Hybrid particle swarm optimizer for a class of dynamic fitness landscape

Abstract

This article presents the use of particle swarm optimization (PSO) for a class of non-stationary environments. The dynamic problems studied in this work are restricted to one of the possible types of changes that can be produced over the fitness landscape. A hybrid PSO approach (called HPSO_dyn) is proposed, which uses a dynamic macromutation operator to maintain diversity. In order to validate the approach, a test case generator previously proposed in the specialized literature was adopted. Such a test case generator allows the creation of different types of dynamic environments with a varying degree of complexity. The main goals of this research were to analyze the ability of HPSO_dyn to react to the changes in the environment, to study the influence of the dynamic macromutation operator on the algorithm's performance and finally, to analyze the algorithm's behavior in the presence of high multimodality.

Keywords:

• Particle swarm optimization
• Dynamic optimization
• Non-stationary environments
• evolutionary algorithms

Acknowledgements

The first author acknowledges the continuous support received from the Universidad Nacional de San Luis and the ANPCYT. The second author acknowledges support from CONACyT project number 42435-Y.