Basic Research on Concept Construction of Artifacts by Collaboration among Multiple View Points

人工制品的概念构建需要基于多种思维的多个视点，而不是基于单一类型思维的对象或事件的一个视点，来解决真实的世界中的各种需求。在本研究中，研究了表示观点的方法，并揭示了当前状态下的三种类型。(1)视点的符号构造，其主要具有由人类抽象的属性组成的知识。利用这种观点和语法分析器的方法，实现了一个基于WWW的知识库，以展示用各种观点描述知识的能力，使其具有更广泛的应用。另一方面，用非符号方式的内部状态表示视点的方法被阐明如下。(2)视点的位构造，其将GA和交换文法应用于由位构成的染色体，并获取谓词和程序;（3）视点的状态构造，其调整连接和程序， ...更多信息 从实例中确定神经网络的权值，并通过自收敛得到最优状态。前者可以被赋予理解状态的语法或规则，而后者很难被赋予这样的语法或规则。应用结果表明，方法（1）适用于宏观处理，方法（2）适用于微观处理，方法（3）适用于直观处理。在这些方法的应用中，基于协作视点控制的概念构建，例如预先给定的或在协作中获得的视点之间的消息通信和协商，是基本有效的，因为使用单个视点的系统的完整性需要足够的准备和大规模的调试。研究了基于多Agent的协同控制方法，在对象及其关系具有部分不确定性的知识库和数据库上建立Agent系统，通过Agent之间的协商消除不确定性，并能够基于知识库和数据库模拟正确的协同运动。通过对二维机械图纸的理解和运动仿真，验证了该系统的有效性。少

The concept construction of artifacts need, to solve various demands in the real world, not one view point of objects or events based on a singlular type thought, but multiple view points based on various thoughts. In this study, methods to represent view points are investigated and revealed as three types shown bellow at the current state. (1) symbolic construction of view points which has mainly knowledge composed of properties abstracted by human. With this method of view points and syntactic parsers, a knowledge base over so-called www is implemented to demonstrate the ability of knowledge described with various view points for more wider application. On the other hand, methods to represent a view point with internal state of non-symbolic ways are clarified as follows. (2) bit construction of view points which applies GAs and an exchange grammar to chromosome composed of bits and acquires predicates and programs, (3) state construction of view points which adjusts connections and t … More heir weights of Neural Networks from examples and results in an optimum state by self convergence. The former can be given grammars or rules to comprehend states, but the latter is difficult to give such grammars or rules. As results of applications of the methods, it proved that method (1) suits for macro processing, method (2) for miro processing, and method (3) for intuitive processing. On application of the methods, concept construction based on collaborative view points control, such as message communication and negotiation, among view points given beforehand or acquired in the collaboration is essntially effective because the completeness of a system using a singular view point needs enough preparations and debug of a large scale. Investigation into methods of the collaborative control based on multi agents results in building, on knowledge and data base having partially uncertainty of objects and their relations, an agent system which eliminats the uncertainty with negotiation among agents and enables to simulate corrective and collaborative motion based on the knowledge and data base. The validity of this system is confirmed in comprehension and motion simulation of 2D mechanical drawings. Less