Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

DIDLM: A SLAM Dataset for Difficult Scenarios Featuring Infrared, Depth Cameras, LIDAR, 4D Radar, and Others under Adverse Weather, Low Light Conditions, and Rough Roads

Weisheng Gong, Kaijie Su, Qingyong Li, Chen He, Tong Wu, Z. Jane Wang

TL;DR

DIDLM presents a comprehensive, publicly available SLAM dataset designed for difficult scenarios, integrating 4D millimeter-wave radar, infrared, depth cameras, LiDAR, RGB cameras, GPS, and IMU across 19 scenes and 18.5 km. The data, collected with both ground robots and roof-mounted vehicles, include robust ground-truth (GPS/INS and tum) and loop closures, enabling rigorous evaluation of fusion-based SLAM under adverse weather, low light, and rough roads. The authors benchmark multiple SLAM approaches (LiDAR, visual, infrared, radar, Gaussian splatting, and R3LIVE) to reveal sensor-specific strengths and limitations, guiding future multi-sensor fusion research. This dataset aims to advance robust autonomous navigation by providing a realistic, diverse platform for testing SLAM algorithms in extreme environments.

Abstract

Adverse weather conditions, low-light environments, and bumpy road surfaces pose significant challenges to SLAM in robotic navigation and autonomous driving. Existing datasets in this field predominantly rely on single sensors or combinations of LiDAR, cameras, and IMUs. However, 4D millimeter-wave radar demonstrates robustness in adverse weather, infrared cameras excel in capturing details under low-light conditions, and depth images provide richer spatial information. Multi-sensor fusion methods also show potential for better adaptation to bumpy roads. Despite some SLAM studies incorporating these sensors and conditions, there remains a lack of comprehensive datasets addressing low-light environments and bumpy road conditions, or featuring a sufficiently diverse range of sensor data. In this study, we introduce a multi-sensor dataset covering challenging scenarios such as snowy weather, rainy weather, nighttime conditions, speed bumps, and rough terrains. The dataset includes rarely utilized sensors for extreme conditions, such as 4D millimeter-wave radar, infrared cameras, and depth cameras, alongside 3D LiDAR, RGB cameras, GPS, and IMU. It supports both autonomous driving and ground robot applications and provides reliable GPS/INS ground truth data, covering structured and semi-structured terrains. We evaluated various SLAM algorithms using this dataset, including RGB images, infrared images, depth images, LiDAR, and 4D millimeter-wave radar. The dataset spans a total of 18.5 km, 69 minutes, and approximately 660 GB, offering a valuable resource for advancing SLAM research under complex and extreme conditions. Our dataset is available at https://github.com/GongWeiSheng/DIDLM.

DIDLM: A SLAM Dataset for Difficult Scenarios Featuring Infrared, Depth Cameras, LIDAR, 4D Radar, and Others under Adverse Weather, Low Light Conditions, and Rough Roads

TL;DR

DIDLM presents a comprehensive, publicly available SLAM dataset designed for difficult scenarios, integrating 4D millimeter-wave radar, infrared, depth cameras, LiDAR, RGB cameras, GPS, and IMU across 19 scenes and 18.5 km. The data, collected with both ground robots and roof-mounted vehicles, include robust ground-truth (GPS/INS and tum) and loop closures, enabling rigorous evaluation of fusion-based SLAM under adverse weather, low light, and rough roads. The authors benchmark multiple SLAM approaches (LiDAR, visual, infrared, radar, Gaussian splatting, and R3LIVE) to reveal sensor-specific strengths and limitations, guiding future multi-sensor fusion research. This dataset aims to advance robust autonomous navigation by providing a realistic, diverse platform for testing SLAM algorithms in extreme environments.

Abstract

Adverse weather conditions, low-light environments, and bumpy road surfaces pose significant challenges to SLAM in robotic navigation and autonomous driving. Existing datasets in this field predominantly rely on single sensors or combinations of LiDAR, cameras, and IMUs. However, 4D millimeter-wave radar demonstrates robustness in adverse weather, infrared cameras excel in capturing details under low-light conditions, and depth images provide richer spatial information. Multi-sensor fusion methods also show potential for better adaptation to bumpy roads. Despite some SLAM studies incorporating these sensors and conditions, there remains a lack of comprehensive datasets addressing low-light environments and bumpy road conditions, or featuring a sufficiently diverse range of sensor data. In this study, we introduce a multi-sensor dataset covering challenging scenarios such as snowy weather, rainy weather, nighttime conditions, speed bumps, and rough terrains. The dataset includes rarely utilized sensors for extreme conditions, such as 4D millimeter-wave radar, infrared cameras, and depth cameras, alongside 3D LiDAR, RGB cameras, GPS, and IMU. It supports both autonomous driving and ground robot applications and provides reliable GPS/INS ground truth data, covering structured and semi-structured terrains. We evaluated various SLAM algorithms using this dataset, including RGB images, infrared images, depth images, LiDAR, and 4D millimeter-wave radar. The dataset spans a total of 18.5 km, 69 minutes, and approximately 660 GB, offering a valuable resource for advancing SLAM research under complex and extreme conditions. Our dataset is available at https://github.com/GongWeiSheng/DIDLM.
Paper Structure (15 sections, 12 figures, 5 tables)

This paper contains 15 sections, 12 figures, 5 tables.

Table of Contents

  1. Introduction
  2. RELATED WORK
  3. The DIDLM Dataset
  4. Sensors setup
  5. File format and topic name
  6. Sensor Calibration
  7. Data Collection
  8. Evaluation
  9. Lidar SLAM
  10. Visual SLAM
  11. Infrared SLAM
  12. 4D millimeter-wave radar SLAM
  13. Gaussian splatting SLAM
  14. R3live SLAM
  15. Conclusion

Figures (12)

  • Figure 1: The figure illustrates the actual ground robot and the vehicle equipped with sensors, with detailed labels indicating the function of each component. Additionally, the coordinates of each sensor are marked, with red representing the x-axis, green representing the y-axis, and blue representing the z-axis.It should be noted that within the scope of this manuscript, the 2D LiDAR and GNSS G-MOUSE depicted in the figure were not employed. Their application is reserved for separate endeavors.
  • Figure 2: Two-view diagram of the ground robot. Our ground robot uses a tracked movement system, which may allow it to better adapt to complex road conditions.unit:mm.
  • Figure 3: The figure provides detailed information about the collected dataset. The dataset includes 19 different scenes and 9 types of sensors (the 7 types that can be visualized). LT: Left Camera, CO: RGB Camera, RT: Right Camera, DH: Depth Image, ID: Infrared Camera, LR: 3D Lidar, RR: 4D Radar. L: Low Speed, H: High Speed (10-30 km/h), B: Bumpy, NB: Non-Bumpy, BS: Bumps, SU: Sunny, R: Rainy, SN: Snowy, D: Day, N: Night.
  • Figure 4: The figure includes the extrinsic calibration results from LiDAR to camera and infrared camera, as well as the extrinsic calibration results from LiDAR to radar. The red dense point cloud represents LiDAR point clouds, while the green sparse point cloud represents 4D millimeter-wave radar point clouds.
  • Figure 5: The left image is a picture of a section of the internal road of a school, with numerous cracks and potholes on the surface. Ground robots traversing this area would encounter significant jolts, and vehicles would experience substantial vibrations. Therefore, we designate this stretch as the bumpy road section. The right image is of a flat road, but with several speed bumps in the middle, which we name as the road with bumps section.
  • ...and 7 more figures