Predição de mapeamento para navegação autônoma de um robô de inspeção em vasos de pressão esféricos

Abstract

Spherical pressure vessels used for the storage of liquefied petroleum gas need to be periodically inspected in order to identify problems in their structure and to prevent future accidents. This work aims at the development of a mapping technique based on map prediction for the navigation of an autonomous inspection robot. The technique developed consists of the following steps: autonomous search of known characteristics, perception of the environment, estimation of the environment, prediction of the characteristics, creation of occupation grid filled by voxels, application of degree of uncertainty in measurements and path planning. In the end, it is predicted the complete mapping of the tank to be inspected, containing its main physical characteristics, such as center, radius, plates and weld bead. This work assists the task of inspection in spherical pressure vessels performed by an autonomous robot providing a complete map of the environment, with the possibility of storing the position of fissures and other problems identified during the inspection.