Monocular simultaneous localization and mapping system for a wheeled mobile robot

Abstract

The present work describes the implementation of a monocular visual SLAM applied to a wheeled mobile robot moving in an indoor environment. It will be described the whole system and details of each part of the current implementation. Parts comprising the system include the estimation filter together with both motion and measurement models, as well as a set of computer vision algorithms for image processing and data association. The implemented visual SLAM make use of the latest techniques for undelayed landmark initialization which are required given the partially observability of bearing only SLAM. Presented results show the performance of the implementation mainly for robot pose estimation, from which it can be observed a highly accurate result in robot orientation estimation.