Toward robust sensor based exploration by constructing reduced generalized Voronoi graph

Abstract

Considers sensor based mobile robot exploration in an unknown indoor environment. One of the big problems in this area is positioning error. To deal with this problem in the exploration task, we proposed a localization method using a partially constructed topological map termed the generalized Voronoi graph (GVG). The GVG-based method performed successfully in some large scale indoor environments, but had problems resulting from poor sensing ability of the sixteen sonar sensors on-board the robot. The poor sensing ability was particularly deleterious at geometric structures in the GVG that were "unstable." To solve the problem, we adopted "the Reduced Generalized Voronoi Graph (RGVG)" for the exploration map instead of the original GVG. The RGVG is a stable subset of the GVG and is sufficient for motion planning. To enable realistic and reliable exploration using the RGVG structure, we developed an edge matching procedure for topology matching. In this paper, we introduce the RGVG structure and the edge matching procedure for robust exploration. Experimental results verify the described work.