Fuzzy controller model construction from sensor data through a new modified GAs approach

In this paper, a fuzzy control rules generation method inclusive of two main learning stages is presented. In the primary stage, automatically extract numerical control rules from the sensor data without the help of experts by means of a Genetic Algorithms (GAs), which add a different bit crossover operator to the standard GAs in order to increase the diversity of individuals and raise convergence speed of tradition GAs. Every generated numerical rule is accumulated in a control table called a numerical rule-based controller. In the secondary stage, find a fuzzy system with fuzzy rules using GAs to approach an identified system which is described by numerical rule above. Both antecedent and consequent variables of the numerical rules are fuzzified, and all training data are directly derived from the numerical rule with simple manipulations to tune the membership functions of the corresponding fuzzy system. An illustrative experiments are successfully made on the computer simulation. The experimental results reveal that the proposed approach is efficient and effective to design a fuzzy system.