Computer Engineering and Applications ›› 2009, Vol. 45 ›› Issue (33): 38-41.DOI: 10.3778/j.issn.1002-8331.2009.33.013

• 研究、探讨 • Previous Articles     Next Articles

Improved multi-objective particle swarm optimization algorithm

ZHOU Liu-xi,ZHANG Xing-hua,LI Wei   

  1. College of Automation,Nanjing University of Technology,Nanjing 210009,China
  • Received:2008-07-02 Revised:2008-09-27 Online:2009-11-21 Published:2009-11-21
  • Contact: ZHOU Liu-xi


周刘喜,张兴华,李 纬   

  1. 南京工业大学 自动化学院,南京 210009
  • 通讯作者: 周刘喜

Abstract: An Improved Multi-Objective Particle Swarm Optimization(IMOPSO) algorithm is proposed,in which elitism archived strategy is used,global best position is provided by non-dominated solutions in the archive and individual best position is updated based on Pareto dominance.The algorithm uses objective solutions linear density to measure non-dominated solutions quality and employs the strategy of deleting low density non-dominated solutions to enhance non-dominated solutions uniformity.To overcome the shortcoming of theoretical index in multi-objective evolution algorithm,a practical index is developed.Simulation results on benchmark functions show the proposed method can obtain a lot of non-dominated solutions,rapidly converge to the Pareto front and uniformly spread along the front.

Key words: particle swarm, multi-objective evolutionary algorithm, Pareto optimal, elitism strategy, archive technique

摘要: 提出一种改进的多目标粒子群优化算法,该算法采用精英归档策略,由档案库中的非劣解提供粒子速度更新时的全局最优位置,根据Pareto支配关系来更新粒子的个体最优位置。使用非劣解目标的线密度度量非劣解前端的均匀性,通过删除小密度的非劣解提高非劣解前端的均匀性。针对多目标进化算法理论型指标的不足,设计了应用型评价指标。标准函数的仿真实验结果表明,所提算法能够获得大量的非劣解,快速地收敛于Pareto最优解前端,且分布比较均匀。

关键词: 粒子群, 多目标进化算法, Pareto最优, 精英策略, 归档技术

CLC Number: