By: Vanness You, Vice President, Huawei Middle East Carrier Business Group 

The emergence of always-connected smartphones around 2007, has massively transformed the wireless network. Since then, Huawei has been leading the wireless network transformation with the SingleRAN concept and built cutting-edge technology and capabilities on the foundations of vertical integration.

We firmly believe vertical integration is the key to true innovation, validated by the success of vertically integrated industry leaders such as Tesla and Intel.  Moreover, the vertical integration model in the wireless network technology allows us to develop capabilities that many customers appreciate, such as the need for convergence, multi-RAT support, deep integration and the ability to support software-based evolution.

Since the first launch in 2007, our solution has evolved to cover many current and future needs of minimizing total footprint and maximum performance per watt. Additionally, in 2018, we introduced green technology into our solution, which is three years ahead of the industry in terms of embracing green ICT.

As we move into 2022, intelligence has emerged as a key cornerstone of the network. Our main objective in developing the 5G network has shifted from building new infrastructure to embrace the Internet Of Everything to enabling the Intelligent World of Everything by combining 5G with big data, cloud, and AI to meet different requirements:

First, enable services fast and agile service roll-out. This ability relies on an 'intelligent connected world' covering 2C and 2B domains. Apart from voice and data, consumers demand advanced services such as AR and VR, while blue ocean industry automation and connectivity use cases are gaining momentum at the enterprise end. The service provisioning cycles need to be 'on-demand' and agile. The time for carrying out requirement analysis, service design, wireless network site survey, the actual network design, parameter configuration and onsite tests has to be less.

Second, meet different wireless networks service requirements, such as latency, rate, and uplink, translating into massively diversified services. Therefore, we optimize cells at both frequency and networking dimensions to accommodate rapidly advancing technologies such as AR, VR, and XR Pro, which require Gbps and low latency. In addition, we are working to resolve complexity for networks still layered with multiple frequency bands, macro sites, small cell sites, and millimeter wave solutions.

Third,  meet the demands  for green ICT. The 100-fold traffic increase expected by 2030 cannot translate into a 100-fold increase in energy consumption. China Mobile has proposed a more moderate increase in energy consumption and envisions a 15% rise in power usage to accommodate the 100-fold traffic increase.

Forth, meet the industry target of O&M automation, go from L0 to L5 autonomy on the network intelligence side. Some network parts have already implemented L2 autonomy, and manual interventions are slowly coming down. Currently, less than 5% of failures can be prevented and predicted; the rest are solved reactively, i.e., faults are only identified and resolved after they occur.

To meet the requirements of enabling the Intelligent World of Everything, Huawei has been working on applying intelligence in the wireless network. We have introduced many AI mechanisms and algorithms for O&M alarm and handover algorithm optimization, advancing basic AI technologies from non-real-time computing to deep learning, including federated learning.

We aim to perform intelligent prediction through data correlation analysis and pass intelligent judgment for future event prediction. To achieve these objectives, it requires that wireless networks be built based on an intelligent network architecture, divided into several layers.

Our IntelligentRAN model starts with introducing Mobile Intelligent Engine(MIE), a solution built upon two fundamental building blocks driven by AI. First, by deploying the BTS intelligence, it brings intelligence to our radio product portfolio. Second, by deploying MIE, it eanbles wireless network intelligence at the EMS and performance management layer, including training larger data models and storing more data.

One key point to understand here is that IntelligentRAN is built as a realization of ADN in the wireless network.

Based on the MIE architecture, various applications and use-cases could be excellent starting points, such as realising network and base station intelligence to improve the user experience significantly. Network intelligence can be used to optimise parameters for mobility, RF, etc., across the network to enhance user experience. Base station intelligence focuses on air interfaces. L3 smart grid is one of the common technologies we use on multi-band networks. Let's first look at how an optimal band is selected without intelligence. Firstly, a device leaves its operating frequency to measure other frequencies and then reports the results to the base station. Then the base station analyses the results and chooses the optimal carrier. The process can be lengthy and complicated. However, with intelligence, it becomes much more straightforward. By analysing the coverage records, grid-level energy efficiency, and real-time statistics, the base station predicts the device's coverage, spectral efficiency, and experience and quickly and accurately selects the optimal carrier. With IntelligentRAN, multi-band and multi-site intelligent three-dimensional coordination improve user experience by 50%.

IntelligentRAN also allows us to implement quick service provisioning and service assurance. Take the example of penetration loss of wireless networks, where glass loses about 10 dB or 30 dB for a concrete wall. In a typical industrial environment, some factories are dense, while others are more sparsely equipped. The best method is to build a model for each factory and scenario. The model can be optimized based on the deployment situation, making it more accurate.

Our mission is to achieve zero faults and zero service interruption during O&M. Here, we need intelligent fault identification, particularly root alarm identification. For example, numerous alarms are generated in case of a faulty transmission link, including base station and transmission alarms. Collaboration of historical data allows us to identify the real cause of the root alarm, which improves the efficiency and accuracy of fault management. The transmission link may be at risk due to an aging optical port and attenuation of laser information. If we can detect this rule several days in advance, we can perform proactive maintenance before system failure.

Today, by implementing intelligent identification and fault detection, we are also accumulating the capability to prevent and predict faults. Verifying these capabilities with our partners enriches our model database and abilities, allowing us to make better and more precise predictions. We look forward to working with the entire industry, including collaborating with key vertical market players, whether carriers or equipment vendors, to agree on the direction of autonomy and intelligence.