1A2-E27 パーティクルフィルタを用いた全方向移動ロボットの自己位置同定における効率的な探索法

概要

論文の詳細を見る
This paper presents an efficient exploration strategy for the self-localization of omnidirectional mobile robots based on particle filtering. It is a challenging task to identify the robot's global position in a workspace based only on local sensory information, especially if there exist multiple locations where similar measurement signals are obtained. A popular existing technique called Active Localization, which chooses the next action so as to minimize the entropy of the expected posterior distribution of the robot's position, cannot be directly applied to particle-based localization, since entropy cannot be evaluated easily when a probability distribution is expressed by particles. Instead, in the proposed method, the robot moves to a position where the variance of" the expected measurement signal is maximized. The variance of a particle distribution is easy to compute and therefore the proposed method is well suited to particle-based localization. The proposed technique is tested in an in-door environment.
一般社団法人日本機械学会の論文