CTE-MLO: Continuous-time and Efficient Multi-LiDAR Odometry with Localizability-aware Point Cloud Sampling

TL;DR

The paper addresses robustness and high frequency localization in GNSS denied environments by developing CTE-MLO, a continuous-time, filter-based multi LiDAR odometry framework. It integrates Gaussian process trajectory representation with a Kalman filter, enabling per-point queries on the trajectory and real-time updates, while a decentralized multi LiDAR synchronization scheme and a localizability-aware sampling strategy manage data volume and degeneracy. Voxel map management provides efficient map handling. Experimental results on NTU VIRAL and real world platforms show competitive accuracy with significantly improved efficiency, and the work provides an open source implementation for the community.

Abstract

In recent years, LiDAR-based localization and mapping methods have achieved significant progress thanks to their reliable and real-time localization capability. Considering single LiDAR odometry often faces hardware failures and degeneracy in practical scenarios, Multi-LiDAR Odometry (MLO), as an emerging technology, is studied to enhance the performance of LiDAR-based localization and mapping systems. However, MLO can suffer from high computational complexity introduced by dense point clouds that are fused from multiple LiDARs, and the continuous-time measurement characteristic is constantly neglected by existing LiDAR odometry. This motivates us to develop a Continuous-Time and Efficient MLO, namely CTE-MLO, which can achieve accurate and real-time estimation using multi-LiDAR measurements through a continuous-time perspective. In this paper, the Gaussian process estimation is naturally combined with the Kalman filter, which enables each LiDAR point in a point stream to query the corresponding continuous-time trajectory using its time instants. A decentralized multi-LiDAR synchronization scheme is also devised to combine points from separate LiDARs into a single point cloud without the primary LiDAR assignment. Moreover, with the aim of improving the real-time performance of MLO without sacrificing robustness, a point cloud sampling strategy is designed with the consideration of localizability. To this end, CTE-MLO integrates synchronization, localizability-aware sampling, continuous-time estimation, and voxel map management within a Kalman filter framework, which can achieve high accuracy and robust continuous-time estimation within only a few linear iterations. The effectiveness of the proposed method is demonstrated through various scenarios, including public datasets and real-world applications. The code is available at https://github.com/shenhm516/CTE-MLO to benefit the community.

Figures (14)

• Figure 1: Decentralized multi-LiDAR synchronization scheme. For centralized synchronization, no point can merge in a period ${\left[t'_{k},t'_{k+1}\right)}$ due to the primary LiDAR failure (LiDAR $\#1$). The decentralized synchronization scheme can obtain synchronized point clouds even when only one LiDAR is functioning.
• Figure 2: Demonstration of proposed multi-LiDAR synchronization and sampling strategy under sequence Port 01. LiDAR $\#1$ experienced data loss at $591s-654s, 675s-801s, 1269-1597s$, and $1609-1664s$ due to overly high temperatures, and the data of LiDAR $\#7$ is totally lost due to hardware failure.
• Figure 3: An illustration of the localizability concept in 2D.
• Figure 4: Benchmark results of CT-LO, CT-MLO, and CTE-MLO under the tnp02 sequence of the NTU VIRAL dataset.
• Figure 5: Localizability and used points of CT-LO, CT-MLO, and CTE-MLO under a degenerate sequence (tnp01) of the NTU VIRAL dataset.

Theorems & Definitions (4)

• Remark 1
• Remark 2
• Remark 3
• Remark 4