The wireless industry stands at the brink of a transformation, driven by the fusion of AI with advanced 5G and upcoming 6G technologies that promise unparalleled speeds, ultra-low latency, and seamless connectivity for billions of AI-powered endpoints. 6G specifically will be AI-native, enabling integrated sensing and communications, supporting immersive technologies like extended reality and holographic interfaces, and delivering new levels of network performance, spectral efficiency and energy efficiency, with AI. Achieving the full potential of AI-native 5G and 6G wireless networks requires innovative approaches for network design, simulation, deployment, and testing.
At NVIDIA GTC 2024, we introduced the Aerial Omniverse Digital Twin (AODT), an AI-native digital twin platform built with NVIDIA Omniverse and designed for cutting-edge research and development (R&D) of 5G and 6G wireless systems.
AODT key features include:
- Simulation of the actual physical (PHY) and medium access control (MAC) layers of a 5G/6G network without abstractions.
- Application of highly-accurate ray-traced channels to the PHY signals.
This year we’re announcing an expansion of AODT capabilities with new feature enhancements and broader accessibility options. Available everywhere, from a laptop, server, public cloud, or as an NVIDIA cloud service, AODT enables users to move seamlessly across diverse computing environments. Developers can work on a wide spectrum of projects, whether they require small, single GPU computation power or larger, multi-GPU computational support.
AODT everywhere
AODT offers flexibility and multiple accessibility options to developers, including:
- Laptops: A cost-effective solution for 6G researchers.
- Workstations: Available with single or dual GPU configuration.
- Servers: Equipped with multiple GPUs for enhanced performance.
- Public cloud: Supported on major platforms including Azure and AWS.
- NVIDIA Cloud Service: AODT supported on NVIDIA DGX Cloud Serverless Inference, enabling access via web portal (see Figure 1).
Doing this, encourages a more inclusive and dynamic environment for innovation, accelerating the development, and deployment of AI-native wireless technologies.
| Laptop (NVIDIA RTX 3500 Ada) | Workstation (NVIDIA RTX 6000 Ada) | NVIDIA Cloud Service | Server (NVIDIA RTX 6000 Ada) | Public Cloud (NVIDIA L40 and NVIDIA L40S) | |
|---|---|---|---|---|---|
| Simulation Scale | Up to 3 RU | Up to 9 RU | Up to 57 RUs (*roadmap) | Up to 57 RUs(*roadmap) | Up to 57 RUs(*roadmap) |
| System Setup | Easy | Equipment procurement | – Immediate access using web portal- No HW & SW installation needed- Multi-GPU (*roadmap) | – Equipment procurement- Multi-GPU (*roadmap) | – No HW installation needed- Multi-GPU (*roadmap) |
| Use Cases | – Basic PHY design – Small datasetgeneration | – PHY and basic MAC design – Medium dataset generation | – PHY and advanced MAC design with fast simulation | – PHY and advanced MAC design with fast simulation – Large dataset generation | – PHY and advanced MAC design with fast simulation – Large dataset generation |
Accelerating AI-native wireless innovations with AODT
AODT enables innovation by offering a high-fidelity digital twin of the radio access network (RAN) and of the radio environment in which it operates. This platform provides numerous benefits to developers, including:
- Accelerated commercialization of advanced 5G and 6G algorithms: The digital twin platform accelerates the development and testing of innovative AI/ML algorithms thanks to its tight integration with NVIDIA Aerial CUDA-Accelerated RAN.
- Enhanced computation speed: AODT leverages NVIDIA RTX technology to deliver faster RF simulations, enabling large-scale simulations or quick runtimes.
- Physically accurate modeling: By using geo-spatial data, live network information, modeling reflections, transmissions, diffractions, and diffusions in detail, AODT produces RF environments that closely replicate the real world.
- Modular architecture: AODT’s modular design, coupled with well-defined APIs, enables ecosystem partners to extend their toolchains and serve their customers across a variety of new use cases.
- Synthetic data generation: AODT offers an SQL interface for all the results derived within the simulations, and an H5 interface to directly dump waveforms. This enables the creation of datasets for training, simulating, and optimizing advanced algorithms to enhance network performance and user experience.
Modeling AI for the physical world in a three-computer solution
AODT is part of the NVIDIA AI Aerial Platform, which provides a complete three-computer solution (see Figure 2) to:
- Develop and train: Design cutting-edge algorithms and train AI models using NVIDIA Aerial AI Radio Frameworks.
- Simulate and fine-tune: Simulate complex network scenarios, test and refine models with realistic and repeatable environments, and optimize performance under various channel conditions.
- Deploy: Deploy trained models on an NVIDIA Aerial Commercial Test Bed (ARC-OTA) for testing and validation. Developers can capture field behavior early on, and increase the quality and quantity of the data used to train and tune algorithms.
