Radio Composer 2.0: Intent-Driven Network Adaptation

5G is more than a new generation of technologies; it marks a new era in which connectivity will become increasingly fluid and flexible. By 2025, 5G networks are likely to cover one-third of the world's population. To provide optimal experience and high performance, 5G networks shall adapt to applications and performance tailored precisely to user needs. Therefore, integrating AI into radio network has become a consensus in the industry.

Radio Composer is the industry's first native-AI based solution that has AI built inside 5G base station for fine-grained management of radio resources to offer better user experience with higher network efficiency and less carbon footprint.

Radio Composer 1.0: Industry's First Network Resources Orchestration

User experience is a key driving force of 5G development. However, 5G still uses some radio resource management technologies that are not really competent for the task, and uses an inflexible radio resource management policy to deal with all 5G users, which can not perceive and adapt to diverse user needs. It is of great importance for operators to deliver a user-centric experience for consumers and industry applications, which is exactly what Radio Composer can achieve.

Radio Composer enables multi-dimensional perception and machine learning of service characteristics, network characteristics and UE capabilities to achieve optimal match between user needs, network resource allocation and UE capabilities. This helps to yield the best user experience and the highest network efficiency, thus increasing 5G user base and promoting the development of 5G industry and ecosystem. Radio Composer offers a real-time and always-smooth user experience anytime and anywhere. AI training and inference are both implemented at the base station, and the AI training is carried out under the low load of CPU to minimize the impact on legacy user experience.

ZTE has deployed Radio Composer in more than 10,000 sites in China, increasing 5G network capacity by more than 30% and user speeds by 300%. In-depth and comprehensive trials of this solution are also being carried out in Thailand, Spain and Italy, which will help to build higher-quality 5G networks.

Radio Composer 2.0: Industry's First Intent-Driven Network

The real value of AI, however, is not limited to the applications connected to the network, but will eventually be realized in the network itself, thus facilitating advanced network intelligence. That is why the intent-driven technology is introduced as an evolution of Radio Composer (Radio Composer 2.0).

Radio Composer 2.0 deeply integrates the intent-driven technology into the most common daily scenarios, such as Tiktok live streaming, Tencent video streaming and WeChat online conferencing to implement automatic strategy optimization in B2B and B2C scenarios. Simply by using natural language as the input to describe the business outcomes customers wish to accomplish, the network can convert these objectives into the configuration necessary to achieve them without having to manually code and execute individual tasks. The network implements a software-enabled automation process that uses high levels of intelligence, analytics, and orchestration to improve network operations.

Taking the intent-driven user experience guarantee as an example. For an intent to “have a smooth TikTok experience around 10:00 tomorrow morning”, the system will automatically translate the intent and handle the service guarantee flow, including automatic millisecond scheduling policy generation, minute-level performance feedback, and automatic service experience optimization. With historical learning of packet size, sending frequency and intervals, transmission protocols used, intervals between uplink sending and downlink receiving and data throughput, an offline model is trained with convolutional neural networks (CNN) and random forest and put in BBU. This is the implementation of BBU-based traffic pattern analysis. When data flow is input, it can be identified in real time, with an accuracy of up to 90%. An application can be identified as multiple services. For example, Wechat can be divided into at least 5 service types, including Wechat voice calls, video calls, payments, moments and videos. As a result, more than 7000 service types can be identified, covering 95% of mainstream services.

The intent-driven network deployed in China Mobile's Fujian Branch shows that the specific service experience has been improved by more than 100%, and the efficiency of operation and maintenance has increased by 500%. More integrated intents will be implemented in the future, such as the combination of user experience and network energy efficiency.

It is also challenging to integrate AI into RAN networks. Taking data and algorithms for an example. Traditional data is discrete, coarse-grained and unidentified, while the data used for AI needs to be structured and identified, so it is necessary to automatically collect, clean, correlate, identify and store data based on a unified data model. Algorithms need to be abstracted for independent evolution, just as experience prediction algorithm will be scheduled by multiple AI-based applications and evolve with the development of applications. The middle platform will flexibly orchestrate data and algorithms to empower diverse AI-based applications, thereby improving user experience.

As 5G develops rapidly, intent-driven networks will be widely deployed. Therefore, a large amount of computing power is needed, which is quite challenging, and the computing power sharing scheme between different base stations is a perfect solution to this problem. The computing power orchestration breaks the limit of computing power of a single site and realizes cross-site computing power sharing. It increases the available computing power of busy sites by 14% and empowers more intelligence-based applications such as energy saving.

As the key infrastructure of digital society, 5G network not only provides consumers with more and more diverse and demanding applications, but also supports digital transformation for a variety of vertical industries. Radio Composer helps to unlock the potential of 5G networks and promote advanced network intelligence.