MSC-LIO: An MSCKF-Based LiDAR-Inertial Odometry with Same-Plane-Point Tracking

Overview

• The paper proposes a new LiDAR-inertial odometry system called MSC-LIO, which is based on the multi-state constraint Kalman filter (MSCKF) framework.
• It introduces an efficient LiDAR same-plane-point (LSPP) tracking method for frame-to-frame data associations, without explicit feature extraction.
• The paper also presents an effective point-velocity-based LiDAR-IMU time-delay (LITD) estimation method.
• Extensive experiments on public and private datasets show that MSC-LIO achieves higher accuracy and efficiency compared to the state-of-the-art methods.

Plain English Explanation

The paper describes a new system for estimating the position and orientation of a mobile robot or vehicle using data from LiDAR (Light Detection and Ranging) sensors and inertial measurement units (IMUs). This type of system, known as LiDAR-inertial odometry, is important for applications like self-driving cars, drones, and robots that need to navigate accurately in 3D environments.

The key innovation in this paper is the use of the multi-state constraint Kalman filter (MSCKF) framework, which has been shown to be more efficient than traditional graph optimization methods for visual-based odometry. The authors apply this MSCKF approach to LiDAR-based odometry for the first time, creating a system called MSC-LIO.

To make this work, the authors developed two new techniques:

1. LiDAR same-plane-point (LSPP) tracking: This method tracks natural features in the LiDAR data (points that lie on flat surfaces) without needing to explicitly detect and extract them, which is computationally expensive. This improves the efficiency of the data association process.

2. LiDAR-IMU time-delay (LITD) estimation: LiDAR and IMU sensors do not perfectly synchronize, so the authors developed a method to accurately estimate the time delay between them, which is important for fusing the data correctly.

Through experiments, the authors show that their MSC-LIO system outperforms other state-of-the-art LiDAR-inertial odometry methods in terms of both accuracy and efficiency. This could lead to more robust and reliable navigation systems for autonomous vehicles and robots.

Technical Explanation

The core of the proposed MSC-LIO system is the multi-state constraint Kalman filter (MSCKF) framework, which has been shown to be more efficient than graph optimization for visual-based odometry. The authors adapt this MSCKF approach to LiDAR-inertial odometry, creating a tightly coupled system that fuses data from LiDAR and IMU sensors.

The key components of the MSC-LIO system are:

1. LiDAR Same-Plane-Point (LSPP) Tracking: Instead of explicitly detecting and extracting visual features from the LiDAR data, the authors propose an efficient LSPP tracking method. This method identifies points that lie on flat surfaces in the environment and tracks them from frame to frame, without the computationally expensive feature extraction step.

2. LSPP Measurement Model: The tracked LSPP features are used to construct an LSPP measurement model, which forms the multi-state constraint in the MSCKF framework.

3. LiDAR-IMU Time-Delay (LITD) Estimation: Since LiDAR and IMU sensors are not perfectly synchronized, the authors develop a point-velocity-based method to accurately estimate the time delay between the two sensor measurements. This is crucial for properly fusing the data in the MSCKF.

The authors evaluate their MSC-LIO system on both public and private datasets, comparing it to several state-of-the-art LiDAR-inertial odometry methods. The results demonstrate that MSC-LIO achieves higher accuracy and efficiency than the baselines.

Critical Analysis

The paper presents a well-designed and thorough evaluation of the proposed MSC-LIO system, including comparisons to several state-of-the-art methods on both public and private datasets. The authors also provide ablation studies to demonstrate the contributions of the key components, such as the LSPP tracking and LITD estimation methods.

One potential limitation of the work is that it has only been evaluated in indoor and small-scale outdoor environments. It would be interesting to see how the system performs in larger-scale, more complex outdoor scenarios, such as those encountered in autonomous driving applications.

Additionally, the paper does not discuss the system's robustness to sensor failures or environmental changes, which are important considerations for real-world deployment. Future work could explore the system's behavior under these challenging conditions.

Overall, the paper makes a compelling case for the effectiveness of the MSCKF framework for LiDAR-inertial odometry and introduces some novel techniques that could be broadly applicable to other sensor fusion problems.

Conclusion

This paper presents a novel LiDAR-inertial odometry system called MSC-LIO, which is based on the multi-state constraint Kalman filter (MSCKF) framework. The key innovations are the efficient LiDAR same-plane-point (LSPP) tracking method and the effective point-velocity-based LiDAR-IMU time-delay (LITD) estimation approach.

Extensive experiments demonstrate that MSC-LIO outperforms state-of-the-art LiDAR-inertial odometry methods in terms of both accuracy and efficiency. This work could lead to significant improvements in the navigation capabilities of autonomous vehicles, robots, and other mobile systems that rely on accurate 3D localization.