Finding Sets of Non-Dominated Solutions with High Spread and Well-Balanced Distribution using Generalized Strength Pareto Evolutionary Algorithm

The paper presents a generalization of the Strength Pareto Evolutionary Algorithm 2 (SPEA2) and its application in selected well-known twoand threeobjective optimization benchmark problems. The proposed solution is referred to as our SPEA3. The generalization consists in the exchange of the environmental selection procedure in SPEA2 for a new original algorithm which aims to determine the final non-dominated solutions with a high spread and well-balanced distribution in the objective space. During the evolutionary optimization process, the non-dominated solutions are gradually incorporated into the resulting set and placed in it in such a way that the distances between them and their nearest neighbors in the objective space are the greatest possible. A comparative analysis with alternative multi-objective optimization techniques shows that our approach is superior with regard to the spread and distribution of solutions while being still competitive with regard to their accuracy.