Asynchronous Multiple LiDAR-Inertial Odometry using Point-wise Inter-LiDAR Uncertainty Propagation (2305.16792v2)

Abstract: In recent years, multiple Light Detection and Ranging (LiDAR) systems have grown in popularity due to their enhanced accuracy and stability from the increased field of view (FOV). However, integrating multiple LiDARs can be challenging, attributable to temporal and spatial discrepancies. Common practice is to transform points among sensors while requiring strict time synchronization or approximating transformation among sensor frames. Unlike existing methods, we elaborate the inter-sensor transformation using continuous-time (CT) inertial measurement unit (IMU) modeling and derive associated ambiguity as a point-wise uncertainty. This uncertainty, modeled by combining the state covariance with the acquisition time and point range, allows us to alleviate the strict time synchronization and to overcome FOV difference. The proposed method has been validated on both public and our datasets and is compatible with various LiDAR manufacturers and scanning patterns. We open-source the code for public access at https://github.com/minwoo0611/MA-LIO.

• The paper introduces a novel method that integrates asynchronous LiDAR and inertial data without requiring strict sensor synchronization.
• It utilizes continuous-time IMU modeling with B-Spline interpolation to align measurements in real time and effectively address temporal discrepancies.
• It employs point-wise uncertainty propagation and scan matching to enhance mapping accuracy and robustness in challenging environments.

Asynchronous Multiple LiDAR-Inertial Odometry Using Point-wise Inter-LiDAR Uncertainty Propagation

The paper, "Asynchronous Multiple LiDAR-Inertial Odometry Using Point-wise Inter-LiDAR Uncertainty Propagation," explores the integration of multiple LiDAR systems for enhanced robotic navigation accuracy. Its focus is on addressing temporal and spatial discrepancies that impede the integration of multiple LiDARs, particularly when employed asynchronously.

Overview and Methodology

The increasing popularity of multiple LiDAR systems is underscored by their ability to enhance accuracy and stability owing to an extended field of view (FOV). However, integrating these systems remains challenging due to temporal asynchrony and spatial discrepancies. This paper proposes mitigating these challenges through a novel approach involving continuous-time (CT) inertial measurement unit (IMU) modeling to evaluate inter-sensor transformation and model associated uncertainties on a point-wise basis.

A salient feature of the approach is its departure from the need for strict sensor synchronization, which is traditionally required for multi-LiDAR integration. Instead, the authors employ a B-Spline interpolation for trajectory estimation and integrate this with IMU readings to align LiDAR measurements accurately in real-time environments while maintaining computational efficiency. This enables the complementing of temporal discrepancies with high accuracy.

The proposed model also addresses spatial disparities by employing a scan matching strategy to merge point clouds across differing scan patterns. Most notably, the paper introduces an uncertainty propagation method, effectively capturing and managing ambiguities arising from merging point clouds through point-wise uncertainty assessments. The model accounts for point acquisition time and range, resulting in more robust odometry and mapping. These elements underscore the adaptability of the framework across various LiDAR types and scanning patterns.

Performance and Evaluation

The paper presents a comprehensive evaluation of the proposed method against several contemporary alternatives, including Fast-LIO2, M-LOAM, and LOCUS 2.0. The evaluation utilizes datasets such as the Hilti SLAM 2021, UrbanNav, and a custom dataset established by the authors. The results underline the proposed method's effectiveness, demonstrating superior performance in terms of accuracy and robustness, particularly in dynamic environments with multiple dynamic objects and high-speed applications.

Key numerical insights highlight the superiority of this approach in handling degenerate spatial environments, such as tunnels and narrow corridors, where traditional methods may falter. The experiments also feature LiDAR configurations with varying fields of view (FOVs), affirming the method's robustness across diverse hardware configurations.

Implications and Future Directions

The findings hold substantial implications for the design and deployment of asynchronous multi-LiDAR systems in real-time navigation and mapping contexts. By alleviating the need for precise time synchronization, the proposed method reduces hardware constraints and expands the applicability of multi-sensor platforms in challenging environments. The incorporation of high-performance point-wise uncertainty modeling has immediate ramifications for enhancing sensor fusion strategies across robotics and autonomous vehicle domains.

Given the foundational nature of this research in addressing LiDAR-inertial odometry discrepancies without synchronization, future studies may focus on extending the method's applicability to other sensor fusion contexts, exploring adaptive uncertainty modeling, and refining computational efficiency for large-scale deployments. Further research might also delve into the integration with machine learning techniques for enhanced model adaptability and generalization across more complex and unpredictable real-world scenarios.

In conclusion, the approach outlined in this paper represents a significant step toward refining multi-sensor integration methodologies by providing robust and adaptable solutions for asynchronized LiDAR systems, potentially transforming multi-sensor navigation and mapping landscapes.