Multiobjective optimization using neighbourhood model genetic algorithms

We propose a method of multiobjective optimization using genetic algorithms. The proposed method doesn't reduce the objective vector to a scalar value, but finds a set of Pareto-optimal solutions using neighborhood model genetic algorithms. In the neighborhood model, population members are distributed on grid. The range of genetic interaction is limited to population members in the immediate neighboring nodes. This maintains the diversity of the chromosomes and avoids the premature convergence. Moreover, our method is suited for massively parallel computers. The results of numerical experiments show that the proposed method can find a set of various Pareto-optimal solutions more effectively compared with usual GA.