Development of a Decision-Support System Based on Emergent Interaction

当考虑大规模复杂系统中的工程问题时，包含双向交互的紧急方法是重要的，它使我们能够以自适应或交互的方式提取决策模型。本研究以设计一种将建模过程嵌入到求解过程中的方法为目标，构建了交互式信息场和紧急计算方法。要点如下。(1)重构紧急计算方法——基于强化和基于遗传的机器学习技术，设计了一种基于仿真的决策规则获取和调整方法。(2)交互式信息场(I)的设计与实现——设计了一个层次自治的去中心化框架，该框架明确考虑了子模型之间的交互，并研究了基于该框架的仿真与决策模型原型。(3)交互式信息场的设计与实现（II）——基于该框架开发了一种决策与学习方法，其中智能体和接口将共同进化，并研究了基于该框架的学习智能体模型原型

When considering an engineering problem-solving in large-scaled and complicated systems, an emergent approach containing bi-directional interactions is important, which enables us to extract a decision-making model in an adaptive or an interactive manner. In this research, aiming at designing a methodology in which a modeling process is embedded in a solution process, an interactive information field as well as an emergent computing method have been (re)constructed. The key points are as follows.(1) Reconstruction of emergent computing method - a simulation-based method for acquiring and adjusting decision-making rules has been designed, based on both reinforcement and genetics-based machine learning techniques.(2) Design and implementation of an interactive information field (I) - an hierarchical autonomous decentralized framework in which an interaction among sub-models are explicitly considered has been designed, and an prototype of a simulation and decision-making model based on the framework has been investigated.(3) Design and implementation of an interactive information field (II) - a decision-making and learning method has been developed based on such framework where an agent and an interface are to be co-evolved, and an prototype of an learning agent model based on the framework has been investigated