Interpretable Fuzzy Rules Acquisition of Coupled System Using Interactive Genetic Algorithms
Dun-Yong Lu, Takehisa Onisawa · Journal of Advanced Computational Intelligence and Intelligent Informatics · 2007
This paper describes the use of the interactive genetic algorithms to acquire fuzzy control rules with good interpretability for a complex system having dependent variables and non-linear property. A chromosome is coded with integer to represent a fuzzy rule, and individuals composed of various numbers of chromosomes are evolved by GA operations. The acquired fuzzy rules are explained with the linguistic expressions for fuzzy sets. These linguistic expressions are determined through comparing with the standard fuzzy sets of linguistic variables designated in advance. To reduce human fatigue in the individual evaluation process, only quantitative evaluation with fitness functions is given at earlier stage. When aso-called better individualappears, not only quantitative evaluation but qualitative one is used to evaluate both the interpretability and control performance of the acquired fuzzy rules. The presented approach is applied to the control of the coupled system having two control objectives with multi-input/output variables. Simulation experiments show that the approach is feasible to acquire the satisfactory fuzzy rules with good interpretability and good control performance.