Rosenbrock搜索与动态惯性权重粒子群混合优化算法  

Hybrid optimized algorithms based on the Rosenbrock search method and dynamic inertia weight PSO

作　　者：贾树晋[1] 杜斌[2] 

机构地区：[1]上海交通大学自动化系,上海200240 [2]上海宝钢研究院自动化所,上海201900

出　　处：《控制与决策》2011年第7期1060-1064,共5页Control and Decision

摘　　要：为了提高复杂优化问题的优化精度和鲁棒性能,提出两种将Rosenbrock搜索与动态惯性权重粒子群(DIPSO)相结合的混合算法,即"协同"与"接力"混合算法.两种算法充分利用了Rosenbrock搜索算法强大的局部搜索能力和DIPSO算法的全局寻优能力,很好地平衡了算法的全局"探索"与局部"开发".通过4个典型基准函数的实验研究,表明了所提出的算法具有优化精度高、鲁棒性强等特点,适合于对高维多峰函数进行优化.In order to improve the accuracy and robustness performance of complex optimization problems,two hybrid algorithms,which combine the Rosenbrock search method and dynamic inertia weight PSO（DIPSO）,are proposed in this paper.The algorithms make full use of the powerful local search ability of the Rosenbrock search method and the global optimization ability of DIPSO algorithm,which well balance the global＂exploration＂and the local＂exploitation＂.The experiment study in four typical benchmark functions show that the proposed algorithms have the characteristics of high accuracy and robustness,and are suitable for optimizing high-dimensional and multimodal functions.

关 键 词：Rosenbrock搜索算法 粒子群优化算法 高维多峰函数优化 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]