2P1-F30 連続な状態行動空間において学習可能なQ-learningの提案

書誌事項

タイトル別名
  • 2P1-F30 Q-learning in Continuous State and Action Spaces

抄録

This paper proposes the new Q-learning that can learn mapping from continue state spaces to continue action spaces. The proposed method estimates the expectation value of actions on a state by using artificial neural networks, and decides an action according to the distribution of the estimated expectation value. In this paper, we investigate the performance of the proposed method through two types of simple experimentations.

収録刊行物

詳細情報 詳細情報について

問題の指摘

ページトップへ