Loop Closure using laser 2D
A fundamental necessity in mobile robotics is the ability of knowing the robot own location. In many applications, a map of the environment is not available or it does not have the accuracy or the information necessary for robot localization. In this cases, it is necessary to build a map simultaneou...
| Autor: | |
|---|---|
| Tipo de documento: | dissertação |
| Data de publicação: | 2021 |
| País: | España |
| Recursos: | Universitat Politècnica de Catalunya (UPC) |
| Repositório: | UPCommons. Portal del coneixement obert de la UPC |
| Idioma: | inglês |
| OAI Identifier: | oai:upcommons.upc.edu:2117/355855 |
| Acesso em linha: | https://hdl.handle.net/2117/355855 |
| Access Level: | Acceso aberto |
| Palavra-chave: | Robot vision -- Mathematical models -- Software Adaptive control systems -- Design and construction Autonomous robots Visió artificial (Robòtica) -- Models matemàtics -- Programari Sistemes adaptatius -- Disseny i construcció Robots autònoms Àrees temàtiques de la UPC::Informàtica::Robòtica |
| Resumo: | A fundamental necessity in mobile robotics is the ability of knowing the robot own location. In many applications, a map of the environment is not available or it does not have the accuracy or the information necessary for robot localization. In this cases, it is necessary to build a map simultaneously (SLAM). In applications where the GPS signal is not reliable, like in indoor environments, the localization has to be based in estimating robot position with the information of the robot movement. This estimation is obtained in local coordinates and it implies that a drift in the estimation accumulates error over time. In order to reduce the drift effects, it is necessary the capacity to recognize places that had already been visited and readjust the localization. This search of loop closures provides global reference points to the localization system. This field of study is already a very discussed topic, but finding an equilibrium between efficiency and robustness in order to be able to work in large environments is still a challenge. The aim of this work is to equip the WOLF SLAM library with an state-of-art loop closure for 2D LIDAR sensors. WOLF SLAM library is integrated with ROS and proposes an interesting environment to manage SLAM and localization problems based in graph-SLAM. In this work, we first performed a review of the state of the art for loop closure that lead us to FALKO. It is an open source project focused in loop closure that implements feature detection and scene matching using descriptors. Furthermore, the FALKO loop closure algorithm has been integrated along with an Iterative Closest Point method for robustness and accuracy purposes. ICP is a point to line metric optimized for range-finder scan matching. The integration and developed algorithms has been tested with experimental data from a real robot |
|---|