This comprehensive approach streamlines the entire process from initial development to final deployment, fostering innovation and accelerating the advancement of AI-native wireless networks for 5G and 6G. NVIDIA AI Aerial provides the tools to train, simulate, and deploy new AI/ML innovations. It enables a data loop across the three computers, for developers to push the boundaries of wireless communication, ultimately leading to more robust and efficient networks.
Empowering the developer ecosystem with AODT
Given its modular architecture as shown in Figure 3, AODT has been integrated by various ecosystem partners to accelerate the development of next-generation networks.
Some notable partnerships include:
Amazon
AWS developed a solution capable of predicting radio frequency coverage for telecom operators. This solution leverages AODT to build a full network digital twin on AWS, utilizing Amazon SageMaker, Amazon RedShift, and Amazon QuickSight. The integration is validated on Amazon Elastic Cloud Compute (EC2) G6e instances, powered by NVIDIA L40S. This technology enables wireless operators to accurately and efficiently design, simulate, and optimize their networks. The capability is being demoed at the AWS booth.
Amdocs
AODT is integrated into Amdocs Network Optimization Suite, a key component of the expanded Amdocs amAlz Suite. This enhancement introduces modular and comprehensive network agents, enabling operators to create a digital twin of their network. This helps the operator optimize network planning, streamline network deployment for greater efficiency, and deliver enhanced customer service. Amdocs announced new product expansion at NVIDIA GTC 2025.
Ansys
The company integrated its Perceive EM real-time, dynamic electromagnetic wave simulation solver API (based on physical optics) into the AODT platform, as an alternative solver to NVIDIA EM Solver (based on geometrical optics). This enables the AODT platform to also address non-communication use cases, such as radar or the mixture of communications and radar. As part of the AODT ecosystem partnership, Ansys customers can also use NVIDIA Aerial CUDA-Accelerated RAN and Aerial AI Radio Frameworks for AI-native wireless network research and development. More details of Ansys’ integration work is provided in their technical blog.
Keysight
Keysight introduced a new product line, the RF Raytracing Digital Twin Solution (RaySim), as part of the Keysight Channel Studio Software Suite. RaySim leverages the AODT platform to enable modeling of wireless deployment scenarios in complex 3D environments with high fidelity. Using the computational power of NVIDIA GPUs, RaySim accelerates complex RF raytracing, delivering highly accurate, site-specific channel models tailored for 6G R&D. Combined with Keysight’s base station, user equipment and channel emulators, RaySim creates a comprehensive Physical AI testing environment for next-generation wireless systems. For more details on Keysight RaySim, refer to their technical blog post.
MathWorks
By integrating AODT physical digital twin with MATLAB modeling and simulation functions, MathWorks users can leverage the rich library of MATLAB algorithms for wireless communications, including site-specific and physically accurate ray-tracing channel models for algorithm design with AODT. The combination results in accurate 5G and 6G simulations with rapid computation time. Read the MathWorks technical blog for more details on the integration.
VIAVI
VIAVI Solutions is expanding its TeraVM AI RAN Scenario Generator (AI RSG) product by integrating with AODT. AI RSG creates a RAN digital twin that replicates real-world conditions in the lab, harnessing data obtained through Open RAN interfaces to train the RIC and r/xApps. The integration enables more realistic and detailed radio access network (RAN) simulations, supporting network equipment manufacturers to leverage network data and AI/ML to optimize network operations. A detailed overview of the integration is available via the VIAVI Perspectives blog. ?A detailed overview of the integration is available via the VIAVI Perspectives blog post.
Conclusion
AI-native wireless networks will be designed in the digital world before being deployed in the physical world. NVIDIA Aerial Omniverse Digital Twin, part of the NVIDIA AI Aerial platform, is a critical enabler for validating and fine-tuning new AI models over the entire RAN, before deploying them in production networks.
At its core, AODT uses physically accurate modeling to deliver high-fidelity simulations of real-world network conditions. This ensures that developers can test and optimize their designs in a highly realistic simulation environment early in the development cycle. Its modular architecture further enhances flexibility and scalability, making it adaptable to a wide range of network design and simulation needs. Its cross-platform availability empowers developers to innovate without being constrained by hardware limitations.
As a key component of the NVIDIA AI Aerial platform, AODT also enables NVIDIA three-computer solutions for AI-native wireless network research and development, paving the way for AI-RAN and 6G. This integrated approach connects training, simulation, and deployment into a unified workflow and dataflow, dramatically accelerating the development process.
Embraced by a wide ecosystem of partners, AODT is at the forefront of AI-native wireless network development. Its ability to bridge the gap between theoretical design and practical implementation, enabled by fully software-defined and NVIDIA CUDA-Accelerated RAN software within NVIDIA AI Aerial, is set to revolutionize the speed and efficiency of wireless advancements. Ultimately, AODT is not just a tool but a driving force for shaping the future of AI-native 5G and 6G technologies.
Get started
Access the NVIDIA Aerial Omniverse Digital Twin by joining the NVIDIA 6G Developer Program.
