Automatic lane-level map generation for advanced driver assistance systems using low-cost sensors

Abstract

Lane-level digital maps can simplify driving tasks for robotic cars as well as enhance performance and reliability for advanced driver assistance systems (ADAS) by providing strong priors about the driving environment. In this paper, we present a system for automatic generation of precise lane-level maps by using conventional low-cost sensors installed in most of current commercial cars. It mainly consists of two modules, i.e. road orthographic image generation and lane graph construction. First, we divide the global map into fixed local segments based on the road network topology. According to the local map segments, we accumulate the bird's eye view images of the road surface by fusing GPS, INS and visual odometry, and subsequently integrate them into synthetic orthographic images with the reference of the local map segments. Furthermore, the information of the driving lanes is extracted from the orthographic images and a large amount of vehicle trajectories, which is used to construct the lane graph of the map based on the lane models we proposed. Such a system can offer increased value as well as promote the automation level for today's commercial cars without being supplemented additional sensors. Experiments show promising results of the automatic map generation of the real-world roads, which substantiated the effectiveness of the proposed approach.