Abstract

In this paper, a method for robotic exploration oriented to the automatic three-dimensional (3D) reconstruction of outdoor scenes is presented. The proposed algorithm focuses on optimizing the exploration process by maximizing map quality, while reducing the number of scans required to create a good-quality 3D model of the environment. This is done by using expected information gain, expected model quality, and trajectory cost estimation as criteria for view planning. The method has been tested with an all-terrain mobile robot, which is also described in the paper. This robot is equipped with a SICK LMS 111 laser scanner attached to a spinning turret, which performs quick and complete all-around scans. Different experiments of autonomous 3D exploration show the suitable performance of the proposed exploration algorithm.