Research On Online Calibration,Simultaneous Localization And Mapping Of Multiple Lidars

Multiple Li DARs can improve detection distance and detection density,effectively preventing degradation of state estimation in unstructured environments,thereby its can enhance the accuracy and stability of simultaneous localization and mapping.However,the pose transformation relationship between multiple Li DARs is easily affected by the external environment,and prolonged turbulence,temperature changes,or collisions can invalidate the originally convergent external parameters.This paper mainly relies on the national key research and development program "Research and Practice of swarm intelligence Consignment Vehicles for Airport Baggage Transfer"(2021YFE0203600).To solve the above problems,a system that can realize simultaneous online external parameter calibration,localization and mapping of multiple Li DARs is proposed.The main work is as follows:(1)Preprocessing the raw data of multiple Li DARs and integrated inertial navigation systems mainly includes ground extraction based on column evaluation,clustering denoising,feature extraction,and motion distortion correction of Li DAR point cloud data,GNSS coordinate conversion of integrated inertial navigation system data,and external parameter calibration of the main Li DAR and integrated inertial navigation system.(2)Associate the feature point cloud of the auxiliary Li DAR with the local map of the main Li DAR,construct a residual model,calculate the information matrix,and use the minimum eigenvalue of the information matrix as the degradation factor to detect the convergence of the external parameters of multiple Li DARs.At the same time,based on this residual model,online external parameter calibration between multiple Li DARs can be achieved;Using odometry information and ground point clouds,online calibration of three rotation angles and z-axis translation between the Li DAR and vehicle coordinate system can be completed.(3)To avoid the impact of excessive number of feature point clouds on the real-time performance of the system,this paper proposes a feature fusion strategy based on the smoothness and distribution of feature point clouds.Then,based on the Ceres library and LM nonlinear optimization method,inter frame matching is achieved to complete the calculation of the odometry.(4)Build a graph optimization model using the GTSAM library,and add loop detection constraint factors based on Scan Context descriptors and ground constraint factors based on planar feature matching to achieve global trajectory optimization and z-axis error limitation.And by using principal component analysis to associate feature point clouds with local maps,the map is constructed and updated in the form of keyframes.(5)Build a real vehicle experimental platform,collect real environmental data at Chang’an University,and design experiments to verify the robustness,real-time performance,and accuracy of the online calibration algorithm in this paper;And through comparative experimental analysis,it is concluded that the laser SLAM algorithm in this paper not only has relatively accurate pose estimation,but also the constructed map is consistent with the real environment;At the same time,it was verified that the multiple Li DARs SLAM system has a significant improvement in accuracy compared to a single Li DAR.