The Method Of Fusion Localization And Mapping Based On Spinning LiDar

LiDAR(Light Detect and Range)has the advantages of strong anti-interference ability,long measuring distance,and high ranging accuracy,and has been widely used in the fields of unmanned vehicles,robotics,geographic mapping and so on.The commonly used LiDAR has limited scanning beams and angles,resulting in a small field of view(FoV)and sparse point cloud,which cannot guarantee reliable inter-frame correlation,and thus it is difficult to meet the requirements of reliable positioning and fine mapping.The spinning LiDAR,which is formed by driving the LiDAR with an additional axis of rotation,can significantly improve the scanning FoV and point cloud density.However,the additional rotational axes of the spinning LiDAR will introduce the new mechanical extrinsic calibration,and also cause the extrinsic parameters to creep due to mechanical wear and tear,self-motion,and thermal expansion and contraction,etc.To improve the fusion localization and mapping performance of the spinning LiDAR,it is not only necessary to efficiently and accurately estimate the mechanical extrinsic parameters,but also need to realize the self-calibration of the extrinsic parameters while localization and mapping.Aiming at the problems of mechanical extrinsic calibration and self-calibration localization and mapping of the spinning LiDAR,the following work is carried out in this dissertation:(1)Based on multi-beam LiDAR,IMU(Inertial Measurement Unit)and motor(including encoder),a single-degree-of-freedom continuous spinning LiDAR system is built,which is the hardware foundation for the subsequent work.Meanwhile,to address the incomplete kinematic model of the spinning LiDAR,two sets of extrinsic parameters are constructed for the LiDAR and encoder motion part and encoder stationary part and IMU by analyzing the encoder characteristics.Combined with the compensation model of motor rotation and platform motion and the global state of IMU,a unified kinematics model of spinning LiDAR is constructed,which builds theoretical foundation for the follow-up work.(2)Aiming at the limitations of the current spinning LiDAR mechanical extrinsic calibration,which still requires specific tools and regular calibration scenarios.Combined with the multigrid region planar extraction based on the Random Sample Consensus,coarse and fine planar correlation and weighted optimization of GRSC(Grid RANSAC),we design the grid region FG(Filtered Grid)strategy through the multi-dimensional curvature computation and normalized probability estimation,and propose a heuristic mechanical extrinsic calibration method FGRSC,which does not use specific calibration tools and regular scenarios and improves the calibration accuracy of the comparison methods by 42.02% on average.(3)To solve the problem of extrinsic parameter creep caused by mechanical wear and tear,self-motion and thermal expansion and contraction of spinning LiDAR with high dynamic attributes,through encapsulating the mechanical extrinsic parameter,the sensor extrinsic parameter,and the IMU state into the unify observation model and using the ESIEKF(Errorstate Iterated Extended Kalman Filter)optimization framework,we firstly propose a multi extrinsic parameters self-calibration fusion positioning and mapping method OMC-SLIO(Online Multiple Calibration Spinning LiDAR Inertial Odometry),which realizes the simultaneous optimization of the mechanical extrinsic parameter,the sensor external parameter,the global position,as well as map update.(4)Based on the proposed FGRSC and OMC-SLIO methods,a static modeling device and a dynamic localization and mapping device are constructed by using spinning LiDAR,and are applied to the automated grain cuttings robotic system and real-time localization and mapping in complex scenes,respectively,which verifies the practical application value of the research work in this dissertation.