Precise self-localization for last mile delivery automated driving in unstructured environments

Abstract

In the research on last mile automated driving, self-localization is an important problem to solve. In this paper, a precise self-localization algorithm is presented, which is based on a given map using LiDAR and camera sensors. The proposed approach is used as a solution for the localization problem within the VanAssist project. Several experiments were carriedout in order to validate the work and compare it to a reference and accurate RTK-GPS data. The evaluation shows that the localization result is within the requirements for last mile automated driving. Moreover, it indicates that the solution is robust to handle limitation in comparison to other approaches in the literature.