My resent focus was on specific issues in highly distributed information systems.
- Agent technology. Multiagent systems are an ideal model for large-scale distributed systems where there is only lose coupling between the individual components and each component acts fairly autonomously. Specifically we study such systems from the viewpoint of a large distributed database. Such a database will definitely be inconsistent, hold logical contradictions, and the stored facts may be uncertain to some degree. The most challenging issue is how to keep the database in a working condition by readjusting the local databases such that a prescribed degree of consistency can be maintain, something that often goes under the heading of uncertainty management.
- A companion issue is agent dependability. We work on layered agent architectures that allow to assign specific fault tolerance and error processing techniques to specific layers and combine transactional recovery techniques with compensation based on uncertainty management.
- We apply agent technology in collaboration with other academic groups and with industry to applications ins production engineering and in road traffic.
- Agent and multiagent systems are by definition goal-oriented. Hence, semantics play an important role. To gain first experience, we specifically address the issue of context, i.e., which semantics apply in a current situation. We study context in two scenarios, one of workplace-oriented eLearning, the other in interpreting historic printed material in the light of a historian’s knowledge of the time of writing.
- Semantics is also a central issue when it comes to interoperability of services. We develop formal, ontology-supported techniques that allow to describe the component functions in component-based software engineering, and the domain semantics of databases that are jointly accessed in virtual organizations.