Towards an Industrial Metaverse : NVIDIA Omniverse has the building blocks

Michael T. Wagner
6 min readOct 24, 2021


A smart factory, or “the machine building the machine”, as Elon Musk puts it, is all about digital twins — these virtual counterparts of real world entities and processes are used all over the place when it comes to designing products, planning and controlling the processes in production and logistics or even after-sales when it comes to providing digital services for the delivered products.

digital factory twin, Image used courtesy of BMW AG

A digital twin of a product is regarded as the single source of truth providing all information about the product and its behaviour along its lifecycle. While a real entity is bound to physical space and time digital twins can be used to aggregate historic states and to predict or simulate behaviour.

The digital twin of a smart factory is a complex system of interconnected digital twins which represent the state and processes of all elements of the manufacturing environment for any point in time in the past and for planned scenarios in the future.

However, as of today, we dont have a smart factory with consistent digital twins but merely something what could be called “digital shadows” — information is scattered among different domains along the smart manufacturing value chain. Examples for different domains of the smart factory may be building- and infrastructure planning, product engineering, assembly planning, logistics planning and maintainance.

Collaboration bewteen different digital factory domains

In every domain there may exist only partial digital twins representing different aspects of real world resources or processes which are important for the domain specific use-cases.

communication with third parties is difficult with incomplete digital twins

While its already hard on a technical level to create a consistent digital factory twin integrating all resources and processes with their digital twins, its also a challenge for the people from different domains working together with just their share of information about their aspect of the smart factory.

They’re working on different locations, are using different expert tools, have different forms of visualizations which leads to much friction in communication and decision making.

Collaboration along the product life cycle and across supply chains

Digital twins need to be shared along the product life cycle and across supply chains.

Its one thing to deal with that complexity in a single factory but things get really interesting if we look into production networks — products of a part supplier are built into the product of the OEM, products of a tool supplier are used in the factory environment to manufacture the product of the OEM.
Using digital twins between suppliers and manufacturers would be beneficial throughout the product life cycle process, from design where the engineers can use the correct 3d part models with their physical attributes like material for CFD simulation to manufacturing where production planners can use the digital twin of the manufacturing resources to optimize the production process using up to date performance attributes of the assets in use.

Synchronization between reality and multiple planning scenarios

In every company there are many parallel planning projects with a potential impact on production — new product derivatives get introduced, the product mix is adapted to changes in projected demand, new logistics concepts are implemented and so on. These kind of projects rely on an up to date representation of the current production-processes and -environment. Using a holistic digital twin would enable to “branch off” the real environment any time into one or multiple virtual environments and simulate all aspects of the different projects scopes with respect to each other.

multiple active planning projects with alternative scenarios require “branching of” into projected realities

Constantly tracking the real production environment against the simulation scenarios would help to verify that the project goals will be reached while the different project plans are executed.

Currently we’re nowhere near to having such a multi-project multi-scenario planning and simulation environment unfortunately.

This is why next generation smart manufacturing networks need a common ground —

  • a common language to describe digital twins of smart factories and their elements
  • an open platform to access, share and connect them non-destructively across domain- and company borders in a secure way
  • collaborative visualization and planning tools enabling users to make better decisions and to work efficiently with their colleagues, suppliers and customers

— the Industrial Metaverse.

Ultimately the industrial metaverse is a key to more sustainability because it enables production networks to design and produce consumer products as efficiently as possible by optimizing manufacturing processes across entire supply chains.

The recently launched NVIDIA Omniverse has a lot of ingredients to make this a reality:

A common language : Universal Scene Description (USD)

The Universal Scene Description format has been developed by Pixar who has a long history of working with complex distributed scene graphs used to render their award winning animation movies.

Unique features like layering and composition arcs allow users from different digital factory domains to create assets which can be referenced and non-destructively modified in a global digital factory twin.

USD enables referencing and layering different assets in a global factory namespace

An open platform: NVIDIA Omniverse

NVIDIA Omniverse builds on USD as native format and adds enterprise backend services like distribution, caching, versioning and rendering with live-collaboration.

Another important building block is KIT, a rapid application development platform which is consequently built around an extension model using Python which makes it easy to create apps for specific use-cases.

Last not least it provides RTX realtime rendering to visualize and simulate the digital factory twin in high fidelity.

Omniverse Platform:

Omniverse used to create the digital factory twin at BMW:

Collaborative Visualization and Planning Tools:

In order to build up the industrial metaverse a critical mass of collaborative applications is needed to create a viral distribution effect — users need to get immediate benefits when using them — for example they dont have to use a myriad of converter tools and import/export processes which may or may not transport their digital twin aspect of their respective domain but are able to access a shared digital factory twin.

Therefore applications of the different factory domains have to be able to speak the common language USD and to connect to the platform — a first step is to create connectors which enable each application to share and consume digital twins in the omniverse — there is a growing amount of application connectors from different backgrounds out there doing that.

connected applications share and consume digitial twin elements Image used curtesy of ipolog

In order to get all benefits from the collaborative nature of the USD language and the omniverse platform it will make sense for applications to adopt the USD format as a native datamodel backend and to use the the advanced platform features like live synchronization and high end visualization.

NVIDIA has created several applications built on KIT which are part of the standard distribution of Omniverse, the most notable ones are “Create” used to build and modify 3d environments and “View“ which is used to review scenes and to interact with other users. Both applications leverage the extension system which has already over 200 extensions ranging from simple additional utilities to complex simulation frameworks.

This may ultimately lead to a new kind of lightweight and highly collaborative apps forming a manufacturing metaverse ecosystem.

NVIDIA Omniverse:

On Building Omniverse Tools and Connectors:

Further Reading:

Lets build it together !

While there is lots of development going on in in all areas from platform to connectors to applications there is still lots of work left to make the industrial metaverse a reality — one aspect is to define schemas which describe a digital twin semantically and to ensure their interoperability on a logical level — the omnigraph efforts may be a way to achieve this or also embedding python logic and user interface elements in the USD components.

These are exciting times for the digital factory enthusiast and I’m happy to be part of it :-)

Edit 10/2022: changed the headline wording to “Industrial Metaverse”

Icons made by wichaiwi



Michael T. Wagner

CTO and Co-Founder &, creating clever solutions for smart factory