Organizations: School of Electrical Engineering, KAIST (Korea Advanced Institute of Science and Technology), Daejeon 34141, Republic of Korea · Hanwha Aerospace, 6, Pangyo-ro 319 beon-gil, Bundang-gu, Seongnam-si, Gyeonggi-do, Republic of Korea
Multiple LiDAR-inertial odometry methods have been widely applied in robotic applications owing to their enhanced accuracy and reliability. However, adopting multiple LiDAR systems can be challenging due to the discrepancies in measurement noise across different LiDARs. Existing methods have typically overlooked the noise discrepancies, which can significantly affect accuracy. In this paper, we propose Noise-aware Multiple LiDAR-Inertial Odometry (NM-LIO) that addresses the noise discrepancies. We integrate a noise model to quantify the measurement noise of each LiDAR. Additionally, we estimate the uncertainty of the residuals based on the measurement noise, allowing the measurement model to capture the noise discrepancies. Our proposed method is evaluated on a public multiple LiDAR dataset and compared with state-of-the-art methods. The experimental results demonstrate that the proposed method can accurately estimate the odometry in various environments by accounting for the noise discrepancies.
Figures & tables
Figure 1: Overview of our proposed Noise-aware Multiple LiDAR-Inertial Odometry (NM-LIO) system. The system consists of five key processes. (a) First, LiDAR measurement noise is modeled after the points are accumulated. (b) Second, IMU propagation is performed to predict the state while the LiDAR points are accumulated. (c) Third, motion compensation is performed to compensate for the motion distortion. (d) Fourth, the residual is computed to form the measurement model. (e) Finally, the state is iteratively updated until it converges.
Figure 2: Illustration of the LiDAR measurement noise model. (a) Measurement noise is modeled based on uncertainties in range and bearing direction. (b) Measurement noise is visualized by a color scale.
Sequence
Method
UZH
Basement
Campus
Construction Site
LAB
M-LOAM [ 5 ]
0.276
0.115
0.386
2.647
0.064
LOCUS 2.0 [ 6 ]
0.177
0.120
0.087
0.290
0.040
MA-LIO [ 7 ]
0.177
0.036
0.046
0.063
0.024
Ours
0.174
0.023
0.038
0.061
0.025
Table 1: Translation component of ATE on the Hilti dataset [ 9 ] .
Figure 3: Accumulated scan with estimated trajectory of the proposed method on the (a) campus sequence and the (b) construction sequence of the Hilti dataset [ 9 ] .