【作者】 潘第弘； 傅宸辉； 夏晨； 陈金水； 卢建刚；

【Author】 Pan Dihong;Fu Chenhui;Xia Chen;Chen Jinshui;Lu Jiangang;State Key Laboratory of Industrial Control Technology, College of Control Science and Engineering, Zhejiang University;

【机构】 浙江大学控制科学与工程学院,工业控制技术全国重点实验室；

【摘要】 同步定位与建图(SLAM)技术在自动驾驶与机器人领域都发挥着重要作用。常见的主流SLAM算法,尤其是以多线激光雷达作为传感器的3DSLAM系统,常存在高程误差,对建图效果与定位精度都存在不利影响;同时,平面检测也是点云处理领域的常见任务。本文提出了一个融合快速平面检测的激光雷达-惯性测量单元(IMU)SLAM系统,在对点云做运动补偿处理后通过快速、鲁棒的平面检测识别地面与墙壁,完成特征点提取后在后端构建激光里程计因子、IMU因子、地面因子,进行全局性的因子图优化。本文利用数据集对平面检测的方法进行实验并与其他方法进行对比,并在实际大场景中完成了建图实验。更多还原

【Abstract】 Simultaneous Localization and Mapping(SLAM) technology plays an important role in both autonomous driving and robotics fields. The common mainstream SLAM algorithms, especially SLAM systems using light detection and ranging(LiDAR)as sensors, often have elevation errors, which have adverse effects on the mapping effect and positioning accuracy; Meanwhile,plane detection is also a common task in the field of point cloud processing. This article proposes a LiDAR inertial measurement unit(IMU) SLAM system that integrates fast planar detection. After a motion compensation processing towards the point cloud,the system identifies the ground and wall through fast and robust plane detection. After feature point extraction, the system constructs laser odometer factors, IMU factors, and ground factors in the back-end, and optimizes them globally through factor graph optimization. This article uses a data set to conduct experiments on the effectiveness of plane detection and compares it with other methods, and completes mapping experiments in actual large-scale scenarios.更多还原