Focusing on Four Key Areas of Intelligence to Build a Full-Stack AI+ Core Network

The world is entering a new era of digitalization, intelligence, and networking, driving major transformations in networks and AI. In the network field, 5G has entered its second phase, while 6G development has begun. In the AI field, generative AI has made breakthrough progress. AI has become the engine of new-quality productivity, shifting industries from “+AI” to “AI+”. As the brain of the mobile network, the core network urgently needs a complete “AI+” transformation to enhance its autonomous and innovation capabilities, enable differentiated services, and create new growth points.

To address the development trends of AI technology and core network challenges, systematic reconstruction is required in four major areas: service, connectivity, operation and maintenance (O&M), and cloud infrastructure. Building on its existing 5G Common Core solution, ZTE has launched the AI+ core network solution: the AI-reshaped core (AIR Core). In the short term (5G-A), the solution focuses on core network experience and efficiency, while in the long term (6G), it aims at evolving the core network’s native intelligent architecture. It creates a full-stack intelligent “new network brain” solution built on an open ecosystem, helping operators accelerate 5G monetization and the deployment of AI applications.

The AI+ core network solution consists of four areas of intelligence, built on a “three-layer, one-domain” architecture (Fig. 1):

• AI+ cloud infrastructure layer: Leverages intelligent computing hardware resources to provide AI capabilities for core network applications.

• AI+ connectivity layer: Provides the capabilities to stimulate traffic and enhance the experience, enabling the shift from “traffic-centric” to “experience-centric” operation.

• AI+ service application layer: Delivers intelligent capabilities for messaging and new calling services.

• AI+ O&M domain: Redefines the O&M paradigm, supporting the evolution of the autonomous network towards unmanned operation.

AI+ Service: Multimodal Interactive New Services

OTT is eroding tradtional voice and messaging traffic. The AI+ service leverages AI technology to enable multimodal, interactive, and immersive service experiences for real-time voice and messaging, rejuvenating traditional voice and messaging communication services.

AI+ 5G New Calling enables intent-driven interaction between humans and machines by integrating AI capabilities. It intelligently orchestrates multiple service capabilities, allowing for more complex services, such as "switch to a cartoon avatar" and "answer the call on my behalf" within a single interaction. This upgrades a simple voice call to multimodal communication, delivering a more intelligent and diversified calling experience.

AI+ messaging introduces an intelligent plane and integrating it with the messaging plane to create a unified AI+ messaging portal. It provides three categories of services: AI+ messaging services, AI+ information services, and AI+ application services—catering to individual consumers (ToC), industries and enterprises (ToB), households (ToH), and other sectors (ToO), creating a new-generation business model. No modifications or upgrades are required for mobile terminals.

AI+ Connectivity: Experience Monetization & Differentiated Network Operation

5G networks must shift from traffic-centric to experience-centric operation. Traditional user service assurance is achieved through QoS subscriptions, which cannot perceive real-time service quality, resulting in a poor user experience and making experience-based operation difficult to implement.

AI+ connectivity builds an intelligent plane on the 5GC side, with the network data analytics function (NWDAF) acting as the brain, the policy control function (PCF) as the policy anchor, and the AI engine embedded in network elements (NEs) as the executor. Collaborating with the wireless network, it constructs end-to-end connection intelligence that enables all-round profiling of users, services, and networks. This provides operators with a means to explore the value of connectivity, achieving real-time user experience perception, experience monetization, and intrinsic security.

In terms of experience monetization, intelligent service identification and service experience measurement are crucial. The introduction of AI technology has greatly improved service identification and experience measurement, with the update cycle for the deep packet inspection (DPI) library also shortened from weekly to hourly, greatly enhancing dynamic identification, experience measurement, and response capabilities for online services. Operators can utilize intelligent service identification and service experience measurement to accurately analyze user experience, adjust data channel QoS accordingly, and implement differentiated services, greatly enhancing user experience, realizing experience monetization, and enabling refined operations.

