives Semantisches Planungssystem’ arbeiten die Projektpartner DFKI und UOS gemeinsam an einem adaptiven semantischen Assistenzsystem für den Feldeinsatz mit mehreren kooperativen Maschinen sowie der Integration einer geometrischen Umgebungsdarstellung.