Utilizing NVIDIA Omniverse™ for Digital Twin Projects at the NTT DATA Innovation Center

What is Nvidia Omniverse?

NVIDIA Omniverse is a computing platform that enables individuals and teams to develop Universal Scene Description (USD)-based 3D workflows and applications. It has been developed by NVIDIA as an open ecosystem that permits artists, designers, and engineers to work together collaboratively on complex projects, also allowing developers to create custom plugins and applications that can be integrated into the platform. Omniverse integrates various technologies, including advanced real-time graphics, physical simulation, ray tracing rendering, VR integration and AI training. The platform allows users to create detailed and realistic virtual environments, where they can design, simulate, and visualize complex objects, scenes, and interactions.

One of the key features of Omniverse is its ability to connect different software and design tools, facilitating collaboration on projects even with tools outside the Omniverse ecosystem. Furthermore, there is the real-time streaming functionality, allowing users to view and interact with their projects on various devices, including PCs, remote servers, and mobile devices.

How Digital Twin Applications can benefit from NVIDIA Omniverse

The integration of NVIDIA Omniverse in digital twin applications brings a multitude of benefits on many levels, from scene realism enhancement to AI enabling use-cases, leveraging virtual environments for training, testing, simulation, and design optimization purposes empowers AI algorithms to learn, adapt, and contribute to advancements in these fields.

One potential application lies in the field of autonomous vehicles, where AI algorithms can be trained and tested in realistic virtual environments before being deployed on actual roads. This approach enables safer and more efficient development, as it allows for the exploration of diverse scenarios, weather conditions, and challenging situations that may be difficult to replicate in the real world. Another application of virtual photorealistic scenes can be found in the healthcare sector, virtual scenes can aid in medical training and surgical simulations. AI-powered virtual environments can replicate intricate medical procedures, allowing medical professionals to practice and refine their skills in a safe and controlled setting. This reduces the risk associated with real-life interventions and enhances patient safety.

Furthermore, in architecture and urban planning, AI algorithms can analyze these scenes to simulate the impact of different design choices, such as building placement, material selection, or lighting conditions. This empowers decision-makers and urban planners to make informed decisions and optimize designs for functionality, aesthetics, and energy efficiency.

The plethora of connectors in NVIDIA Omniverse enables seamless interoperability with various third-party tools. This means that digital twin applications can easily integrate with existing popular 3d software such as Autodesk Maya, 3ds Max, Unreal Engine, through connectors that facilitate bidirectional data exchange, enabling developers to seamlessly synchronize 3D models with Omniverse. This capability provides a significant advantage, as it allows developers to leverage existing assets and models without the need for extensive manual rework. Moreover, Omniverse connectors facilitate a collaborative workflow between different professionals involved in the digital twin development process. Usually, architects, engineers, designers, and domain experts use different specialized 3D modeling software for their respective tasks, making it harder to merge their work but, through connectors, these professionals can operate on their preferred software environment by creating and refining 3D models, and then easily transferring and integrating their work to the Omniverse platform.

NVIDIA Omniverse also boosts photorealistic environments, thanks to real-time ray tracing technology (RTX), enabling stunning visual fidelity and thus representing digital twins that closely resemble their physical counterparts. Indeed, just by importing an existing 3D model into Omniverse we can achieve much more realism thanks to better lighting, reflections and texture rendering.

Figure 1: Process Simulate 3D environment (on the left) imported on Omniverse (on the right)

This high level of realism, aside from improving the overall visual quality of digital twin scenes, enables the possibility of training artificial intelligence (AI) algorithms in the digital world that would also perform effectively when deployed in a real-world setting, overcoming the most common problem when training AI: datasets to refine the model. Omniverse synthetic data generation simplifies and accelerates AI development by eliminating manual data collection, addressing data limitations, enabling comprehensive testing and offering cost efficiency, faster iterations and improved model robustness. In a single environment, AIs can be built, trained, and finally deployed seamlessly from the virtual to the real environment.

This ability to leverage virtual training data and environments accelerates the AI development process and ensures that AI models are well-equipped to handle real-world scenarios. This allows customers to plan and optimize their assets without risking any harm to their real-world production settings thus minimizing downtime, costs and improving overall productivity.

