Localization and Mapping

Group Leader: M.Sc. Jan-Hendrik Pauls

Automated driving still depends on highly accurate maps that provide enriched layered knowledge, including lane geometries, traffic rules, and traffic lights. This information is not always perceivable or the surroundings do not allow safe detection - whether it is due to bad weather or occlusions. The research group Mapping & Localization investigates challenges related to map-based automated driving and addresses the future vision of mapless driving.

When using highly accurate maps, multiple challenges arise. First, the vehicle's ego-motion needs to be determined (Visual/Lidar/Inertial Odometry) to properly reconstruct the world around it. We create rich and dense 3D representations as well as sparse, abstract, or semantic feature maps that are compact in storage and low in maintenance. These maps not only contain physical elements but also higher-order information, such as lanes, routes, or traffic rules. The process of deriving a suitable representation from sensor data is called Mapping.

To use a map for automated driving, the automated vehicle has to be localized in the map (Localization). This is achieved by associating static map data with detected dynamic information. Since seemingly static map elements can change over time, their alterations must be detected to ensure safe driving using only correct information (Map Verification).

In cases where a map is outdated or the localization is misleading, an online Map Perception must be established as a fallback solution. Deep learning-based map perception methods have proven to be suitable for this task but require a lot of labeled data. Using accurate localization in verified maps, we can generate scalable ground truth data without additional labeling effort (Map Learning). Furthermore, we learn bird's-eye view representations and aim to leverage the strengths of various sensor types by fusing multimodal data. To update vectorized HD maps, we utilize graph neural networks, as they have been proven effective in representing semantic map elements and their relationships.

Projects: