Multi-objective particle swarm optimization based on global margin ranking

It is prevalent that the Pareto-based dominant framework is inefficient in non-dominated sorting because the performance sharply deteriorates when there are numerous weak dominance relations. In order to address this issue, the paper presents a novel ranking strategy called Global Margin Ranking (GMR) which deploys the position information of individuals in objective space to gain the margin of dominance throughout the population. The method not only considers the distribution of population, but also incorporates the associated information of individuals, without incurring user-defined parameters. Moreover, in view of the challenges faced by Multi-objective Particle Swarm Optimization (MOPSO) in selection of gBest and pBest, we present an innovative strategy for selecting gBest and pBest by integrating the GMR and the individual's density information. We compare the GMR method with a variety of other ranking methods in terms of the distribution of ranks, the ranking landscape and convergence of the evolutionary process. The relatively extensive experimental results on some benchmark functions show that MOPSO/GMR performs better than those specialized MOEAs.