ＧＮＳＳ測位の高度化と自動運転への応用

精确定位是实现城市城区自动驾驶的关键。通常的理解是，GNSS接收机、激光雷达和视觉传感器相结合，以车道级的精度提供定位服务。车道的宽度一般在3米左右。根据最近的研究，可以通过应用二维地图和三维模型来改进GNSS、LiDAR和相机的定位方法。自2007年以来，使用3D建筑物模型作为减轻/排除/纠正多路径和非视距效应的辅助手段已成为增强全球导航卫星系统接收器的流行办法。提出了一种基于三维建筑物模型和光线跟踪模拟的GPS误差改正方法。最后，该方法成功地实现了定位误差小于3米的性能。

Precise positioning is essential to realize autonomous driving in city urban area. The common understanding is GNSS receiver, LiDAR and vision sensor are integrated to provide localization service with lane-level accuracy. In general, the width of lane is about 3 meters. According to recent studies, localization methods of GNSS, LiDAR and camera can be improved by applying 2D maps and 3D models. Since 2007, using 3D building models as aids to mitigate/exclude/correct the multipath and non-line-of-sight effects has become a popular approach to enhance GNSS receivers. We propose an innovative GPS error correction method by the use of 3D building model and ray-tracing simulation. Finally, the proposed method successfully achieves a performance with less than 3 meters of positioning error.