UrbanLoco

测绘与定位是自动驾驶的关键模块，在该领域取得了显著成果。在全球导航卫星系统（GNSS）之外，点云配准、视觉特征匹配、惯性导航等方面的研究极大地提高了不同场景下建图定位的准确性和鲁棒性。然而，高度城市化的场景仍然具有挑战性：基于激光雷达和摄像头的方法在处理大量动态对象时表现不佳；基于 GNSS 的解决方案会遇到信号丢失和多路径问题；惯性测量单元 (IMU) 会出现漂移。不幸的是，当前的公共数据集要么不能充分解决这一城市挑战，要么不能提供足够的与地图绘制和定位相关的传感器信息。在这里，我们展示 UrbanLoco：一个在高度城市化环境中收集的地图/定位数据集，具有完整的传感器套件。该数据集包括在旧金山和香港收集的 13 条轨迹，总长度超过 40 公里。我们的数据集包括各种各样的城市地形：城市峡谷、桥梁、隧道、急转弯等。更重要的是，我们的数据集包括来自激光雷达、相机、IMU 和 GNSS 接收器的信息。

Mapping and localization are critical modules for autonomous driving, and significant progress has been achieved in this domain. Beyond Global Navigation Satellite System (GNSS), studies on point cloud registration, visual feature matching, inertial navigation and other related fields have substantially enhanced the accuracy and robustness of mapping and localization across diverse scenarios. However, highly urbanized scenarios remain challenging: LiDAR and camera-based methods perform poorly when dealing with a large number of dynamic objects; GNSS-based solutions suffer from signal loss and multipath effects; and Inertial Measurement Units (IMUs) exhibit drift. Unfortunately, current public datasets either fail to adequately address this urban challenge or do not provide sufficient sensor information relevant to mapping and localization. Herein, we present UrbanLoco: a mapping and localization dataset collected in highly urbanized environments with a complete sensor suite. This dataset includes 13 trajectories collected in San Francisco and Hong Kong, with a total length exceeding 40 kilometers. Our dataset covers a wide variety of urban terrains: urban canyons, bridges, tunnels, sharp turns, and more. More importantly, our dataset contains sensory data from LiDAR, cameras, IMUs, and GNSS receivers.

搜集汇总

背景与挑战

背景概述

UrbanLoco是一个专注于高度城市化场景的自动驾驶地图与定位数据集，旨在解决城市环境中动态对象多、GNSS信号丢失等挑战。它包含在旧金山和香港收集的13条轨迹，总长度超过40公里，覆盖城市峡谷、桥梁、隧道等多种复杂地形，并提供了激光雷达、相机、IMU和GNSS的完整传感器数据，适用于点云配准和定位算法研究。