ZTE drives core network intelligence towards future networks evolution


With the large-scale commercial use of 5G, the world has started the research and exploration of the next stage of 5G evolution, and 5G-Advanced (5G-A) will start the first year of commercial use in 2024. At the same time, the telecom industry is also exploring artificial intelligence (AI) technology, and its applications in key scenarios of 5G and future networks are gradually moving towards commercial use.

In the following interview, Weibin Wang, chief scientist of product planning of ZTE, shares his views on the necessity and urgency for telcos to introduce intelligence to the existing and future networks, especially the core network. 

Since last year, the rise of ChatGPT and Large Language Model (LLM) has become a hot topic of widespread discussion in the industry, what is the impact on telecommunication networks, especially core network?

Weibin Wang: The launch of ChatGPT is a milestone in the great achievements of Large Language Model (LLM), and we will accelerate into the era of AI. Telecommunication networks should embrace the new opportunities and challenges brought by AI.

On the one hand, AI will deeply empower telecom networks, improve O&M efficiency and service innovation, and specifically focus on core networks, including:

  • Accelerate the evolution of the network to the high-level autonomous stage, and implement high-level intelligence such as traffic prediction, intelligent decision-making, root cause analysis, self-healing, and intelligent customer service.

  • Promote service innovation and experience improvement, empower 5G New Calling, New Messaging, and ubiquitous real-time communications, and realize the transformation and upgrading of basic communication services.

On the other hand, the large-scale application of AI also poses new challenges to telecom networks, including:

  • Diversified network transmission requirements for AI applications. While AI drives network traffic consumption, it also puts forward new requirements for network transmission. As the brain of the network, the core network should sense the transmission requirements in real time and provide differentiated network policy guarantees.

  • Ubiquitous AI computing power requirements. Telecom networks need to evolve to computing power networks and further provide endogenous AI computing power in the network. The core network will play an important role in providing computing power at the edge of the network.

In the future, the deep integration of telecom networks and AI will become a new digital infrastructure that empowers thousands of industries. ZTE's self-developed intelligent computing solution can provide full-stack products from AI computing power, AI platform, large models and applications. In the field of telecom networks, it has launched a number of products such as network large model + intelligent core network O&M, anti-fraud large model + SMS anti-fraud, CV and LLM + New Calling/intelligent customer service.

The telecom industry believes that "native intelligence" is one of the characteristic of 6G evolution of future networks. In terms of the core network, what are the main aspects of "native intelligence"?

Weibin Wang: The deployment characteristics of 6G core network intelligence are mainly distributed, native, and ubiquitous. The distributed and native methods mainly solve the nearby real-time analysis and processing of data, reduce the additional network resource overhead caused by data centralization, reduce the delay, and improve the efficiency of data processing. Ubiquity is reflected in the use of AI to optimize the management of each node and process of the core network, enhance the security and reliability of the network, and achieve self-organization and autonomy of the network itself, at the same time, it is reflected in the key nodes (such as UPF) of the core network to process user business data, and the frame granularity intelligently guarantees the service experience, service continuity and ultimate determinism of connectivity, etc.

The capabilities of native intelligence mainly include three aspects:

  • Intelligent perception and analysis: Network anomaly detection (early perception), traffic analysis (intelligent DPI), user behavior analysis (mobile terminal trajectory prediction), security threat analysis (IoT terminal security threat perception), etc.

  • Intelligent decision-making and control: Exception handling (autonomous decision-making), network optimization (network autonomous optimization), etc.

  • Intelligent services and applications: Personalized service experience assurance (user service assurance), etc.

At present, most of the operators have regarded intelligence as an important requirement of the network, so what aspects should be considered for the core network to introduce intelligence?

Weibin Wang: With the continuous evolution of cloud-native and full convergence of core networks, the core network is becoming more and more complex, and the introduction of intelligence is particularly important, which can be reflected in three aspects: service intelligence, network intelligence, and O&M intelligence.

  • Service intelligence:

Intelligence can be applied to services such as New Calling and New Messaging. For example, in New Calling, AI can be used to complete functions such as background replacement, avatar replacement, gestures & voice expressions, and translation, so as to ensure the ultimate user experience of calling services. In New Messaging, ZTE and Chinese operators have completed the practice of anti-fraud SMS based on large model, and the interception rate has been greatly improved compared with the traditional model.

  • Network intelligence:

Based on the NWDAF-centric intelligent system architecture, it collaborates with network elements such as PCF, SMF, and UPF to complete closed-loop control of experience perception and assurance of key customers and key services, forming a hierarchical network quality assurance system, supporting operators' differentiated business models and assisting operators in creating more business value. In addition, to solve the pain points such as the emergence of new services in an endless stream, the difficulty of perception due to frequent changes in key service features, the intelligent DPI service identification capability on the UPF NE can be implemented and quickly sense service changes.

  • O&M intelligence:

The cloud-based and fully converged core network architecture leads to increasing O&M difficulties and costs. Too many network elements and complex cloud infrastructure hardware and software lead to potential catastrophic impacts in the event of a network failure. In the new O&M model, new technologies such as large model, digital twins, and intent-driven technologies are integrated into all stages of the network O&M lifecycle, such as service provisioning, monitoring and troubleshooting, and fault handling, to meet operators' requirements for high O&M efficiency and network stability and reliability. ZTE's core network is based on the “ZTE Nebulas large model”, and cooperates with professional intelligent O&M small models such as FM, Metrics, and Log to form an end-to-end O&M model system, enabling the continuous evolution of core network and cloud network intelligent O&M to L4+ high-level autonomous networks.

In the future, ZTE will launch more intelligent applications, helping the operators build future-oriented native intelligence core network and accelerate the success of 5G commercialization.