Foraging Strategic Genetic Algorithms to Obtain Multiple Acceptable Designs

Abstract

These days, requirements of the machine become more and more complicated and we need to satisfy many of those requirement simulataneously. In such cases, it would be rational and reasonable to carry out multi-objective optimization. However, in the conventional decision making problems, we need to give preference according to local information such as trade-off. But, it is not that easy to give proper preference from such local information. If we can give multiple acceptable designs to the designers, it will be great help for them to give preference, because they will give something like map of the design space. However, to give multiple Pareto solutions, we need to carry out a great number of scalar optimizations, and it is unrealistic. In order to give multiple acceptable designs by single optimization process, we simulate adaptation strategies of wild lives combined with genetic algorithms in the previous study. We have succeeded in keeping variation among population and directions to give multiple acceptable designs, but we failed in controlling of the number of population. As a sequence of the previous study, we focus our attention mainly to adaptation of foraging and try to simulate the evolution of species as the changes of searching ranges for design variables. The goals of the proposed method lie in finding multiple acceptable designs, keeping variation of population and see the changes of searching range for each objective functions to give clear map to the design space to the designers. In this study, the proposed method are analyzed using a simple two-variable, two-objective problems whose analytical solution is available.