FAST-LIO2: Fast Direct LiDAR-inertial Odometry (2107.06829v1)

Abstract: This paper presents FAST-LIO2: a fast, robust, and versatile LiDAR-inertial odometry framework. Building on a highly efficient tightly-coupled iterated Kalman filter, FAST-LIO2 has two key novelties that allow fast, robust, and accurate LiDAR navigation (and mapping). The first one is directly registering raw points to the map (and subsequently update the map, i.e., mapping) without extracting features. This enables the exploitation of subtle features in the environment and hence increases the accuracy. The elimination of a hand-engineered feature extraction module also makes it naturally adaptable to emerging LiDARs of different scanning patterns; The second main novelty is maintaining a map by an incremental k-d tree data structure, ikd-Tree, that enables incremental updates (i.e., point insertion, delete) and dynamic re-balancing. Compared with existing dynamic data structures (octree, R*-tree, nanoflann k-d tree), ikd-Tree achieves superior overall performance while naturally supports downsampling on the tree. We conduct an exhaustive benchmark comparison in 19 sequences from a variety of open LiDAR datasets. FAST-LIO2 achieves consistently higher accuracy at a much lower computation load than other state-of-the-art LiDAR-inertial navigation systems. Various real-world experiments on solid-state LiDARs with small FoV are also conducted. Overall, FAST-LIO2 is computationally-efficient (e.g., up to 100 Hz odometry and mapping in large outdoor environments), robust (e.g., reliable pose estimation in cluttered indoor environments with rotation up to 1000 deg/s), versatile (i.e., applicable to both multi-line spinning and solid-state LiDARs, UAV and handheld platforms, and Intel and ARM-based processors), while still achieving higher accuracy than existing methods. Our implementation of the system FAST-LIO2, and the data structure ikd-Tree are both open-sourced on Github.

Overview of FAST-LIO2: A Fast Direct LiDAR-Inertial Odometry Framework

The paper presents FAST-LIO2, a LiDAR-inertial odometry (LIO) framework that emphasizes speed, robustness, and versatility. Built upon a tightly-coupled iterated Kalman filter, FAST-LIO2 introduces two key innovations for efficient and accurate LiDAR navigation and mapping.

Key Innovations

1. Direct Registration of Raw Points: FAST-LIO2 registers raw point clouds directly to the map, bypassing the need for feature extraction. This approach exploits subtle environmental features, enhancing accuracy, and removes the necessity for a hand-engineered feature extraction module, allowing adaptability to various LiDAR scanning patterns.
2. ikd-Tree Data Structure: The framework employs an incremental k-d tree (ikd-Tree) for managing map data. ikd-Tree supports efficient incremental updates and dynamic rebalancing, outperforming existing data structures like octrees and R$^*$-trees. It also facilitates on-tree downsampling, maintaining computational efficiency.

Benchmark Comparisons and Results

FAST-LIO2 was evaluated on 19 sequences from various LiDAR datasets, demonstrating higher accuracy and lower computational loads compared to state-of-the-art LiDAR-inertial navigation systems. The system can operate in real-time, achieving up to 100 Hz odometry and mapping even in large outdoor environments.

Robustness Across Platforms

The system's versatility was tested through various real-world experiments. It proved reliable in cluttered indoor environments and adaptable to different platforms, including UAVs and ARM-based processors. The implementation minimizes the computational demand, achieving consistent real-time performance across different hardware.

Implications and Future Prospects

• Practical Applications: FAST-LIO2's efficiency and adaptability make it suitable for a wide range of robotic applications, from autonomous vehicles to handheld mapping devices. Its real-time processing capabilities are particularly beneficial for dynamic environments and fast-moving platforms like drones.
• Theoretical Advancements: By eliminating feature extraction modules, FAST-LIO2 simplifies the odometry pipeline, potentially influencing future research in sensor fusion and computational geometry. The ikd-Tree presents a novel approach to dynamic data structure management, promising further research into efficient map operations.
• Future Directions: Continued advancements in LiDAR technology may further leverage FAST-LIO2's adaptable framework. Potential developments could focus on integrating additional sensors for improved robustness and exploring machine learning techniques to enhance mapping accuracy.

Overall, FAST-LIO2 represents a significant contribution to the field of LiDAR-inertial odometry, balancing speed and accuracy with innovative data management approaches. Its open-source availability encourages wider adoption and potential enhancements by the research community.