logo_header
  • Topics
  • Research & Analysis
  • Features & Opinion
  • Webinars & Podcasts
  • Videos
  • Event videos

Using AI to empower 5G with intelligent networks and operations

This Catalyst explores how AI can empower network automation to help operators boost network efficiency, improve customer experience and optimize operating expenses.

Annie Turner
27 Oct 2020
Using AI to empower 5G with intelligent networks and operations

Using AI to empower 5G with intelligent networks and operations

In the era of 5G, as networks become larger and more complicated, and run more complex applications, AI is necessary to help automate them. A new proof of concept Catalyst explores how AI can empower network automation to help operators in various ways. They include: to plan, construct, operate and maintain networks to boost their efficiency; to improve customers’ experience and energy consumption; and to optimize operating expenses for telecom operators. The many simultaneous and challenging issues the project set to address are summarized in the graphic below. It was in recognition of these looming challenges that China Mobile became one of the founders of the TM Forum Autonomous Networks (AN) project in 2019 (see related white paper, Autonomous networks: Empowering digital transformation for the telecoms industry). The AN project was also set up to address business issues, such as developing and enabling new business models, as well as the operational and technical aspects of automation.

China Mobile is the champion of the proof of concept Catalyst project called AI empowered 5G intelligent operations, with support from participants AsiaInfo, BOCO Inter-Telecoms, Huawei and Nokia. The Catalyst team intends to feedback the experience and knowledge it gains from the proof of concept into the AN project, which is involved – in a controlled way – in the actual maintenance of its live, commercial network.

The team referenced the AN framework, set out in the original white paper, which includes three layers of single-domain closed loops for automation, and multi-domain closed loops for end-to-end network automation. The Catalyst looked at various use cases for operations, network and business.

AI for operations

For operations, the team ran a use case for 4G/5G intelligent incident management, which includes alarm aggregation, root cause analysis and accurate tickets. In the real world, there are 500,000 alarms every day on a local network or more than 300 per second, which generates a ticket storm. Filtering the alarms using root cause analysis relies on specialists’ experience, but this is slow and error prone. Tickets are despatched repeatedly, adding to the burden, and low priority alarms and the accuracy of tickets are difficult to establish, making it hard to locate and fix faults across the network’s layers and different vendors’ equipment. In the Catalyst, by using AI, the team compressed the alarm rate by 99.9%, reducing tickets by 35%.

Yao Yuan is Project Manager at China Mobile, and leader of the Catalyst project. He explains that the operator is using AI to manage between 50% and 60% of the alarms, and believes that over time, this will rise to 80% or 90%. He says it is difficult to assess how much money this already saves and could save his company in future, but it could run to millions of dollars.

AI in the network

One of the network use cases is the intelligent optimization of a 5G network. Using massive MIMO to deliver services involves more than 1,000 patterns with horizontal and vertical beam widths, down-tilt angles and azimuth (an angular measurement in a spherical coordinate system). This is complicated by the environment the network operates in, which might have high buildings, and other sources of interference. Usually, setting up the network coverage depends on drive tests, which results in overshooting, overlapping or poor coverage. The Catalyst improved coverage by 15.8% on 91% of roads, picking up problem areas using a propagation model and AI.

AI to improve customer perception

The traditional approach to handling users’ perception was through passive work mode, which means all the processes are handled manually by technicians. This takes a huge amount of effort and resolves few issues. This Catalyst’s use cases are based on big data, enabled by AI. All customers with low levels of satisfaction customers are identified automatically. The optimal solution can be delivered as they are identified, improving customer service efficiency by at least 20%. In other words, 88% of users who are complaining can be dealt with directly versus 50% previously.

Business benefits

For business, one of the Catalyst’s use cases was how AI could be deployed to offer premium broadband with guaranteed, superior performance. In real life, many subscribers receive far slower transmission speeds than the theoretical top speed of their connection. Many factors can affect performance, within both the network and the home. Using AI to offer guaranteed performance increased the marketing/sales conversion rate by five to seven times.

Success and hard lessons

Yao Yuan says, “The biggest success of the first phase of the Catalyst is practicing the application of AI in the real network and for network planning. It has given us a chance to better understand AI and helped us figure out models. AI is a great but not an easy technology, and sometimes training the AI is as difficult as raising a baby”.

The work undertaken in the Catalyst shows how AI can be applied and already China Mobile is committed to the project becoming multi-phase. The next one will be on show in summer 2021 and Yao Yuan predicts that it will focus more about implementation than learning and training. Watch the team presenting its work from the first phase of the Catalyst on this video, part of the Catalyst Digital Showcase.