Performance analysis of a particle swarm optimization algorithm to solve multiobjective optimization problems

Érica C. R. Carvalho¹; Afonso C.C. Lemonge²; José P.G. Carvalho³; Patrícia H. Hallak²; Heder S. Bernardino⁴

¹ Graduate Program of Computational Modeling, Federal University of Juiz de Fora - UFJF; ² Department of Applied and Computational Mechanics, Federal University of Juiz de Fora - UFJF; ³ Graduate Program of Engineering Civil, Federal University of Juiz de Fora - UFJF; ⁴ Department of Computer Science, Federal University of Juiz de Fora - UFJF

doi:10.20906/CPS/CILAMCE2017-0127

Resumo

The interest in multiobjective optimization algorithms has grown in recent years due to its applicability in problems from several areas, especially those from engineering. In general, the objectives considered in these problems are conflicting and a Pareto Front curve composed by the non-dominated solutions is searched as the expected solution. In the context of evolutionary computation there are many of algorithms applied to this type of problem, such as genetic algorithms, differential evolution, and particle swarm. In addition, multiobjective problems may present constraints turning them more complex, and requiring effective strategies that can direct the search for solutions that are in the feasible search space. This paper aims to evaluate the ability of a multiobjective optimization algorithm, called Multiobjective Craziness based Particle Swarm Optimization (MOCRPSO) in a set of unconstrained and constrained benchmark functions. An Adaptive Penalty Method (APM), which has been successfully applied to solving mono and multiobjective optimization problems, is used to handle the constraints. The results found by MOCRPSO illustrate the efficiency of the algorithm when compared to the results found in the literature.

Palavras-chave: Particle Swarm Optimization; Multiobjective Optimization; Adaptive Penalty Method