Self-Localization is a crucial task for mobile robots. It is not only a requirementudfor auto navigation but also provides contextual information to supportudhuman robot interaction (HRI). In this paper we present an active vision-basedudlocalization method for integration in a complex robot system to work in humanudinteraction scenarios (e.g. home-tour) in a real world apartment. The holisticudfeatures used are robust to illumination and structural changes in the scene. Theudsystem uses only a single pan-tilt camera shared between different vision applicationsudrunning in parallel to reduce the number of sensors. Additional informationudfrom other modalities (like laser scanners) can be used, profiting of an integrationudinto an existing system. The camera view can be actively adapted and theudevaluation showed that different rooms can be discerned.
展开▼