移動体の意思決定のための情報量基準に基づく観測対象選択戦略

抄録

This paper proposes a method for constructing a decision tree and prediction ones of the landmarks that enable a robot with a limited visual angle to make decisions without self-localization in the environment. Since global positioning from the 3-D reconstruction of landmarks is generally time-consuming and prone to errors, the robot makes decisions depending on the appearance of landmarks. By using the decision and the prediction trees based on information criterion, the robot can achieve the task efficiently.