Datasets recorded onboard an autonomous UAV system (described in [1, 2]) deployed within the environment of the DARPA Subterranean Challenge Finals (post-event testing). Contains mid-range flights in extremely narrow subterranean environments, with perceptual degradation from whirling dust, no dynamic obstacles, one mine-cave-mine loop, average speed ~0.5 m/s.
The rosbags contain unsynchronized LiDAR points and image streams of two cameras.
The groundtruth trajectory was estimated (code) during post-processing: registration of on-board data onto the map of the environment using ICP set with high-precision parameters.
Dataset | UAV name | Length (m, s) | Environment | Loop | Dust noise | Autonomy | GoPro |
---|---|---|---|---|---|---|---|
uav21_2 | uav21 | 306, 757 | metro | no | no | link | N/A |
uav21_3 | uav21 | 23, 58 | cave | no | no | link | N/A |
uav21_4 | uav21 | 196, 432 | mine | no | yes | link | N/A |
uav21_5 | uav21 | 377, 914 | metro | no | no | link | N/A |
uav21_6 | uav21 | 164, 432 | mine and cave | yes | yes | link | link |
uav22_1 | uav22 | 131, 332 | mine | no | yes | link | N/A |
uav22_2 | uav22 | 347, 347 | mine | no | yes | link | link |
uav22_3 | uav22 | 266, 644 | metro | no | no | link | N/A |
uav24_1 | uav24 | 162, 475 | urban storeroom | no | no | link | link |
- OS0-128 LiDAR (128 rows, 90 deg vFoV) topics:
- raw points:
/UAV_NAME/os_cloud_nodelet/points
of typesensor_msgs/PointCloud2
- to filter out the UAV frame, use points with minimal distance of
0.5 m
from the sensor origin
- to filter out the UAV frame, use points with minimal distance of
- processed points:
/UAV_NAME/os_cloud_nodelet/points_processed
of typesensor_msgs/PointCloud2
- 32 rows (every 4-th row of raw data, evenly spaced), filtered UAV frame, filtered dust
- raw points:
- Basler RGB cameras:
- resolution:
600 x 800 px
- two cameras [
basler_left
,basler_right
] - camera_info:
/UAV_NAME/basler_X/image_raw/camera_info
of typesensor_msgs/CameraInfo
- image:
/UAV_NAME/basler_X/image_raw/compressed
of typesensor_msgs/CompressedImage
- resolution:
- baselink:
UAV_NAME/fcu
- points:
UAV_NAME/os_lidar
- camera
basler_X
[basler_left
,basler_right
]:- optical frame:
UAV_NAME/basler_X_optical
- ROS frame:
UAV_NAME/basler_X
- optical frame:
UAV_NAME/fcu
└───> UAV_NAME/os_sensor
└───> UAV_NAME/os_lidar
└───> UAV_NAME/basler_X_optical
└───> UAV_NAME/basler_X
subt_finals
│ download.sh
│ subt_finals.pcd
│
└───DATASET
│ trajectory_groundtruth.txt
│ rosbag.bag
│ fcu_in_map.mat
└───...
- Script to download large data:
download.sh
- ground truth map of the environment:
subt_finals.pcd
- datasets:
rosbag.bag
- ground truth map of the environment:
- Ground truth trajectory:
trajectory_groundtruth.txt
- in local origin (use
fcu_in_map.mat
to transform to the map frame): zero-pose initialization - format:
timestamp x y z qx qy qz qw
- in local origin (use
- General mission information:
mission_data.txt
- Initial transformation
map->UAV_NAME/fcu
:fcu_in_map.mat
- 4x4 transformation matrix
- [1] Petracek, P.; Kratky, V.; Petrlik, M.; Baca, T.; Kratochvil, R.; Saska, M. Large-Scale Exploration of Cave Environments by Unmanned Aerial Vehicles, IEEE Robotics and Automation Letters 2021, 6, 7596–7603.
- [2] V. Kratky, P. Petracek, T. Baca and M. Saska, An autonomous unmanned aerial vehicle system for fast exploration of large complex indoor environments, Journal of field robotics, May 2021.