METHOD AND SYSTEM FOR UPDATING LOCALIZATION MAPS OF AUTONOMOUS DRIVING VEHICLES

摘要

In one embodiment, a self-aware self-evolved self-contained localization map system for an autonomous driving vehicle (ADV) is disclosed. The localization map system collects ADV localization data and real-time poses for analysis in view of a first localization map of the ADV. The analysis determines confidence scores based on rankings of similarities scores of candidate poses, consistency of the real-time pose, and spatial or temporal completeness of a map feature space surrounding the ADV. The subset of the collected localization data is aligned with the first localization map based on a highest ranked candidate pose and is applied onto the first localization map based on the determined confidence scores to generate a second localization map to subsequently determine a location of the ADV.