HeLiPR: Heterogeneous LiDAR Dataset for inter-LiDAR Place Recognition under Spatiotemporal Variations

TL;DR

The HeLiPR dataset is proposed, the first heterogeneous LiDAR dataset supporting inter-LiDAR place recognition with both non-repetitive and spinning LiDARs, accommodating different field of view (FOV)s and varying numbers of rays.

Abstract

Place recognition is crucial for robot localization and loop closure in simultaneous localization and mapping (SLAM). Light Detection and Ranging (LiDAR), known for its robust sensing capabilities and measurement consistency even in varying illumination conditions, has become pivotal in various fields, surpassing traditional imaging sensors in certain applications. Among various types of LiDAR, spinning LiDARs are widely used, while non-repetitive scanning patterns have recently been utilized in robotics applications. Some LiDARs provide additional measurements such as reflectivity, Near Infrared (NIR), and velocity from Frequency modulated continuous wave (FMCW) LiDARs. Despite these advances, there is a lack of comprehensive datasets reflecting the broad spectrum of LiDAR configurations for place recognition. To tackle this issue, our paper proposes the HeLiPR dataset, curated especially for place recognition with heterogeneous LiDARs, embodying spatiotemporal variations. To the best of our knowledge, the HeLiPR dataset is the first heterogeneous LiDAR dataset supporting inter-LiDAR place recognition with both non-repetitive and spinning LiDARs, accommodating different field of view (FOV)s and varying numbers of rays. The dataset covers diverse environments, from urban cityscapes to high-dynamic freeways, over a month, enhancing adaptability and robustness across scenarios. Notably, HeLiPR includes trajectories parallel to MulRan sequences, making it valuable for research in heterogeneous LiDAR place recognition and long-term studies. The dataset is accessible at https://sites.google.com/view/heliprdataset.

Figures (9)

• Figure 1: (Top row) LiDAR place recognition challenges (i) Variance in resolution between high and low ray count LiDARs affects sensing abilities. (ii) While some LiDARs perform 360-degree scans, others have limited FOV due to occlusion or sensor limitations. (iii) Most LiDARs scan the repetitive area, whereas non-repetitive LiDAR densely scan by stacking individual scans. However, each scan tends to be sparse, as depicted in the left red box. (iv) Ground truth, crucial for executing LiDAR place recognition, is challenging to determine due to varying LiDAR coordinates and scan acquisition times. (Bottom row) HeLiPR dataset provides heterogeneous LiDARs and additional channels, thereby granting opportunities to utilize texture information from LiDAR.
• Figure 2: Sensors coordinate information between sensors. (a) and (b) represent the transformation with the xy-plane and z-axis. After the extrinsic calibration, the inter-sensor transformation can be found in the Calibration folder.
• Figure 3: \ref{['fig:LiDARCalib_a']} Extrinsic calibration trajectory: A circular path used for map construction and calibration purposes. \ref{['fig:LiDARCalib_b']} Post-calibration LiDAR alignment: A sky plot view illustrating the contours overlap between individual LiDAR scans.
• Figure 4: File structure of the HeLiPR dataset, illustrating the organization of LiDAR scans, ground truths, calibration, and inertial sensor measurements for each sequence
• Figure 5: INS-based trajectories for sequences 01, 02, and 03. The left shows trajectories on aerial images for 01, while the right visualizes 02 (bottom) and 03 (top) with a color gradient. Notably, red indicates the start point, while blue designates the endpoint.