From a 3D city to a decision: geospatial AI for digital twins

A detailed 3D city makes a place easier to inspect. It does not, by itself, explain which neighborhoods lack services, which assets need further investigation or how a proposed development relates to its surroundings.

Those questions require analysis as well as visualization. LGM’s proposed role in digital-twin workflows is to provide geospatial context and computation that an existing city, infrastructure or engineering platform can use.

Add an analytical service, not another source of truth

A customer may already maintain an authoritative asset model, project database or operational platform. Integrating LGM does not require replacing those systems.

Instead, define a question that needs external geographic context or spatial computation. The customer platform supplies the relevant area or asset references; LGM works with the available and permitted datasets; the application receives a result it can display or review.

For example, an infrastructure team might investigate the surroundings of several assets before deciding where a more detailed assessment is needed. The output is a screening layer with evidence, not an automatic engineering approval.

Connect context at the scale of the decision

Regional terrain can support an initial territorial assessment. Building or parcel information may support a more localized investigation. Operational records can explain conditions that neither dataset captures.

The analysis must make those scales explicit. A coarse raster should not be presented as a precise measurement at an individual asset. Likewise, a detailed 3D object does not make the underlying environmental information more accurate.

Useful context may include land cover, transport, population, soil, facilities and customer-supplied collections. Availability and fitness for the task need to be checked for each region.

Return results that can travel between systems

A proposed integration should agree on an output contract before building the visual experience. At minimum, define:

• Geometry or geographic extent, with the coordinate reference system.
• Stable feature references where they are available.
• Indicator names, units and missing-value behavior.
• Source dates and analysis assumptions.
• The distinction between an observation, a calculation and a recommendation.

The receiving application may need a geometry layer, a table linked to assets, a raster product or a combination. Integration work includes adapting those outputs to the destination, rather than assuming one format fits every engine.

Keep exploratory analysis separate from simulation

Showing a flood-related layer is not the same as running a validated hydrological model. Comparing road accessibility is not equivalent to simulating traffic. A useful product makes these boundaries understandable to its users.

Where specialist simulation is needed, LGM can be considered as part of a workflow connecting data preparation, analytical tools and expert review. The required model, calibration and validation remain a separate scope.

Start with one complete integration loop

LGM’s distribution direction includes digital-twin and engineering ecosystems such as Cesium, Autodesk, Bentley and NVIDIA Omniverse. These are target integration pathways, not a claim of released native connectors or endorsed partnerships with each vendor.

A practical pilot follows one question from the customer’s interface to a spatial result and back. Verify the result, its provenance and its usefulness before adding more capabilities.

Explore LGM integrations and the API to frame that connection. For users who want to investigate directly, LGM GIS provides the product’s map-and-chat access surface.