Robust and fast self localization by 3D point cloud

Abstract

Localization and mapping are very important for autonomous robot and human-robot communications. This paper propose a self localization method by using a range sensor. The range sensor can get point clouds. Then, we achieve robust 3D alignment of the point clouds. The iterative closest point (ICP) algorithm is famous for 3D alignment, however, in general, before applying the ICP algorithm, point clouds must be registered to roughly correct positions. Therefore, we solve this problem by using exhaustive search. In addition, to reduce computational cost, we use non-extremum suppression and distance field. In order to evaluate the proposed method, we apply for real environment. From the result, the computational time are very fast and alignment accuracy is very good.