How 5G networks can get smarter

Mobile network service adoption is reaching the point of saturation in many markets worldwide. As a result, network operators are increasingly under pressure to reduce operational expenditures, gain competitive advantage and grow revenue. In order to succeed in this high-stakes environment, mobile network operators (MNOs) need to be laser-focused on optimizing the speed and quality of 5G service delivery.

Recent advancements in machine learning (ML) and artificial intelligence (AI) technologies offer significant potential for improvements across a range of network functions. But exactly how can MNOs apply these types of intelligent technologies to substantially improve service delivery for greater profitability?

Ignite network intelligence

Evolution of the radio access network (RAN) is driving increased virtualization to create a transformational shift in network infrastructure. As networks become more open and virtualized, this enables a capable, resilient network with simplified operation and flexible resource availability in the cloud.

Taking this a step further, the power of AI enables the network to autonomously make appropriate decisions on events that are difficult to handle manually, allowing the network to effectively respond to diverse service demands and unexpected events in real-time. Yet, digital transformation of legacy networks doesn’t happen overnight — particularly with today’s complex, multi-vendor, multi-cloud networks.

The development of AI-powered models like neural networks, Large Language Models (LLM) and generative AI provides operations teams with the necessary tools to implement virtualized initiatives linked to positive business outcomes and revenue recognition. With the capabilities of sophisticated AI-powered network applications and connected clouds, operations teams can quickly solve complicated network problems in minutes, rather than days or weeks, to ensure that optimized service quality is consistently maintained. Intelligent network applications also help MNOs more easily identify trends to be monetized with individualized service combinations for direct and indirect customers.

Tame operational complexity

As 5G networks continue to evolve, the combined physical and multi-cloud components in multi-vendor 5G+ network architecture will become ever more challenging to manage. Intelligent applications with network-focused ML models can alleviate that complexity, helping these networks function at full capacity and capability, while providing expanded functionalities powered by generative AI, neural networks and LLM systems.

In order to make the most of this intelligence, however, MNOs need to provide AI/ML applications with as much historical network data as possible. This includes hardware IDs, severity and timestamps of alarms, as well as resolution tickets. Armed with this data, intelligent applications have the capability to build a simulated network for predictive modeling, adaptive control, and to train other applications, drastically simplifying operations.

Once the information from hundreds of thousands of nodes and end points has been ingested, these intelligent applications can then use neural networks to recognize new combinations of known behaviors and identify new behaviors. With this type of intelligent environment in place, operations teams are better able to deliver accurate root cause diagnoses and efficient resolutions, responding more rapidly to network issues and threats.

Moreover, new datasets provide valuable information for evidence-based analysis and responses, as well as improved business forecasts. This data allows operations teams to quantify the relationship between user experience and network quality of service (QoS) with quality of experience (QoE) models, encapsulating network measurements such as throughput, delay and loss. These advanced capabilities enable MNOs to change the competitive landscape with differentiated service offerings, based on more realistic QoS and QoE predictions.

Train for flexibility

With the advent of AI-powered network intelligence, MNOs are in a position to take full advantage of the resource flexibility introduced with today’s open and virtualized architecture. This intelligence enables a better understanding of the converged network’s entire data landscape, delivering actionable insights with specific countermeasures that can be tailored to individual network scenarios and conditions. For example, generative AI can be used to create customer solutions that investigate specific network conditions, with LLM technology and natural language interactions then used to generate responses to users.

As recommendations are implemented with intelligent applications, results are fed back into the network model for continuous improvement, leading to more accurate investigations and remediation. These intelligent applications will then become smarter as they learn from network data such as resolution tickets, log files and past failures, allowing a constant focus on enhancing network productivity and service delivery.

Likewise, configuration parameters optimized by AI-driven analysis enable automated maintenance and orchestration of the RAN for optimized resource management and flexible control. This empowers improved energy efficiency, increased performance and reduced costs, as well as differentiated service delivery through network slice management.

In fact, during the Fall 2023 O-RAN ALLIANCE Global PlugFest, an AI-powered network application demonstrated significant power-saving capabilities. The intelligent application leverages user equipment data and traffic estimates powered by AI and ML to switch network capacity on or off as needed while maintaining service continuity, demonstrating confirmed power savings of more than 20 percent compared to conventional methods of estimating communication traffic for individual base stations.

Furthermore, network intelligence allows MNOs to improve proactive security by leveraging data to monitor legitimate network use and stop cyberattacks when anomalies are detected, rather than after a breach is detected. Plus, automated intelligence helps decrease hands-on access to the network, reducing opportunities for rogue users and phishing threats.

Build value for today and tomorrow

Evolution to a smarter RAN built on ML models, generative AI, LLM systems and neural network modeling will unleash a self-healing infrastructure capable of continuous learning and responding in real-time to unexpected events and changing environments. This allows MNOs to reduce problem resolution times, eliminate repetitive tasks and make faster, more accurate decisions, setting the stage for optimized speed and quality of 5G service delivery for real competitive advantage.

Blake Hlavaty is Director of Global Network Software at Fujitsu Network Communications. He has deep expertise in the telecommunications industry, growing out of roles in network design, software engineering and consulting, and product management. Most recently, Blake has focused on network management software including open networking, SDN, cloud, AI/ML and Open RAN technologies. The ever-changing nature of software products motivates Blake and his team to realize the vision of these technologies.