Indoor SLAM based on line observation probability using a hand-drawn map

Abstract

In this paper, we propose a novel method of building indoor map information under the condition that a hand-drawn map as prior information is given. So far, previous studies using the hand-drawn map have been limited to robot pose estimation and navigation. Therefore, we propose a novel method to build a map by finding the correspondence between the shape of the real environment and the shape of the hand-drawn map. In addition, even if the estimation of the robot pose on the hand-drawn map fails, our method can continue to construct the map by re-estimating the robot pose on the hand-drawn map based on the previous corresponding information. In the simulations, we verified the accuracy of the built map in a simulation environment using three hand-drawn maps.