Research On Real-time Location Mapping Method Of Indoor AGV Based On Lidar

With the rapid development of e-commerce and retail industry,as well as the continuous improvement of automation level in the factory production process,there is a growing demand for Automated guided vehicle(AGV)in the process of goods handling and assembly.At present,the technical means of most mature AGV products rely on artificial road signs,such as magnetic nails,magnetic tape,two-dimensional code,etc.,but these traditional positioning and navigation methods have been unable to meet the rapid change of the working environment and the complexity of the work task.Autonomous positioning and navigation AGV based on lidar is the most mainstream development direction at present.By using lidar to generate environment map,AGV autonomous planning action plan can be realized,which has the characteristics of high flexibility and high adaptability.Simultaneous Localization and Mapping(SLAM)is an important part of them.Based on the indoor application and the self-built platform,this paper studies the SLAM method of AGV based on lidar.The main work of this paper is as follows:1.The hardware and software system and sensor model of the experimental platform are introduced.Hardware components include control board,driver,encoder,lidar,etc.The software system uses the framework based on Robot Operating System(ROS),which is rich in resources and open,making the test platform with good expansibility and universality.The sensor model introduces the odometer measurement model and the lidar data conversion model.2.SLAM algorithm based on particle filtering.Firstly we analyzed the probability model of the SLAM problem,raises RBPF algorithm.The core of RBPF algorithm is to calculate the location and mapping of SLAM algorithm separately.RBPF is derived on the basis of Bayesian estimation,and then the construction method of raster map is introduced.In thisexperiment,the Gmapping algorithm for RBPF optimization was used.This algorithm has high speed,strong practicability and wide application in industrial field.3.Improvement of point cloud registration algorithm.At present,the most commonly used registration algorithm is Iterative Closest Point(ICP).Aiming at the slow speed of existing methods,a new phase matching based point cloud registration method is proposed,which mainly transforms point cloud from Eulerian coordinate system to Hough Domain(HD)by Hough Transform(HT).For rotation and translation,phase matching algorithm is used to get the offset of two point clouds,which greatly improves the speed and accuracy of the final effect.4.According to the method proposed in this paper,SLAM experiments are carried out using hardware platform.At the same time,the AGV follow-up function based on different sensors studied during the demonstration period is also presented.