On January 26, 2023 (local time), Finland Nokia released a commentary on AI (artificial intelligence) and ML (machine learning), which are being used in next-generation mobile networks. Below is an overview.
Related blog:AI/ML unleashes the full potential of 5G-Advanced
AI-native mobile system realized with 6G after 5G-Advanced
After introducing AI/ML technology for network automation in 5G, 3GPP is promoting the use of AI/ML for devices and wireless RAN in Release 18, the first standardization specification for 5G Advanced. Release 19 will further promote the use of AI/ML in the radio interface, RAN, and system architecture, and Release 20 will make AI and ML integral to the system. Advanced technologies such as distributed learning will work with AI deeply embedded in networks and devices to improve performance and usability.
Prospects and Challenges of AI/ML
The effects of introducing AI/ML technology to networks are as follows.
- Device and network enhancements for more efficient radio transmission and radio resource management: 3GPP has already seen performance improvements of up to 30% in cell-edge throughput and other items.
- AI/ML based baseband solution: Nokia has already reported up to 3dB receiver performance improvement and 30% throughput improvement
- Energy saving: Accelerate reduction of power consumption of network nodes and devices and improvement of energy efficiency
- Improved end-user experience: Reduce latency and network link failures to improve user experience when using AR (Augmented Reality), XR, etc.
- End-to-end network management automation: Network management automation including faster prediction and response to network conditions and faults
In addition, MLOps (coordination of machine learning, development, and operations), which is an extension of DevOps, in which developers and operations staff collaborate, and operational efficiency improvements by integrating ML model training and development/deployment into highly automated processes. is also possible.
On the other hand, there are also challenges to overcome in order to maximize the potential of AI/ML. For standardization, it is necessary to define an efficient and flexible operational framework for utilizing ML. Training an ML model must ensure high-quality input while ensuring manageable data collection. We also need to define appropriate performance requirements that take into account the adaptability of ML-powered solutions.
AI/ML is already becoming a fundamental building block of mobile communication networks, but its potential has yet to be fully explored. Proceeding with the above research will likely create a foundation for AI/ML utilization in mobile communications from 5G-Advanced to 6G and beyond.