A study on a cooperative character modeling based on an improved NSGA II

Different animation characters are loved by people at different ages, in different countries and from different background, which implies the difficulties of designing characters that might be loved by many. Grouping different designers together by computer networks or the Internet is surely a good solution to design a character with expected popularity. This paper presents a character modeling method based on an improved non-dominated sorting genetic algorithm II (CMIN). CMIN borrows ideas from biological evolution, especially from multi-objective genetic algorithm (MOGA) and is formalized as a procedure for character modeling. CMIN adopts binary tree data structure to express transformation rules which are used to diversify character models and uses crossover and mutation operators of genetic algorithm to generate new rules. CMIN also adopts cooperative multi-objective evaluation on generated characters. The objectives are designed to embody both qualitative and quantitative aspects of character personalities, which are assigned by different cooperative designers and calculated automatically by computers respectively. The incorporation of qualitative and quantitative evaluation is formally realized by introducing a MOGA framwork. A multi-objective evaluation-based cooperative character modeling system (MOCMS) was developed to verify the proposed CMIN. Representative case studies demonstrate that the proposed method can evolve character models according to the designers’ intentions and preferences and generate creative character models far beyond man’s own imagination.