Research Of Laser SLAM Algorithm Based On Pose Estimation Closed-loop Detection

SLAM is a technology in which a robot collects data according to its own sensors,locates itself in an unknown environment,and establishes incremental maps for unknown environments.It can complete instant positioning and build maps.One task,so SLAM is the core issue of robot autonomous navigation technology.As a real-time three-dimensional spatial information acquisition technology,laser radar can effectively measure physical and environmental information.LiDAR as a new real-time access to three-dimensional spatial information technology,can quickly and effectively measure the physical and environmental scene information.The use of a laser scanner as a sensor to accomplish SLAM technology without GPS signals in the room enables high-precision indoor 3D reconstruction tasks.In the process of building the indoor environment map by laser SLAM,it is necessary to estimate the position and pose for the point cloud and realize the data matching.At the same time,due to the accumulated pose error,the closed-loop optimization of the map is also needed.According to the closed-loop results,the map position and pose information inside the closed loop is optimized to correct the SLAM map data and reduce the accumulated error caused by the increase of the map range,and ultimately improve the positioning accuracy and mapping quality.Based on the analysis of the point cloud data feature,the scan matching SLAM algorithm and the pose-probability estimation closed-loop detection algorithm,this paper proposes a laser SLAM algorithm based on pose estimation closed-loop detection,the specific work is as follows:(1)On the basis of determining the laser SLAM construction coordinate system and the representation of environmental information,modeling the sensors used in the laser SLAM technology.By introducing the working principle of laser radar data and the process of coordinate transformation,a method of using ICP point cloud matching algorithm to estimate the point cloud data scanning pose and completing the point cloud data matching is proposed.(2)In order to solve the problem of accumulative error of pose and pose in point cloud data matching,combined with the rotation invariability of laser point cloud data,the laser SLAM algorithm of this paper use bilinear interpolation to rasterize the whole point cloud data.Based on the analysis of the distribution of occupancy probability of grid graph,by calculating the probability distribution similarity score,adopting the branch-and-bound algorithm to accelerate the detection of closed loop.The optimization of pose graph is approximated to the nonlinear least-squares fitting problem,and the closed-loop constraints are used to adjust the local map and point cloud data pose.(3)The laser SLAM algorithm in this paper is validated by two datasets,Chinese mapping building CASM1712 and open source Revo-LDS.By comparing the results of building the map with the actual indoor scene,the accuracy of the algorithm is analyzed,the accuracy before and after the closed-loop optimization,the efficiency of closed-loop detection,and the factors of missed detection and false detection are obtained.The availability of the algorithm is proved,and the accuracy under the closed-loop constraint can reach 5 cm.(4)In the end,the work of this paper is summarized and putting forward several problems that need attention when using the laser SLAM algorithm to perform indoor three-dimensional reconstruction.At the same time,the shortcomings of the algorithm and the direction of improvement are given.