Evolutional shape modeling by extended genetic algorithm: multi-step GA

In a computer aided design (CAD) system, shape design or shape modeling is one often a bottleneck because shape design is a trial and error process often depending on a designer´s intuition and experience. Authors have already presented a method, named evolutional shape modeling (ESM), to generate suitable shapes automatically using genetic algorithm (GA) and free-form deformation (FFD). This ESM method, however, has a limitation on its ability of shape generation because of the limitation on FFD´s shape deformation ability and the limitation on standard GA´s ability. This paper proposes a new evolutional shape modeling method. It includes new genetic algorithm, named multistep GA, which can avoid premature convergence, and new constructing method of FFD by which shape deformation ability is highly improved