{
  "paper_id": "464a33f2235dc236575438ccd85688aa",
  "paper_title": "D-LIOM: Tightly-Coupled Direct LiDAR-Inertial Odometry and Mapping",
  "paper_authors": [
    "Zhong Wang",
    "Lin Zhang",
    "Ying Shen",
    "Yicong Zhou"
  ],
  "publication_date": "2022-04-19",
  "publication_venue": "IEEE Transactions on Instrumentation and Measurement",
  "abstract": "Simultaneous localization and mapping via LiDAR-Inertial fusion is a crucial technology in many automation-related applications. Recently, a number of approaches based on geometric features have evolved, yielding impressive results via tightly-coupled estimation. This sort of feature-based techniques, however, are inextricably linked to the scanning mechanism of the LiDAR, relying on stable feature detection, and thus are difficult to adapt to multi-LiDAR systems. A few \"direct\" solutions, on the other hand, register the raw point cloud with the built probability map, which is more computationally efficient and easy to be extended. But, the existing direct approaches are all loosely-coupled, lacking correction of the IMU biases, and thus only work well in 2D cases. To this end, we present D-LIOM, a tightly-coupled Direct LiDAR-Inertial Odometry and Mapping framework. In D-LIOM, a scan is directly registered to a probability submap, and the LiDAR odometry, the IMU pre-integration, and the gravity constraint are integrated to build a local factor graph in the submap's time window, allowing the system to perform real-time high-precision pose estimation. Furthermore, to eliminate accumulated errors in time, we detect loops and adjust the sparse pose graph based on mutual matching of projected 2D submaps, allowing D-LIOM to run stably in large-scale scenes. In addition, to improve its flexibility to varied sensor combinations, D-LIOM supports multi-LiDAR inputs and facilitates the initialization with a common 6-axis IMU. Extensive experiments demonstrate that D-LIOM largely outperforms the existing state-of-the-art counterparts in mapping effect and localization accuracy as well as with high time efficiency. Lastly, to ensure that our results are entirely reproducible, all necessary data and codes are made open-source available.",
  "keywords": [
    "LiDAR-Inertial odometry",
    "SLAM",
    "loop detection",
    "data fusion"
  ],
  "paper_institutions": [],
  "output_language": "zh",
  "provider": "doubao",
  "model": "glm-5.3",
  "prompt_template": "simple",
  "extracted_at": "2026-09-17T09:39:22.937429Z",
  "template_tag": "simple",
