Article Preview
Top1. Introduction
Particle swarm optimization (PSO) was first introduced by Kennedy and Eberhart in 1995 (Eberhart & Kennedy, 1995) based on a social-psychological model of social influence and learning. Like most evolutionary algorithms (EAs), PSO is a population-based stochastic search technique. Each member of the PSO swarm, also called a particle, represents a candidate solution in the search space. During the optimization process, each particle iteratively adjusts its flying direction according to its velocity, which is dependent on the best experiences of both the swarm and the particle itself. A fitness value is used to estimate the quality of each particle’s position, and accordingly determine its subsequent flying direction.
PSO is easy to implement, and has been successfully applied to solve various optimization problems such as nearest neighborhood classification (Blackwell & Bentley, 2002), software development (Chang & Huang, 2012), receive-diversity-aided STBC systems (Liu & Li, 2008), mixed discrete nonlinear programming (Nema, Goulermas, Sparrow, & Cook, 2008), economic load dispatch in power systems (Ratnaweera, Halgamuge, & Watson, 2004), and Value-at-Risk based fuzzy random facility location models (Wang & Watada, 2012). However, PSO easily gets trapped in local optimal when solving complex multimodal problems. PSO converges quickly in that information transmits throughout the swarm rapidly, while the search scope is also shrinking rapidly, which usually leads to the lack of population diversity and premature convergence. If the particles happen to be initialized in good positions, PSO can reach a preeminent solution; otherwise it would likely converge to an inferior solution and result in mediocre performance.
Many PSO variants have been proposed to solve the problem by increasing the population diversity (J. J. Liang, Qin, Suganthan, & Baskar, 2006; Peram, Veeramachaneni, & Mohan, 2003; Poli, Kennedy, & Blackwell, 2007; Van den Bergh & Engelbrecht, 2004). In this paper, a reflecting bound-handling scheme is imposed on a local best PSO (LPSO) using von Neumann topological neighborhood, which helps to achieve a better balance between diversity and convergence speed. On top of that, a novel Crown Jewel Defense (CJD) strategy is proposed to direct the algorithm toward a superior solution when particles get trapped. The proposed algorithm is a combination of CJD and LPSO with reflecting bound-handling (rfl) scheme, called LCJDPSO-rfl for short. The performance of LCJDPSO-rfl is evaluated through a series of experiments on benchmark functions. The experimental results demonstrate the stability and efficiency of LCJDPSO-rfl on most of the functions.
The rest of this paper is structured as follows. Section 2 provides a brief review of PSO and its variants. Section 3 describes the details of the proposed LCJDPSO-rfl. In Section 4, four groups of experimental results are analyzed to illustrate the performance of the proposed algorithm. Finally, the paper is concluded in Section 5.