基于滑动窗口优化的激光雷达惯性测量单元紧耦合同时定位与建图算法

针对现有的激光里程计在面临室外大场景建图时,普遍会出现定位精度低、鲁棒性差的问题,本文提出一种16线激光和IMU惯性测量单元紧耦合的SLAM算法。首先对IMU进行估计位姿,通过线性插值矫正激光点云的运动畸变；接着通过曲率提取场景特征,并根据不同特征性质进行分类；然后利用帧间匹配模块在滑动窗口内构建局部地图；最后利用帧与局部地图匹配得到的距离和IMU数据构建联合优化函数。借助KITTI数据集和自行录制的园区数据集,对改进算法与主流的Lego-LOAM和同样使用紧耦合方案的LIO-Mapping进行分模块和整个系统的精度评定,实测结果表明,在符合里程计实时性的要求下,改进激光里程计精度高于Lego-LOAM和LIO-Mapping方案。

SLAM based on sliding window optimization for tightly coupled LiDAR and IMU

Aiming at the problems of low precision location and poor robustness of the existing laser odometer in the outdoor scene mapping. Odometer localization plan with 16 lines lidar and IMU tightly coupled are proposed in this paper. Firstly, estimated the pose of IMU and correction of motion distortion of laser point clouds by linear interpolation; Then, the scene features are extracted by curvature and classified according to different feature properties; The local maps are built within sliding windows using the interframe matching module; Finally, the distance obtained are used by matching the frame with the local map and the IMU data to construct the joint optimization function. By using the KITTI dataset and self-recorded datasets, Accuracy evaluation is performed of the improved algorithm Lego-LOAM and using tightly coupling scheme LIO-Mapping. The experimental result has showing that the Laser odometer precision location is better than Lego-LOAM scheme and LIO-Mapping under the odometer real-time.