Semantic 3D twin · Situational picture · Planning · Test and training data for AI
Sovereign map and geographic information system.
Aerial imagery, cadastre and buildings in 3D, plus warnings, weather and air traffic in one map. From Jena, with our data in Germany.
Short video from SovAtlas: UAS geographical zones as 3D volumes, with a reported manned aircraft inside (ADS-B as of 28 September 2026, 18:16). City centre in a valley: the 3D city model on the aerial image, building by building. Training world through the changing seasons. UAS geographical zones as 3D volumes, with a reported manned aircraft inside (ADS-B as of 28 September 2026, 18:16).
Crisis management starts with situational awareness.
Sovereign: data in Frankfurt am Main.
Our map and building data is held there; the base map comes from the federal government.
Semantic: a dataset, not a picture.
Buildings carry height, roof pitch and solar suitability; control zones, their altitude band.
One map instead of many portals.
A geographic information system (GIS) for data from state and federal bodies and open networks.
WorldCrafter: test AI before it flies.
Colour, depth, class and material per capture, for autonomous systems in the air and on the ground.
- 3.14 millionMap tiles with about 16 GB of our own labels, from Europe down to street names in Germany
- 1,676Radiation monitoring probes nationwide, continuously updated on the map
- 53Layers and tools, freely selectable for each map
Generative world models produce plausible worlds. SovAtlas builds one from official measurements and open data.
Seen, played through, planned, trained.
Approach to a city, layer by layer, in one map.
Sources
State data
Aerial imagery · Laser scan · Cadastre · DGM1 elevation model (via terrain tiles)
Nationwide services
Warnings via the BBK · DWD · BfS · GFZ · DFS · BKG
Nationwide data, set up regionally
Census grid · Motorway status
Open networks
OpenStreetMap · Air traffic (open receiver network)
Other sources
Thermal hotspots (NASA) · Terrain tiles · Flight routes on click
SovAtlas 3D GISSemantic 3D twin
Building model
Height · Footprint · Solar suitability · Roof pitch
Air traffic checks
Aircraft in drone zone · GPS interference
Current situation
Warnings · Radiation · Weather · Earthquakes
Play it through
Viewshed · Sun and shadow · Fire (Demonstrator)
Water level
Scenario as a horizontal surface
Labels
our own map labels
Applications
One map per organisation
Layers and tools as configured
Access
Password · personal link · open
Embedding
on approved web addresses
Protected layers
KRITIS and power grid on protected maps only
From the same company
KRITISplan (opens in a new tab)
Crisis plans modelled on BSI Standard 200-4
Training worlds for autonomous systems
Three applications from one company, built on the same state and map data.
- today
- Outlook
- set up with the state’s data
Example: a hospital on the edge of town, surrounded by schools and nurseries, each in the colour of its sector.
The site is secured. Its surroundings are not.
Energy, water, hospitals, telecommunications: anyone responsible for keeping them running needs to know what is happening around the site – on the ground, in the air and in the weather. SovAtlas brings it together in one picture.
Example: a hospital on the edge of town, surrounded by schools and nurseries, each in the colour of its sector.
The KRITIS Umbrella Act (KRITIS-Dachgesetz) has been in force since 17 March 2026.
| Obligation | Contribution |
|---|---|
| § 13 Abs. 3 · Monitoring the surroundings | UAS geographical zones, reported air traffic, warnings, weather and radiation monitoring in one picture. |
| § 13 Abs. 3 · Training and exercises | KRITISplan documents exercises with their type, participants and findings. |
| § 18 · Notification to the BBK without undue delay and no later than 24 hours after becoming aware, stating the affected geographical area | KRITISplan records notification duties in the plan, with deadline and legal basis. |
Which buildings nearby are themselves critical?
Based on the use recorded in the cadastre, we assign buildings to a critical infrastructure sector, and the map colours them accordingly: health, education and care, utilities, government and administration, security and rescue. Cultural buildings appear in violet; all other buildings recede into grey. At a glance, you can see which hospital, which school and which utility facility are next to the site.
Example: viewshed from a castle above the valley. Where can the site be seen from?
From any point, including a rooftop, the viewshed calculates which areas of terrain can be seen: observer at a height of 2 to 500 metres, range up to 150 kilometres. A pre-selection for search operations and site planning.
What protects stays protected.
Our critical infrastructure layers and the power grid appear only on maps with a password or a personal access link. The database rejects any other assignment, with no exceptions for any role.
- Open map: rejected
- Map with password: allowed
- Personal access link: allowed
- Third-party embedding: blocked
From picture to plan.
KRITISplan records what follows from the picture: who acts, when, for how long and with what resources. The plans are modelled on BSI Standard 200-4 and run from preparation to follow-up. Each measure has seven mandatory fields; approval is only possible once everything is complete.
Document, exercises and notification duties
The result is a Word document with a map image for each phase, version details and an approval record.
KRITISplan documents exercises with their type, participants and findings. Notification duties are recorded in the plan with deadline and legal basis; KRITISplan helps with reporting duties, for example under NIS2.
