AI to Boost 5G Mobile Network Operations for Future

Release Date:2019-01-24 By Tang Liang

 

With the cloud and NFV transformation of telecom networks, integration of 5G and IoT, and development of diverse industrial applications, the operations of telecom networks will face unprecedented challenges in the 5G era. These challenges involve complex networking, diverse services and personalized experience.

  • Complex networking. Coexistence of 2G, 3G, 4G and 5G networks brings difficulties in network synergy and interoperability, difficulties in fault demarcation and location under the hierarchical decoupling architecture, and also challenges of unified resource scheduling and operations due to the dynamic change of cloud and virtual networks.
  • Diverse services. The single man-to-man communication mode has gradually evolved into a full-scenario communication mode that involves man-to-man, man-to-machine, and machine-to-machine communications. The business scenarios will be more complicated and thus bring about differentiated SLA requirements such as high bandwidth, massive connections, ultra-high reliability and low latency, and the associated complex network management.
  • Personalized experience. Relying on 5G network capabilities and abundant business modes, 5G service experience will also tend to be diverse and personalized, such as immersive experience, real-time interaction, and accurate perception of emotions and intentions. The network's support for the experience will subvert the traditional model and usher in new challenges.

Therefore, the challenges of network operations that come with the 5G era will be significant. The advanced automated operations are gradually forming a gap with the traditional operations based on expert experience. Automated and intelligent network operations will be the just need in the 5G era. Artificial intelligence (AI) technology has intrinsic advantages in solving high-volume data analysis, cross-domain feature mining, and dynamic strategy generation, and will create new modes and capabilities for 5G network operations.

In the future, based on the cloud infrastructure, the network that combines 5G, AI and IoT will gradually become the intelligent center of digital society and promote intelligent interconnection of all things.

As the world's leading telecom solution provider and 5G leader, ZTE has actively combined the AI technology with 5G to carry out automation and intelligence in related fields covering 5G wireless, cloud, slicing, bearer and operations services. Moreover, ZTE has also actively participated in the development of relevant standards and contributions to open source technologies.

A 5G network serves as the fundamental infrastructure that can support digital development of various industries and provide differentiated services for all industry scenarios. To support a variety of industry applications and business scenarios in a flexible and on-demand manner, the 5G network will be built with cloud service-based architecture (SBA) to meet the future long-term development needs. 5G RAN implements CU/DU separation. The CU can support cloud or dedicated hardware deployment, flexibly adapting to various scenarios. The service-based architecture allows for a converged core of 2G, 3G, 4G and 5G networks, meeting the needs of smooth evolution, collaborative development and long-term coexistence.

5G cloud and service-based architecture is the foundation to support various industry applications and business scenarios. Enabling efficient, flexible, low-cost, easy operations with openness and innovation will be the core competitiveness of operators in the 5G era. This is also a key trend of 5G network intelligence and its main requirements involve:

  • Flexible radio and cloud resource management: The 5G network needs to support on-demand allocation of radio air interface resources including spectrum, frame structure, physical layer, and high-level processing flow. It needs to implement software and hardware decoupling, dynamic allocation of processing resources, and agile creation and adjustment of network capabilities. Dynamic allocation of cloud and bearer network resources, intelligent management of global strategies, and automated management of end-to-end slices are also required.
  • Air interface coordination and site collaboration: It is necessary to optimize the interference and site collaboration in dense 5G networks. There is also a pressing need for designing a more efficient and more intelligent mobility management mechanism in ultra-dense 5G networks.
  • Flexible function deployment and edge computing: AR/VR, industrial internet, and internet of vehicle (IoV) have placed higher requirements on communication latency, reliability, and security. The 5G network moves some functions from the core layer to the edge to form edge computing capabilities. By shortening link distance and improving intelligence of the edge network, operators can save backhaul bandwidth, reduce network latency and intelligently ensure user experience.
  • Enhanced network intelligent management: 5G network requires multi-system coexistence and coordination, cloud-based hierarchical decoupling and fault location, SBA-based holographic perception, and on-demand scheduling of bearer resources. This greatly increases complexity and difficulty in network management and optimization. It is therefore necessary to introduce AI to improve automation and intelligence of network management, reduce artificial interference, save costs, and thus enhance network QoS and user experience.

Faced with the challenges of 5G development and the needs to introduce intelligence, the combination of AI and telecom networks will be ubiquitous in 5G networks. An intelligent 5G network can be achieved by introducing algorithm models and intelligent engines into different levels of the network, as shown in Fig. 1.


Based on the cloud and service-based architecture, 5G network has distinct differences at different network levels. The upper layer is more centralized and has higher requirements for cross-domain analysis and scheduling capabilities such as E2E slice orchestration and management and global cloud resource coordination that rely on the centralized smart engine (SE) for centralized global strategy training and reasoning. The lower layer closer to the end side focuses on intelligence enhancement of professional subnets or single network elements. Access network, bearer network, and core network introduce Lite SE (LSE) to enhance intelligence of subnets or the sub-slice domain such as management strategies and smart operations. Edge devices such as MEC and 5G gNB introduce real-time SE (RSE) to achieve real-time or quasi-real-time intelligence at the edge.

The AI algorithm model and smart engines at various levels can be deployed based on the hardware computing environments in 5G networks. The combination of engines, model components, and application algorithms with different network functional entities enables 5G network intelligence.

The basic hardware environments where AI capabilities are deployed in a 5G network can be centralized GPU clusters, general-purpose servers or blade servers, or 5G base stations. The intelligent capability layer (AI layer) contains the engine layer, model layer and application layer (Fig. 2).


  • Engine layer: It supports smart engines at different levels including centralized AI and big data SE, LSE, and real-time or quasi-real-time SE, as well as visual modeling component AI Explorer and machine learning, and deep learning framework, flexibly meeting the needs of different deployment scenarios.
  • Model layer: It supports abundant general-purpose capability model components in a 5G network, such as alarm correlation model, capacity prediction model, user behavior model and traffic model, flexibly supporting the calls of the application layer.
  • Application layer: It is oriented to 5G intelligent applications, flexibly supporting a variety of application scenarios such as intelligent prediction, RF fingerprinting, intelligent slicing, and root cause analysis.

 

The combination of various AI capabilities can be integrated into ZTE 5G products including 5G NR, 5G cloud core, 5G UME, VMAX and BigDNA to meet specific needs of network deployment. ZTE's AI-assisted intelligent networks will help operators plan their networks more scientifically with more accurate fault location capabilities, lower operation costs, and business capabilities that better meet user needs, so that they can survive the fierce competition in the coming 5G era.

 

[Keywords] Artificial intelligence (AI), 5G network operations