In the era of rapid product update and intense competition, aesthetic design has been increasingly important in various fields, as aesthetic feelings of customers largely influence their purchase preferences. However, the quantification of aesthetic feeling is still a very subjective process due to vague evaluations. The determination of form parameters according to aesthetics is difficult hitherto. Aesthetic measure recently arises as a prominent tool for this purpose using formulas derived from aesthetic theory. But as revealed by existing studies, it needs to be customized with deterministic and objective methods to be reliable in practice use. To facilitate this application, this paper proposes an evolutionary form design method, integrating aesthetic dimension selection and parameter optimization. After summarizing initial aesthetic dimensions, aesthetic dimension selection based on expert decision-making and particle swarm optimization (PSO) is carried out. With filtered aesthetic dimensions, design parameters are optimized with NSGA-II (non-dominated sorting genetic algorithm). The quality of pareto solutions obtained to be design schemes is assessed by three criteria to conduct sensitivity analysis of cross and mutation probability and population size. Our experiment using bicycle form design shows that the proposed evolutionary form design method can generate numerous and variant aesthetic design schemes rapidly. This is very useful for both product redesign and innovative new product development.

This paper deals with design characteristics-oriented approach to architectural design based on the combination of three methods – recognition, generation, and evaluation. Design characteristics are understood as a set of specific features which constitute a discriminant of a class of architectural forms. The Biederman recognition-by-components theory is used to recognize the design structure. An evolutionary algorithm, which serves as a generative tool, is driven by the fuzzy evaluation based on Birkhoff's aesthetic measure. Phenotypes of architectural objects are seen as configurations of Biederman's basic components essential for visual perception. Genotypes of these objects are represented by graphs with bonds, where nodes represent object components, node bonds represent component surfaces, while graph edges represent relations between surfaces. Graph evolutionary operators, that is, crossover and mutation, are defined in such a way that they preserve characteristic features seen as design requirements specified for designed objects. The fitness function is determined by the fuzzy evaluation of designs based on Birkhoff's aesthetic measure for polygons adapted for three-dimensional solids. The approach is illustrated by examples of designing objects with the use of a fuzzy evaluation mechanism, which takes into account both aesthetic criteria and the degree to which design requirements corresponding to object characteristic features are satisfied.