| 引用本文: | 章恩泽,陈庆伟.改进的r支配高维多目标粒子群优化算法[J].控制理论与应用,2015,32(5):623~630.[点击复制] |
| ZHANG En-ze,CHEN Qing-wei.Improved r-dominance-based particle swarm optimization for multi-objective optimization[J].Control Theory and Technology,2015,32(5):623~630.[点击复制] |
|
| 改进的r支配高维多目标粒子群优化算法 |
| Improved r-dominance-based particle swarm optimization for multi-objective optimization |
| 摘要点击 3763 全文点击 2206 投稿时间：2014-09-30 修订日期：2015-02-01 |
| 查看全文 查看/发表评论 下载PDF阅读器 |
| DOI编号 10.7641/CTA.2015.40924 |
| 2015,32(5):623-630 |
| 中文关键词 高维多目标优化 偏好 粒子群优化 |
| 英文关键词 multi-objective optimization preference particle swarm optimization |
| 基金项目 国家自然科学基金项目(61074023), 江苏省科技支撑计划项目(BE2012175), 江苏省普通高校研究生科研创新计划项目(KYZZ 0121)资助. |
|
| 中文摘要 |
| 高维多目标优化问题是广泛存在于实际应用中的复杂优化问题, 目前的研究方法大都限于进化算法. 本文 利用粒子群优化算法求解高维多目标优化问题, 提出了一种基于r支配的多目标粒子群优化算法. 采用r支配关系进 行粒子的比较与选择, 并结合粒子群优化算法收敛速度快的优势, 使得算法在目标个数增加时仍保持较强的搜索能 力; 为了弥补由此造成的群体多样性的丢失, 优化非r支配阈值的取值策略; 此外, 引入决策空间的拥挤距离测度, 并 给出新的外部存储器更新方法, 从而进一步防止算法陷入局部最优. 对多个基准测试函数的仿真结果表明所得解 集在收敛性、多样性以及围绕参考点的分布性上均优于其他两种算法. |
| 英文摘要 |
| Multi-objective optimization problems (MOPs) are complex optimization problems existing in practice, for which most of the modern research methods are focused on evolutionary algorithms. In this paper, a multi-objective particle swarm optimization algorithm based on the r-dominance is proposed for investigating the behavior of the particle swarm optimization (PSO) in MOPs. The combination of the r-dominance with the fast convergence properties of PSO maintains strong search capabilities of the algorithm when the number of objectives increases. In particular, the value of the nonr- dominance threshold is varied in an improved way in order to keep desired population diversity. Furthermore, a new updating strategy of the external repository, which incorporates the crowding distance in the variable space, is presented to get rid of the local optimum. Effectiveness of the proposed algorithm is validated by several benchmark test functions. Results indicate that the proposed algorithm outperforms two other existing algorithms in terms of convergence, diversity and distribution over the reference point. |
|
|
|
|
|