GNSS(Global Navigation Satellite System) is the most popular positioning system for vehicle localization. Especially, RTK(Real-Time Kinematic) algorithm is used for precise positioning in real time. The limitation of GNSS based positioning is huge dependency of visible satellite. In addition, RTK algorithm requires wireless data link to receive satellite measurement from reference station, and this is big infrastructure. Positioning accuracy for ITS is lane-level positioning. In order to express lane-level positioning system, precise map is required. 3D LIDAR consists of multi-layer LRF(Laser Range Finder). This sensor provide not only range but also intensity(reflectivity) information simultaneously. We can make precise probabilistic map using this intensity. In this paper, position error and heading angle error is estimated using probabilistic map by 3D LIDAR and low-cost GPS receiver. Histogram filter and ICP(Iterative Closest Point) were used to estimate error estimation. As a result, histogram filter was better than ICP in the area with clear lane-paint and ICP was better in the area with dim lane-paint.

Keywords: pose estimation, 3D-LIDAR, map, localization