@techreport{oai:ipsj.ixsq.nii.ac.jp:00050979,
 author = {平山, 勝敏 and 山田, 誠二 and 豊田, 順一 and Katsutoshi, Hirayama and Seiji, Yamada and Jun-Ichi, Toyoda},
 issue = {86(1993-ICS-090)},
 month = {Sep},
 note = {分散人工知能の組織に関する研究の1つとして，分散制約充足問題　()　を大域的な情報を持たない複数エージェントが動的に組織を形成しながら解く方法を提案する．DCSPは，分散人工知能の問題を形式的に記述できる枠組みであり，その上での組織形成に関する議論には，かなりの一般性が期待できる．本稿では，まず，組織形成方法として，LMO　(cal　Minimum　driven　Organizati)　について説明する．これは，エージェントが局所最適解に陥ったときに組織を形成するという方法である．また，個々のエージェントの処理から導かれるマクロな挙動の特徴として，健全性と完全性を証明する．最後に，エージェント全体が，問題の難易度に応じて組織を形成し，集団としての適応性があることを実験的に示す．, We propose a method to solve Distributed Constraint Satisfaction Problem (DCSP) in which agents solve their own problems by organizing. DCSP gives us a framework for Distributed Artificial Intelligence. Thus, implementing the organizing in DCSP makes it possible to discuss the problems of organization independent of specific domains. We present LMO (Local Minimum driven Organizing) in which the agents organize when they get caught in local minima. This paper describes agent's behaviors and shows emergent properties resulting from individual agents' behaviors. One property is completeness and soundness. We prove it analytically. The other is that the more difficult DCSP agents solve, the larger groups they organize, i.e. they adapt themselves to the degree of difficulty. For verifying this property, we compare several societies which organize differently. As a result, the society with LMO better than the others.},
 title = {山登り法を用いた分散制約充足における組織化},
 year = {1993}
}