M3DSS-dataset

Driving

Introduction

Multi-sensor fusion for localization and mapping in urban is still a challenging problem. Urban scenes are characterized by numerous dynamic objects and frequent lighting variations, which adversely affect the performance of cameras and lidar systems. Furthermore, the presence of high-rise buildings poses a serious obstacle to the accuracy of GNSS. To address these challenges, M3DSS-Driving introduces a comprehensive dataset aimed at accelerating SLAM research in urban scenes. In addition to lidar and RGB cameras, this dataset incorporates four event cameras to enhance the collection of visual information. It includes sensor measurements from LiDAR, RGB cameras, event cameras, and IMUs, along with precise ground truth data obtained from the SPAN-CPT system. Unlike existing event camera-based driving datasets, M3DSS-Driving specifically emphasizes urban scenes, particularly those that are more challenging in nature.


Sensor list

(1)3D LiDAR sensor (HDL 32E Velodyne)

Frequency:

10Hz

FOV:

360°(Horizontal),-30°∼10°(Vertical)

Beam:

32

Distance:

up to 80m


(2)Right slant LiDAR (Velodyne VLP-16)

Frequency:

10Hz

FOV:

360°(Horizontal),-15°∼15°(Vertical)

Beam:

16

Distance:

up to 100m


(3)Left slant LiDAR (Lslidar C16)

Frequency:

10Hz

FOV:

360°(Horizontal),-15°∼15°(Vertical)

Beam:

16

Distance:

up to 150m


(4)IMU (Xsens Mti-10)

Frequency:

400Hz

Accelerometer in-run Bias Instability:

15μg

Gyroscope in-run Bias Instability:

18°/h


(5)GNSS receivers

Quantity:

u-blox ZED-F9P(X2)

    

Frequency:

u-blox ZED-F9P(1Hz,10Hz)

EVK-M8T(x1)

    

EVK-M8T(1Hz)

NovAtel Flexpak6(x1)

    

NovAtel Flexpak6(1Hz)


(6)RTK GNSS/INS (NovAtel SPAN-CPT)

Frequency:

1Hz

localization RMSE:

5cm

Characteristic:

post-processing using inertial explorer software


(7)RGB Camera (FLIR BFLY-U3-23S3C)

Quantity:

X2

Frequency:

20Fps

Resolution:

1920×1200 pixel

FOV:

96.8°(Horizontal),79.4°(Vertical)


(8)Event Camera

Inivation DAVIS346-color:

Quantity:

X2

    

Inivation DVXplorer:

Quantity:

X2

Frequency:

1000Hz

    

Frequency:

1000Hz

Resolution:

346×260 pixel

    

Resolution:

640×480 pixel

IMU:

MPU6150, 6-axis

    

IMU:

MPU6150, 6-axis



Sensor setups

Image description Image description

Calibration rosbag and Our calibration result

title1

title2

title3

title4

Image Image Image Image

Calibration:rosbag

Calibration:rosbag

Calibration:rosbag

Calibration:rosbag


Our calibration result:txt


Data

Urban_day_loop

Urban_day_unloop

Image Image

Data: rosbag

Groud-Truth: txt

Data: rosbag

Groud-Truth: txt



Urban_night_loop

Urban_night_unloop

Image Image

Data: rosbag

Groud-Truth: rosbag

Data: rosbag

Groud-Truth: rosbag



All topics published by automatic driving platform

Sensors Topics
Event cameras /DAVIS346_left/events
/DAVIS346_left/image
/DAVIS346_left/imu
/DAVIS346_right/events
/DAVIS346_right/image
/DAVIS346_right/imu
/DVXplorer_left/events
/DVXplorer_left/imu
/DVXplorer_right/events
/DVXplorer_right/imu
RGB cameras /camera_left/image_raw
/camera_right/image_raw
IMU /imu/data
Mechanical LiDARs /left/lslidar_point_cloud
/right/velodyne_points
/velodyne_points
Time synchronization /ptp/trigger_time