Topological map building and exploration based on concave nodes

Abstract

This paper addresses the problem of topological mapping and exploration of unknown environments for mobile robot. The topological map is constructed by defining the spatial relationship between adjacent nodes and each node that has geometric information about its circumference. A laser range finder is used as the major sensing modality to collect the geometric data. The node point is extracted from the data in two stages: image-based skeletonization and node point decision. The geometric data are transformed into a skeleton image and the node point is determined among the branch points in the image. We adopt an image processing method to skeletonize the geometric data. The key problem of the topological exploration is to determine the target node among the current node's neighboring nodes. We also propose an efficient target node decision algorithm based on a concept of concave node. An experiment shows that our approach can explore indoor environments more efficiently.