| Resumo: | Localization has always been one of the fundamental problems in the field of robotic navigation. The emergence of GPS came as a solution for localization systems in outdoor environments. However, the accuracy of GPS is not always sufficient and GPS based systems often fail and are not suited for indoor environments. Considering this, today there is a variety of real time localization technologies. It is quite common to see magnetic anomalies in indoor environments, which arise due to the presence of ferromagnetic objects, such as concrete or steel infrastructures. In the conventional ambient magnetic field based robotic navigation, which uses the direction of the Earth’s magnetic field to determine orientation, these anomalies are seen as undesirable. However, if the environment is rich in anomalies with sufficient local variability, they can be mapped and used as features for localization purposes. The work presented in this dissertation aims at demonstrating that it is possible to combine the odometric measurements of a mobile robot with magnetic field measurements, in order to effectively estimate the position of the robot in real time in an indoor environment. For this purpose, it is necessary to map the navigation space and develop a localization algorithm. First, the issues addressed to create a magnetic map are presented, namely data acquisition, employed interpolation methods and validation processes. Subsequently, the developed localization algorithm, based on a particle filter, is depicted, as well as the respective experimental validation tests.
|
|---|