UAV

PASTEL

Aerial lidar SLAM tuning and robustness

Seventeen reported manually flown UAV sequences with three simultaneous lidar and three IMU streams, tracked reference positions, registered reference maps and calibration parameters.

Metadata: Complete Metadata curator reviewed Released 2025

Code

Not reported

Supplementary

Not reported

Quick facts

Vehicle type: UAV
Environment: Outdoor
Data origin: Real-world
Sensor count: 6
Ground truth: Available
Calibration: Parameters Only
Annotations: Not reported

UWB

Not reported

Ground-truth references for PASTEL
Scope	Method	Reference system	Count	Rate	Coverage	Reference provenance External Measured using equipment independent of the platform’s normal onboard sensor pipeline. Derived Computed primarily from the dataset’s own recorded sensors. Hybrid Combines independent reference equipment with onboard measurements.
3D position	Total station	Leica total station · Nova MS60	1 total station	20 Hz	Partial · Flight Trajectories With Temporary Gaps From Prism Occlusion.	Reference provenance External Measured using equipment independent of the platform’s normal onboard sensor pipeline. Derived Computed primarily from the dataset’s own recorded sensors. Hybrid Combines independent reference equipment with onboard measurements. External
Accuracy notes The introduction describes millimeter-accuracy Leica equipment; no independent sequence-level pose accuracy is specified. Clock alignment and prism-to-sensor transforms matter. A PoseStamped message does not establish external orientation measurements. Reference notes Tracks one prism on the UAV. Topic /leica/pose, frame map; trajectory CSV is compatible with NTU VIRAL evaluation. Transform sensor estimates into prism frame for comparison; the evaluation interpolates only gaps below 0.1 s and aligns trajectories by Umeyama. This same Leica apparatus also supplies the separate map product; counts do not imply another physical station.
Other	Total station	Leica total station	Not reported	Not reported	Full	Reference provenance External Measured using equipment independent of the platform’s normal onboard sensor pipeline. Derived Computed primarily from the dataset’s own recorded sensors. Hybrid Combines independent reference equipment with onboard measurements. Hybrid
Accuracy notes Merged scans report 2–4 cm ICP RMS; SLAM-map comparisons in Table 8 have RMSE below 0.1 m. These are registration/evaluation statistics, not independent global map-error guarantees. Reference notes Static reference maps in LAS format, formed from twelve Leica scan positions using cleaned overlap regions for ICP and original scans for merging. Multiple map configurations reflect moved panels. Table 5 assigns Seq. 9 two files and repeats one filename across two rows; preserve the mapping conflict and uncertain distinct-map count.

Level: Parameters Only
Processed parameters: Available
Raw calibration data: Not reported
Calibration targets: Not reported

Parameters only

Processed calibration parameters are reported, but raw calibration data is not reported.

Notes

URDF coordinate transforms and per-sensor YAML transformations/IMU intrinsics are released. Translations were measured with 0.5 mm precision; rotations used lidar_imu_calib, intrinsics used Allan variance, and Ouster lidar-to-IMU transforms came from documentation. Raw calibration recordings are not established.

Temporary prism occlusion causes trajectory truth gaps; independent reference orientation is not established.

Payload cabling, frame and landing gear create lidar occlusions; no-return points were removed.

Wind can move the safety net and increase map-evaluation error.

Three reported data-table issues need clarification: sequence count, Seq. 9 map assignment and matching statistics for Seqs. 14/17.

Recorded flights are manual at one structured testbed; vegetation-rich scenes, dynamic urban obstacles and autonomous flights are future extensions.

Statistics retain the analysed paper and version-1 Zenodo scope. The repository points to a newer release (22145476), whose additional numbered bags are not silently merged into these totals.

Not reported

Ouster OS0-128 integrated IMU

Model: InvenSense ICM-20948
Count: 1
Accel / gyro: 100 / 100 Hz

Ouster OS1-16 integrated IMU

Model: InvenSense ICM-20948
Count: 1
Accel / gyro: 100 / 100 Hz

CubePilot Cube Black flight-controller IMU stream

Model: Not reported
Count: 1
Accel / gyro: 25 / 25 Hz

Not reported

Ouster OS0-128 (Rev 6)

Model or name: Ouster OS0-128 (Rev 6)
Count: 1
Dimensions: 3
Rate: 10 Hz
Channels: 128
Range: 50 m

Ouster OS1-16 (Rev A)

Model or name: Ouster OS1-16 (Rev A)
Count: 1
Dimensions: 3
Rate: 10 Hz
Channels: 16
Range: 110 m

Velodyne VLP 16 (Puck Lite)

Model or name: Velodyne VLP 16 (Puck Lite)
Count: 1
Dimensions: 3
Rate: 10 Hz
Channels: 16
Range: 100 m

Not reported

Availability: Not reported
Format: Not reported
Class count: Not reported

Semantic segmentation Not reported

Instance segmentation Not reported

Object detection Not reported

Optical flow Not reported

Depth ground truth Not reported

Citation key: milijas2025pastel

@article{milijas2025pastel,
  title={PASTEL: An Aerial Multi-LiDAR Dataset for Research in SLAM Tuning and Robustness},
  author={Milijas, Robert and Martínez-de Dios, Jose Ramiro and Bogdan, Stjepan},
  journal={IEEE Access},
  volume={13},
  pages={153010--153023},
  year={2025},
  doi={10.1109/ACCESS.2025.3603733}
}

Dataset corrections

Notice a specific issue after reviewing the page?

Report an inaccuracy

PASTEL

Access links

Website

Download

Paper

Code

Supplementary

Quick facts

Sensor overview

Cameras

IMUs

GNSS

LiDAR

Radar

Magnetometers

Barometers

UWB

Ground truth

Calibration and synchronization

Parameters only

Notes

Known or intentional limitations

Cameras

IMUs

Ouster OS0-128 integrated IMU

Ouster OS1-16 integrated IMU

CubePilot Cube Black flight-controller IMU stream

GNSS

LiDAR

Ouster OS0-128 (Rev 6)

Ouster OS1-16 (Rev A)

Velodyne VLP 16 (Puck Lite)

Radar

Additional sensors

Annotations

Citation

BibTeX

Dataset corrections