[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. © 2024 Press of Tsinghua University. All rights reserved.
