PRESS RELEASE: As networks continue to evolve, new services and application scenarios are emerging. However, 5G-A communications technologies are encountering a range of challenges, including increasingly complex channel models and intricate data processing requirements. In response, AI technologies offer robust capabilities in data analysis, task processing, flexible adaptation, and intelligent decision-making, all of which can significantly enhance the capabilities of 5G-A networks. China Mobile has proposed the “5G-A Network AI Application” technical solution, which seeks to integrate 5G-A communications technologies with AI applications. This solution aims to establish core “native intelligence” capabilities, ultimately addressing critical challenges in the development of 5G networks and advancing the evolution of intelligent networks.
Technical Challenges Faced by 5G Network Development
5G-A offers enhanced data transmission rates, reduced latency, broader connectivity, more reliable performance, and more intelligent capabilities. While these advancements address the increasing digital demands, they also introduce new technical challenges for networks. Firstly, the complexity of channel models has escalated. Traditional terrestrial channels must be adapted to cover dense indoor environments and low-altitude areas, making channel models more complex. Secondly, data processing has become more intricate. Continuous influxes of service experience data and user behavior data are necessary for network. Furthermore, the data collection interval has been reduced to just one minute, leading to an exponential increase in the volume of data. Thirdly, network requirements are highly variable. 5G-A networks must accommodate diverse needs, including data rates, latency, the number of connected users, and SLAs. Different services and scenarios impose distinct network requirements. Lastly, there is a significant demand for cross-domain convergence. This involves horizontal coordination among networks, terminals, and services, as well as vertical alignment between sensing capabilities, service characteristics, and communication capabilities.
5G-A Network AI Solution
With the evolution of wireless AI applications, global communication standard organizations have engaged in active discussions to define AI application architectures for RAN (radio access network) elements and management systems. Within this framework, the AI Intent Wireless Agent (AIWA) operates in the wireless operation and maintenance center (OMC), overseeing network intelligence within RAN domain. Meanwhile, the AI Integrated Wireless Unit (AIWU) at base station manages network element (NE) intelligence specific to that base station. This layered architecture facilitates the implementation of more efficient wireless AI applications throughout the entire network.
Three Aspects of Key Technologies
Typical application scenarios: AI-driven real-time service assurance addresses differentiated service requirements, enhancing overall user experience. For network optimization, intelligent technologies are employed to refine networks and allocate resources properly. Moreover, China Mobile leverages intelligent energy-saving technologies to promote sustainable development while implementing RAN AI applications across three typical scenarios.
Two-level BBU architecture: Intelligent network technologies present new challenges for the computing resources of base stations, which must now offer both computing and storage resources for AI models. To address this, new computing resources are deployed within several base stations, creating a two-level BBU architecture that delivers the necessary hardware support for AI applications in RAN.
Open interface capabilities: AI applications within base stations produce a wealth of valuable data. The standard open capabilities of intelligent NEs powered by AI enable fine-grained data analysis and evaluation.
Summary and Future Outlook
In the future, standard protocols will continue to evolve, enhancing the vitality of 5G-A. The ongoing expansion of service applications will create broader business opportunities. As networks develop, addressing specific scenario-related issues will require cross-domain collaboration. Additionally, AI capabilities need collaboration across multiple network elements, particularly with the cooperation of RAN and core networks to resolve network challenges in an end to end manner.