Five phases, one plan.
Prolonged power outage (blackout) – southern supply area.
Sample plan from the KRITISplan demo; names, places and organisations are fictitious. Map: © OpenStreetMap contributors.
Risk analysis, planning of measures, awareness, exercises and continuous improvement before an incident occurs.
Update the grid control centre’s emergency manual
Annual review of escalation chains, switching authorisations and spare transformers.
Success criterion: Manual v2026 approved and communicated to all shifts
Black-start exercise, generator park
Quarterly load pick-up exercise for diesel-backed island grid formation.
Success criterion: Load pick-up < 90 s, frequency stable ±0.2 Hz
WorldCrafter: real places as test worlds for physical AI.
Anyone who operates autonomous systems in the air or on the ground needs the space as it is, and a world in which to practise safely. Both start with official and open geodata.
The same riverside location, the same camera path, from spring to winter.
Four seasons, 37 weather conditions, one call.
WorldCrafter turns the geodata of a real place into a 3D world for Unreal Engine 5, in which drones and ground vehicles can be tested and trained. The synthetic sensor data comes from the same world.
- Real places with edge lengths of 500 m and 1 km. Other places and delivery format on request.
Technical details of the training world
- Elevation model on a one-metre grid (DGM1), aerial imagery (DOP20) and map data from OpenStreetMap, georeferenced in ETRS89/UTM 32N. Same input, same world.
- Foliage, crops, snow, wet surfaces, fog, cloud and sun position change with a single call, without a rebuild.
- Grass, wheat, shrubs and trees move in five size classes, with gusts; wind strength and direction are parameters.
- Every surface carries a material and a class: 21 materials, each with its own physical material, and 54 classes. Almost all objects have collision.
- With every colour image (PNG): depth in centimetres, normals, class, material and instance as EXR, plus camera pose and camera data as JSON.
Every capture comes with its data, down to the thermal image.
From the material map of each capture, WorldCrafter calculates the apparent temperature in the long-wave infrared (8 to 14 µm). Over a summer day, asphalt heats up to more than 44 °C, grass cools the most at night, and water is slow to change. Calculated, not measured.
The airspace as it is.
Share of aircraft with disrupted satellite positioning per grid cell over Germany (ADS-B as of 28 September 2026, 19:50 to 20:04).
Aircraft report their position and how precisely they know it. From this, SovAtlas works out the share of aircraft with disrupted satellite positioning per area of about 1,900 km², over a rolling window of 15 minutes, updated every minute.
14.8%0.7%
reported positioning interference without and with equipment filter, nationwide
How it is counted
An equipment filter separates device quirks from the finding: on 19 September 2026, without the filter, 94 of 634 aircraft counted as affected, 89 of them gliders and microlights. With the filter, 4 of 538 remained.
Play it through before anything happens.
Water level scenario on a river, adjusted from 137 to 155 m above sea level.
Time-lapse over an old town with a cathedral: shadows of the 3D buildings over the course of a summer day.
Demonstrator: a hypothetical fire spreads across a wooded hillside at the simulation’s own pace; wind and rain at default values.
Your use case, the right application.
- Nationwide: drone zones, warnings, radiation monitoring network, air traffic, GPS interference, map labels.
- Regionally denser, with the state’s data: buildings, cadastre, noise, water.
- Beyond the borders: terrain, air traffic, GPS interference and map labels at overview scale.
Municipalities and planning
Which roofs suit solar, and where is there room for something new?
Energy and utilities
A grid connection has been requested. Which line, which substation is nearest?
KRITIS operators
What lies around the site, and who lives there?
Civil protection
An official warning comes in.
Real estate and insurers
Acquisition, listing or risk review: what applies at this address?
Drones and training
Before take-off: where may the drone fly?
Protected, traceable, open.
For requests from Germany, the server side of the map runs in Frankfurt am Main.
ISO/IEC 27001:2022 for rooom AG.
rooom AG, the maker of SovAtlas, is certified for the development, maintenance and operation of its SaaS platform, including cloud infrastructure and data centres. The certificate is on the Operations and assurance page.
One map per organisation.
Access by password or personal access link. Our critical infrastructure layers and the power grid appear only on such maps; the database rejects everything else.
Write access only with a second factor.
Every change in the administration is logged with person, action, target and time.
Open formats, no format lock-in.
GeoJSON, JSON tiles, 3D Tiles, PMTiles, PostgreSQL with PostGIS.
The world model is in place. So is the synthetic data.
Measured, not invented.
Terrain, buildings, population grid and protected areas: the objects carry their attributes.
Synthetic data from the same world.
Images with known ground truth, in sunshine as well as in fog and snow: WorldCrafter provides the material to test and train the vision of autonomous systems.
Which place should we show you?
In a demo, we show SovAtlas at a place that matters to you, with the data that comes together there today.
- A meeting with a demo, on your premises or online.
- A map for your organisation, password-protected, with the layers you need.
- A pilot with a defined term; scope, roles and rights are set out in the agreement.