摘要
为了满足室内服务机器人定位能力要求,在考虑室内环境结构化特征和激光雷达独特优势的基础上,提出了从雷达点云中提取面元表达环境以实现室内精确定位和建图的方法。对激光雷达的扫描线进行分割,通过计算分割段内各点的凸系数的连续性条件确定面元覆盖范围。定位和建图分为两个阶段,首先求解点云帧间运动,作为全局位姿的估计;其次对全局位姿进一步优化,依概率将帧面元融入地图。两个阶段均优化隐式移动最小二乘距离代价,而面元不确定模型则由激光雷达测量误差确定。最后,分别在空旷和狭窄两种场景下进行试验,结果表明,所提方法相对定位精度分别为0.022%和0.11%,地图一致性好、满足实时性要求。
In order to meet the requirement of localization capability of indoor service robot, considering the structural characteristics of the indoor environment and the unique advantage of the lidar, a novel approach for accurate indoor localization and mapping is proposed, which extracts surfels from radar point cloud to express indoor environment. Firstly, the scan lines of the lidar are segmented, the continuity condition of the convex coefficients of the points in the segments is calculated to determine the coverage range of the surfel. Then, localization and mapping are carried out in two stages. In the first stage, the inter-frame motion of the point cloud is solved to estimate the global pose. In the second stage, the global pose is further optimized and the frame surfels are fused into the map according to probability. The distance costs of implicit moving least squares are optimized in both stages, and the uncertainty model of the surfel is determined according to the measurement error of the lidar. Finally, experiments in empty and narrow circumstances were conducted, and the results show that the relative localization accuracies of the proposed method are 0.022% and 0.11%, respectively, the consistency of the map is good, which meets the real-time requirement.
作者
刘今越
唐旭
贾晓辉
徐文枫
李铁军
Liu Jinyue;Tang Xu;Jia Xiaohui;Xu Wenfeng;Li Tiejun(School of Mechanical Engineering,Hebei University of Technology,Tianjin 300132,China)
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2020年第7期99-106,共8页
Chinese Journal of Scientific Instrument
基金
国家重点研发计划(2019YFB1302702)
国家自然科学基金(U1813222)
河北省教育厅重点项目(ZD2018220)资助
关键词
激光雷达
室内定位和建图
面元
概率融合
lidar
indoor localization and mapping
surfel
probability fusion
