Dynamic Objectives Aggregation Methods in Multi-objective Evolutionary Optimization

Several approaches for solving multi-objective optimization problems entail a form of scalarization of the objectives. This chapter proposes a study of different dynamic objectives aggregation methods in the context of evolutionary algorithms. These methods are mainly based on both weighted sum aggregations and curvature variations. Since the incorporation of chaotic rules or behaviour in population-based optimization algorithms has been shown to possibly enhance their searching ability, this study proposes to introduce and evaluate also some chaotic rules in the dynamic weights generation process. A comparison analysis is presented on the basis of a campaign of computational experiments on a set of benchmark problems from the literature.