基于三方竞争机制的反向多目标粒子群优化算法  

Opposition-based multi-objective particle swarm optimization algorithm based on tripartite competition mechanism

作　　者：韩飞[1,2] 郑明鹏 HAN Fei;ZHENG Mingpeng(School of Computer Science and Communication Engineering,Jiangsu University,Zhenjiang,Jiangsu 212013,China;Jiangsu Key Laboratory of Security Technology for Industrial Cyberspace,Jiangsu University,Zhenjiang,Jiangsu 212013,China)

机构地区：[1]江苏大学计算机科学与通信工程学院,江苏镇江212013 [2]江苏大学江苏省工业网络空间安全技术重点实验室,江苏镇江212013

出　　处：《江苏大学学报（自然科学版）》2021年第6期685-693,共9页Journal of Jiangsu University：Natural Science Edition

基　　金：国家自然科学基金资助项目(61976108,61572241)。

摘　　要：针对传统多目标粒子群优化算法容易早熟的问题,提出一种基于三方竞争机制的反向多目标粒子群优化算法(MOPSO-TCOL).该算法利用当前种群在每一代中选择的三方竞争者来引导种群进化,这能够有效减少维护外部存档时的计算成本.在每次竞争中,MOPSO-TCOL从种群中随机挑选3个粒子进行比较,并基于不同的策略分别进行更新,这有利于保持种群的多样性.提出了一种基于反向学习策略的渐进式粒子更新方式,部分粒子进行反向学习以避免算法陷入局部最优,其他粒子通过向指定的更优粒子学习进行更新以加强收敛性.将所提出算法与8个多目标优化算法在14个标准测试函数上进行了性能比较试验.结果表明MOPSO-TCOL算法在多样性和收敛性上具有显著优势,且具有更快的收敛速度.To solve the problem of premature in traditional multi-objective particle swarm optimization algorithms,an opposition-based multi-objective particle swarm optimization algorithm was proposed based on tripartite competition mechanism(MOPSO-TCOL).The tripartite competitors were selected from the current population in each generation to guide the population evolution in MOPSO-TCOL,which could effectively reduce the computational cost.In each competition,three particles were randomly selected from the population for comparison and updated by different strategies,which could help to maintain the diversity of population.A novel progressive particle update strategy was proposed based on opposition-based learning(OBL).Some particles were updated by OBL strategy to avoid the algorithm from falling into local optima,and other particles were updated by learning from the specified better particles to improve the convergence.The experimental results on 14 benchmark test instances verify the superiority of the proposed algorithm in terms of diversity and convergence over 8 multi-objective optimization algorithms,and it has faster convergence rate.

关 键 词：粒子群优化算法 多目标优化 进化算法 三方竞争机制 反向学习 PARETO前沿 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]