Regarding intrinsic security, the solution leverages the built-in intelligent engine, network operation data, and AI algorithms to predict the states of terminals and NEs, enabling efficient node selection and service orchestration, end-to-end load management, energy-saving management, and signaling storm prevention. For energy savings, improved prediction accuracy of idle states in NEs enables dynamic and precise energy management. For example, the power-saving strategy is initiated before an idle period and deactivated before a busy period. For signaling storm prevention, the solution intelligently identifies abnormal terminals, preventing network attacks and signaling storms. Additionally, based on the digital twin network, it simulates signaling storm scenarios and takes preventive measures proactively.

AI+ O&M: A New Paradigm in Advanced Autonomous O&M

The core network features a large variety and quantity of NEs, as well as decentralized deployments. The adoption of cloudification technology, combined with the stringent requirements for high network stability, has intensified the pressure and challenges on core network O&M. AI technology presents unprecedented opportunities to transform core network O&M. ZTE's AI+ O&M combines large AI models with digital twins, promoting the evolution of network O&M towards a higher level of autonomy.

• A large language model (LLM) is used to build a core network signaling model. By analyzing network signaling in real time, the model can detect anomalies, identify potential problems, and automatically adjust signaling policies to optimize network performance and user experience.

• By integrating large-model technology, AI for IT operations (AIOps) implements end-to-end management—from fault detection to repair—through multi-agent coordination during the O&M process. No longer limited to specific application scenarios, AIOps demonstrates powerful generalization capability, allowing it to  automatically adapt to different network environments and diagnose and resolve network faults through agent coordination.

• By introducing an intent-based network and integrating it with digital twin technology for intent verification and network interaction, precise execution of operational intents—such as SLA assurance and energy efficiency—can be achieved, paving the way for an advanced autonomous network.

AI+ Cloud: High-Efficiency, High-Stability New Infrastructure

The cloud infrastructure layer is the foundation for running the core network, consisting of cloud operating systems, servers, storage devices, network devices, and other software and hardware. The AI+ cloud encompasses two aspects of intelligence: on the one hand, it provides an efficient and diverse computing power resource pool to empower the development of the core network's AI capabilities; on the other hand, it introduces AI technology into infrastructure management.

• Provide an efficient and diverse computing power resource pool to empower the construction of the core network's scenario-based AI capabilities. With evolution of cloud infrastructure from the traditional CPU-centric general computing to CPU+GPU-centric general-intelligent collaborative computing, it is necessary to support technologies such as computing power pooling, orchestration and scheduling, high-performance parallel storage access, and high-channel lossless networks to ensure the efficiency and stability of resource supply. At the product level, to meet the demands of fine-tuning and real-time inference scenarios, ZTE has launched a training and inference integrated machine that is ready to use out of the box (see Fig. 2).

• Introducing AI technology in infrastructure management enables multi-dimensional intelligent resource scheduling and orchestration from a business perspective, as well as intelligent O&M featuring automatic pre-event prediction and early warnings, intelligent in-event discovery, delineation and restoration, and post-event review and optimization. This leads to high infrastructure stability, smarter O&M, and efficient resource utilization.

ZTE: Empowering Customers to Continuously Create New Value

Through cooperation with leading operators, ZTE's full-stack AI+ core network solution addresses both the rapid monetization of 5G-A networks and the implementation of AI applications, while also laying the foundation for future 6G networks.

In service intelligence, ZTE released "Smart Safeguard", the industry's first SMS anti-fraud system based on a large model, and successfully deployed the world's first modular AI+ 5G new calling network.

In connectivity intelligence, ZTE, in partnership with operators, created the industry's first commercial project for hierarchical VIP user guarantees.

In O&M intelligence, ZTE pioneered a quantitative evaluation system for resource pool switching based on digital twins. By considering the relationship between NEs and resource pools, it simulates and quantitatively analyzes the resource pool switching process.

ZTE will closely follow AI and network advancements, actively innovate, and fully support operators in building AI+ core networks, to develop a future-oriented "new network brain".