Mobile connectivity pervades every aspect of today’s society. Changing lifestyles, remote work, smart automation and the proliferation of cloud-native applications are all driving fundamental changes in mobile network usage patterns. At the same time, technology advances are continually enabling innovative capabilities, empowering exciting new use cases.
Yet, even as technology enables greater innovation, users demand more — functionality, speed, performance, simplicity and access — everywhere and all the time. In an effort to keep pace with this virtuous cycle, mobile network requirements continue to diversify and evolve.
As a result, the entire network ecosystem is becoming increasingly flexible. Ultimately, the goal is to achieve an intelligent network capable of responding in real-time to unexpected events and changing environments. Underpinning this paradigm shift is the growth of artificial intelligence (AI) in the radio access network (RAN) architecture.
Why AI in the RAN?
As network architecture complexity escalates, the ability to make appropriate decisions on a multitude of simultaneous events becomes impossible to handle manually. This is particularly critical in the RAN where mobile network operators (MNOs) face significant operational challenges maintaining service quality, improving efficiencies, managing multi-vendor integration and decreasing power consumption.
Sudden, short-term changes in RAN connectivity usage patterns and data traffic flow cause rapid fluctuations in quality of service (QoS) from moment to moment. Meeting user expectations and partner service level agreements (SLAs) requires the optimal combination of various components, traffic volumes and service types, all of which interact in complex relationships to support diversified service delivery. Achieving this task quickly to ensure prompt service restoration involves intelligent, real-time response, requiring greater reliance on AI.
Moreover, as traffic demand intensifies and MNOs further densify their 5G networks, an enormous amount of RAN equipment will be needed. Not only does this increase complexity and capital expenditures (CapEx), but power consumption also is rapidly rising in the RAN, which already accounts for 70 to 80% of overall network energy usage. Addressing this sustainability challenge goes beyond just reducing power consumption of base station equipment. The question lies in how to significantly transform overall RAN orchestration and management to holistically optimize operations.
Why AI now?
The ongoing evolution and disaggregation of the RAN means that networks are becoming more open, virtualized and small-cell, enabling more flexible resource availability in the cloud. MNOs are now in a position to realize this flexibility with key elements of Open RAN architecture, including the RAN Intelligent Controller (RIC). Harnessing the power of AI and machine learning (ML), the RIC enables the network to effectively leverage this intelligence for a number of advanced use cases and technologies.
With the power of AI, the RIC platform can optimize RAN resource management and automate operations, either in the non-real time RIC (Non-RT RIC) or the near-real time RIC (Near-RT RIC). Non-RT RIC processing functionality is deployed in the service management and orchestration (SMO) framework, providing intelligent orchestration for flexible control of the entire RAN.
Using configuration parameters optimized by AI-driven analysis, the SMO generates policies related to RAN control, allowing it to perform automated maintenance and orchestration of the RAN. In this way, the SMO enables optimal resource management, improving energy efficiency, increasing performance and reducing costs, as well as speeding differentiated service delivery through network slice management.
How to get proactive
With conventional RAN control, MNOs typically measure network quality, adjust parameters and respond to failures based on traffic volume and packet loss statistics during periodic inspections. Traditionally, this approach results in poor quality of service (QoS), inefficient resource utilization and slow recovery from failures.
On the other hand, AI enables MNOs to automatically detect QoS degradation to forecast the QoS score in real-time and adjust network resources accordingly. This proactive automatic RAN optimization allows AI to release over-used network resources, reducing power consumption and improving efficiencies while preventing QoS degradation.
However, in order to ensure the best possible outcomes, the practical use of AI technology in the RAN requires an understanding of how, in addition to why. Proactive automatic operation requires accurate quality predictions and appropriate operating control of network element parameters to meet quality expectations and key performance requirements. Going forward, network managers will need new solutions to simplify the application of AI, helping them better understand and trust AI predictions and learn to confidently use these technologies.
Embrace the new normal
As the pace of network evolution marches on, pressure intensifies to reduce costs, increase sustainability and simplify complexity. With the continuous improvement of 5G, AI is certain to play a more prominent role in network and service management and orchestration. In fact, adoption of 5G-Advanced specifications, expected in mid-2024, will only accelerate reliance on intelligent orchestration.
Looking ahead, increasingly sophisticated networks will cover a variety of functions and services while also considering the environment. AI will be an essential technology to dynamically control these networks, becoming the new normal for RAN architectures in 5G, 6G and beyond.
Table of Contents
