LiDAR Inertial SLAM

LiDAR Odometry estimates accurate trajectories of a sensor system by aligning LiDAR point clouds captured at different positions. Frequently, this process is supported by additionally integrating inertial measurements. The provided angular velocity and linear acceleration already define a coarse motion and help to enhance the accuracy and robustness of the aspired pose estimation especially for high dynamic movements. LiDAR Inertial Odometry can effectively support trajectory estimation from GNSS measurements especially at scenarios, where GNSS occlusions and multi-path effects limit the quality of these results. This is the motivation of our research project SLAMantic, funded within the ZIM program of the Federal Ministry for Economic Affairs and Climate Action. It is a cooperation between the ifp and the company IGI, a provider for GNSS/IMU navigation systems for airborne and land applications. 

As an important processing step of LiDAR Odometry, scan matching aligns point clouds captured from different locations. This serves as a prerequisite for determining the respective sensor position during combination of multiple scans into a consistent scene. Scan matching is realized for consecutive scans, but is also required to provide loop closures, when the scanner platform revisits known places. Within the project, we developed a Dense Multi Scan Adjustment (DMSA) approach, which enables fine registering of multiple point clouds simultaneously. DMSA considers the points from these point clouds in a common frame as one set of multivariate normal distributions. During evaluation, the normal distributions are updated in each iteration and serve as weights for the optimization process. By these means, non-matching point cloud sections e.g. at dynamic objects are automatically weighted lower. This avoids the need for an additional outlier rejection mechanism, which is an important advantage for our aspired application aiming at the control of construction machines. In such scenarios moving objects like cars or pedestrians might disturb alternative scan matching approaches. The approach only requires correspondences from each point to its respective normal distribution, no direct matches between the single point clouds have to be provided. This considerably reduces the number of correspondences to be generated and evaluated during evaluation, which enables an time-efficient optimization of multiple point clouds. Experiments demonstrate a superior robustness and accuracy of our system, which additionally applies pre-integrated IMU measurements during optimization based on our DMSA results.

Skuddis, D. & Haala, N. (2022). Globally Optimal Point Cloud Registration for Robust Mobile Mapping, Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLIII-B2-2022, 273–281, https://doi.org/10.5194/isprs-archives-XLIII-B2-2022-273-2022, 2022

Skuddis, D.; Haala, N. (2024) DMSA - Dense Multi Scan Adjustment for LiDAR Inertial Odometry and Global Optimization arXiv preprint https://doi.org/10.48550/arXiv.2402.19044

https://github.com/davidskdds/DMSA_LiDAR_SLAM.git