基于最优觅食理论的新型生物启发式计算方法研究及应用

CORC > 沈阳自动化研究所 > 中国科学院沈阳自动化研究所 > 信息服务与智能控制技术研究室

题名	基于最优觅食理论的新型生物启发式计算方法研究及应用
作者	刘洋
学位类别	博士
答辩日期	2015-05-28
授予单位	中国科学院沈阳自动化研究所
授予地点	中国科学院沈阳自动化研究所
导师	胡琨元
关键词	生物启发式计算最优觅食理论人工蜂群算法菌群优化算法
其他题名	Research and Application of New Bio-inspired Computing Method Based on Optimal Foraging Theory
学位专业	机械电子工程
中文摘要	在计算智能领域，生物界某些个体或群体的行为特征、演化特性给予研究人员很多启示，许多模拟生物行为和现象的优化算法应运而生，上述研究统称为生物启发式计算方法。生物行为表现在多个方面，其中：觅食行为是生物的生存及繁殖重要的基本特性，不同类型的生物，从低等的单细胞细菌到高等的动物都具有不同的觅食行为模式，有关模拟生物觅食行为规律的启发式方法自提出以来，一直受到国内外学者和工程技术人员的广泛关注。尽管基于生物觅食行为的启发式计算研究日趋成熟，但通过分析现有工作可以看出，在求解实际复杂问题过程中，如何实现算法的多样性保持策略、兼顾全局与局部搜索的均衡策略以及算法参数自适应优化策略，有效克服早熟收敛、提高搜索效率和收敛精度等方面尚存在较大的改进空间。本文利用自然生物最优觅食理论、复杂自适应系统等成果，在国内外生物启发式计算相关工作的基础上，从生物建模、算法设计和工程应用层面，针对基于觅食行为的生物启发式算法展开了深入系统地研究，并结合数据聚类分析、彩色图像处理等典型实际问题设计了新的求解方法。论文开展的主要工作如下：（1）针对传统基于单层生物启发式优化模型的原始蜂群算法(Arti?cial Bee Colony,ABC)存在“早熟收敛”问题，将层次型信息交流拓扑结构引入人工蜂群觅食模型中，提出基于层次型信息交流机制的多蜂群协同进化算法，实现在搜索过程中能够维持整个群落种群多样性的群落级进化。通过仿真试验表明，该方法能够有效的保持整个群体的多样性，在一定程度上平衡了探索开发能力，有效提升了算法的收敛速度与收敛精度。（2）从能量变化角度出发构建生物生命周期优化模型，在此基础上，针对传统的菌群优化算法（Bacterial Foraging Algorithm，BFA）进行改进，设计了一种基于生命周期的菌群觅食自适应优化方法。将E.coli种群按照生命周期进行演化，即E.coli个体在觅食过程中根据其能量获取与消耗状态动态地分裂、死亡和迁徙，种群规模随环境变化而进行适应性变化。通过仿真实验表明，本文建立的E.coli菌群优化模型符合微生物生命周期变化规律，函数测试结果验证了算法具有较好的优化性能。（3）针对传统模糊C均值（Fuzzy c-Means，FCM）聚类算法存在易陷入局部极小值，对初始值和噪声数据敏感等不足，引入基于层次型信息交流机制的多蜂群协同进化思想，提出基于MCABC-FCM的聚类优化算法，并应用于求解教学评估问题中。实例仿真表明，相对于传统FCM聚类算法，该方法的寻优能力、收敛速度得到显著提高，与此同时，评价效果更具有代表性。（4）将基于生命周期的菌群觅食自适应优化算法用于彩色图像处理中，提出一种新的多阈值分割算法，融合群体并行性搜索且不易陷入局部最优的特点，以寻找图像分割的最优阈值组合，并最大限度的提高寻优精度和效率。通过给定图像的实例仿真证明，该方法的分割结果更加精确，并且极大地降低多阈值分割的计算时间，为解决类似工程问题提供了新的思路。综上所述，本文从机理建模、算法设计和工程应用层面针对典型的生物觅食行为启发式计算方法进行研究，取得了具有创新性和应用价值的成果，所提出的改进策略和优化方法对于拓展相关领域的研究、指导实际应用都将具有一定的借鉴意义。
索取号	O241/L76/2015
英文摘要	In the field of computational intelligence, behavioural characteristics and evolution characteristics of biological certain individuals or groups give researchers a lot of inspiration, a lot of optimization algorithm which simulate biological behaviour and phenomenon emerged, all these above studies referred to as bio-heuristic calculation method. Biological behaviour shows in many aspects, including: foraging behaviour is organism’s important basic characteristics for survival and reproduction, different types of creatures, from low single-cell bacteria to higher animals have different foraging behaviour patterns, and related simulation of biological foraging behaviour heuristics method has been paid extensive attention by domestic and foreign scholars, engineers and technicians since proposed. Although heuristic computing research based on biological foraging behavior more and more matures, by analyzing the existing work can be seen that in the process to solve complex practical problems, how to implement the algorithm diversity preservation strategy, taking into account the balance of global and local search strategies and adaptive algorithm parameter optimization strategy, overcome premature convergence, improve search efficiency and convergence precision, all above aspect still have large room for improvement. This paper use natural biological optimal foraging theory, complex adaptive systems result, based on foreign biological heuristic computing, from biological modeling, algorithm design and engineering application, to in-depth research biological heuristic computing based on foraging behavior, combined with data clustering analysis, color image processing, designed a new method for solving. The main