引入补偿采样和平滑算子的概率集群优化算法.pdfVIP

第27 卷 第4 期 控 制 与 决 策 2012 年 4 月 Vol. 27 No. 4 Control and Decision Apr. 2012 文章编号: 1001-0920 (2012) 04-0519-06 引入补偿采样和平滑算子的概率集群优化算法 王 楠, 胡天江, 陈 璟, 沈林成 ( 国防科学技术大学机电工程与自动化学院，长沙410073) 摘 要: 探讨一种新颖的分布式多智能体优化算法及其在求解约束优化问题上的实现方式. 通过分析该算法在有限 采样数下寻优能力所受到的影响, 提出在迭代优化过程中引入补偿采样机制和平滑算子的改进方案, 在保留原算法 优点的基础上提高了采样过程对决策空间的覆盖度, 增强了方法的全局和局部邻域搜索能力. 实验结果表明, 引入补 偿采样和平滑算子后的概率集群优化算法在收敛速度、解质量和稳定性等方面均得到了明显改善. 关键词: 多智能体系统；混合决策变量；概率集群；补偿采样；平滑算子 中图分类号: TP301.6 文献标识码: A Probability collectives with compensate sampling and smooth factors WANG Nan, HU Tian-jiang, CHEN Jing, SHEN Lin-cheng (College of Mechatronic Engineering and Automation，National University of Defense Technology ，Changsha 410073, China ．Correspondent ：WANG Nan ，E-mail ：acnwang@) Abstract ：：：This paper explores a novel distributed multi-agent optimization algorithm for solving constrained optimization problems. The performance of the algorithm is analyzed and the search capability is found to be influenced by the limitation of the sample number during iteration. Therefore, an improved algorithm is proposed with compensate sampling technique and smooth factors, which maintains the benefits of the original algorithm while improves its global and local search capability. Experimental results show that the proposed algorithm achieves obvious improvement in convergence speed, solution quality and long term stability. Key words ：：：multi-agent system ；mixed decision variables ；probability collectives ；compensation sampling ；smooth factor 1 引引引 言言言 策变量的概率分布而非策略值为主要优化对象, 能较 复杂系统的决策与优化问题在理论研究和工程 好地处理连续、离散及混合型决策变量, 并在收敛速 实践上均受到了广泛而持久的关注, 各种智能决策及 度、跳出局部极值能力和稳定性等方面具