Lane confidence assessment and lane change decision for lane-level localization

This paper presents the confidence assessment algorithm of lane mark detected by a vision sensor for lane change decision and its application to lane-level localization. The lane mark information is critical to determine whether driving maneuver is either lane following or lane change. Furthermore, it should be considered that the detected lane mark could be wrong due to its worn-out or soil/snow/obstacles on the mark. Thus it is necessary to assess the confidence of lane. By using detection results of each side of lane from vision sensor and vehicle state measured by in-vehicle sensor, lane confidence assessment algorithm based on probabilistic data association filter (PDAF) is proposed. And by using estimation results from PDAF, lane change decision algorithm is also proposed. The proposed algorithms are validated experimentally via field test data.