Lane-level vehicle self-localization in under-bridge environments based on multi-level sensor fusion
Lane-level vehicle self-localization in under-bridge environments based on multi-level sensor fusion
复制标题
DOI：
10.1109/itsc.2017.8317815
复制
发表时间：
2017-10
期刊：
影响因子：
--
通讯作者：
Lijia Xie;Yanlei Gu;S. Kamijo
中科院分区：
文献类型：
--
作者：
Lijia Xie;Yanlei Gu;S. Kamijo
复杂场景下车辆的自定位是自动驾驶系统的关键问题。目前的定位技术在开放天空区域具有良好的定位效果，而在封闭天空区域，由于卫星信号被拒、误差累积和交通拥堵等不利因素，定位精度会迅速下降。因此，我们提出了一种新的系统，可以在没有全球导航卫星系统（GNSS）的情况下，在桥梁或高架桥下的特殊道路上实现车道级车辆的自定位。在系统中，所有数据由立体摄像机和惯性传感器采集，并在三个不同的层次进行融合，以减少数据的不确定性，获得更准确的定位结果。通过在东京新宿的实验，证明了该系统的有效性和鲁棒性。
Vehicle self-localization in complex scenarios is a critical problem for autonomous driving system. Current localization techniques perform well in open-sky areas, while in closed-sky areas, localization accuracy is rapidly degenerated due to unfavorable factors as denied satellite signal, accumulated error and traffic congestion. Therefore, we proposed a novel system in this work, which is able to achieve lane-level vehicle self-localization in special roads under bridges or viaducts without the help of Global Navigation Satellite System (GNSS). In the system, all data is collected by stereo camera and inertial sensors, and fused in three different levels for reducing the uncertainty of data and acquiring a more accurate localization result. The proposed system is proved to be effective and robust through experiments conducted in Shinjuku, Tokyo.