Universität Karlsruhe
Soft Computing

Description

Processing of vague and uncertain knowledge.

Really complex systems can generally not be exactly modeled in every detail. The other way round, a single datum loses its significance with rising complexity. Nowadays, many applications are known that became possible only by neglecting needless details and an adequate consideration of uncertain data.

Reasonable and precise decisions can be based on vague and uncertain knowledge. The best example for systems that successfully deal with such information are human beings. However, in areas like data processing and technical automation more and more people realize that it is not only too costly to consider all relevant information but often impossible. Generally, there are two different areas for the application of processing fuzzy information:

  • applications that are too complex to be realized with classic approaches
  • increase of the efficiency in existing applications

We especially research reasoning methods that

  • intuitively model fuzzy knowledge,
  • allow different kinds of information sources (vague statements, uncertain data, positive/negative information ...),
  • consequently operate on a high level of abstraction and, therefore, are very efficient.

These techniques are applied in Fynesse (FuzzY-NEuro-SyStEm), for example, a hybrid architecture for sequential decision problems like control of dynamic systems or scheduling problems (job shop scheduling, instruction scheduling in compiler construction). In Fynesse, fuzzy knowledge is used to restrict the search space of potential solutions on the one hand, and to describe found solutions in a comprehensible form on the other hand. In this way, it becomes possible to deal with really complex systems.

Projects in this area

Former

Fynesse
Autonomous learning of control strategies.
GK Naturkatastrophen
Fuzzy modeling of natural disasters
Reasoning With Vague And Uncertain Terms

Related publications

2002
Wagner, Statistical Based Fuzzy Sets
Wagner, Risikoabschätzung mit Fuzzy Methoden
 
2001
Wagner, Risk assessment with Possibility Measures
Wagner, Risk Assessment in Natural Disasters with Fuzzy Probabilities
 
2000
Riedmiller, Spott, Weisbrod, Fynesse: A hybrid architecture for selflearning control
Login
Links