AI & ML interests

None defined yet.

Recent Activity

Organization Card

Automatum Data

Drone-based trajectory data of real traffic — bird's-eye view, all tracked road users, OpenDRIVE road models. Automatum Data GmbH, Ingolstadt, Germany · method published at IEEE IV 2021 · in operation since 2021

Rendering videos of the five datasets (3D digital twins rendered from our trajectories; last clip: pipeline video of the IEEE IV 2021 paper). Datenquelle: Bayerische Vermessungsverwaltung – www.geodaten.bayern.de, Daten verändert (CC BY 4.0).

Open datasets

Free excerpts from the Automatum data pool for research and teaching. Every dataset card follows the same structure: rendering video, key figures, recordings, data format, quick start, example scripts.

Dataset Content License
T-Crossing 2 recordings · 683 vehicles · urban/rural T-junctions CC BY-ND 4.0
Roundabout 2 recordings · 212 vehicles · single-lane roundabouts CC BY-ND 4.0
Highway with Ramps 4 recordings · 7,178 vehicles · on-/off-ramps, merging CC BY-ND 4.0
Full Highway 114 recordings · ~200,000 vehicles · 30 h of A9 motorway (IEEE IV 2021) CC BY-ND 4.0
Highway Area Pools 3 multi-drone pools · 3,951 vehicles · ~1.2 km per scene, Parquet viewer CC BY-NC 4.0

Custom drone recordings for research projects — your site, your time window: how it works · info@automatum-data.com

Data for your own research question

The public datasets are excerpts. The same validated pipeline has produced 1,670 recordings (368 hours) at 128 sites as of November 2026, and it records and processes new data on request:

  • Your site, your time window. Junctions, roundabouts, motorway sections, construction-site layouts, service areas — recorded where your research question lives.
  • All tracked road users, full area. Cars, vans, trucks, buses, motorcycles and vehicles with trailers in one consistent trajectory set with an OpenDRIVE road model. Several drones (up to 18 so far) can be fused in time and space into continuous trajectories over more than one kilometre.
  • Timing. Recording and processing can be scheduled for the start of your project, so the first project phase can be spent on analysis rather than on data collection.
  • For proposals. Feasibility note, cost estimate and, if required, a letter of intent as subcontractor or partner for your funding application. Several groups can share one recording campaign.
  • For theses and publications. Student theses and joint method papers are welcome.

Write to info@automatum-data.com with a short description of your research question and site, or use the contact form on automatum-data.com. Please include the site, the time window, the road users of interest and whether the data are for a proposal or a running project. We reply with a feasibility note and a cost estimate.

What the data contain

  • Kinematics for every vehicle: position (UTM), velocity, acceleration, jerk, heading, size and class, at 29.97 Hz.
  • Road model: OpenDRIVE (XODR) per site; lane assignment, neighbour relations, TTC/THW and lane-marking distances in most releases; OpenSCENARIO exports on request.
  • Validated processing: < 0.2 % velocity error; every processing step is versioned and recorded in each file (CalculationStepVersions), so results stay traceable and reproducible.
  • Python access: pip install openautomatumdronedata — one loader for all datasets.

Typical uses: traffic engineering and traffic surveys, ADAS/AD scenario extraction and validation, behaviour and interaction modelling, trajectory prediction.

Quick start

from openautomatumdronedata.dataset import droneDataset

dataset = droneDataset("path/to/recording")      # a folder from Sample_Data/ or the archive
for obj in dataset.dynWorld.get_list_of_dynamic_objects()[:5]:
    speed = (obj.vx_vec[0]**2 + obj.vy_vec[0]**2) ** 0.5 * 3.6
    print(obj.UUID, obj.type, f"{speed:.1f} km/h")

Citation

@inproceedings{spannaus2021automatum,
  title     = {AUTOMATUM DATA: Drone-based highway dataset for development and validation of automated driving software},
  author    = {Spannaus, Paul and Zechel, Peter and Lenz, Kilian},
  booktitle = {IEEE Intelligent Vehicles Symposium (IV)},
  year      = {2021}
}

Contact

automatum-data.com · info@automatum-data.com · Documentation · PyPI

Geodata attribution

The 3D renderings (video and preview images) use official open geodata (DOP20, DGM1, LoD2, airborne laser data): Datenquelle: Bayerische Vermessungsverwaltung – www.geodaten.bayern.de, Daten verändert (CC BY 4.0, https://creativecommons.org/licenses/by/4.0/).

Automatum Data GmbH · Ingolstadt, Germany

models 0

None public yet