一种基于杂草克隆的多目标粒子群算法  

A New and Efficient Multi-Objective Particle Swarm Optimization(MOPSO) Algorithm Based on Invasive Weed Cloning

作　　者：鲁鹏[1] 章卫国[1] 李广文[1] 刘小雄[1] 李想[1] 

机构地区：[1]西北工业大学自动化学院,陕西西安710072

出　　处：《西北工业大学学报》2012年第2期286-290,共5页Journal of Northwestern Polytechnical University

基　　金：航空科学基金(20090753008)资助

摘　　要：多目标粒子群算法(MOPSO)在优化函数时,尤其对于Pareto前沿是分段不连续的优化函数,存在收敛速度慢,种群多样性差的缺陷。针对此问题,将杂草克隆机制引入MOPSO,提出了一种新的多目标粒子群算法,称之为IWMOPSO。该算法利用改进的档案维护策略和不可行解增强多样性和均匀性,通过标准测试函数验证,能够使所求得的Pareto最优解逼近整个Pareto真实前沿,收敛性和多样性明显优于MOPSO和NSGA-Ⅱ,具有较强的应用性。When the existing MOPSO algorithm is applied to optimizing the functions with the discontinuous Pareto front,its convergence and the diversity of its population are poor.To solve the problem,we propose our new IWMOPSO（Invasive Weed MOPSO） algorithm,which we believe is more efficient than existing ones.Sections 1 through 2 of the full paper explain our new IWMOPSO algorithm.Section 1 presents the defects of the MOPSO algorithm.Section 2 explains how to reduce such defects to a minimum.Section 3 uses five benchmark test functions to compare the performance of our new IWMOPSO algorithm with those of the existing MOPSO and NSGA-Ⅱ algorithms.The test results,given in Tables 1 and 2 and Fig.7,and their analysis show preliminarily that both the convergence of our IWMOPSO algorithm and its diversity are enhanced by the improved file maintenance strategy and the unfeasible solutions,with the Pareto front obtained with our new algorithm very close to the real Pareto front,thus being more efficient than both the MOPSO and NSGA-Ⅱ algorithms.

关 键 词：多目标算法 粒子群算法 PARETO前沿 杂草克隆 MOPSO NSGA-Ⅱ 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]