1A1-B07 未知物体認識のためのPOMDPを用いた探り動作計画(3次元計測/センサフュージョン(1))

Abstract

The authors aim at finding and modeling unknown objects in human life environment using robot technologies. In this paper, the authors proposed a method of unknown object segmentation using manipulation. Partial Observation Markov Decision Process (POMDP) is used for the planning. POMDP requires states for the estimation of the unknown objects. However, it is hard to define the states of the unknown objects. Therefore, the authors used primitive shape (e.g., quadrangular prism, ball) as the states. Use of primitive shape allows the estimation of unknown objects by using POMDP. We show a simulation result of the unknown object finding.