The comparison of efficiency of the population formation approaches in the dynamic multi-objective optimization problems

Dynamic multi-objective optimization problems are challenging and currently not-well understood class of optimization problems but this class is important since many real-world optimization problems are changing over time. In such problems, the objective functions, their parameters and restrictions imposed on the search space can change over time. This fact means that solutions of the problem change too. When changes appear in the problem, an optimization algorithm needs to adapt to the changes in such a way that the convergence rate is sufficiently high. The work is devoted to the comparison of the different approaches to formation of a new population when changes in the dynamic multi-objective optimization problem appear: using solution, which obtained in the previous step; using a random generating of the population; partial using solutions which obtained in the previous step. In the first part of the article the classification of the changes in the problems is provided; the currently existing approaches to solving the problems based on evolutionary algorithms are considered. During the research NSGA-2 and SPEA2 algorithms are used to solving the dynamic optimization problems, the benchmark problems set is used to the comparison of the approaches. Obtained results being processed by Mann-Whitney U-test. It was obtained that changes rate in the problem is affect to the efficiency of the application of the solutions which obtained in the previous step in the forming of the new population.