Creation of Hybrid Hierarchical Models by Using Omnidirectional Vision and Machine Learning Techniques

Abstract

Over the past few years, the presence of mobile robots has significantly increased. Nowadays, they can be used for a wide range of applications and they can be found in diverse kinds of environments, such as industrial, household, educational and healthcare. Regarding mobile autonomous robots, these systems need a high degree of autonomy to develop their tasks. This means that they must be able to localize themselves and to navigate through environments, which are a priori unknown. Therefore, the robot will have to carry out the mapping task, which consists in obtaining information from the environment and creating a model. Once this task is done, the robot will be able to address the localization task, i.e., estimating its position within the environment with respect to a specific reference system. This thesis presents the analysis and design of mapping and localization methods in indoor environments. On the one hand, the thesis presents a work that focuses on solving these problems in underfloor voids with the aim of tackling a spray foam insulation task. On the other hand, a hierarchical localization framework is proposed and evaluated, considering severe visual effects that can influence the accuracy of the proposed method. The present thesis carries out a work, which focuses on solving the mapping and localization problems in voids between floor and foundations. Solving these tasks in such environments is especially challenging concerning visual information because the environment is dark and the terrain is uneven as stones, bricks fragments or sand are often present. Within these environments, the robot should be able to localize itself and apply insulation foam to the underside of the floor. Hence, the localization process is solved by estimating the position of the robot with respect to previously known position. This is done by using the alignment between point clouds (depth information). The robot is equipped with a 2D laser sensor, which permits building point clouds from several positions of the underfloor environment. This thesis describes several algorithms to obtain robustly the alignment between two positions. The proposed algorithms are tested with a set of point clouds captured with a laser scan under real working conditions. The results show that the localization problem can be solved and the accuracy obtained is enough to develop the insulation task.