Tuning membership functions, tuning AND and OR operations, tuning defuzzification: which is best?

Fuzzy control is a methodology that transforms linguistic IF-THEN rules formulated by an expert into a precise control strategy. There are many fuzzy control models that are able to implement such transformations. For a given situation, one model may perform better than another; as the state changes, switching between models can improve the quality of the resulting control. However, the switching algorithm can be computationally expensive. In this paper, we show that a similar increase in the quality of the resulting control can be obtained (with a much smaller computation time) by dynamically changing defuzzification exclusively