Association of Straight Radar Landmarks for Vehicle Self-Localization

In man-made surroundings, many structure elements appear as straight line segments in a radar gridmap representation. If they do not move over time, they are well-suited as landmarks for environment-based vehicle self-localization. For use in localization, current observations need to be associated to known landmarks. This paper investigates this association step. Two information sources for association (geometric distance and feature signature) are compared and combined into a common association method.