Qualitative Modeling and Controller Design using Dynamic Fuzzy Systems

Qualitative modeling may be applied when knowledge about a system is only available in linguistic form. The knowledge might be processed by a dynamic fuzzy system consisting of a rule base and an inference method modeling human reasoning. Conventional fuzzy inference methods do not consider this association to human reasoning and therefore are not suitable for the dynamic processing of linguistic knowledge. Inference has to provide both, quantitative and qualitative information about the model output. In this paper a new inference method based on the concept of interpolating rules is presented. It results in a mapping of fuzzy inputs onto fuzzy outputs with a feedback to the input, considering in particular the informational content of the fuzzy sets. Thus, a dynamic process model is built from given linguistic knowledge in form of rules. Finally, a controller design method is presented based on the qualitative process model.