パノラマ画像に基づく移動ロボットのデュアル地図構築と自己位置推定

概要

論文の詳細を見る
To develop truly autonomous robots, it is essential that they learn an environmental map and localize themselves on the map. This paper proposes a dual (grid-based and topological) map building and localization architecture based on panoramic images. In the architecture, range information is estimated from panoramic images by Sobel edge detection. Rao-Blackwellized particle filters are used to conduct grid-based mapping and localization. Hidden Markov Models (HMMs) are used to recognize environmental features such as straight corridor, crossroads, L-shape, and T-shape junctions. Based on this architecture, a topological map is expected to be built. Experiments are carried out by a Webots simulator. Simulation results demonstrate that the proposed architecture can realize dual map building and localization simultaneously.
社団法人電子情報通信学会の論文
2010-03-02