A major prerequisite for autonomous driving is precise and reliable vehicle self-localization. Currently available solutions for this problem are mostly based on satellite navigation systems which suffer from several inherent deficiencies especially in challenging urban traffic scenarios. In recent years, research has therefore been conducted on matching sensor data from the vehicle’s environment perception systems with high-definition digital maps in order to achieve accurate self-localization. One unsolved problem with this approach concerns the obsolenscence of the map data in dynamic traffic environments. Particularly in urban scenarios, environment features detectable by the vehicle’s sensor systems are regularly subject to change so that a digital map suitable as a reference for localization must be updated continuously. The focus of this dissertation lies in the development of a system for automated updating of digital maps for the highly accurate localization of autonomous vehicles. The system requires a connected traffic environment in which vehicles and infrastructure are able to communicate. On the infrastructure side, a data processing system automatically evaluates sensor data provided by the connected vehicles in order to continuously update a digital map of the respective traffic area. The map should include all objects that remain stationary for at least several minutes and which are thus useable for self-localization of following vehicles. The key methods used here are graph-based optimization and automatic alignment of lidar point clouds as well as grid-based probabilistic mapping approaches. The second emphasis of this work is the development of the vehicle system that achieves accurate and robust real-time localization based on the digital map. The main sensor on the vehicle side is a scanning lidar with a horizontal field of view of 360°. In addition, the system fuses data from an inertial sensor and wheel speed sensors in order to provide estimations for the vehicle position and orientation at a rate of 50 Hz. The key method for the vehicle localization is an adapted particle filter with a dynamic vehicle model. After discussing the methods for mapping and localization, this dissertation also includes an experimental validation of the connected system in an urban test scenario.


    Zugriff

    Download

    Verfügbarkeit in meiner Bibliothek prüfen


    Exportieren, teilen und zitieren



    Titel :

    Automatische Aktualisierung digitaler Karten für die hochgenaue Lokalisierung autonomer Fahrzeuge


    Weitere Titelangaben:

    Automated updating of digital maps for the highly accurate localization of autonomous vehicles


    Beteiligte:
    Quack, Tobias Michael (Autor:in) / Abel, Dirk (Akademische:r Betreuer:in) / Eckstein, Lutz (Akademische:r Betreuer:in)

    Erscheinungsdatum :

    2020-01-01


    Format / Umfang :

    1 Online-Ressource (xvi, 135 Seiten) : Illustrationen, Diagramme pages


    Anmerkungen:

    Dissertation, Rheinisch-Westfälische Technische Hochschule Aachen, 2020; Aachen 1 Online-Ressource (xvi, 135 Seiten) : Illustrationen, Diagramme (2020). = Dissertation, Rheinisch-Westfälische Technische Hochschule Aachen, 2020



    Medientyp :

    Sonstige


    Format :

    Elektronische Ressource


    Sprache :

    Deutsch






    Hochgenaue Fahrzeuglokalisierung und kollektives Erlernen hochgenauer digitaler Karten

    Rauch,S. / Savkin,A. / Schaller,T. et al. | Kraftfahrwesen | 2012


    Hochgenaue Fahrzeugeigenlokalisierung und kollektives Erlernen hochgenauer digitaler Karten

    Rauch, Sebastian / Savkin, Artem / Schaller, Thomas et al. | Tema Archiv | 2012


    Hochgenaue Positionsbestimmung für Fahrzeuge

    KASSLATTER FRITZ / RITTER THOMAS / WUNDER ROLAND | Europäisches Patentamt | 2018

    Freier Zugriff