Hierarchical map representation using vector maps and geometrical maps for self-localization

Map-based self-localization estimates the pose of the self-driving vehicle in an environment, becoming an essential part of autonomous driving tasks. Generally, maps used in self-localization have detailed geometric information on an environment in formats such as point cloud maps and Gaussian mixture model (GMM) maps. As other maps are widely developed for autonomous driving, vector maps store more object-focused information, such as buildings and road facilities, for navigation and scene understanding in autonomous driving tasks. However, it is not compatible with self-localization due to the lack of detailed geometric information. The two different map formats of vector maps and maps for self-localization complicate the management, preventing the development of the area where a self-driving vehicle can drive stably. This paper proposes a unified map format with a hierarchical structure that enables both vector maps and self-localization maps (i.e., GMM maps) to be managed more easily. Because proposed maps can be treated as vector maps at the high-level layer, various tasks related to navigation and scene understanding in autonomous driving can utilize. A GMM map is stored at the low-level layer associated with a vector map component, enabling accurate self-localization in an environment. The proposed map format is compatible with vector maps widely developed by mapping companies on the surface and facilitates future map management. The experimental results of self-localization in urban areas showed that the proposed map gives the competitive self-localization accuracy compared with the GMM map even with fewer cells that link to vector components. The proposed maps enable self-localization with sufficient accuracy for safe autonomous driving operations.     

Towards lane-level traffic monitoring in urban environment using precise probe vehicle data derived from three-dimensional map aided differential GNSS

Today's urban road transport systems experience increasing congestion that threatens the environment and transport efficiency. Global Navigation Satellite System (GNSS)-based vehicle probe technology has been proposed as an effective means for monitoring the traffic situation and can be used for future city development. More specifically, lane-level traffic analysis is expected to provide an effective solution for traffic control. However, GNSS positioning technologies suffer from multipath and Non-Line-Of-Sight (NLOS) propagations in urban environments. The multipath and NLOS propagations severely degrade the accuracy of probe vehicle data. Recently, a three-dimensional (3D) city map became available on the market. We propose to use the 3D building map and differential correction information to simulate the reflecting path of satellite signal transmission and improve the results of the commercial GNSS single-frequency receiver, technically named 3D map-aided Differential GNSS (3D-DGNSS). In this paper, the innovative 3D-DGNSS is employed for the acquisition of precise probe vehicle data. In addition, this paper also utilizes accelerometer-based lane change detection to improve the positioning accuracy of probe vehicle data. By benefitting from the proposed method, the lane-level position, vehicle speed, and stop state of vehicles were estimated. Finally, a series of experiments and evaluations were conducted on probe data collected in one of the most challenging urban cities, Tokyo. The experimental results show that the proposed method has a correct lane localization rate of 87% and achieves sub-meter accuracy with respect to the position and speed error means. The accurate positioning data provided by the 3D-DGNSS result in a correct detection rate of the stop state of vehicles of 92%.     

面向智能车定位的道路环境视觉地图构建

为实现智能车视觉定位,提出了一种基于多视角、多维度道路环境表征的高精度视觉地图构建方法,该方法明确了视觉地图的表征模型,包括视觉特征、场景结构信息以及轨迹信息等。在视觉特征中,运用前视场景全局特征描述道路环境,视觉特征不局限于某一种特征描述子;在场景结构信息中,运用俯视路面的2D结构信息进行描述,该特征与前视视觉特征构成多视角;轨迹信息则通过视觉维度以及地理维度的多维度方式完成计算,在视觉维度中,通过平面单应性计算节点间的轨迹;地理维度中,通过高精度经纬度信息消除累积误差问题。试验选取武汉理工大学内长约700 m的半开放式环形路段进行试验。试验结果表明：制图的单节点平均误差为3.1 cm,标准差为2.3 cm,最大节点误差为9.3 cm,累积误差率为0.5%。运用所制地图进行定位检测,平均定位误差约为11.8 cm,因此,研究所提出的方法可应用于半开放式路段或固定场景的视觉地图构建,为实现智能车在上述场景的定位打下基础。同时,研究提出的制图方法不需使用双目摄像机,在降低数据存储量以及制图成本的前提下,实现了对道路环境的充分表征;此外,运用路面2D特征结构信息计算轨迹,解决了视觉3D重建精度不稳定的问题,为视觉地图构建提供了新的构建思路。     

