Datenbestand vom 13. März 2019
Tel: 089 / 66060798
Mo - Fr, 9 - 12 Uhr
Fax: 089 / 66060799
aktualisiert am 13. März 2019
978-3-8439-0301-1, Reihe Informatik
Integrating Sensing, Localization and Visualization in the Robot Control Loop
156 Seiten, Dissertation Eberhard-Karls-Universität Tübingen (2011), Softcover, B5
It can be observed, that the perception of the environment and its mapping to some kind of mathematical model, is a common field of interest for computer graphics, computer vision, and robotics. The research on this topic is not limited to direct means of localization, but relates to other disciplines as well. In this thesis, we present our work on different aspects of this problem that can be described consistently from a robotics-centric point of view.
We begin with the presentation of our work on appearance-based pose estimation comprising orientation estimation and localization of a mobile agent using noisy low-resolution panoramic images. Inspired by a correlation-based approach, we propose a method to orientation estimation that is insensitive to low-quality input data. This technique is then combined with our novel appearance-based localization framework that allows for metric position estimation and efficient mapping.
Structural information of the surroundings is necessary for a mobile agent for visualization purposes and to perform higher-level tasks that involve interaction with the environment. We concentrate on two aspects of scene acquisition that deal with the denoising of range data and the registration of successively acquired two-dimensional laser scans into a common model of the scene.
Since it is desirable to provide visual feedback of the relevant data of an agent's state and of the environment it is located in, we develop an interactive application for the visualization of pose information along with additional sensor data in three-dimensional scene models and an installation that allows for intuitive navigation in virtual environments.
We consider the work on spherical image analysis, rotation estimation, and appearance-based localization in particular as the major contributions of this thesis.