Rule pairing methods for crossover in GA for automatic generation of fuzzy control rules
Hiroyuki Inoue, K. Kamei, Kazuo Inoue · 2002
We (1996) have previously presented fuzzy rule generation methods by genetic algorithm (GA). In this paper, we propose three methods to determine rule pairs for crossover in GA for fuzzy rules generation in order to improve the search efficiency and reduction of the number of rules. In the first two methods rule pairs are determined based on the distance between the rules of two individuals to be crossed. In the third method rules of each individual are sorted based on the distance between the origin and a rule center in input space. We apply these methods to generate fuzzy rules for a trailer truck back up control, and show that the rule sorting method can generate a compact and high performance fuzzy system.