双目视觉辅助的激光惯导SLAM算法
CSTR:
作者:
作者单位:

1. 南开大学 机器人与信息自动化研究所,天津 300350;2. 天津市智能机器人技术重点实验室,天津 300350

作者简介:

通讯作者:

E-mail: zhangxuebo@nankai.edu.cn.

中图分类号:

TP273

基金项目:

国家自然科学基金项目(62073178,62003176);工业物联网与网络化控制教育部重点实验室开放基金项目;天津市自然科学基金项目(22JCZDJC00810);天津市杰出青年基金项目(20JCJQJC00140);中央高校基本科研业务费专项资金项目(ZB23003422);中国博士后科学基金项目(2020M670628).


Stereo vision aided lidar-inertial SLAM
Author:
Affiliation:

1. Institute of Robotics and Automatic Information Systems, Nankai University,Tianjin 300350,China;2. Key Laboratory of Intelligent Robotics,Tianjin 300350,China

Fund Project:

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    激光同步定位与地图构建(simultaneous localization and mapping,SLAM)算法在位姿估计和构建环境地图时依赖环境结构特征信息,在结构特征缺乏的场景下,此类算法的位姿估计精度与鲁棒性将下降甚至运行失败.对此,结合惯性测量单元(inertial measurement unit,IMU)不受环境约束、相机依赖视觉纹理的特点,提出一种双目视觉辅助的激光惯导SLAM算法,以解决纯激光SLAM算法在环境结构特征缺乏时的退化问题.即采用双目视觉惯导里程计算法为激光扫描匹配模块提供视觉先验位姿,并进一步兼顾视觉约束与激光结构特征约束进行联合位姿估计.此外,提出一种互补滤波算法与因子图优化求解的组合策略,完成激光里程计参考系与惯性参考系对准,并基于因子图将激光位姿与IMU数据融合以约束IMU偏置,在视觉里程计失效的情况下为激光扫描匹配提供候补的相对位姿预测.为进一步提高全局轨迹估计精度,提出基于迭代最近点匹配算法(iterative closest point,ICP)与基于图像特征匹配算法融合的混合闭环检测策略,利用6自由度位姿图优化方法显著降低里程计漂移误差并构建环境地图.最后,将所提出方法在公开与自制数据集上进行实验验证,并与主流开源的SLAM算法进行对比.实验结果表明,所提出算法可以在结构特征缺乏环境下稳定运行,并且相较于对比算法具有更高的位姿估计精度和算法鲁棒性.

    Abstract:

    Lidar-based simultaneous localization and mapping(SLAM) algorithms rely on the structural features to estimate the poses and construct the environmental map. Therefore, the pose estimation accuracy and robustness of such algorithms will decline or even fail in structureless environments. To solve this problem, combined with the characteristics that an inertial measurement unit(IMU) is not constrained by the environment and the camera depends on visual texture, a stereo vision aided lidar-inertial SLAM algorithm is proposed to solve the degradation problem of a pure lidar SLAM algorithm in structureless environments. This paper uses a stereo visual-inertial odometry algorithm to provide a priori pose for the lidar scan matching module, and further combines the visual pose constraint with the lidar structure feature constraints for joint pose estimation. In addition, a combined strategy of a complementary filtering algorithm and a factor graph optimization algorithm is proposed to align the lidar odometry reference frame with the inertial frame. Based on factor graph optimization, the lidar pose and IMU data are fused to constrain the IMU bias, so as to provide an alternative relative pose prediction for lidar scanning matching in the case of visual odometey failure. In addition, in order to further improve the accuracy of global trajectory estimation, this paper proposes a hybrid loop-closure detection strategy based on an iterative closest point matching algorithm(ICP) and an image feature matching algorithm, and the six degree of freedom pose graph optimization method is used to significantly reduce the odometey drift error and construct a global consistency environment map. Finally, experiments are carried out on public and self-made data sets, and the proposed method is compared with the mainstream open source SLAM algorithms. The experimental results show that the proposed algorithm can work stably in structureless environments, and achieve higher accuracy and robustness than the comparison algorithm.

    参考文献
    相似文献
    引证文献
引用本文

刘辉,张雪波,李如意,等.双目视觉辅助的激光惯导SLAM算法[J].控制与决策,2024,39(6):1787-1800

复制
分享
文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:
  • 最后修改日期:
  • 录用日期:
  • 在线发布日期: 2024-05-11
  • 出版日期: 2024-06-20
文章二维码