基于因子图优化的众包高精度地图云端融合方法

针对高精度地图传统云端融合方法生成的地图置信度较低、误差较大的问题,提出了一种基于因子图优化的众包高精度地图云端融合方法。利用RTK-GPS数据对车端上传的局部语义地图进行全局化处理;对地图片段进行匹配,并利用一致性筛选流程提高匹配精度;以车道线匹配对构建约束地图间变换关系矩阵的因子图优化模型;利用某城市道路真实多车轨迹数据进行测试。结果表明,相较于传统方法,该算法对于优化初值的依赖性较低,对于地图间聚集度提升了44.7%;与车道线数据真值相比,该算法绝对误差均在1 m以内,具有较高的实用价值。     

多线索融合的结构化道路检测算法研究

为实现结构化道路的检测，提出了一种用于道路检测的激光雷达和视觉融合法，通过提取激光雷达在道路边缘的三维点云信息，将其投影到视觉图像上，形成激光点云与图像的映射关系，生成激光雷达似然图。通过改进提取道路的颜色、纹理、水平线等特征的方法，生成相对应的视觉似然图。在贝叶斯框架下将激光雷达和视觉生成的似然图进行融合。在KITTI数据集上测试可知，精度达到94%，准确率达到86%，表明该道路检测法具有较好的道路检测果。     

基于点特征的车载LiDAR与GNSS/IMU联合标定方法研究

激光雷达(LiDAR)、全球导航卫星系统(GNSS)与惯性测量单元(IMU)之间的外部参数标定精度是影响多传感器融合及高精度地图的主要因素。基于此，提出一种适用于无人车的LiDAR与GNSS/IMU标定方法，该方法可实时提取标定板点云中心坐标，并利用点特征进行外部参数标定。首先，分析了LiDAR、GNSS、IMU及通用横墨卡托格网系(UTM)坐标系的变换关系；其次，基于IMU安装平面与地面平行的假设，通过车辆前方地面的法向量计算LiDAR俯仰角和滚转角的初值，并利用标定板中心偏移量计算偏航角初值；然后，假设车辆在平面上保持直线运动，采用恒定姿态运动将旋转角度的求解问题转化为最优化问题，并根据标定板UTM坐标不变的约束求解出平移参数；最后，通过分析算法误差、传感器测量误差以及中心点匹配误差验证了方案的可行性，并通过无人矿车采集LiDAR、GNSS和IMU的同步数据，对所提出的外部参数标定方法进行测试。试验结果表明：提出的方法只需要一定范围的平坦区域和标定板就可以解决3自由度运动估计6自由度外部参数的退化问题，且标定后的外部参数可以保证点云地图在20 m内的拼接误差小于20 cm。     

基于激光雷达的自动驾驶同步定位与建图方法综述

同步定位与建图（Simultaneous Localization and Mapping,SLAM）技术可使自动驾驶车辆在未知环境中根据车载传感器采集到的数据估计自身位姿,建立环境地图,为车辆的规划、决策提供定位信息,是近年来自动驾驶技术研究的热点之一。基于车载激光雷达的点云数据,聚焦SLAM技术在自动驾驶领域的应用,围绕前端里程计、后端优化和回环检测技术,对国内外相关研究进行综述。考虑到单一传感器的局限性,结合目前多传感器融合研究的热点与难点,展望了自动驾驶多传感器融合SLAM技术在自动驾驶领域的机遇与挑战。     

基于多传感信息融合的智能车辆定位方法

针对高精度定位系统中地图的重要性问题,将定位问题分为无地图定位与基于地图定位,分别对智能车辆的定位问题进行探索.对研究的智能车辆、传感器及其定位问题进行建模分析,再对该平台实施传感器校准以减小系统误差.对于无地图定位问题,利用扩展卡尔曼滤波算法将里程计与惯性测量单元(IMU)数据相融合,通过试验证明航迹推测法存在累计误...     

