基于连续时间样条约束的改进激光里程计

Improved Lidar Odometry Based on Continuous-Time Spline Constraints

针对激光雷达因为颠簸等路面原因突然转向和做摇摆运动所造成的运动漂移问题,提出一种基于连续时间样条约束的改进激光里程计。在运动非连续性假设下,基于扫描关键帧和样条分割来提高点云匹配精度,然后添加样条约束,基于改进迭代最近邻(ICP)算法进行帧与地图匹配,有效抑制运动轨迹的漂移。在KITTI里程数据集和实验室采集的里程数据集上的实验结果表明,所提激光里程计改进算法分别使运动轨迹的全局平均误差降低了12.43%、29.40%。与目前基于几何特征的方法相比,所提改进激光里程计稳定且有效抑制了运动轨迹漂移,改善了激光里程计性能。

An improved Lidar odometer based on continuous-time spline constraints is proposed to address the problem of motion trajectory drift caused by sudden turning and swaying of the laser radar owing to road bumps and other road conditions.By assuming motion discontinuity,the accuracy of point cloud matching is improved by scanning keyframes and spline segmentation,and then adding spline constraints.By using the improved iterative nearest neighbor(ICP)algorithm for frame-to-map matching,the drift of motion trajectories can be effectively suppressed.The experimental results of using the KITTI mileage and laboratory-collected mileage datasets show that the proposed Lidar odometer algorithm reduces the global average error of the motion trajectory by 12.43%and 29.40%,respectively.Compared with existing methods that are based on geometric features,the proposed Lidar odometer is stable and effectively suppresses motion trajectory drift,improving the performance of the Lidar odometer.