The implementation of autonomous driving requires particularly precise localization in urban environments. To complement a common fusion of GNSS and IMU, this research project also relies on a cost-effective LiDAR sensor, which provides position and orientation for discrete time steps through point-to-map mapping. In the filter-based approach, the continuously measured points are mapped onto a high-resolution, accurate map consisting of planes (building and road segments) and cylindrical pole objects, and assigned to components. The existing deviations are used for epoch-by-epoch improvement of the prediction step. To keep runtime low, an algorithm is used that selects a small subset of LiDAR points relevant for positioning. Additionally, collaboration with neighboring vehicles is achieved via an image-based method, in which feature points are extracted using an AI-based detection approach. This can stabilize the estimation, for example, in dense traffic conditions when the LiDAR platform is largely obscured. For this approach, the exchange of states and individual features is essential.

This research was supported by the DFG through the i.c.sens Research Training Group (RTG 2159) until November 30, 2025.