Kansei Design by Interactive Reduct Evolutionary Computation (With Attention Paid to Favored Feature of Design)

Abstract

We propose a method to evolve designs based on a user's personal preferences. The method works through an interaction between the user and a computer system. The objective of the method is to help a customer set design parameters by simple evaluation of displayed samples. An important feature of this method is that the design attributes which a user pays more attention (favored features) are estimated with Reduct in Rough Sets Theory and are reflected while refining the design. New design candidates are generated by the user's evaluation of design samples generated at random. While values of attributes estimated as favored features are fixed in the refined samples, the others are generated at random. This interaction continues until the samples converge to a satisfactory design. Thus, a personal and subjective feature is evaluated by this efficient design process. This method is applied to design a 3 D cylinder model such as a cup or a vase. This method is then compared with an Interactive GA.