F-LOAM: Fast LiDAR Odometry And Mapping (2107.00822v1)

Abstract: Simultaneous Localization and Mapping (SLAM) has wide robotic applications such as autonomous driving and unmanned aerial vehicles. Both computational efficiency and localization accuracy are of great importance towards a good SLAM system. Existing works on LiDAR based SLAM often formulate the problem as two modules: scan-to-scan match and scan-to-map refinement. Both modules are solved by iterative calculation which are computationally expensive. In this paper, we propose a general solution that aims to provide a computationally efficient and accurate framework for LiDAR based SLAM. Specifically, we adopt a non-iterative two-stage distortion compensation method to reduce the computational cost. For each scan input, the edge and planar features are extracted and matched to a local edge map and a local plane map separately, where the local smoothness is also considered for iterative pose optimization. Thorough experiments are performed to evaluate its performance in challenging scenarios, including localization for a warehouse Automated Guided Vehicle (AGV) and a public dataset on autonomous driving. The proposed method achieves a competitive localization accuracy with a processing rate of more than 10 Hz in the public dataset evaluation, which provides a good trade-off between performance and computational cost for practical applications.

Overview of "F-LOAM: Fast LiDAR Odometry and Mapping"

"F-LOAM: Fast LiDAR Odometry and Mapping" introduces an efficient approach to LiDAR-based Simultaneous Localization and Mapping (SLAM), addressing the dual challenges of computational efficiency and localization accuracy. SLAM is crucial for many robotic applications such as autonomous driving and drone navigation where precise localization in unknown environments is essential. Traditionally, LiDAR-based SLAM approaches have relied on iterative algorithms that can be computationally expensive and therefore, not suitable for real-time applications, particularly for systems with limited computational resources.

Methodology

The authors propose a novel two-stage distortion compensation method incorporated into a LiDAR SLAM framework. This method eschews the need for iterative calculation commonly used in traditional SLAM systems by adopting a non-iterative approach for the initial distortion compensation and scan alignment. This greatly reduces the computational overhead while maintaining significant levels of accuracy.

1. Feature Extraction: For each LiDAR scan, edge and planar features are extracted. The framework partitions features into edge and surface categories based on their local smoothness, aiming for computational efficiency and robust feature matching.
2. Distortion Compensation: The system compensates for motion-induced distortion in two stages. Initially, a constant velocity model assumes uniform motion during each scan, which allows for immediate distortion correction. Following pose estimation, the second stage involves recalculating and updating feature distortions to enhance accuracy without the intensive iterative overhead.
3. Pose Estimation and Mapping: The LiDAR odometry seeks to tightly couple current scan features with a global map. Optimization minimizes distances between matched features using an innovative weighting strategy that accounts for local geometric properties, thereby enhancing pose accuracy.
4. Global Map Maintenance: A keyframe-based map update strategy ensures computational efficiency and memory conservation. Only significant changes in pose (over a certain threshold) result in keyframe updates to the map.

Experimentation and Performance

The proposed F-LOAM was benchmarked extensively, including evaluations against established datasets such as KITTI for outdoor environments and tests on indoor warehouse logistics scenarios leveraging both simulations and actual hardware implementations.

• Numerical Results: On the KITTI dataset, F-LOAM demonstrated less than 0.80% average translational error and a 0.0048°/m average rotational error, positioning itself as one of the fastest and most accurate open-sourced SLAM systems in this benchmark.
• Robustness and Scalability: F-LOAM shows significant promise in scalability from indoor to outdoor settings, overcoming common challenges like environmental dynamics (e.g., changes from static to dynamic surroundings).

Implications and Future Directions

F-LOAM can be instrumental in pushing forward practical autonomous systems applications by providing a reliable real-time LiDAR SLAM framework with efficient computation. The system's open-source nature invites further improvements and extensions, such as integration with additional sensors (e.g., cameras, IMUs) for enhanced resilience in complex environments. Moreover, its architecture allows seamless adaptation for diverse robotic platforms, possibly leading to more widespread adoption in field robotics and autonomous navigation. Future work could explore tighter sensor fusion strategies and even incorporate deep learning methods for feature extraction to further improve robustness and adaptability across varied scenarios.