Multi-objective evolutionary generation process for specific personalities of artificial creature

Abstract

An artificial creature has its own genome in which each chromosome consists of many genes that contribute to defining its personality. The large number of genes allows a highly complex system. If these gene values are manually assigned for the individual genome, it becomes increasingly difficult and time-consuming to ensure reliability, variability and consistency for the artificial creature`s personality. To overcome this difficulty, this dissertation proposes an Evolutionary Generative Process for an Artificial Creature`s Personality (EGPP). EGPP evolves genome population such that it customizes the genome, which meets a simplified set of personality traits desired by the user. However, there need to assign the preference values and it is hard to anticipate the resulting personality. To get the set of personalities without providing the assignment of preference values, a Multi-Objective Evolutionary Generation Process for artificial creatures` specific Personalities (MOEGPP) is proposed where the dimension of personality model is defined as that of optimization objectives. Key components of MOEGPP are as follows: i) the complement of (1-k) dominance, ii) the pruning method considering objective deviation for all genomes, and iii) the mutation using biased normal distribution, to obtain a set of nondominated genomes with personalities according to the personality dimension. By using the proposed MOEGPP, nondominated genomes having specific personalities can be obtained. There are many number of objectives. Hence, it requires much computation time to evaluate genome in virtual environment simulation. Moreover, it needs a large number of population although there is a limitation for the population number. To solve this problems, parallelization of MOEGPP is proposed where many computing resources can be utilized. In this scheme, binary tree topology and two migration methods are employed to maintain a proper balance between explor...