Development of genetic range genetic algorithms (3rd report, proposal of fitness function to search multiple local optima)

In some cases of developing a new product, response surface of an objective function is not always single peaked function, and it is often multi-peaked function. In that case, designers would like to have not only global optimum solution but also as many local optimum solutions as possible, so that he or she can select one out of them considering the other conditions that are not taken into account priori to optimization. Although this information is quite useful, it is not that easy to obtain with single trial of optimization. In this study, we will propose a screening of fitness function in genetic algorithms, and give rough search for local optima. Once after they are obtained we will carry out local search to obtain multiple quasi-optimum solutions. In this paper, we demonstrate the effectiveness of the proposed method through a simple benchmark test problem.