地铁一侧项目施工对已建地铁结构的安全影响研究

随着城市建筑和地铁工程的快速发展,经常会出现建筑物紧邻地铁一侧施工,造成对地铁结构的不良影响,因而必须评估其施工过程中地铁结构的安全性。本文运用MIDAS/GTS有限元软件,以厦门某项目紧邻已建地铁结构施工为例,建立了地铁结构二维和三维数值分析模型,研究了建筑物在整个施工过程中对已建成的车站主体结构、附属结构、盾构法和矿山法区间结构的内力和变形影响。结果表明,项目施工对地铁结构的变形和内力影响均在可控范围内,满足地铁结构的安全性要求。     

基于模型试验的浅埋暗挖隧道施工过程中地表建筑物变形规律研究

为研究城市浅埋暗挖隧道施工过程中地表建筑物的变形规律，通过物理模型试验并采用3D扫描对地表建筑物沉降进行监测，同时采用数字散斑技术（DIC）对建筑物模型墙体的应变进行监测，得到浅埋暗挖隧道施工过程中建筑物墙体的应变规律。研究结果表明： 1）当建筑物中轴线与隧道中轴线平行，建筑物模型位于隧道正上方时，模型的各项应变随着与开挖掌子面距离的增大而减小； 2）建筑物模型受扭转产生的剪切应变主要受模型角点的不均匀沉降差值影响，不均匀沉降差值越大，受扭剪切应变越大； 3）建筑物模型受弯产生的拉伸应变和剪切应变主要受各角点沉降值的影响，沉降越大（即挠曲越大），产生的挠曲拉伸应变和剪切应变也越大。     

A hybrid dead reckoning error correction scheme based on extended Kalman filter and map matching for vehicle self-localization

In this research, a hybrid dead reckoning error correction scheme is developed based on extended Kalman filter (EKF) and map matching (MM) to improve the positioning accuracy for vehicle self-localization. The developed method aims at obtaining accurate positions when the GPS signals are occasionally unavailable or weakened. First, the heading data collected from an odometer and an optical fiber gyroscope are integrated by an EKF to reduce the random errors in dead reckoning. Then a modified topological MM algorithm is developed to reduce the systematic errors in dead reckoning. In this work, both cross-track errors and along-track errors are considered to improve positioning accuracy of MM. The errors are finally corrected using the results achieved from both the dead reckoning and the MM when the driving distance of a vehicle exceeds a predefined length or the vehicle turns in an intersection. Experiments have been conducted to evaluate the developed method and the results show that the maximum error and average error of dead reckoning can be respectively reduced to 15.4?m and 5.2?m during the experiment with total distance of 43?km. This positioning accuracy is even better than the accuracy of the low-cost GPSs which are usually at the order of 15–20?m (95%). The developed method is effective to achieve the positions of the vehicle when the GPS signals are occasionally unavailable or weakened.     

车辆导航系统中基于三维电子地图的单GPS车道级定位方法研究

在对三维电子地图建模的基础上,提出了一种基于单GPS的虚拟差分、高程辅助定位和碰撞检测误差消除相结合的车道级车辆定位方法.该方法将车辆定位分为两次地图匹配过程,即首次通过虚拟差分技术和高程辅助定位技术对GPS定位参数进行修正,提高定位精度,然后在三维电子地图中结合碰撞检测技术实现车道级匹配.通过实地DGPS跑车对比实验,证明该方法完全能够满足车道级路径引导的要求.     

城市客运场站交通影响分析及设计

阐述了城市客运场站交通影响分析的意义及作用,着重分析客运场站的交通发生特点,给出了对现状已有客运场站和待规划客运场站进行交通影响分析的步骤和研究重点,利用交通流车辆穿插理论进行了客运场站的出入口以及相邻路段动态交通影响分析计算方法研究,并在此基础上提出交通组织设计的方法。最后,结合江苏省昆山市城市客运场站交通影响及交通组织设计进行了实例分析。     

基于改进D-S证据理论的车辆导航地图匹配

对传统D-S证据理论算法进行了改进,解决了证据的可信度和决策规则问题,使新的组合推理规则能够有效地处理证据中的冲突信息.并结合当前城市道路网日益复杂的实际特点,采用Pignistic概率的决策理论和基于权值D-S证据推理的地图匹配新算法,建立车辆位置信息和方向信息判决规则,有效地利用现有信息扩大各匹配道路之间的信度差异,提高了算法的鲁棒性.通过仿真和试验表明,该算法可以较好地解决城市复杂路网地图匹配问题.     

