マルチエージェント環境における遡行的信念推定アルゴリズム A Regressive Belief Estimation Algorithm in Multi - agent Environments

この論文にアクセスする

この論文をさがす

著者

抄録

マルチエージェント環境では,各エージェントが世界のごく一部しか知らず,ほかのエージェントもそれぞれの判断で活動しているので,他人がプランの実行を妨害することもあれば,自分の仕事を遂行するために他人の助けが必要なこともある.したがって,他人の知識(信念)を推定する手法は,交渉や協力において重要な役割を果たす.我々の観測は限られているので,論理的には膨大な数の可能性が考えられ,すべての可能性を調べる方法は現実的でない.しかし,人間はもっともらしい結論を導ける.これは何を考慮して何を無視すべきかの判断が,推定結果の品質と効率を大きく左右することを意味する.著者らはこれまでにいくつかの手法を提案したが,データの入力が困難であったり,単純な推論しかできなかったり,計算が複雑すぎたりした.本論文では,データを与えるのが比較的容易で,しかもある程度複雑な推論が可能で,効率の良い信念推定アルゴリズムを提案する.このアルゴリズムは,命題や事象がどのような条件のもとで観測できるかを記述した「観測条件」を利用して,他者の信念を推定する.実験によれば,このアルゴリズムの推論結果は多くの場合に直観的で分かりやすい.そして我々は,このアルゴリズムの扱う信念推定の論理的意味を明らかにする.エージェントの信念は論理的に不完全なので,論理的に不完全な信念の変化を表現するのに適した非標準的意味構造を用いる.In multiagent environments,each knows only a small part of the world,and other agents are working by following their own decisions.Sometimes others interrupt the execution of one's plan,and sometimes one needs others' help to achieve one's own task.Hence,a method for estimating others' bilefs plays an important role in order to negotiate and cooperate with them.Since limitation of our observations logically allows numerous possibilities,it is impractical to check all the possibilities.However,people derive plausible conclusions.It means that the quality and efficiency of an algorithm for estimating others' beliefs heavily depends on what is regarded and what is disregarded.We proposed several algorithms in the past,but they were unsatisfactory because of difficulty in data entry,poor inferential ability,or inefficiency.In this paper,we propose an efficient algorithm that is applicable to nontrivial cases without difficulty in data entry.This algorithm estimates others' beliefs by using observability conditions under which one can observe a proposition or an event.Moreover,we shed light on a logical meaning of the algorithm.Since agents are logically incomplete,we employ a nonstandard semantic structure that can represent changes of logically incomplete beliefs.

In multiagent environments, each agent knows only a small part of the world, and other agents are working by following their own decisions. Sometimes others interrupt the execution of one's plan and sometimes one needs others' help to achieve one's own task. Hence, a method for estimating others' beliefs plays an important role in order to negotiate and cooperate with them. Since limitation of our observations logically allows numerous possibilities, it is impractical to check all the possibilities. However, people derive plausible conclusions. It means that the quality and efficiency of an algorithm for estimating others' beliefs heavily depends on what is regarded and what is disregarded. We proposed several algorithms in the past, but they were unsatisfactory because of difficulty in data entry, poor inferential ability, or inefficiency. In this paper, we propose an efficient algorithm that is applicable to nontrivial cases without difficulty in data entry. This algorithm estimates others' beliefs by using obcervability conditions under which one can observe a proposition or an event. Moreover, we shed light on a logical meaning of the algorithm. Since agents are logically incomplete, we employ a nonstandard semantic structure that can represent changes of logically incomplete beliefs.

収録刊行物

  • 情報処理学会論文誌

    情報処理学会論文誌 38(3), 429-442, 1997-03-15

    一般社団法人情報処理学会

参考文献:  28件中 1-28件 を表示

被引用文献:  6件中 1-6件 を表示

各種コード

  • NII論文ID(NAID)
    110002721495
  • NII書誌ID(NCID)
    AN00116647
  • 本文言語コード
    JPN
  • 資料種別
    Journal Article
  • ISSN
    1882-7764
  • NDL 記事登録ID
    4159269
  • NDL 雑誌分類
    ZM13(科学技術--科学技術一般--データ処理・計算機)
  • NDL 請求記号
    Z14-741
  • データ提供元
    CJP書誌  CJP引用  NDL  NII-ELS  IPSJ 
ページトップへ