In 2019, the 5G network was put into commercial use in Korea, the United States, Switzerland, the United Kingdom and Spain. In June of the same year, the Ministry of Industry and Information Technology of China issued four 5G commercial licenses. For 5G commercial networks, operation and management of networks have become a focus in the industry. Not only key features such as cloud-based networks, SBA architecture, and CU separation bring infinite possibilities to 5G networks, but also unprecedented challenges to 5G network operation and maintenance. Traditional manual passive O&M is not sustainable, thus automatic and intelligent O&M mode becomes an inevitable choice. The 5G cloud-based infrastructure, full software-based application systems, and massive data are the best ways for cultivating AI applications. The world’s leading operators and equipment manufacturers are working closely with each other to implement automatic and intelligent network O&M.

Facing to 5G complicated and diversified services, different networks and management layers have different requirements for intelligence. ZTE proposes the three-layer network intelligence architecture to gradually implement 5G intelligent network in a hierarchical mode. For the intelligent function with high real-time requirements, a real-time intelligent engine is embedded in the NE layer, and the NE implements intelligent self-loop.
Figure 1. ZTE Proposes Three-Layer Closed-Loop Network Intelligence Architecture
For the intelligent O&M function with low real-time requirements and medium amount of calculations, a small intelligent ring in a single network domain is implemented by embedding the light intelligent engine on the management and control layer. For the lowest real-time requirements and large amount of calculations, including the platform supporting AI training, a large intelligent ring across the entire network and network domains is implemented by the operation layer.

5G network intelligent O&M covers two keys: the AI platform construction and the implementation of value scenarios.
The problems of data collection, storage, analysis, and processing should be solved by the AI platform construction, which provides basic training data to design and a series of tool chains for AI application design and development and AI training. Meantime, the AI platform is used to develop practical applications in 5G network O&M to solve the problem of the highest ranking and highest value in O&M.

ZTE has carried out joint innovation and cooperation with some leading operators to explore the intelligent O&M of 5G networks. As for the construction of the AI platform, ZTE provides a visual AI application programming and training platform “uSmartInsight”, which has three core capabilities:

  • Data lake: The uSmartInsight platform integrates the big data capability. It can set up data standard according to the service needs, so as to solve the data structure and standardization problems of different network management, manufacturers, regions and applications, and construct the communication network data lake. By providing data service capability, the uSmartInsight platform can decouple data from upper-layer applications.
  • The AI training platform: provides a complete tool chain based on the DevOps, and supports visual AI application modeling, training and verification. The platform supports both traditional machine learning related algorithms and deep learning algorithms, helping partners to empower traditional O&M engineers to be AI application development engineers.
  • The hierarchical AI reasoning engine: can be applied in the NE layer, management and control layer and operation layer with the reasoning engine in the running state to provide the reasoning service for various intelligent services.

As for the implementation of value scenarios, ZTE has launched the UniSeer service solution including six application scenarios combined with its rich experience in communication network practice. The solution provides operators with intelligent and predictable network planning, deployment, O&M and optimization, thus reducing O&M costs and improving O&M efficiency. When used for fault locating and recovery, the intelligent O&M proposes a series of intelligent applications based on uSmartInsight.

  • Turn passive responded maintenance into active preventive O&M

By analyzing network alarms, performance data and logs based on the actual situation of network faults, it realizes semi-automatic marking. The neural network and other algorithms are used to obtain the prediction model of typical faults in the wireless network. By providing the fault prediction function, the solution can predict potential network faults in advance, and convert passive troubleshooting to active preventive maintenance, greatly improving network reliability.

  • Focus on the accurate location of network faults

The UniSeer solution provides two key applications: end-to-end fault delimiting and locating, and intelligent RCA, which can rapidly delimit and locate faults from two dimensions: services and networks. End-to-end fault location collects data from wireless, core network, and bearer, and obtains end-to-end sample data based on active service prediction. Based on intelligent determination, the problems of terminals, wireless networks, core networks, bearers, and services across domains can be rapidly delimited. Based on the problems in the wireless field, it can further locate the specific sites and root causes of the problems. To solve the problems in the core network domain, the intelligent RCA technology can be used to locate three-layer virtualized decoupling nodes.

  • Focuses on the quick recovery of network faults

The UniSeer provides a set of closed-loop mechanism based on policy engine. For example, the automatic diagnosis script can be configured and different self-healing policies can be executed in accordance with the diagnosis result, thus faults can be rapidly fixed. Based on O&M practices, it can be customized and provides a wealth of troubleshooting policies, so that O&M experience can be accumulated continuously.

The UniSeer service solution, based on ZTE’s uSmartInsight AI platform, has worked closely with operators on 5G network intelligence, including joint innovation based on uSmartInsight platform, and carrying out 5G network O&M efficiency improvement based on the UniSeer has made positive progress. In a province of China Mobile, through the application of UniSeer intelligent RCA and the work order reduction function, the work order reduction rate exceeds 30%, greatly saving the O&M cost.

5G network Intelligent O&M is a process of long-term evolution. The AI platform and AI application are also a step-by-step improvement process. The AI platform must have the basic AI application design tool chain, data processing capability, and AI training capability. On the other hand, the access of various types of network data and the accumulation of training sets are a long-term process. Data accumulation, in turn, raises new requirements for platform capabilities, thus achieving spiral upgrades. The implementation of valuable scenarios will be accompanied by the commercial use of 5G networks and the continuous enrichment and improvement of 5G vertical application scenarios. Meanwhile, the O&M team will gradually become an intelligent O&M team with AI application development capabilities. ZTE is willing to work with partners to jointly practice intelligent operation and maintenance of 5G networks and realize autonomous closed-loop 5G networks.