Infinite Geometry Streaming from the cloud

3D Repo is pioneering the development of a new cutting-edge solution for infinite geometry streaming massive 3D models in the cloud using Unreal Engine.

Recently, there has been an increased interest in digital twins and vast open-world simulations for engineering and other purposes and even for creating the ‘metaverse’ itself. 3D engineering models continue to get bigger and more complex by the day with some boasting millions of objects and hundreds of millions of polygons.

AEC professionals are increasingly finding it difficult to visualise large-scale projects with existing solutions. Unfortunately, due to memory limitations imposed by web browsers, some extremely large models simply will not load regardless of the power of the underlying computer being used.

Due to the COVID-19 pandemic, cloud computing and remote working have also become more common and the need for more collaborative solutions has risen dramatically.

The 3D Repo infinite geometry streaming solution with Unreal Engine aims to remove those technological barriers to collaboration to enable the rendering of massive 3D models in the cloud for engineering, real-world simulations, or possibly even the metaverse.

This project was made possible after 3D Repo received an Epic MegaGrant in 2021.

The 3D Repo Infinite Geometry Streaming Solution

Instead of loading the entire model, the 3D Repo geometry streaming solution uses pre-optimised assets which are dynamically loaded directly into a running game, giving users the ability to log in via their 3D Repo credentials and seamlessly retrieve any of their assets on the fly.

Via open APIs, objects are grouped into small clusters of super meshes that allow priority loading of geometry by the engine and client, allowing users to explore conceivably vast open worlds.

This process operates in four stages:

1. Small clusters of objects are grouped into super meshes
2. Clusters are organised and structured
3. Information about the clusters is passed to the client so it can request specific clusters on demand
4. The client loads these clusters according to priority

Using a federation from the Canary Wharf, Wood Wharf development in London – the image gallery below shows an example of the clusters of objects being loaded into view as the viewer approaches closer to the tower. For this example, we have set the threshold exceptionally low to make it obvious how the geometry is being streamed in and out as the viewer moves around.

40 Leadenhall Street Commercial Office

Led by Mace, 40 Leadenhall Street is a commercial office project and one of the biggest schemes to ever receive planning permission in the City of London. In fact, it is the most digitally advanced project in Mace’s portfolio!

This project was selected as a test for the new infinite streaming solution as it is one of the largest projects 3D Repo has ever come across, with over 1.8 million objects, 700 million polygons, and two billion vertices – spanning some 100 GB of data!

Once again, for example purposes, we have set the threshold exceptionally low to make it obvious how the geometry is being streamed in and out as the viewer moves around. The empty bounding boxes, shown in blue, depict the geometry that is available in the server but has not yet been loaded into the GPU memory.

In this example, the federation has been uploaded from Revit directly to the 3D Repo cloud for processing. From there, the data is made available through APIs where it can be used in the Unreal Engine and business intelligence tools such as Microsoft Power BI.