AODT City-Scale Modeling Methodology (e.g., Tokyo Scene) and Automation Approaches

Accelerated Computing Aerial Forum
1

Hello Aerial （ AODT） Team,

We are very impressed by the large-scale digital twin environments demonstrated in AODT, especially the detailed modeling of cities such as Tokyo. We would like to better understand the underlying methodology and engineering effort involved in building such environments.

We have several questions:

1. City-Scale Modeling Methodology

In AODT, the Tokyo digital twin appears to be highly detailed (including buildings, streets, materials, etc.).

  • What modeling approach is used to construct such a large-scale city scene?

    • Is it based on:

      • GIS data (e.g., OpenStreetMap, city GIS databases)?

      • Photogrammetry reconstruction?

      • Manual 3D modeling?

      • Or a hybrid pipeline?

  • How are material properties (e.g., reflection, diffraction characteristics for RF simulation) assigned in the scene?

2. Engineering Effort (Time & Human Resources)

For a city-scale model like Tokyo:

  • Approximately how much manual effort is involved?

  • What is the typical time scale to build such a model?

  • Is most of the process automated, or does it still require significant manual refinement?

Understanding this would help us evaluate feasibility for our own deployments.

3. Automated Modeling for New Cities (e.g., in the US)

We are very interested in building our own digital twin environments.

For example, if we want to model a city in the United States:

  • What are the recommended automated pipelines?

Specifically:

  • Can AODT / Omniverse workflows ingest:

    • LiDAR scans (e.g., from mobile mapping or drones)?

    • Depth camera / RGB-D reconstruction?

    • Satellite or aerial imagery?

  • Are there supported tools or best practices for:

    • Converting LiDAR / point clouds → USD scenes?

    • Automatically generating buildings and environments from GIS data?

4. Recommended Toolchain

Could you share a typical toolchain for city-scale modeling in AODT?

For example:

  • Omniverse + USD Composer

  • Sionna RT integration

  • Any data preprocessing tools (e.g., for point clouds or GIS)

We are actively working on AI-RAN and digital twin–based wireless system validation, and we are very interested in building realistic RF environments for our own deployments.

Any guidance or references would be greatly appreciated.

Thank you!