基于全局位姿优化的移动机器人3D激光融合定位与建图  

3D Laser Fusion Positioning and Mapping of Mobile Robots Based on Global Pose Optimization

作　　者：杨鸥 章文誉 汪步云[2,3] 程军 许德章[2,3] YANG Ou;ZHANG Wenyu;WANG Buyun;CHENG Jun;XU Dezhang(Xuzhou XCMG Special Machinery Co.,Ltd.,Xuzhou 221116,China;School of Artificial Intelligence,Anhui Polytechnic University,Wuhu 241000,China;Wuhu Yunqing Robotics Technology Co.,Ltd.,Wuhu 241007,China)

机构地区：[1]徐州徐工特种机械有限公司,江苏徐州221116 [2]安徽工程大学人工智能学院,安徽芜湖241000 [3]芜湖云擎机器人科技有限公司,安徽芜湖241007

出　　处：《计算机测量与控制》2025年第4期209-216,共8页Computer Measurement &Control

基　　金：国家自然科学基金项目(61741101);安徽省高校协同创新项目(GXXT-2023-076);安徽省高校自然科学研究项目(2023AH050926);机器视觉检测安徽省重点实验室开放基金项目(KLMVI-2024-HIT-14);安徽未来技术研究院企业合作项目(2023qyhz35)。

摘　　要：基于工厂车间无人化智能仓储的移动机器人自主定位和导航需求,设计了基于全局位姿优化和改进LOAM算法的3D激光SLAM导航方法;前端里程计算法融合了激光雷达和IMU数据,采用ICP点云配准算法进行激光点云特征点匹配,通过初始定位流程及线段匹配技术,实现机器人在全局地图中的动态定位,包括惯性导航位姿推算,地图匹配的位姿计算和动态重定位;后端优化算法包括位姿图构建,基于Scan Context的回环检测,全局位姿优化等;利用整个轨迹上的所有观测数据来构建机器人的位姿图,比较不同时间点的Scan Context确定机器人是否回到了之前访问过的位置,将里程计、回环检测和RTK数据作为约束,对全局位姿进行优化;通过无人叉车型移动机器人定位建图与导航实验验证了改进算法的有效性,实验结果显示,所设计的SLAM算法能够实现室内环境下机器人自主定位和导航,重复定位误差和偏移量均小于±10 mm,达到了工业应用的要求,系统鲁棒性好。In order to meet autonomous positioning and navigation requirements of unmanned intelligent warehousing mobile robots for in factory workshops,a 3D laser SLAM navigation method based on global pose optimization and improved LOAM algorithm is designed.The front-end mileage calculation method integrates laser radar and IMU data,and uses ICP point cloud registration algorithm to match laser point cloud feature point.Adopting initial positioning process and line segment matching technology to achieve dynamic positioning of the robot in the global map,including inertial navigation pose calculation,pose calculation for map matching,and dynamic repositioning;The back-end optimization algorithm includes pose graph construction,loop detection based on Scan Context,global pose optimization,etc.Using all observation data along the entire trajectory to construct the pose map of the robot,comparing the Scan Context at different time points to determine whether the robot has returned to the previously visited position,and taking odometer,loop detection,and RTK data as constraints to optimize the global pose.The effectiveness of the improved algorithm is verified through the positioning,mapping,and navigation experiment of an unmanned forklift type mobile robot.Experimental results show that the designed SLAM algorithm can achieve autonomous positioning and navigation of the robot in indoor environments,the repeated positioning error and offset are both less than±10 mm,meeting the requirement of industrial application,and the system has good robustness.

关 键 词：3D激光雷达 三维点云 位姿估计 移动机器人 同时定位与建图 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]