As we approach the deployment of 6G networks, it is becoming increasingly clear that Artificial Intelligence (AI) will be a fundamental cornerstone in shaping their architecture. Unlike 5G, where AI is often applied as an optimization tool, 6G networks are being designed with AI deeply integrated into their core functionalities. This shift represents a revolutionary change in how networks are conceived, deployed, and managed.
AI-native architecture refers to a design approach where AI capabilities are deeply embedded into the system, and AI is seen as a natural part of its functionality. This deep integration allows for a level of proactive network management that was previously unachievable, enabling real-time, intelligent decision-making and automation. Through AI-driven predictive analytics, the network can foresee congestion, security threats, and performance issues before they happen. Machine learning algorithms enable the network to automatically adjust resources, reroute traffic, and apply necessary security measures, ensuring optimal performance and robust security.
At the EuCNC & 6G Summit 2024, researchers extensively discussed AI-native architectures for 6G. For example, in the paper presentation “AI-native architecture for 6G networks and services with model dependencies,“ Nassima Toumi proposed a solution for managing the lifecycle of AI/ML models, considering possible dependencies between different models. Specifically, when models are trained using Transfer Learning, the degradation of the pre-trained model can lead to the degradation of the fine-tuned model. Therefore, efficient orchestration of AI/ML models is necessary to collect monitoring data and manage different versions of these models.
The importance of DataOps and MLOps has been emphasized at the conference as a way to address the challenges of deploying AI, specifically in managing model deterioration with inter-model relationships. In the paper presentation “The Role of AI Enablers in Overcoming Impairments in 6G Networks,” the authors provided a comprehensive overview of the impairments for designing AI-native 6G architectures. These challenges include adhering to fair but strict AI regulations regarding the use of private or sensitive data and the need for sustainable and efficient model training and inference. Additionally, the work discusses AI enablers, such as architectural enhancements and AIaaS (Artificial Intelligence as a Service).
As mentioned, the journey toward AI-native 6G networks is not without its challenges. Ensuring data privacy and security for the data used in training AI models is of prime importance. Moreover, the development and deployment of AI systems must adhere to principles of transparency, fairness, and freedom from bias to maintain trust. Finally, interoperability with existing network infrastructures and technologies is crucial to seamless integration.
The presentations at the EuCNC & 6G Summit also highlighted energetic sustainability as a key optimization goal in AI-native systems. For instance, the BeGREEN project focuses its research on designing an Intelligent Plane for AI/ML control and management to decrease the energy consumption of core components in 6G networks. Another significant contribution can be found in the paper presentation “Energy Efficiency in AI for 5G and Beyond: A DeepRx Case Study,“ which introduces various techniques to help evaluate the energy consumption of AI/ML models in networks and minimize it throughout different stages of the AI/ML lifecycle.
As an active actor in the telecommunication innovation, Martel did not just participate in the event as a viewer but also as a consortium member of the 6G-NTN and HORSE projects, which are contributing to AI-native research by leveraging AI/ML for physical and digital resource orchestration (6G-NTN) and integrating AI at all levels of a platform for autonomous threat detection and healing (HORSE). The two projects had multiple opportunities to disseminate their work at several sessions throughout the conference. Specifically for this event, Martel contributed to the poster “Enablers for E2E integration in 6G-NTN” showcasing AI/ML solutions for the dynamic orchestration of Virtual Network Functions.
In conclusion, the deep integration of AI into 6G networks is going to transform the telecommunications landscape. AI-native architectures will enable smarter, more efficient, and more secure networks, unlocking new possibilities for innovation across various industries. As we prepare for this next generation of connectivity, embracing AI-native design principles will be key to realizing the full potential of 6G, however, we must be aware of a good balance between the accuracy of models and energy efficiency, for a sustainable way of doing innovation.
- Iovene et al., ”Defining AI native: A key enabler for advanced intelligent telecom networks”, Ericsson Whitepaper, 2022, https://www.ericsson.com/en/reports-and-papers/white-papers/ai-native.
- https://www.eucnc.eu/
- Toumi, N., & Dimitrovski, T. (2024). AI-native architecture for 6G networks and services with model dependencies. Zenodo. https://doi.org/10.5281/ZENODO.10522709
