Creative 3D Designs Using Interactive Genetic Algorithm with Structured Directed Graph

摘要

We propose a methodology for representing artificial creatures like 3D flowers. Directed graph and Lindenmayer system (L-system) are commonly involved in Al-based creativity research for encoding ceatures. It is difficult for L-systems to directly feed back real morphologies structurally from their genotypes, since they are a grammatical rewriting system and also use parameters such as loops, procedure calls, variables, and primitive parameters for representing their genotypes. In this paper flower genotypes are manifested by a knowledge-based structured directed graph (SDG) and phenotypes are represented a flower morphology resulting from the derivation and graphical representation of the genotypes. Evolution is simulated using an interactive genetic algorithm (IGA), where a SDG is useful for genotypic representation of creatures and IGA uses human evaluation for the fitness function. We have applied the creation of 3D flowers using the knowledge-based SDG and IGA. Experimental results show that realistic flower morphologies can be created by the proposed method.