CarMap: Fast 3D Feature Map Updates for Automobiles
CarMap: Fast 3D Feature Map Updates for Automobiles
复制标题
DOI：
--
复制
发表时间：
2020
期刊：
影响因子：
--
通讯作者：
Fawad Ahmad;Hang Qiu;Ray Eells;F. Bai;R. Govindan
中科院分区：
文献类型：
--
作者：
Fawad Ahmad;Hang Qiu;Ray Eells;F. Bai;R. Govindan
自动驾驶车辆需要准确、最新的 3D 地图来根据周围环境进行自身定位。如今，地图收集工作已经很少进行，并且需要使用专门的车辆车队。在本文中，我们探索了一种不同的方法：从具有先进传感器（LiDAR、立体摄像机）的车辆中进行近实时的众包 3D 地图收集。我们的主要技术挑战是找到 3D 地图的精益表示，以便新的地图片段或现有地图的更新足够紧凑，可以通过蜂窝网络近乎实时地上传。为此，我们开发了 CarMap，12，它找到了特征地图的简约表示，包含新颖的对象过滤和基于位置的特征匹配技术以提高定位鲁棒性，并结合了一种新颖的拼接算法来组合来自多个车辆的地图片段以获取未映射的道路片段，以及用于更新现有片段的高效地图更新操作。评估表明，CarMap 更新地图的时间不到一秒，相对于竞争策略，地图大小减少了 75 倍，具有更高的定位精度，并且能够在其他方法失败时在极端情况下进行定位。
Autonomous vehicles need an accurate, up-to-date, 3D map to localize themselves with respect to their surroundings. Today, map collection runs infrequently and uses a fleet of specialized vehicles. In this paper, we explore a different approach: near-real time crowd-sourced 3D map collection from vehicles with advanced sensors (LiDAR, stereo cameras). Our main technical challenge is to find a lean representation of a 3D map such that new map segments, or updates to existing maps, are compact enough to upload in near real-time over a cellular network. To this end, we develop CarMap, 12 which finds a parsimonious representation of a feature map, contains novel object filtering and position-based feature matching techniques to improve localization robustness, and incorporates a novel stitching algorithm to combine map segments from multiple vehicles for unmapped road segments and an efficient map-update operation for updating existing segments. Evaluations show that CarMap takes less than a second to update a map, reduces map sizes by 75 × relative to competing strategies, has higher localization accuracy, and is able to localize in corner cases when other approaches fail.