Research On Localization And Mapping Methods For Degenerated Scene Based On Fusion Of LiDAR And IMU

Simultaneous Localization and Mapping(SLAM)is considered as an effective method to solve the problem of mobile robot's environment perception,localization and navigation,which means a robot equipped with certain sensors estimates its own pose during the movement and completes the building of the environment model at the same time.3D light detection and ranging(Li DAR)is widely used in SLAM research because of its accurate and reliable measurement performance,with fruitful progress achieved.However,in some scenes where the structural features are not rich enough,traditional Li DAR SLAM based on point cloud information will degenerate,which will lead to the failure of robot positioning and mapping,and then the failure of subsequent navigation tasks.In this paper,aiming at the problem of inaccurate positioning and mapping of Li DAR SLAM in degenerated scenes,a SLAM system is designed by fusing inertial measurement unit(IMU)and Li DAR information.Firstly,aiming at the problem of inaccurate estimation of Li DAR degeneration state,the causes of degeneration phenomenon were analyzed and studied according to the principle of lidar pose estimation.The analysis and judgment of deg eneration direction and degeneration degree were completed by using perturbation model,and the degeneration feature direction and degradation factor are established.The normalization of degeneration factor was completed based on the scale unification of equations.The uncertainty model of pose estimation was established based on the degeneration feature direction and degeneration factor.It provided a good foundation for the follow-up research of sensor information fusion.Secondly,aiming at the problem of insufficient information constraint in the process of localization and mapping of lidar in degenerated scenes,an internal sensor IMU which is independent of environmental information was introduced.The time and space coordinates of the sensors were aligned through the unified clock source and external parameter calibration.The error model of IMU pose estimation was established through the analysis of IMU measurement error and pose estimation.Then,a two-step filtering pose information fusion scheme is established by using extended Kalman filter based on the robot degenerate state and IMU error model.On this basis,combined with the general process of Li DAR positioning and mapping,a hierarchical fine positioning and mapping method including pre positioning,coarse positioning and mapping,fine positioning and mapping is proposed.Finally,the applicability of the proposed algorithm in degenerated scenes was verified by simulation and physical experiments.This paper makes a key exploration in the application of robot positioning and mapping in deg enerated scenes,and proposes a SLAM system based on information fusion,which provides a useful reference for improving the applicability of SLAM.