摘要
文中提出一种将激光雷达和轮式里程计相融合的方法,用于提升SLAM (Simultaneous Localization and Mapping)算法前端点云配准效率和点云畸变矫正效果。在点云配准方面,对PL-ICP算法做出改进,首先对激光点云进行预处理去除无效点,再利用自适应体素滤波方法对激光点云进行下采样,在保留点云特征的同时将点云稀疏化,从而减少点云配准的计算量,利用轮式里程计的测量值为点云配准提供初值,提高点云配准的效率和定位效果。在点云畸变矫正方面,按照轮式里程计测量的机器人位姿的时间戳,利用拉格朗日线性插值法对点云配准配得到的机器人位姿进行线性插值,利用EKF算法融合轮式里程计测量的位姿和点云配准插值得到的位姿对轮式里程计的测量误差做出矫正,然后为激光点云提供运动补偿,从而去除点云畸变提升SLAM算法定位和建图效果。利用ROS搭建仿真环境验证了本文提出的算法的有效性。
This paper proposes a method for fusing data from a lidar and wheel odometry to improve the effi-ciency of front-end point cloud registration and distortion correction in SLAM (Simultaneous Locali-zation and Mapping) algorithms. Specifically, in terms of point cloud registration, the paper im-proves the PL-ICP algorithm by first preprocessing the lidar point cloud to remove invalid points, and then using an adaptive voxel filtering method to down sample the lidar point cloud, which re-duces the computational load of point cloud registration while preserving the point cloud features. In addition, the paper uses the measurement values from the wheel odometry to provide an initial value for point cloud registration, which improves the efficiency and accuracy of point cloud regis-tration. In terms of point cloud distortion correction, the paper uses the timestamp of the robot pose measured by the wheel odometry, and applies Lagrangian linear interpolation to linearly in-terpolate the robot pose obtained from point cloud registration. The paper then uses the EKF algo-rithm to fuse the robot pose measured by the wheel odometry and the pose obtained from point cloud registration to correct for the measurement errors of the wheel odometry. Finally, the paper provides motion compensation for the lidar point cloud to remove point cloud distortion and im-prove the localization and mapping effects of the SLAM algorithm. The paper verifies the effective-ness of the proposed algorithm by using ROS to build a simulation environment.
出处
《建模与仿真》
2023年第4期3299-3318,共20页
Modeling and Simulation
