SIA OpenIR  > 信息服务与智能控制技术研究室
Alternative TitleResearch of Bee Colony Optimization Model, Algorithms and Application Based on Hierarchical Co-evolution Mechanism
Thesis Advisor胡琨元 ; 陈翰宁
Keyword协同进化 生物群落 群体智能 人工蜂群算法 动态优化
Call NumberTP18/M16/2014
Degree Discipline机械电子工程
Degree Name博士
Degree Grantor中国科学院沈阳自动化研究所
Place of Conferral沈阳
Abstract本文根据生物系统中普遍存在的共生进化、层次化涌现、种群动态迁移等现象,利用基于个体模型(Individual Based Modle, IBM)建模方法,抽象出基于层次协同演化模式的统一计算框架及模型,并利用其改进传统的人工蜂群算法,打破传统优化算法的局限性,最后将新模型及算法应用于实际复杂问题求解,以验证其有效性。本文的主要研究工作概括如下: (1)借鉴复杂适应系统层次结构和生态群落共生进化思想,提出了基于层次协同演化模式的多种群进化模型(Hierarchical Co-evolution Model, HCOM)。HCOM模型从协同进化过程的本质出发,定义了个体-种群-群落层次,描述了协同进化共性机制,其模型框架涵盖了生物群落进化过程中的种群演化模式,以及个体层、种群层、群落层的基本操作与进化策略. (2) 在三层HCOM框架基础上,提出一种基于群落层次演化的混合多巢人工蜂群优化算法(Hybrid-Multi-Hive Artificial Bee Colony, HABC)。针对大规模复杂问题,HABC算法采用基于分治(Divide-And-Conquer)策略的维度分解方法,将复杂高维优化问题分解为相对简单的子问题,由并行的子群体进行协同进化。采用交叉操作与精英策略增强种群间信息交流,以保证群体多样性。采用随机分组策略解决在没有先验知识条件下的强耦合变量分解问题。 (3) 引入外部档案和拥挤策略处理多目标非支配解,将单目标HABC算法扩展为多目标混合蜂群优化算法(MOHABC)。MOHABC算法将协同进化机制和非支配解记忆存储机制思想相结合,使之具有较好的搜索能力和较强的收敛性。同时采用外部档案用来保存获得的帕累托(Pareto)解集,计算拥挤距离以淘汰外部档案中的冗余解。 (4) 针对复杂动态优化环问题特征,引入了基于生命周期的种群动态迁移、强化学习、基于Powell方法的局部精细搜索等策略,设计了一种能够快速适应环境变化且性能优良的基于种群动态迁移的混合蜂群算法PMHABC),并应用于动态无线网络的读写器部署规划。。 (5) 以复杂微滴图像的智能分割与全印制电子喷射系统中的微滴图像特征提取为应用背景,提出了基于多阈值聚类的微滴图像分割算法。在基于传统的Otsu与Kapur熵判据估计基础上,通过引入群落演化与迁移协同优化机制,包括基于增强学习策略与局部精细搜索操作,平衡了算法在探索与开发之间的平衡。
Other AbstractBased on the symbiotic evolution, the hierarchial emergence, the population dynamics and the migration phenomenon in the the biological system, by using the individual based modelling (IBM) method, this paper abstracts unified computing framework and model, and improves the performance of the Artifical Bee Colony (ABC) algorithm to overcome the limitations of the traditional optimization algorithms. Finally, this paper applies these new models and algorithms to resolve the practical complex problem to verify their validiy. The main contents of this paper can be generalized as follows: (1) By incorporating the concepts of the hierarchical structure of CAS and symbiotic evolution of the ecological community, this paper proposes a new multi-population evolutionary model, namely Hierarchical Co-evolution Model, HCOM. (2) On the basis of the three-level HCOM framework, this paper proposes a hybrid-multi-hive artificial bee colony algorihm based on the community-level evolution pattern, called HABC. For the large-scale complex problems, HABC uses the divide-and-conquer approach to decompose the high-dimensional problem into relatively simple sub-tasks, which will be tackled by involved sub-groubs in parallel. At the same time, the elitist selection and crossover operators are incorporated to enhance information exchange between populations, maintaining diversity of the population. The random grouping strategy is used to solve the strong coupling variable conditions without prior knowledge of the decomposition problem. (3) By applying the external archive and crowded distance strategies for dealing with multi-objective non-dominated solutions, this paper extends the single objective HABC algorithm to multi-objective hybrid artificial bee colony algorithm (MOHABC). MOHABC combines the co-evolution mechanism and the non-dominated solution memory storage concept, getting good search ability and strong convergence. Meanwile, the external archive is ued to store Pareto solution set and the crowded distance is employed to eliminate redundant solutions of external archive. (4) According to the characteristics of the complex dynamic optimization problems, this paper incorporates the lifecycle-based dynamic population migration, reinforcement learning, and local search based on Powell’s method to design a new optimization algorithm, namely Population-migrating-based Hybrid Artificial Bee Colony (PMHABC) algorithm, which can quickly adapt to the environmental change. Then the PMHABC is applied to the reader deployment planning of the dynamic wireless network. (5) Aiming to the application background of the intelligent segmentation of complex droplet images and the image feature extraction of printed electronic system, this paper proposes a new droplet image segmentation algorithm based on the multi-threshold clustering. On the basis of the entropy evaluation criteria including the traditional Otsu and Kapur, by incorporating the community-level co-evolutation optimization mechanism, the enhancement learning and local fine search strategies can provide appropriate balance between exploration and exploitation of the algorithm.
Contribution Rank1
Document Type学位论文
Recommended Citation
GB/T 7714
马连博. 基于群落层次协同演化的蜂群优化模型、方法及应用研究[D]. 沈阳. 中国科学院沈阳自动化研究所,2014.
Files in This Item:
File Name/Size DocType Version Access License
基于群落层次协同演化的蜂群优化模型、方法(9360KB) 开放获取CC BYApplication Full Text
Related Services
Recommend this item
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[马连博]'s Articles
Baidu academic
Similar articles in Baidu academic
[马连博]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[马连博]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.