Research On Positioning And Mapping Of Mobile Robots Based On LiDAR

Simultaneous localization and mapping(SLAM)is one of the key technologies for mobile robots to achieve unmanned control and autonomous exploration and navigation.At present,SLAM solutions are mainly divided into two categories,namely LiDAR sensor-based solutions and camera-sensor-based solutions.Among the methods based on multi-sensor fusion,the SLAM scheme based on lidar is widely used because of its high accuracy and little influence by ambient illumination.In order to improve the localization robustness and accuracy of mobile robots,a positioning algorithm based on hybrid solid-state radar and IMU(Inertial Measurement Unit)fusion is proposed and studied,and the specific work is as follows:Firstly,the relevant theoretical basis of laser SLAM is studied,the coordinate system of mobile robot is constructed,the working principle of hybrid solid-state radar is introduced,and the IMU pre-integration formula is derived on this basis by analyzing LiDAR and IMU models,which is convenient for subsequent work.Secondly,aiming at the problem of inferior points and distortion due to motion in the original data of point clouds,an adaptive voxel filtering method is proposed for downsampling processing,and the inferior points caused by dynamic obstacles and mirrors are filtered,and the point cloud distortion is eliminated through IMU backpropagation,which improves the quality of point cloud data.Aiming at the real-time and robustness of front-end point cloud registration,a point-to-map registration method is proposed,and the incremental octree data structure is used to manage the map to improve the real-time search of adjacent points.The iterative error Kalman filter is used for local optimization,and the IMU pre-integration is corrected while suppressing the cumulative error generated by the lidar.Thirdly,aiming at the accuracy of mobile robot positioning,a keyframe strategy is proposed in the backend,and the IMU pre-integration,lidar observation information and loopback constraint are used as factors based on factor map optimization to carry out global optimization,reduce the cumulative error generated by the front-end,and correct the point cloud to generate the final point cloud map and trajectory information.Finally,in order to verify the accuracy,real-time and robustness of the algorithm proposed in this paper,this paper uses the public dataset and the self-built experimental platform respectively for testing,and the end-to-end average positioning error under the trajectory and short trajectory in the public dataset reaches centimeter-level positioning accuracy,and is better than Fast＿lio in positioning and mapping accuracy.