Emergent Theory for System Design

结合人工制品系统设计的突现性，研究了人工制品设计者角色的数学表达和设计过程的支持系统。主要研究结果如下：1.设计过程的数学模型作为Yoshikawa等人提出的一般设计理论的扩展，Barwise和Seligman基于信息流的概念发展了一种抽象设计理论。给出了设计框架（性能指标空间、约束空间和参数空间）的数学定义，并讨论了设计知识重用的有效性.在设计过程中引入突现方法为辅助人类设计者，将遗传规划扩展到多目标问题。该系统支持设计人员通过显示计算机生成的多个候选方案，设计人员可以根据自己的偏好选择候选方案，并对其进行改进以获得最终的解决方案.作为交互工程的设计过程分析提出了设计过程中的创造力模型，并利用该模型进行了计算机辅助设计的尝试.系统突现设计理论的评价与验证本文提出的设计理论分别针对以下三个问题进行了评价与验证：基于多目标遗传规划的线性无源滤波器设计、基于遗传机器学习的调度规则生成以及突现-分散决策模型的构建。

Relating the emergent property to systems design of artifacts, the mathematical formulation of human designers' role and the supporting systems for design procedure were studied. The followings were the results obtained.1. Mathematical models for design procedure As an extension of the general design theory proposed by Yoshikawa et al, an abstract design theory was developed based on the concept of information flow by Barwise and Seligman. Also the design framework(e.g. the space of performance index, constraints and parameters) was mathematically defined and the effectiveness of reuse of design knowledges were discussed.2. Introduction of emergent methods to design procedure For assisting human designers, the genetic programming was extended for multi-objective problem. The system supports the designers by showing many solution candidates generated in computer, where the designers can select a candidate by their preference and improve it to get a final solution.3. Analysis of design procedure as interactive engineering A model for the creativity in design procedure which has been one of the important role of human designers, was proposed and computer aided design was tried by using this model.4. Evaluation and verification of emergent system design theory The design theory developed here were evaluated and verified for following three problems : the design of linear passive filters by multi-objective genetic programming, the generation of scheduling rules by genetic machine learning, and the construction of autonomous-decentralized decision model.