基于LiDAR-IMU松耦合的同时定位与建图方法

SLAM method based on LiDAR-IMU loose coupling

同时定位与建图(SLAM)是无人车自主导航的基础,基于单一传感器的SLAM算法易受数据关联错误而导致算法跟踪失败。本文提出了一种激光雷达-惯性测量单元(LiDAR-IMU)传感器松耦合的同时定位与建图方法。提出了基于关键帧和基于普通帧的局部地图匹配方法,引入M估计修正代价函数的形状减少错误数据关联的影响,避免了信息损失维持了后端非线性优化的低计算资源需求,同时也能有效处理错误特征关联的问题。采用了基于Scan-Context的回环检测方法消除长期运行的定位漂移累积。实验结果表明本文方法的精度比单一传感器和其他松耦合方法更高。

Simultaneous localization and mapping(SLAM)is the prerequisite for autonomous navigation,SLAM algorithm based on LiDAR is easy to lead to tracking failure due to erroneous data correlation.A simultaneous localization and mapping method based on loosely coupled LiDAR-inertial measurement unit(IMU)fusion is proposed.A local map matching method based on keyframe and ordinary frame is proposed.M-estimation is introduced to modify the shape of the cost function to reduce the impact of wrong data association,which avoids the loss of information,maintains the low computational resource demand of back-end nonlinear optimization,and can effectively deal with the problem of wrong feature association.The Scan-Context-based loop detection method is adopted to eliminate the error accumulation of long-term operation.Simulation results show the effectiveness of the proposed method.