This year, there is an anticipation of significant progress in one of the technology shifts necessary for the implementation of 6G: the emergence of AI-native RANs.
Currently, AI is primarily used as an add-on, serving as an optimization tool rather than a central component of the network. The AI-native architecture of the 6G network will utilize AI as the central nervous system of the network, providing intelligence for every decision. This transition will enable autonomous operations across complex, heterogeneous environments managed by programmable platforms like the RAN Intelligent Controller (RIC).
However, this shift is not without challenges, as AI models can be susceptible to drift and become unreliable when scaled across dynamic, real-world conditions. This degradation occurs because models are based on the world as it existed during training, while real-world threats and conditions continue to evolve. Even with comprehensive training data that represents the entire network, data remains historical and may not capture current network dynamics accurately.
RAN digital twins facilitate a hybrid data strategy, allowing AI algorithms to be continuously recalibrated and tested using a combination of emulated traffic and real-world data.
Through this approach, it becomes possible to prepare for potential upcoming threats, such as DoS attacks, and anticipate changes before they occur in the actual network.
Real versus synthetic training data
The initial step in creating the twin and AI models is training. The quality of the data used for training significantly impacts the effectiveness of the model. Reliable, high-quality data is crucial to ensure that the twin accurately reflects the real-world network.
While the traditional approach involves using data extracted from the live network itself, this method has limitations. Live network data is historical and limited in scope, often providing only a partial snapshot of network traffic from specific divisions within the operator. To address these limitations, synthetic data generated by an AI RAN Scenario Generator (RSG) can be utilized. This emulated data approach offers immediate access, eliminates security and privacy concerns, and can simulate various scenarios that may not have been encountered in the live network.
![\"\"](\"https://www.5gtechnologyworld.com/wp-content/uploads/2026/04/Figure-1-2.png\")
Figure 1. The AI RAN scenario generator uses the hybrid data layer to train and challenge xApps and rApps for use in digital twins. (Image: VIAVI)
The hybrid data layer
By leveraging this method, a high-fidelity digital twin can be created to accurately replicate the physics and behavior of the specific real-world network.
This twin environment can generate forward-looking data, enabling the training of AI-based workflows and models for near-real-time and non-real-time applications used by the RIC to execute hypothetical scenarios created by the RSG.
It is essential to consider not only emulating user equipment and traffic profiles but also the geographical context in which they operate. Therefore, a hybrid data layer approach should incorporate ray-tracing tools to model the position of towers in a specific area and predict signal propagation patterns around various obstacles, enhancing the accuracy of the simulation.
![\"\"](\"https://www.5gtechnologyworld.com/wp-content/uploads/2026/04/Figure-2-1024x531.jpg\")
Figure 2. A digital twin with ray tracing captures, shown by VIAVI at MWC 2026 in partnership with NVIDIA as part of its Aerial Omniverse Digital Twin platform. (Image: VIAVI)
Addressing AI drift
Networks are dynamic entities, and any AI model or digital twin left unattended will quickly fall out of sync with reality. This AI drift can lead to poor decision-making with unintended consequences. To prevent this, continuous validation frameworks are essential. Closed-loop digital twins can be used to test proposed AI changes, with the RSG assessing the impact on network KPIs in real time. The feedback generated from these simulations informs the AI’s subsequent decisions in an iterative loop.
An additional layer of oversight is provided by an app validation engine, which monitors the interaction between the application and the twin to ensure that the AI avoids undesirable trade-offs. By comparing primary objectives against guardrail KPIs like service quality and coverage, the validation engine quantifies trade-offs and prevents the AI from deviating into inefficient or unstable states.
Stress testing AI models
It is essential to evaluate how the network and AI RAN will perform under non-standard conditions, such as edge cases and attacks. The RAN digital twin plays a crucial role in preparing AI applications for unforeseen circumstances.
Using the RSG to conduct what-if experiments within the digital twin’s sandbox environment helps identify weaknesses in the live network and enhance resilience by simulating specific scenarios. This testing can include scenarios like network congestion, hardware failures, or security threats, such as DoS and DDoS attacks, which are challenging to replicate on a live network at scale.
Similarly, this approach can inform future development, such as planning for cell tower locations or predicting the impact of RF propagation in specific environments. These efforts contribute to the testing and advancement of 6G networks.
As evidenced by recent events at MWC Barcelona, 2026 is poised to be a pivotal year for 6G, with field trials commencing and standards being established. The use of RSG for hypothesis testing will bridge the gap between laboratory research and field implementation, providing reliable data for testing 5G-6G spectrum sharing algorithms and expediting the time to market for 6G technology.
A case study: energy-efficient 6G pre-deployment
Digital twins and RSGs are employed to model various scenarios, such as balancing energy efficiency with Quality of Experience (QoE) and implementing Massive MIMO technology.
For instance, a novel approach utilizing AI models and agents has been developed to optimize energy savings on 5G networks, as demonstrated by NVIDIA. Another noteworthy application of RSG is in self-aware networks, where wireless resources are reallocated for data transmission, leading to improved system throughput.
![\"\"](\"https://www.5gtechnologyworld.com/wp-content/uploads/2026/04/diagram-1024x576.jpg\")
Figure 3. Architecture and key features of the Agentic AI blueprint for intent-driven 5G energy savings by VIAVI and NVIDIA. (Image: VIAVI and NVIDIA)
A joint study highlighted the effectiveness of a methodology that uses AI-based predictions and network quality measurements from the digital twin to optimize base station beam control. This approach led to a 20% increase in uplink throughput by reducing radio control overhead, showcasing the potential of digital twins and AI technologies to enhance network performance.
![\"\"](\"https://www.5gtechnologyworld.com/wp-content/uploads/2026/04/Figure-2-1-1024x305.png\")
Figure 4. Simplified system setup for optimal beam selection evaluation based on the DOCOMO Self-awareness Network. (Image: VIAVI)
The successful application of digital twins and AI technologies in optimizing network operations underscores their potential to revolutionize the deployment of 6G infrastructure and expedite the research and development process to bring 6G technology to market.
Next Steps in 6G Development
The advancements in AI-native RANs and the integration of digital twins in network operations mark significant milestones in the evolution towards 6G technology. With ongoing field trials and standardization efforts, 6G is on the horizon, poised to redefine connectivity and communication in the digital landscape.
As we navigate the complexities of transitioning to 6G, the use of AI-driven technologies, such as digital twins and RSGs, will play a pivotal role in shaping the future of wireless networks. By leveraging these innovative tools, network operators can optimize performance, enhance security, and deliver unparalleled user experiences in the era of 6G.
