Improving Autonomous Exploration Using Reduced Approximated Generalized Voronoi Graphs

Lin Li, Xinkai Zuo, Huixiang Peng, Fan Yang, Haihong Zhu, Dalin Li, Jun Liu, Fei Su, Yifan Liang, Gang Zhou

Abstract: Autonomous robotic exploration has been extensively applied in many tasks, such as mobile mapping and indoor searching. One of the most challenging issues is to locate the Next-Best-View and to guide robots through a previously unknown environment. Existing methods based on generalized Voronoi graphs (GVGs) have presented feasible solutions but require excessive computation to construct GVGs from metric maps, and the GVGs are usually redundant. This paper proposes an improving method based on reduced approximated GVG (RAGVG), which provides a topological representation of the explored space with a smaller graph. Additionally, a fast and robust image thinning algorithm for constructing RAGVGs from metric maps is presented, and an autonomous robotic exploration framework using RAGVGs is designed. The proposed method is validated with three known common data sets and two simulations of autonomous exploration tasks. The experimental results show that the proposed algorithm is efficient in constructing RAGVGs, and the simulations indicate that the mobile robot controlled by the RAGVG-based exploration method reduced the total time by approximately 20% for the given tasks.