Genetic algorithm-aided dynamic fuzzy rule interpolation

Fuzzy rule interpolation (FRI) is a well established area for reducing the complexity of fuzzy models and for making inference possible in sparse rule-based systems. Regardless of the actual FRI approach employed, the interpolative reasoning process generally produces a large number of interpolated rules, which are then discarded as soon as the required outcomes have been obtained. However, these interpolated rules may contain potentially useful information, e.g., covering regions that were uncovered by the original sparse rule base. Thus, such rules should be exploited in order to develop a dynamic rule base for improving the overall system coverage and efficacy. This paper presents a genetic algorithm based dynamic fuzzy rule interpolation framework, for the purpose of selecting, combining, and promoting informative, frequently used intermediate rules into the existing rule base. Simulations are employed to demonstrate the proposed method, showing better accuracy and robustness than that achievable through conventional FRI that uses just the original sparse rule base.