摘要
自动引导车(automated guided vehicle,AGV)叉车是工业领域重要的物料运输装备,其定位精度和路径跟踪精度是提高物料运输效率、工厂自动化及智能化水平的重要基础。该文以医药行业室内结构化环境下的单舵轮AGV叉车为对象,利用基于密度的有噪声空间聚类(density-based spatial clustering of applications with noise,DBSCAN)算法和快速迭代最近点(fast iterative closest point,FICP)算法实现了基于反光柱的激光雷达导航定位,并提出了基于距离的异常点剔除规则,保证定位结果的稳定性及算法的鲁棒性。利用Kalman滤波算法,融合激光雷达、角度传感器和惯性测量单元(inertial measurement unit,IMU)的定位数据,提升了定位精度。对于AGV叉车的路径跟踪问题,建立了比例积分(proportional-integral,PI)控制器,设计了基于三阶Bézier曲线的圆弧路径,并对Bézier曲线的参数进行了优化,避免了曲率突变,提高了路径跟踪精度。实验结果表明,基于DBSCAN和FICP算法的激光雷达定位算法能够实现±3 mm的定位精度,实现了AGV叉车的准确定位;基于Bézier曲线改进后的圆弧路径跟踪误差降低了72%,直线和圆弧路径动态跟踪精度控制在25 mm以内,工作站点的重复定位精度达±12 mm,达到了预期设计要求。
[Objective]automated guided vehicle(AGV)forklift is an important material transportation equipment in the industrial field.Its positioning and path-tracking accuracy is an important basis for improving material transportation efficiency,factory automation,and intelligence.Thus,this paper uses a single steering wheel AGV forklift in an indoor structured environment of the pharmaceutical industry as an object,realizing the lidar positioning based on the reflector using the density-based spatial clustering of applications with noise(DBSCAN)and the fast iterative closest point(FICP)algorithms,and designing a proportional-integral(PI)controller to address the path-tracking problem of the AGV forklift.[Methods]First,the kinematics characteristics of the single steering wheel AGV forklift are analyzed,and its kinematics equations and state space equations are established.Subsequently,the DBSCAN and FICP algorithms were used to implement a reflector-based lidar positioning method for an accurate positioning problem.Moreover,a distance-based outlier elimination rule is proposed to address the problem of outliers interfering with the positioning process,which ensures the stability of the positioning results and the robustness of the algorithm.The Kalman filter algorithm is used to fuse the measurement data of the inertial measurement unit(IMU)and the angle sensor to improve the accuracy of the lidar positioning algorithm of the AGV forklift.This study establishes the position error and attitude error in the two core paths of straight lines and arcs based on the geometric relationship for the path-tracking problem.Following that,a PI controller is designed to realize the path tracking of the AGV forklift.Considering curvature discontinuity when the arc of equal curvature is connected with the straight-line path,the arc path based on the third-order Bézier curve was designed in this study.Furthermore,according to the limitation of the AGV forklift in the arc movement process,the parameters of the Bézier curve are analyzed and optimized to avoid the decrease of the path-tracking accuracy caused by the abrupt change of the path curvature.[Results]The experimental verification showed that the lidar positioning algorithm based on DBSCAN and FICP algorithms could achieve±3 mm positioning accuracy.Stable AGV forklift positioning could be achieved when combined with the outlier elimination rules.Furthermore,the Kalman filter-based fusion of IMU and angle sensor data resulted in accurate AGV forklift positioning.The improved arc path based on the Bézier curve reduced the arc path tracking error by about 72%compared with the equal-curvature arc path.The AGV's position and attitude errors were controlled based on the PI controller,which could control the dynamic tracking accuracy to within 25 mm.Furthermore,the repeated positioning accuracy of the work site reached±12 mm,meeting the expected design requirements.[Conclusions]This paper studies the lidar positioning and path-tracking technology of a single steering wheel AGV forklift in an indoor structured environment.An accurate and stable lidar positioning algorithm based on DBSCAN and FICP algorithms is realized by introducing outlier elimination rules and the Kalman filter.The AGV forklift's path tracking is realized using the PI controller,and the tracking accuracy of the arc path is improved using the Bézier curve.Finally,the positioning accuracy,path-tracking accuracy,and repeated positioning accuracy of the work site all met the expected design requirements.
作者
姚铭
段金昊
邵珠峰
苑韶伦
苏云州
YAO Ming;DUAN Jinhao;SHAO Zhufeng;YUAN Shaolun;SU Yunzhou(Department of Mechanical Engineering,Tsinghua University,Beijing 100084,China;State Key Laboratory of Tribology,Tsinghua University,Beijing 100084,China;Beijing Key Laboratory of Precision/Ultra-Precision Manufacturing Equipment and Control,Tsinghua University,Beijing 100084,China;Beijing Chieftain Control Engineering Technology Co.,Ltd.,Beijing 102600,China)
出处
《清华大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2024年第1期117-129,共13页
Journal of Tsinghua University(Science and Technology)
基金
国家重点研发计划资助项目(2020YFB1710700)
国家自然科学基金联合基金项目(U19A20101)