基于Imageware和UG的某轿车车身侧围逆向设计研究

通过测量某实物轿车,得到车身点云图。应用Imageware的点云处理功能和UG的曲面造型方法,提出了利用逆向设计来快速构建侧围模型的方法。本文对用逆向工程方法进行车身建模进行了初步探讨,并最终建立了侧围的模型,得到了逆向工程的一般步骤和建模技巧,为其他车型的逆向设计提供了参考。     

TLS在大型钢梁虚拟预拼和检测应用研究

钢结构桥梁的传统预拼接工作存在数据获取效率低、实地操作困难和成本高等问题,针对上述问题,以长益高速公路扩容工程为例,结合地面固定式三维激光扫描技术,详细阐述了基于三维激光点云数据的多跨π型钢混组合结构桥梁的虚拟预拼和检测方法。主要步骤为扫描设计、点云获取、数据处理及建模、虚拟预拼及误差检测。     

Robust inspection technique for detection of flatness defects of oil pans

This paper discusses two different methods for the detection of flatness defects present on the mounting surfaces of oil pans using laser-scanned point clouds. The first method involves registration, which is a widely used method in the field of 3D data inspection: scanned point clouds are registered with CAD data and the iterative closest point (ICP) algorithm is used for further comparison. The second method is our proposed method, a simple yet effective method for measuring the flatness of an oil pan mounting surface. The process is based on the construction of a reference plane on the scanned surface. The oil pan mounting surface is scanned by a 3D laser scanner, obtaining point cloud data that is then further processed to reduce noise. Using this processed data, a reference plane parallel to the direction of the mounting surface is defined at the mean position of the mounting surface. The direction of the reference plane is determined by the normal vector of the mounting surface. Construction of the reference plane is carried out by the singular value decomposition (SVD) technique. The deviation of the surface from the reference plane is measured by calculating the error distance between the points of the surface to the reference plane using the least-squares method.     

基于车轴图谱的城市道路汽车运行轨迹特性分析

为探究城市道路行车轨迹与路侧之间的横向距离对车辆运行的影响,提高驾驶员行车安全,在某市滨海路进行汽车运行轨迹样本采集试验,使用AxleLight RLU11系列路侧交通数据采集系统分车道采集试验路段汽车运行轨迹样本,利用SPSS Statistics对试验路段不同车道车辆运行轨迹样本进行数据处理,绘制不同行车道运行车辆横向距离的累积频率曲线,计算得到汽车运行轨迹与路侧的横向距离D85,通过绘制行驶车辆距路侧的横向距离直方图,得到不同车道的车辆分布规律。结果显示,驾驶员大多数偏向选择在内侧车道运行。根据试验路段内外2条车道车辆横向距离和运行轨迹特性,可为城市道路交通安全设施的设置提供理论依据,以期提高城市道路交通运行安全。     

装配式地铁车站预制构件加工质量检验与安装精度预测

为解决装配式地铁车站工程预制构件常因质量检验工作任务繁重且误差较大而造成质量不合格及无法精准安装的问题，将BIM+三维激光扫描技术应用于装配式地铁车站工程预制构件生产质量检验中，通过建立预制构件BIM模型与三维点云模型进行虚拟预拼装，检验加工精度，避免现场安装过程中因构件质量问题而造成浪费。运用灰色预测方法构建GM（1，1）灰色预测模型，预测装配式地铁车站多个预制构件拼装后的累计误差，并以此对虚拟预拼装检验结果进行对比验证。结果表明： 1）本次质量检验的7块构件加工误差均在规范允许范围内； 2）整环累计拼装误差为13.920 mm，其中，C1块与D块间拼装误差达4.150 mm，其他相邻块纵缝间隙均小于3 mm； 3）虚拟预拼装误差检验结果与GM（1，1）模型灰色预测结果拟合较好，最大偏差仅为0.407 mm。     

三维激光扫描技术在隧道变形与断面检测中的应用研究

传统的隧道变形监测手段主要是对部分点和断面的数据进行提取,存在工作量大、效率低、数据少等缺点。三维激光扫描技术可集成隧道安全与质量信息,一次扫描即可准确建立隧道三维矢量模型,精确得到隧道的整体变形和轮廓信息。以桐庐隧道为依托,研究三维激光扫描技术在山岭隧道变形监测中的应用,得到以下结论:当激光扫描入射角大于60°时,扫描误差急剧增大,可根据入射角及隧道内径确定最大测站间距;为降低隧道累计拼接误差导致的模型整体偏移,应采用首尾控制点双控技术;采用单点面域分析法,可提升监测数据的可靠性;采用点云及模型套接技术,可快速实现对隧道三维变形、支护侵限、二次衬砌厚度的评估,大大提升隧道数据采集与分析的效率。