Differences in behavior of dynamic parameterized operators

Finding the optimal parameter values for parameterized operators such as the S-OWA-AND aggregation operator or BADD defuzzification method can be difficult. Also, as the reasoning situation changes, the optimal parameter values may change. This paper first demonstrates how much the output of a given fuzzy system, one that controls an inverted pendulum, can vary depending on which aggregation operator and/or defuzzification method is used. These variations can be more significant than varying the rules or membership functions. Next, examples are given of how dynamic switching of parameterized operators can be used to determine appropriate parameter values as the reasoning situation changes. Our technique involves the dynamic generation of switching tables from which hierarchical meta-rules can be created. These meta-rules can then be used to dynamically change the appropriate parameter values as warranted