Early approaches for localization relied on abstract features detected in e.g. camera images or Lidar scans. These features enable precise localization but have no other use.
However, semantic HD maps offer information required for autonomous driving tasks such as planning or behavior generation at a resolution and level of precision that also enables accurate localization. Hence by using facades, poles, road markings, traffic light, and road signs, the additional complexity of a localization layer can be made redundant. At the same time, semantic representations are sensor-agnostic, allowing to share the same localization map across whole fleets from L2 ADAS systems to L4 robotaxis.
Our publications (see below) showed that this generalization can be achieved without a loss in accuracy or availability.
Publications
Frank Bieder, Haohao Hu, Johannes Schantz, Oguzahn Kirik, Florian Ries, Martin Haueis, Christoph Stiller. Map Learning: Ein Skalierbarer Ansatz zur Automatisierten Erstellung von Trainingsdaten unter Verwendung von HD Karten und Mehrfachbefahrungen. In 15. Workshop Fahrerassistenz Und Automatisiertes Fahren (FAS), Berkheim, Germany, October 2023. (Best Paper Award).
Miguel Ángel Muñoz Bañón, Jan-Hendrik Pauls, Haohao Hu, Christoph Stiller. DA-LMR: A Robust Lane Marking Representation for Data Association. In 2022 International Conference on Robotics and Automation (ICRA), Seiten 2193--2199, 2022.
Miguel Ángel Muñoz Bañón, Jan-Hendrik Pauls, Haohao Hu, Christoph Stiller, Francisco A. Candelas, Fernando Torres. Robust Self-Tuning Data Association for Geo-Referencing Using Lane Markings. IEEE Robotics and Automation Letters, 7(4):12339--12346, 2022.
Jan-Hendrik Pauls, Kürsat Petek, Fabian Poggenhans, Christoph Stiller. Monocular Localization in HD Maps by Combining Semantic Segmentation and Distance Transform. In 2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Seiten 4595--4601, 2020.
Julius Kümmerle, Marc Sons, Fabian Poggenhans, Tilman Kühner, Martin Lauer, Christoph Stiller. Accurate and Efficient Self-Localization on Roads using Basic Geometric Primitives. In IEEE International Conference on Robotics and Automation (ICRA), Montreal, Canada, 2019.
Fabian Poggenhans, Niels Ole Salscheider, Christoph Stiller. Precise Localization in High-Definition Road Maps for Urban Regions. In IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Madrid, Spain, October 2018.