A Gaussian approximation of the posterior for digital map-based localization using a particle filter

In this paper, we propose an importance sampling-based localization method that approximates the belief to a Gaussian distribution after update for digital map-based localization. It uses the planned route information for constraining the map possibilities and matches the digital map features and higher features detected by the vehicle. This approach does not constrain the vehicle's sensor setup, does not require as much human effort for mapping when compared to High Definition (HD) maps, is relatively more resilient than low level point cloud approaches since it can rely on features that are relative to the city structure, and is self-aware of when it is uncertain between two or more modes, which is useful when fusing with other state estimation methods. The conducted experiments suggest that the method is adequate for a localization pipeline. The code and dataset used were uploaded to an online repository and are open-access.