Applications of AI in Radio Access Networks

AI assists in making situation-dependent smart decisions, which can be applied to connectivity sites, thereby a huge benefit for radio access networks (RAN).

FREMONT, CA: 3GPP initiated a study on the applications of AI/ML to radio access networks called the enhancement of data collection. The study involved end-user devices creating reports for network equipment. This project is turning into a work item for release 18. With the use of self-organising network principles and moving from a reactive to a proactive stance, algorithms train base stations to deactivate when they do not receive location updates from terminal devices, which indicates that no one in the area needs the network.

Training base stations to power down when they are not required has implications for energy use management. Energy management is the immediate advantage of AI in wireless networks. In addition, energy savings and beam management are the most promising areas for AI/ML to add significant value soon. Some industry trials have already observed tremendous savings obtained from intelligent energy savings algorithms that power down a few base stations or antennas in the network during non-peak hours. The dynamic shaping of beans for leveraging coverage and capacity is also promising.

Beamforming is another AI-related initiative at 3GPP which brings AI/ML applications to the air interface. The enhancement of data collection work involves end-user devices making reports to intelligent base stations, whereas this new project includes two-way communication between base stations and end-user devices using AI to collaboratively solve nonlinear problems.

The research studies different use cases for AI/ML in the air interface, like beamforming, channel state information (CSI), and positioning. Moreover, companies have introduced dedicated hardware for this type of AI and ML. This will result in more AI resources being invested in CPUs and GPUs.

 Many companies are realising the significance of AI in semiconductor solutions for 5G and are referring to AI and ML as game changers for 5G evolution and the future of every wireless network standard. These technologies will assist with link adaptation, service-level agreements, traffic steering, and energy savings.

AI in radio access networks will offer service providers more granular and live control of network operations. ML will allow networks to configure resources to meet workload requirements effectively automatically. This real-time predictive analysis has shown a drastic improvement in spectral efficiency.

As network administrators consider RAN virtualisation, AI is an important element of the value proposition as network operators consider RAN virtualisation. AI can enable network slicing, or the dynamic allocation of network resources to a customer to ensure that they receive a consistent service level. Network slicing is a potential new revenue source for operators enabled by 5G and AI. AI is also applicable to open RAN, which means opening interfaces between RAN elements to help operators mix and match supplier equipment.

 There are progressing works in forums like the O-RAN alliance on the advantages, such as massive multi-user MIMO optimization, energy savings, traffic steering, etc. The evaluation of such optimisation algorithms based on the RIC (RAN intelligence controller) framework requires the use of large-scale network digital twins. More advanced AI-based methods are proposed for late 5G and 6G, which will directly impact the physical and MAC layers. Such advancements include distributed or federated learning techniques and AI-native modules.