A hybrid design methodology for conventional fuzzy control

A hybrid methodology is introduced to design and tune the conventional fuzzy controller. To overcome the difficulties caused by the coupling of parameters in the knowledge base the process is carried out in two separate stages: nominal design and optimal tuning. Each stage integrates both quantitative and qualitative methods. The nominal design intends to figure out the nominal model of the rule base membership functions (MF), and scaling gains in a top-down approach; while the optimal tuning attempts to adjust the parameters optimally based on the nominal model in a bottom-up approach. The nominal rule base should be designed by qualitative approaches; while nominal scaling gains are designed by quantitative approaches. Some recent progress and outstanding problems are discussed.