基于3D点云地图和描述子的蒙特卡罗定位方法  

Monte Carlo Localization Method Based on 3D Point Cloud Map and Descriptor

作　　者：吴官东 宋勇磊 WU Guandong;SONG Yonglei(Nanjing University of Science&Technology,Nanjing 210094)

机构地区：[1]南京理工大学,南京210094

出　　处：《计算机与数字工程》2023年第8期1848-1855,共8页Computer & Digital Engineering

摘　　要：为了实现机器人充分利用多线激光雷达感知信息进行位姿跟踪,论文提出了一种基于已知3D点云地图和描述子匹配观测模型的蒙特卡罗定位方法。首先设计了一种全新的以三维分箱形式统计障碍物结构信息的BOSS描述子,及高效的相似度计算方法;然后设计了基于扫描帧与局部点云地图BOSS描述子相似度匹配的观测模型;最后提出BOSS-MCL定位方法,借助描述子采样集预生成策略完成粒子的高效更新,实现机器人实时位姿跟踪。在实车平台上的测试结果表明,与传统蒙特卡罗定位相比,该方法在定位精度和稳定性上得到显著提升,具有实际应用价值。In order to make full use of the 3D Lidar perceptual information for robot pose tracking,this paper proposes a 3D point cloud map and global descriptor based Monte Carlo localization method.Firstly,a new BOSS descriptor is designed with an efficient similarity calculation method,which describes structure information of obstacle in the form of three-dimensional binning.Then an measurement model based on the BOSS descriptor similarity matching of the scan frame and the local point cloud map is designed.Finally,the BOSS-MCL localization method is proposed,which uses the pre-generation strategy of the description sampling set to complete the efficient update of the particles,and realizes the real-time pose tracking of the robot.The test results on the actual vehicle platform show that,compared with the traditional Monte Carlo localization,this method improves in localization precision and stability,and has practical application value.

关 键 词：点云地图 描述子 蒙特卡罗定位 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]