101 強化学習によるロボットの知的制御 : Lego Mindstorms への実装

• 宮崎 和光

  大学評価学位授与機構

書誌事項

タイトル別名

• Theory of Reinforcement Learning based on Profit Sharing : An Application for (LegoMindstorms)^<TM>

抄録

Reinforcement Learning is a kind of machine learning. It aims to adapt an agent to a given environment with a clue to rewards. If we apply reinforcement learning systems to real world, we cannot bear to repeat numerous trials. We want the agent to acquire a rational policy very quickly even if it does not know all the environment. Furthermore, it is important to apply to non-Markovian environments such as Partially Observable Markov Decision Processes (POMDPs). In this paper, we focus on Profit Sharing (PS) and discuss the rationality of PS in POMDPs. By applying the algorithm to legoMindstorms^TM, its effectiveness is shown.

収録刊行物

• Dynamics and Design Conference : 機械力学・計測制御講演論文集 : D & D

  Dynamics and Design Conference : 機械力学・計測制御講演論文集 : D & D 2000 (0), 1-, 2000-09-01

  一般社団法人日本機械学会

キーワード

• Reinforcement Learning
• Profit Sharing
• Q-learning
• Rationality Theorem
• (LegoMindstorms)^<TM>

詳細情報

• CRID: 1570291227238115456
• NII論文ID: 110002482366
• NII書誌ID: AA11901770
• ISSN: 13480235
• 本文言語コード: ja
• データソース種別:
  • CiNii Articles