A Space Classification-based Hybrid Exploration Strategy for Narrow and Wide-Open Spaces

Abstract

This paper addresses a problem of mobile robot exploration and map building for narrow space and wide-open space with 20 metric sensor. We estimate the type of environment with three steps of criteria, and apply different target pose extraction methods for each space. For the wide-open space, we use a concept of inverse skeleton based strategy which is proposed in this paper, while the narrow space, we use the generalized Voronoi diagram (GVD) for target point extraction. For the overlapped region of narrow space and wide-open space, we suggest a node extraction method using the uncertain region detection strategy and the GVD. With the experiments in the real environment, we show that the proposed method can be a solution for exploration with hybrid map, in indoor environments with narrow space and wide-open space.