Generation of a fuzzy logic controller using evolutionary strategies

In this paper, the authors propose a fuzzy logic controller (FLC) generation scheme using a modified (μ+λ) evolution strategy (ES) with parental population sizing. The object variable portion of each ES individual represents a control rule of the FLC to be optimized. Therefore, the parental population along constitutes one and only one candidate FLC rulebase improved over the generations. The improvement is attributable to an increase of better-fitted individuals in the parental population. Fitness values of individual rules are calculated symbiotically. At each generation, offspring rules compete with their parents to form a new rulebase, which then strives to replace the original candidate rulebase according to elitist principle. In this aspect, the proposed FLC generation scheme practices (1+1) ES on the rulebase level. Furthermore, size of the candidate rulebase is simultaneously adjusted via parental population sizing to reflect system demands. The resultant scheme not only can construct FLC with better performance but also provide additional flexibility to their rulebase structures. Simulations on several control examples have been conducted to demonstrate the virtue of this FLC generation scheme