检索规则说明:AND代表“并且”;OR代表“或者”;NOT代表“不包含”;(注意必须大写,运算符两边需空一格)
检 索 范 例 :范例一: (K=图书馆学 OR K=情报学) AND A=范并思 范例二:J=计算机应用与软件 AND (U=C++ OR U=Basic) NOT M=Visual
作 者:史鹏程 李加元[2] 刘欣怡 张永军[2] SHI Pengcheng;LI Jiayuan;LIU Xinyi;ZHANG Yongjun(School of Computer Science,Wuhan University,Wuhan 430072,China;School of Remote Sensing and Information Engineering,Wuhan University,Wuhan 430079,China)
机构地区:[1]武汉大学计算机学院,湖北武汉430072 [2]武汉大学遥感信息工程学院,湖北武汉430079
出 处:《武汉大学学报(信息科学版)》2024年第7期1088-1099,共12页Geomatics and Information Science of Wuhan University
基 金:国家自然科学基金(42271444,42201474);武大-华为空间信息技术创新实验室开放基金(TC20210901025-2023-06)。
摘 要:针对移动机器人在大范围室内环境的定位难题,提出了一种基于圆柱特征的全局定位方法。首先,设计一种参数化地图,采用随机采样一致性算法和几何模型分割出地图中的圆柱点云,利用栅格地图描述环境中稳定人工构筑物的分布。其次,采用轻量级二进制文件记录圆柱和地物分布。然后,基于圆柱独特的几何特性(离群性、对称性和显著性),提出一种实时LiDAR点云圆柱分割方法。最后,设计两种位姿求解策略:第一种是启发式搜索,在地图与实时数据中搜索出最佳匹配圆柱,进而分别解算平移量和旋转量;第二种是优化求解,利用圆柱之间的拓扑关系构建约束条件并计算最优位姿。为验证所提方法的可行性,采用16线激光雷达在大厅、走廊及混合场景3种典型室内环境进行全局定位和回环检测实验。实验结果表明,该方法可有效实现典型空旷室内环境中机器人的全局定位,可达到90%的定位成功率以及0.073 m定位误差,部分数据可达到毫米级定位精度,最快速度在100 ms内,位置识别性能达到主流方法水平。该方法基本满足实际应用中自动驾驶对全局定位的精度和效率要求。Objectives:Localization is an important module of the light detection and ranging(LiDAR)si-multaneous localization and mapping(SLAM)system,which provides basic information for perception,control,and planning,further assisting robots to accomplish higher-level tasks.However,LiDAR localiza-tion methods still face some problems:The localization accuracy and efficiency cannot meet the require-ments of the robot products.In some textureless or large open environments,the lack of features easily leads to dangerous robot kidnappings.Consequently,aiming at the localization problems of mobile robots in large indoor environments,a global localization method based on cylindrical features is proposed.Methods:First,an offline parameterized map is designed,which consists of some map cylinders and a raster map.Be-cause the point cloud map contains a large number of 3D points and complete cylinders,random sample consensus(RANSAC)and geometric models are combined to directly segment the cylindrical points.The raster map is employed to describe the distributions of stable artificial structures.Then,some lightweight binary files are used to offline record the geometric model of cylinders and the feature distribution of the map.Next,based on three unique geometric characteristics of the cylinder(outlier,symmetry,and salien-cy),a real-time LiDAR point cloud cylinder segmentation method is proposed.Finally,two pose computation strategies are designed.The first is an optimization model based on heuristic search,which searches for the best matching cylinder between the map and real-time point cloud,and calculates the translation and rotation,respectively.The second is an optimization model based on multi-cylinder constraints,which employs both the topological relation(point-to-point and point-to-line constraints)and geometry attributes to find approximately congruent cylinders,then computes optimal pose.Results:To verify the feasibility of the proposed method,we use a 16-line LiDAR to collect the experimental data in three real-world in
关 键 词:自动驾驶 全局定位 回环检测 圆柱分割 参数化地图 激光雷达
分 类 号:P237[天文地球—摄影测量与遥感]
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在链接到云南高校图书馆文献保障联盟下载...
云南高校图书馆联盟文献共享服务平台 版权所有©
您的IP:216.73.216.62