Performance-Enhancing Techniques

Abstract

This chapter presents techniques that have proved to be very useful in enhancing the performance of PSO in various optimization problem types. They consist of transformations of either the objective function or the problem variables, enabling PSO to alleviate local minimizers, detect multiple minimizers, handle constraints, and solve integer optimization problems. The chapter begins with a short discussion on the filled functions approach, and then presents the stretching technique as an alternative for alleviating local minimizers. Next, we present the deflection and repulsion techniques, as a means for detecting multiple global minimizers with PSO, followed by a penalty function approach for constraint handling. The chapter closes with the description of two rounding schemes that enable the continuous, real-valued PSO to solve integer programming problems. All techniques are thoroughly described and graphically illustrated whenever possible.