【作者】 张桂娟；

【导师】 刘希玉；

【作者基本信息】 山东师范大学 ， 计算机软件与理论， 2006， 硕士

【摘要】 协同进化算法是在近十几年来在协同进化论基础上发展起来的一类新的进化算法。协同进化算法与传统进化算法的区别在于:协同进化算法在进化算法的基础上,考虑了种群与种群之间、种群与环境之间在进化过程中的协调。由于协同进化算法的诸多优越性,越来越多的学者对它进行了研究,协同进化已成为当前进化计算的一个热点问题。随着市场竞争的日趋激烈,社会的消费观念不断发生变化,产品的功能已不再是消费者决定购买的最主要因素。产品的创新性、环保性、宜人性、外观造型等因素愈来愈受到重视,在竞争中占据突出地位。在这种情况下,许多企业都意识到产品设计将成为抢占市场的决定性因素,因而试图通过产品的创新性来获得核心竞争力。创新设计能够满足求新和多变的市场需求,有效地提高产品的市场竞争力。因此,创新设计研究倍受国内外相关学者的关注。到目前为止,尚未有一种系统化的方法引导设计者有条理的进行创新设计。大多数的设计实例不是凭空创造新产品,而是在现有产品的基础上对其进行修改,以改进产品的质量和性能,满足用户需求。进化计算技术的兴起,为我们探索一种支持进化的概念创新设计途径提供了很好的支持。在此基础上产生的进化设计已成为最重要的创新设计技术之一。本文致力于研究自适应协同进化模型的建立及其应用问题,目的是通过研究新的、适用于创新设计的协同进化机制,将已有的进化设计的研究从单体进化拓展到多群体协同进化,为分布式环境下的设计人员的协作和创新思路的开拓提供支撑平台。主要工作为:1.提出一种基于学习的自适应协同进化机制传统的自适应遗传算法重在研究进化参数的自适应性,但是参数选择的复杂性与问题的目标函数有直接联系,而在自然界中生物个体的适应方式是多元化的。以此为启示,并以协同进化的生物学理论为指导,本文提出一种基于学习的自适应协同进化机制。引入三个自适应学习算子,将学习和进化两种不同的自适应过程结合起来,改善了算法的性能。仿真实验结果表明了该机制的有效性。2.提出一种适于创新设计的最佳匹配合作者选择机制现有的合作型协同进化算法,在进化中选择合作个体问题上具有随机性、经验性并且与特定问题域有关。本文针对创新设计的特点,设计了最佳匹配合作者选择机制。利用已有的设计经验,为设计个体的部分解组成完全解提供有效支持,提高了设计的智能化水平。3.构造了一个支持创新设计的自适应协同进化模型针对创新设计的独特性,本文在自适应协同进化机制的基础上,构造了一个支持创新设计的自适应协同进化模型。设计了相应的自适应学习算子及个体评测标准,并利用最佳匹配合作者选择机制组合设计主体。通过自适应学习的引入,该模型为自动而有效地使用更多还原

【Abstract】 Co-evolution algorithm has been developed as a kind of new evolution algorithm extensivelyon the base of co-evolution theory for a decade. The difference between co-evolution algorithmand traditional evolution algorithm is that co-evolution algorithm takes the adjustment ofpopulations as well as that of populations and environment into account in evolving process. Forthe advantages of co-evolution algorithm, more and more scholars contribute to this field, whichmakes it a hotspot in evolution computation.Because of the ever sharpness of the market competition, the expenditure idea has changed alot. People no longer treat product function as the main factor of whether to buy something ornot. Much attention has been paid to the attributes of creativity, outline, delightfulness as well asenvironment protection, which stand the important roll in competition. Facing this reality, manyenterprises begin to realize the importance of product design and pay more attention to creativeproducts. Creative design can satisfy the novel and ever-changing market demands and help tomake the product more competitive effectively. Thus, creative design has been a popularresearch item attracts many researchers in relevant fields home and abroad.By far, no systematic method can direct designers to carry out creative design orderly. In fact,most of the instances are created by modifying the products already existed rather than designwithout any foundation. The rise of the evolution computation provides us a new approach forconceptual creative design;therefore, it has become one of the most important techniques forcomputational and creative design.Much attention has been paid to the construction and application of an adaptive co-evolutionmodel in this paper. So that it can extend evolution design from single population tomulti-populations and provide a supporting platform of collaboration and creation for distributeddesigners by studying a new co-evolution mechanism especially for creative design. The mainwork is as follows:1. Propose an adaptive co-evolution mechanism based on learning.Traditional Adaptive Genetic Algorithms pay too much attention to the adaptability of theirevolution parameters, while the complexity of choosing parameters is directly depends on theobject function. The variety of adaptive styles in nature and the biology theory of co-evolutionenlighten us to bring forward the adaptive co-evolution mechanism. Three adaptive learningoperators are introduced and two adaptive processes of learning and evolution are combined inthis mechanism, with which the performance is improved effectively in experiments.2. Put forward a best-matching selection mechanism for designCooperative Co-evolution Algorithms show some randomicity and rely on the speciality ofgiven problems in choosing collaborating individuals. The best-matching selection mechanismespecially for creative design is put forward in this paper. It can provide effective informationfor composing of a complete solution;thus, enhance the intelligence level in design process.3. Construct an adaptive co-evolution model supporting creative designIn view of the uniqueness of creative design, an adaptive co-evolution model is constructedon the base of previous mechanisms. With the best-matching selection mechanism utilized, aseries of operators as adaptive learning and individual evaluation are designed. It affords thefeasibility of using design experience automatically and effectively. Also it shows theimpartibility of the artistic attributes as well as the natural adaptability of a design works via theadaptive operation.4. Implement an architectural creative design system based on the co-evolution modelThe adaptive co-evolution model is applied to architectural design field and a design system,on the base of the model, is realized. With HOOPS/NET adopted as the main skeleton, andentity modeling engine ACIS utilized as the sculpt kernel, the system is designed onWindowsXP platform with VC++.NET 2003. In addition, SQL Server 2000 is used as databasesystem. Finally, the analysis of instances shows that the design results are quite satisfying whichis based on adaptive leaning among populations and that of populations and environment.更多还原