Kansei design using genetic algorithms

Genetic algorithms are assumed precise mathematical or numerical models of their analyses. However, these models are needed only in the calculation of fitness function, and fitness function only determine the possibility of leaving offsprings to the next generation, or the possibility of survival in the next generation. Thus, we need to give in fitness functions an approximate difference among each individual in the population. In this study, we stress these characteristics of fitness function and carry out an optimization of Kansei which has difficulties in making up mathematical or numerical models. To demonstrate the effectiveness, we first show how much robustness do genetic algorithms have in searching a true optimum solution by treating the simple multi-peaked problem with adding white noises to its fitness function. As a demonstrative example, we treat a problem which will make up neutral expression of human faces with given grammar for portraying their faces. Through this example, we show the possibility of using genetic algorithms in Kansei design