  "summary": "### 摘要\n\n本文针对LiDAR-惯性SLAM领域的核心矛盾——基于几何特征的紧耦合方法（如LOAM、LIO-SAM）与LiDAR扫描模式深度耦合、难以适配多LiDAR系统，而现有直接配准方法（如Cartographer）均为松耦合、缺乏IMU偏差校正、仅在2D场景有效——提出D-LIOM紧耦合直接LiDAR-惯性里程计与建图框架：前端将扫描点云直接配准到以八叉树存储的概率占据体素子图，核心机制为最小化残差 Res(ξ̂)=Σᵢ[1−M(Tᵢ(ξ̂))]²（M为体素占据概率、Tᵢ(ξ̂)为经位姿变换后的点），经一阶泰勒展开由高斯-牛顿法求解位姿增量，并将该LiDAR里程计因子、IMU预积分因子与在线重力先验因子在子图时间窗口内构建局部因子图联合优化，实时估计状态并校正IMU偏差；后端将重力对齐的3D子图投影至2D水平面，利用SURF特征与FLANN匹配进行子图间回环检测，经RANSAC估计3自由度变换、分支定界搜索垂直平移后精确配准并优化全局位姿图以消除累积误差。实验选用三类互补数据集以验证不同平台与尺度下的鲁棒性与灵活性：自采集TONGJI手持数据集（16线ROBOSENSE LiDAR@10Hz+400Hz消费级IMU，含最大角速度超223°/s的快速旋转序列及结构化/非结构化校园场景）、VIRAL无人机数据集（水平+垂直双Ouster OS1-16 LiDAR互补视角、厘米级真值，用于绝对定位精度与多LiDAR验证）及Complex Urban车载数据集（Urban-09/10约1小时、近2平方公里高楼遮挡GPS的城市环境，用于大尺度长期一致性验证）；系统以C++实现，前端采用GTSAM构建并优化局部因子图，后端采用OpenCV提取匹配特征、Ceres-Solver优化全局位姿图，运行于双2.4GHz Intel Xeon E5-2620V3与32GB RAM工作站。评估指标涵盖绝对定位误差、重访误差与时间开销：VIRAL上双LiDAR输入加权平均误差0.10m（较LIO-SAM的0.15m提升5cm，较LIOM的1.82m高一个数量级），TONGJI上重访误差0.58–0.81m（LIO-SAM为1.08–40.21m，LOAM/LIOM/Carto3D更差或失败），16线LiDAR单帧里程计27.5ms+建图33.5ms（64线为54.5ms+60.1ms），回环检测精度90.94%、召回率95.45%，快速旋转下重建轮廓更清晰、大尺度场景轨迹与GPS基本一致，这些结果共同验证了直接配准与紧耦合优化结合、重力先验因子及子图间回环检测三项创新的有效性。未来方向包括显式估计IMU与LiDAR时间偏移以提升里程计精度，以及引入语义等高层特征缓解回环误匹配。该框架适用于GPS拒止的工业自动化、自动驾驶与测绘等真实场景，但受直接配准对初始位姿敏感、概率子图分辨率制约精度、低层特征误匹配导致少量回环误检/漏检等约束。\n\n### 关键要点\n\n- 方法核心：前端将扫描直接配准到概率占据体素子图，无需特征提取与匹配，与IMU预积分、重力先验共同构成局部因子图进行紧耦合联合优化，实时估计载体状态并在线校正IMU偏差。\n- 核心机制：扫描到子图配准通过最小化残差 Res(ξ̂)=Σᵢ[1−M(Tᵢ(ξ̂))]² 实现，经一阶泰勒展开与链式法则由高斯-牛顿法求解位姿增量，LiDAR里程计以一元因子（雅可比为单位阵、协方差σ²H⁻¹）加入局部因子图。\n- 地图表示：子图以八叉树形式的占据体素地图表示，占据概率按对数占据概率更新（仅需加减常数C_hitting/C_missing），初始仅维护虚拟空间划分、观测到真实数据才分配内存，兼顾存储与更新效率。\n- 重力先验因子：利用局部窗口内已优化节点在线估计重力并作为先验约束参与后续优化，保证横滚/俯仰低漂移与2D子图投影几何精度；消融实验显示去掉该因子后TJ-1/TJ-4重访误差恶化至7.14m/11.45m。\n- 回环检测：后端将重力对齐的3D子图投影到2D水平面，用SURF特征（SIFT计算复杂度过高、ORB误报过多）与FLANN匹配检测回环，RANSAC估计3自由度变换，分支定界搜索垂直平移后精确配准，优化全局位姿图消除累积误差。\n- 初始化：支持静态（加速度均值对齐重力方向）与动态（NDT扫描匹配与IMU预积分联合估计陀螺偏差、速度与重力）两种模式，仅需普通6轴IMU，使估计状态重力对齐并便于上层应用衔接。\n- 多LiDAR支持：得益于直接配准本质，多LiDAR点云经时空同步与去畸变后融合处理，无需为每个传感器单独设计特征提取模块，可充分利用互补视角数据且无明显效率损失。\n- 数据集：实验涵盖自采集TONGJI手持数据集（16线ROBOSENSE LiDAR@10Hz+400Hz IMU，含最大角速度超223°/s的快速旋转序列）、VIRAL无人机数据集（双Ouster OS1-16互补视角、厘米级真值）与Complex Urban车载数据集（Urban-09/10约1小时、近2平方公里）。\n- 定位精度：VIRAL上双LiDAR输入加权平均绝对误差0.10m，较LIO-SAM（0.15m）提升5cm、较LIOM（1.82m）高一个数量级；TONGJI上重访误差0.58–0.81m，显著优于LIO-SAM（1.08–40.21m）、LOAM、LIOM与Carto3D（后者多序列失败）。\n- 时间效率：16线LiDAR单帧里程计27.5ms、建图33.5ms，64线为54.5ms、60.1ms，后端效率与前端相当；对比的LOAM/LIOM/LIO-SAM需为每个LiDAR单独提取特征，多LiDAR时效率下降。\n- 回环检测性能：在TJ-1~TJ-4上子图间回环检测精度90.94%、召回率95.45%（TP 231、FP 23、FN 11）；去掉该模块后TJ-4重访误差从0.58m恶化至61.37m，验证其必要性。\n- 局限与展望：低层特征误匹配仍导致少量回环误检/漏检（误正影响较小，可在精确配准阶段过滤）；未来可显式估计IMU与LiDAR时间偏移、引入语义等高层特征提升鲁棒性。"
}