Automation, geolocation and ML – three critical issues addressed in SCF papers
New papers from SCF working groups are always a hot topic of conversation at Plenary meetings, and our next Plenary – to be held in London on September 16 and 17 – will be no exception. Among the new publications will be two that address key issues in the planning, management and optimization of very dense small cell networks.
The position of small cells, relative to each other and the user traffic, has always been important to achieve the best cost/performance ratio. But in low density networks, manual tools can often suffice for planning, which becomes impractical when large numbers of cells are involved. Far greater precision, and automated, intelligent methods of positioning cells, are essential to avoid interference and inefficient distribution of resources.
Once the cells are in place, there is a continuous process of optimization, which will become more dynamic, in response to changing traffic patterns, as 5G develops. While automation has been an element of small cell management from the start, via SON (self-optimizing network) tools and APIs, a whole new level of automatic operations is required when access points are deployed in huge numbers. Increasingly, operators are interested in how machine learning (ML) can help add intelligence to their tools.
SCF has published a joint paper on the topic of precise location with 5G Americas. This calls for improved geolocation capabilities when compiling network quality reports to support planning and optimization. It also demonstrates the feasibility of applying ML techniques to improve precision further, and relate the position of the cell not just to existing access points and traffic, but to the predicted changes in user behaviour in future, based on analysis of trends.
Another paper from the TECH group addresses recommendations for how to define and implement SON and other RAN automation techniques in very dense 4G/5G networks. This builds on a rich body of work, created over many years, on topics such as the SON API for small cells.
These open interfaces and specifications will become even more critical to the economics of densification as deployers move to disaggregated networks with hundreds or thousands of cells. In that environment, manual approaches will be unable to keep track of all the elements, and how they are changing in terms of traffic load, traffic type and use cases. Automation will become more extreme, and vendors and providers are starting to experiment with the use of ML to support predictive optimization and fault detection, rather than only conducting optimization based on current conditions.
According to the findings of a recent survey of 76 mobile operators for SCF (Figure 1), the period between 2022 and 2025 will see a sharp upsurge in deployment of AI/ML systems to support intelligent RAN automation.
Figure 1. MNOs planning to deploy AI/ML systems to support SON and RAN automation in various timescales
At the Plenary, these two seminal papers will be discussed in detail, so you will be fully updated on their recommendations and how those will affect your own organization’s decisions and plans for dense networks. Feedback and further debate will be encouraged, and it will also be very important to discuss how to maximize the impact of both publications on the industry at large, in order to help shape the priorities of operators and suppliers and accelerate innovation.
Together with the Marketing group, there will be plans for ensuring that everyone is aware of these papers and their findings, and for getting their highly significant messages to the right audiences, in order to increase consensus and understanding, but also to spark practical actions in the near term.
Do sign up for these sessions to help address this key challenge for achievable densification, and have your views and ideas included in a critical debate. Pre-registration is essential, so don’t delay, book your place at the plenary today.