Robust Vehicle Localization Based on Detecting of Trajectory and White Line Shapes

Abstract

Vehicle localization based on matching between on-vehicle camera images and high-precision map has been developed for automated driving. We propose a method to associate a map with images by stochastically generating a white line section corresponding to appearance from white line types. Moreover, we propose a robust method by estimating not only the latest location but also the trajectory of the vehicle. The proposed method was tested in Shuto expressway, and the vehicle position was identified with sub-meter accuracy.