Furthermore, VR integration in Omniverse is incredibly user-friendly and offered out-of-the-box without any need to make modifications to the 3D model. This enables users to create immersive simulations that replicate real-world scenarios, allowing them to practice and refine their skills in a safe and controlled virtual environment. This is particularly beneficial for job training, as it provides a realistic setting without the risks associated with real-world tasks. It also allows multiple users to engage in the same VR environment simultaneously, promoting teamwork, knowledge sharing, and interactive training sessions, enhancing the overall learning experience.

Success Story: Digital Twin of a Data Center

Customers relying on large infrastructures for their business increasingly need to centralize in a single platform the different complex tools they use to perform various tasks necessary for data center management, such as infrastructure monitoring and simulation of new scenarios and configurations. In this regard, developing a digital twin for a data center represents a significant step in the field of intelligent infrastructure management since it is the best way to address these needs.

NTT DATA Italy tackled all these challenges when building a digital twin of a data center for an important customer in the media industry. The customer was looking for a complete and synchronized virtual replica of their data center with a high level of detail, including specific data such as temperature andavailable devices, that could allow for real-time monitoring and visualization of alarms and physical-electrical aspects as well as testing different operating scenarios using Computational Fluid Dynamics (CFD) simulations.

The use of NVIDIA Omniverse in the development of this digital twin application proved to be a key contribution to the success of the project. Thanks to the connectors to 3d modeling tools such as Autodesk Maya and Blender, it was easy to import a 3d high fidelity replica of the data center into Omniverse, with a level of detail and realism that allows operators to visualize every architectural aspect and piece of equipment, enabling them to explore and interact with the virtual environment as if they were physically present.

As also described in the previous section, this allowed the customer to train operators on simulated scenarios, allowing the staff to familiarize themselves with equipment, procedures, and emergency protocols in a safe and immersive environment enhancing the effectiveness of training programs and improving the response times and overall preparedness for critical situations.

The collaboration aspect of NVIDIA Omniverse proved to be another key advantage, allowing multiple stakeholders, including data center operators, engineers, and decision-makers, to collaborate and exchange insights within the virtual environment.

The deployment of Pixar's USD file format, which stands at the base of Omniverse, is one of the development benefits that this platform can offer. This format allows the digital twin application to store various metadata directly on the 3D models of the data center components.

This allows information such as temperature, humidity, CPU consumption, and PUE (Power Usage Effectiveness) to be stored and then linked with real devices so that they can be synchronized in real time with their physical counterparts.

Furthermore, thanks to being able to build custom extensions, it was possible to easily define interactive behaviors that depend on the metadata values. For example, a device could be highlighted by a bright red color if the value of the PUE reaches a critical threshold or, again, an alarm banner could appear near a device if its overall temperature reaches a critical value.

Finally, with Omniverse it was possible to make use of the vast amount of data collected and synchronized from various sources to realistically simulate temperatures, air flow, cooling and power consumption in the data center, thanks to the connectors available with advanced simulation tools like Paraview, that allowed for the visualization of the results through readable heatmaps, stream tracers or labels. For instance, it is possible to simulate how the cooling system will be affected or how the power consumption of the facility will increase when the data center is expanded with a new device.

In conclusion, thanks to the digital twin built with omniverse, the customer was able to centralize all the tools used to monitor, plan and optimize the data center, making it easy to make informed decisions about preventive maintenance strategies, resource allocation, and operations optimization, reducing downtime and enhancing overall efficiency.

Using data center digital twins, AI can leverage the photorealistic environment to simulate and optimize various operational scenarios. For example, robots can be trained to autonomously perform tasks such as server maintenance, cable routing, or equipment installation. The photorealistic environment provides a realistic setting for AI algorithms to learn and adapt to different situations, improving their ability to handle real-world challenges within the actual data center. In this scenario, AI algorithms can analyze the digital twin's photorealistic representation to understand the spatial arrangement of equipment, identify optimal paths, and develop efficient strategies for completing tasks. By iteratively training and fine-tuning the AI models within the data center digital twin, robots can acquire the necessary skills to perform complex operations accurately, quickly, and safely.

This is an initiative of the Innovation Center, Research and Development Headquarters.