Advancing Towards High-Level Autonomous Networks to Realize Digital-Intelligent Transformation

Updated:2024/5/13 11:47

during the21st Huawei Analyst Summit (HAS), Wang Chen, from China Mobile Zhejiang branch(Zhejiang Mobile for short) delivered a keynote speech entitled "Striding Towards a High-Level Autonomous Network to Realize Digital-Intelligent Transformation " at the "Leading Innovation in Telecom Foundation Model and Advancing the Value of Autonomous Networks" session. In his speech, he shared Zhejiang Mobile's achievements and practice in Autonomous Networks (AN). He emphasized how AI+ provides a powerful approach for transforming Operations and Maintenance (O&M) towards AN L4, and intelligent agents in conjunction with large models will boost the advancement.

Zhejiang Mobile's Wang Chen delivering a keynote speech

As a frontrunner in 5G network development, Zhejiang Mobile actively explores the adoption of AI technologies to achieve optimal network quality, minimize O&M costs, accelerate service provisioning, streamline frontline O&M processes, and enhance resource efficiency. It introduced a novel architecture, with one unified support system, one network twin set, one talent development framework, and process reconstruction across six domains, to elevate the level of native intelligence.

"AI+ is key to advance towards core network AN L4." "Large models and the agents must work together", said Wang Chen. He mentioned that large models should be responsible for knowledge generalization and understanding as a knowledge system, and the agents concentrate on collaborative scheduling as the smart brain. By effectively utilizing large models and enhancing perception and action systems, agents can emulate human-like capabilities in knowledge-intensive and labor-intensive O&M scenarios, making core network O&M more manageable with the power of AI+.

Underpinned by seven key technologies including process collaboration, large- and small- model collaboration, prompts, Chains engine, external knowledge base, and multi-round Q&A, the core network agents developed by Zhejiang Mobile and Huawei reshape traditional O&M approaches, exemplify a practical application of AI-driven core network O&M, and have achieved three significant breakthroughs.

One breakthrough involves the development of signaling and configuration models based on Large Language Models (LLMs) to comprehend signaling and execute configurations effectively.

Large models possess the capability to generalize knowledge and serve as a natural knowledge system. By training these models with signaling and configuration instructions, optimizing their thought processes through CoT corpus refinement, and enhancing their reasoning abilities, we enable the large models to comprehend signaling and perform configurations effectively.

The second breakthrough lies in agents. The large model-based compliant analysis and alarm diagnosis agents are built to instruct operations so as to implement scenario-specific solutions.

The sensing system requires the power of agents. The capabilities of an agent are expanded in the terms of planning, actions, tools, and memory, which empower the large models.Huawei's exclusive network O&M experience and expertise as well as the core network tool library make it possible for the large model-based core network agent to level up its execution capability. As the smart brain, the AI agent collaborates with multiple agents to streamline the whole process of a complex task.

The third is the breakthrough in production. Intelligence and O&M reconstruction concerning complaint handling and troubleshooting (by moving some processes forward) empower a lighter production flow.

Traditional complaint handling and troubleshooting operated by both O&M experts and tools are upgraded to Q&A-based interaction with digital employees.As for complaint handling, complaints are now analyzed in the monitoring department instead of the specific professional department, reducing the time for handling one ticket from 13 hours to 5.3 hours. In the light of the monitoring and troubleshooting process, trouble tickets are now auto-filled, surprisingly lowering the handling duration from 13 hours to just half an hour. It is equivalent to employing 30+ experienced digital employees, effectively boosting up O&M quality and efficiency.

Agents are the key with data as the base and knowledge as the core. Only by pooling automation and intelligence capabilities can we accelerate intelligent transformation and ease the core networks maintenance. In the future, Zhejiang Mobile will continue to enhance digital-intelligent capabilities, explore more scenarios, improve promotion strategies, strengthen operation capabilities, and maintain our leading advantages.

For press release services, please email us at english@c114.com.cn.


    Copyright© 2014 C114 All rights reserved.