work and innovations are carried out as follows:（1）To solve traditional Artificial Bee Colony(ABC) based on single biological heuristic optimization model have "premature convergence" problem, introduce the hierarchical information exchange topology structure to artificial bee colony foraging model, and proposed multi-bee colony convolution algorithm based on hierarchical information exchange mechanism, implement to maintain community parent population diversity community optimization during the search process. The simulation results show that this method can effectively maintain the diversity of the entire group, to a certain extent, the balance of the exploration and development capabilities, effectively increasing the convergence rate and convergence precision.（2）From energy change expect to construct bacterial biological cycle optimization model, to verify the consistency of the model and the actual microbial systems through simulation experiment, improve the traditional Bacterial Foraging Algorithm (BFA), and design bacterial foraging adaptive optimization methods based on the life cycle. E.coli population follow the life cycle of evolution, means that the individual E.coli in process of foraging dynamic split, death and migration according to their energy obtain and consumption status, population size adaptive change by following the environment change. The simulation results show that, E.coli bacterial population optimization model established in this paper match the changing patterns of microbial life, function test results verify the algorithm has better optimize performance.（3）To solve the traditional Fuzzy c-Means(FCM) clustering algorithm is easy to fall into local minima, less sensitive to initial value and noise data, introduce multi-bee colony convolution ideas based on multi-level type of information exchange mechanisms, put forward the clustering optimization algorithm based MCABC-FCM, and used to solve problems in teaching evaluation. The simulation showed that, compared with the traditional FCM clustering algorithm, the optimization searching capability and the convergence rate of this method has been significantly improved, at the same time, the evaluative effectiveness is more representative.（4）Introduce bacteria foraging adaptive optimization algorithm based on the life cycle into color image processing, propose a new multi-threshold segmentation algorithm, integrate the feature of group parallelism searching and avoiding local optimum to find the best threshold combination of image segmentation, and to maximized improve optimization accuracy and efficiency. A given image simulation shows that this segmentation method result is more accurate, and greatly reduce the computation time of multi-threshold segmentation，for solving similar engineering problems offers a new way of thinking. In summary, this paper research typical creatures foraging behavior heuristic calculation method covered mechanism modeling, algorithm design and engineering application, and achieve innovative and application value result, the proposed improved strategy and optimized methods have a certain significance meanings to expand research and to guide the practical application in related fields.
语种	中文
产权排序	1
页码	113页
内容类型	学位论文
源URL	[http://ir.sia.ac.cn/handle/173321/16752]
专题	沈阳自动化研究所_信息服务与智能控制技术研究室
推荐引用方式 GB/T 7714	刘洋. 基于最优觅食理论的新型生物启发式计算方法研究及应用[D]. 中国科学院沈阳自动化研究所. 中国科学院沈阳自动化研究所. 2015.