Fast 3D laser inertial navigation coupled SLAM algorithm based on IEKF
A fast 3D laser inertial navigation coupled SLAM algorithm based on iterative extended Kalman filter(IEKF)was proposed to address large cumulative error and long computing time in laser SLAM.Firstly,an inertial measurement unit(IMU)kinematic model was constructed in the front-end,and the observation model was obtained by directly matching point clouds with local maps.The ikd-Tree structure was introduced to store point clouds to reduce the time for searching neighbor-ing points.Then IEKF was used to rapidly estimate the initial pose of the robot and get the laser inertial odometry.Meanwhile,a loop detection module was added,and the factor graph optimization was used in the back-end to integrate IMU integration factors,laser odometry factors and loop factors to eliminate the accumulated error and further improve the system's accuracy.This algorithm was validated by M2DGR dataset.The results show that,compared with A-LOAM,LIO-SAM,and FAST-LIO2 algorithms,the accuracy of the proposed one applied to large outdoor scenario is improved by 21.738%,9.112%,and 6.750%,respectively.Experiments in real environments also show that the map constructed by this algorithm can accurate-ly and completely reflect the geometric structural features of the surrounding buildings.Furthermore,the map construction effi-ciency of this algorithm is significantly better than that of A-LOAM and LIO-SAM,and the average running time per frame of the mapping module is less than 19 ms.
万方数据
维普
