Research And Application Of 3D Map Construction Of Mobile Robot Based On Vision And LiDAR

Simultaneous Localization and Mapping(SLAM)is a key technology for robot autonomy.With the complexity of robot functions and application scenarios,the requirements for system accuracy,stability and ability to handle complex environments are increasing.The current application of SLAM technology is mainly based on singlesensor implementation,but in complex environments,such as scenes with slopes and varying obstacle heights,single-sensor has certain defects in positioning accuracy,stability and map integrity.In addition,the 2D occupancy raster map structure constructed by 2D Li DAR SLAM is incomplete,and the point cloud map constructed by pure vision SLAM has low accuracy and cannot be effectively used for environmental analysis and processing.Therefore,to address the above problems,this paper investigates the SLAM method of fusing Li DAR data with visual data for multi-sensor SLAM,and constructs a3 D point cloud map.Based on the 3D point cloud map,the analysis and processing methods about the environment structure are proposed.The main research contents of this paper are as follows:1.The development history and research status of mobile robot and simultaneous localization and mapping technology are studied,the mathematical model of simultaneous localization and mapping technology are analyzed,the motion model of mobile robot is analyzed,the software and hardware architecture of robot are built;2.For the shortcomings of single-sensor SLAM in system stability and localization accuracy,a graph optimization-based SLAM scheme for fusion of 2D Li DAR and RGBD cameras is proposed.The characteristics of 2D Li DAR data and RGB-D camera data are investigated and complemented with each other.Firstly,a joint error function is constructed based on the scan matching error of the laser data and the reprojection error of the visual data,secondly,a bag-of-words model of visual feature points is trained for loopback detection,and finally,a joint optimization is performed by a graph optimizer,and a globally consistent 3D point cloud map of the environment is constructed.It is verified that the fused SLAM scheme proposed in this paper outperforms the vision-only SLAM scheme in terms of localization accuracy,system stability and map building effect;3.In order to solve the problem that the maps constructed by single sensor in complex working environment cannot be better applied to the analysis and processing of environmental structure,a fused SLAM scheme is used to construct a 3D point cloud map with high accuracy and complete structure,and based on the point cloud data,a calculation scheme of slope angle is proposed to analyze the passability of slope and implement motion planning.In addition,for the problem that the point cloud map cannot be directly applied to navigation,a scheme is proposed to map the 3D point cloud map to2 D occupancy raster map.Compared with the 2D Li DAR SLAM construction,the structure of the occupied raster map obtained by conversion is more complete and can be better applied to the autonomous navigation of robots.