A Study of Genetic Network Programming with Reinforcement<br>Learning and its Application

カテゴリ: 部門大会

論文No: GS15-1

グループ名: 【C】平成17年電気学会電子・情報・システム部門大会講演論文集

発行日: 2005/09/06

タイトル(英語): A Study of Genetic Network Programming with Reinforcement
Learning and its Application

著者名: 間普真吾 (早稲田大学),Thu Thu Moe (早稲田大学),平澤 宏太郎(早稲田大学),古月 敬之(早稲田大学)

著者名(英語): Shingo Mabu(Waseda University),Thu Thu Moe (Waseda University),Kotaro Hirasawa(Waseda University),Takayuki Furuzuki(Waseda University)

キーワード: Evolutionary Computation|Genetic Network Programming|Reinforcement Learning|Khepera robot

要約(日本語): Genetic Network Programming, GNP, a new graph-based evolutionary algorithm, was proposed. The solution of GNP represents as graph structures. That can improve the expression ability and performance. And then, GNP with Reinforcement Learning (RL) has been proposed to search for solutions efficiently. GNP with RL can use the current information and change its program while task execution. Then, it has an advantage over the evolution-based algorithms in case much information can be obtained during task execution. The GNP we proposed in the previous research deals with discrete information, but from now we extend the conventional GNP with RL method to deal with continuous information. In this paper, the proposed method is applied to the controller of Khepera simulator and its performance is evaluated.

PDFファイルサイズ: 5,726 Kバイト