Based Exact IMU Pre-Integration
文献1基于
1. IMU 固有补偿 - IMU intrinsic compensation
文献1作者首先要在 IMU 预积分阶段之前添加 IMU 固有补偿。当在视觉惯性 SLAM 系统中使用 IMU 信息时,通常会将原始加速度计和陀螺仪测量值建模为
仅考虑测量噪声密度(即
其中,
2. IMU 预积分 - IMU pre-integration
按照上一节的方法,我们对 IMU 轴偏差和比例因子进行补偿,并使用符号
现在继续介绍核心贡献:一种新颖、精确的 IMU 预积分公式,基于矩阵李群
假设
在实际应用中,无噪声项是不可用的,因此用相应的估计值代替,用
由此推导出预积分项演化的矩阵表示
其中,
有
这里的
3. 工程实现
根据上文中的公式与推到,易将具有 IMU 固有补偿 (scale-misalignment 模型) 和自同构指数函数的 IMU 预积分公式移植到 VINS 系列开源算法中。个人的代码实现见 LSXiang/VINS-Fusion : This is an extended version of VINS-Fusion and migrates to ROS2, adding RGB-D mode, IMU preintegration in the form of SE2(3), line features and gpu acceleration 。
Reference
Yifu Wang, Yonhon Ng, Inkyu Sa, Alvaro Parra, Cristian Rodriguez, Tao Jun Lin, Hongdong Li. “MAVIS: Multi-Camera Augmented Visual-Inertial SLAM using SE2(3) Based Exact IMU Pre-integration”. arXiv preprint arXiv:2309.08142 ↩︎ ↩︎
J. Rehder, J. Nikolic, T. Schneider, T. Hinzmann, and R. Siegwart, “Extending kalibr: Calibrating the extrinsics of multiple imus and of individual axes,” in 2016 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2016, pp. 4304–4311. ↩︎
P. Furgale, J. Rehder, and R. Siegwart, “Unified temporal and spatial calibration for multi-sensor systems,” in 2013 IEEE/RSJ International Conference on Intelligent Robots and Systems, 2013, pp. 1280–1286. ↩︎
P. van Goor, T. Hamel, and R. Mahony, “Constructive equivariant observer design for inertial navigation,” arXiv preprint arXiv:2308.11124, 2023. ↩︎
C. Forster, L. Carlone, F. Dellaert, and D. Scaramuzza, “On-manifold preintegration theory for fast and accurate visual-inertial navigation,” IEEE Transactions on Robotics, pp. 1–18, 2015. ↩︎
Nikolay Atanasov: Sensing & Estimation in Robotics Lecture 12: SO(3) and SE(3) Geometry and Kinematics ↩︎