The use of lane assistance systems can reduce the stress levels experienced by drivers and allow for better lane keeping in narrow, bus-dedicated lanes. In 2008, the Intelligent Vehicles (IV) Lab at the University of Minnesota developed such a system for this purpose. The IV Lab lane-assist system uses dual frequency differential GPS (DGPS) for high accuracy position information. This position information is used in conjunction with a geospatial database containing the road geometry and lane boundary positions required for a lane-assistance system. In urban environments, where tall buildings, overpasses, and other obstructions to the sky are present, DGPS suffers from inaccuracies and outages. This report proposes a method for replacing DGPS sensing with a high accuracy vehicle positioning system which fuses data from RFID (Radio Frequency IDentification) and LiDAR (Light Detection and Ranging) curb detection.
A Vehicle Positioning System (VPS) was originally developed by the IV Lab to provide the lane level ("which lane on the road") position of a vehicle with respect to a known reference (i.e., a mile marker or start of roadway) by the use of encoded position information in RFID tags on the roadway, read by the vehicle. The lateral position resolution of VPS is constrained to one lane width, which is insufficient for lane-assistant systems. Thus, in-lane level ("where in the lane") lateral position estimation was supplemented by a LiDAR unit that generates an accurate position of the vehicle with respect to the curb, which is cross referenced with a map database that provides the distance from the lane center to the curb, thus providing the vehicle's lateral offset from the lane center. On-board odometry is used to maintain accurate longitudinal position in between tag reads. By fusing the information from the VPS, LiDAR, and on-board odometry, high accuracy, "where in lane" level vehicle positioning can be maintained from this enhanced VPS during DGPS outages.