Scholarlink

<p>Existing LiDAR–inertial SLAM systems often suffer from instability in sparse-feature environments, significant cumulative drift, and reduced mapping accuracy during long-range operation. To address these issues, we propose a tightly coupled SLAM framework that integrates multi-line surface feature enhancement with improved LiDAR–IMU fusion. First, a multi-beam curvature computation method is introduced to strengthen feature extraction by aggregating geometric information across adjacent laser scan lines, thereby improving robustness in low-structure regions. Second, an Iterated Error-State Kalman Filter (IESKF) is employed as the core fusion engine, where IMU pre-integration is performed using a fourth-order Runge–Kutta (RK4) method. This design provides more accurate motion distortion compensation and reduces trajectory errors under highly dynamic motion. Furthermore, a multi-factor graph is constructed to jointly optimize poses using RK4-based IMU factors, an enhanced ICP-based LiDAR odometry factor, and loop closure constraints, enabling globally consistent mapping. Extensive experiments in Gazebo simulations and real-world vehicle tests demonstrate that the proposed approach achieves high adaptability in large-scale scenes and accurately reconstructs fine structural details. The system delivers reliable real-time performance for autonomous navigation and localization applications.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A LiDAR-inertial SLAM system based on multi-line feature enhancement and tightly coupled optimization

  • 期刊论文

  • 发表日期：

    2026年3月2日

查看全文
  • Hongzhou Chen,
  • Pingqing Fan,
  • Xipei Ma,
  • Jiangqiao Zhang

摘要

Existing LiDAR–inertial SLAM systems often suffer from instability in sparse-feature environments, significant cumulative drift, and reduced mapping accuracy during long-range operation. To address these issues, we propose a tightly coupled SLAM framework that integrates multi-line surface feature enhancement with improved LiDAR–IMU fusion. First, a multi-beam curvature computation method is introduced to strengthen feature extraction by aggregating geometric information across adjacent laser scan lines, thereby improving robustness in low-structure regions. Second, an Iterated Error-State Kalman Filter (IESKF) is employed as the core fusion engine, where IMU pre-integration is performed using a fourth-order Runge–Kutta (RK4) method. This design provides more accurate motion distortion compensation and reduces trajectory errors under highly dynamic motion. Furthermore, a multi-factor graph is constructed to jointly optimize poses using RK4-based IMU factors, an enhanced ICP-based LiDAR odometry factor, and loop closure constraints, enabling globally consistent mapping. Extensive experiments in Gazebo simulations and real-world vehicle tests demonstrate that the proposed approach achieves high adaptability in large-scale scenes and accurately reconstructs fine structural details. The system delivers reliable real-time performance for autonomous navigation and localization applications.