基于运动预测的改进激光里程计

Improved Lidar Odometer Based on Motion Prediction

针对室外大范围建图场景中激光里程计输出的运动轨迹漂移问题,在仅使用激光雷达构建里程计条件下,基于正态分布变换提出一种连续运动预测算法来提高点云匹配初始值估计精度,然后在点云匹配时使用帧与局部地图匹配代替帧间匹配,有效抑制了运动轨迹的漂移。仿真中采用Kitti数据集的2个不同场景进行验证,结果表明:所提的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27.93%、36.66%,其中三维运动轨迹竖直方向(Z轴)上的偏移最大值分别降低了70.29%、82.52%。所提的改进激光里程计有效抑制了运动轨迹漂移。

To address the lidar odometer output trajectory drift problem for a wide range of outdoor building map scenes,a continuous motion prediction algorithm based on the normal distribution transformation is proposed to improve the estimation accuracy of the initial value of point cloud matching under the condition that only lidar is used to construct the odometer.Frame and local map matching is used instead of inter-frame matching.The drift of the motion trajectory is then effectively suppressed.The simulation results are verified by two different scenarios of the Kitti dataset.The improved lidar odometer algorithm reduces the global average errors of two trajectories by 27.93%and 36.66%,while the maximum Z-axis deviation of the two trajectories is reduced by 70.29%and 82.52%.The improved lidar odometer can stably and effectively suppress the motion trajectory drift.