Presentation slides of ICRA2022 publication about 3D LiDAR IMU mapping. Globally Consistent and Tightly Coupled 3D LiDAR Inertial Mapping Kenji Koide, Masashi Yokozuka, Shuji Oishi, and Atsuhiko Banno Proc. IEEE International Conference on Robotics and Automation (ICRA2022), pp. 5622-5628, Philadelphia, USA, May, 2022 https://staff.aist.go.jp/k.koide/

1. Globally Consistent and Tightly Coupled
3D LiDAR Inertial Mapping
Kenji Koide, Masashi Yokozuka, Shuji Oishi, and Atsuhiko Banno
National Institute of Advanced Industrial Science and Technology (AIST), Japan
https://twitter.com/MR2T_AIST
https://staff.aist.go.jp/k.koide/
k.koide@aist.go.jp

2. Key Concepts
Global Matching Cost Minimization
LiDAR-IMU Tight Coupling
• Directly minimize the matching cost over the entire map
with GPU acceleration (unlike pose graph opt.)
• Enables to accurately constrain frames with a very small
overlap
PGO [SuMa, RSS2018] Proposed
• Employ the LiDAR-IMU tight coupling scheme for
all the estimation stages (from frontend to backend)
• Robust to quick sensor motion and environments
without sufficient geometrical features
• Reduce estimation drift in 4 DoF

3. Global Trajectory Optimization
Global Matching Cost Minimization
• We directly minimize the matching cost over the entire map = Global multi-scan registration
• It enables to accurately constrain the relative pose between frames with a small overlap
• This approach can easily be extended to the LiDAR-IMU tight coupling scheme
Koide et al., “Globally Consistent 3D LiDAR Mapping with GPU-accelerated GICP Matching Cost Factors”, IEEE RA-L, 2021
PGO [SuMa, RSS2018] Proposed Constraints between small overlapping frames

4. LiDAR-IMU Fusion
Tight Coupling of LiDAR and IMU constraints
• We build all the estimation stages (from frontend to backend) with tightly coupled LiDAR
and IMU constraints
• For global optimization, IMU factors are created between endpoints of submaps to strongly
constrain submap poses
Frontend factor graph (Fixed-lag smoothing)
Only factors related to the latest frame (𝑥9) are shown
Backend factor graph (iSAM2)

5. LiDAR-IMU Tight Coupling
The proposed method aggressively creates loop constraints
Constraints between small overlapping frames
Environment without sufficient geometries
KAIST Urban Dataset

6. Conclusions
※最新フレーム(𝑥9)に関わるファクタのみ表示
• Global matching cost minimization for backend optimization
• LiDAR-IMU tight coupling is employed for all the estimation stages
• The proposed method can be applied to any kind of range sensors
We are planning to release an extended version of this work as open source