The user’s view of the 5G RAN

Optimizing the RAN to get the best 5G user experience 

Are subscribers getting the user experience that 5G promises? Plenty of reports tell us how impressive the speed and latency of the 5G RAN can be. But the subscriber experience can be less stellar – or at least less uniform. Subscriber experience is to some extent subjective and more difficult to quantify than RAN performance but, for operators, it is what matters. A high-capacity, low-latency RAN is not helpful if subscribers cannot benefit from it.

Obviously, RAN performance and user experience are correlated. Still, there are ways to optimize the RAN performance to maximize the user experience that go beyond the optimization of individual RAN parameters. By improving the user experience, the wireless infrastructure becomes more cost-efficient, and the operator can squeeze more out of its network for the same investment.

Yet, the RAN optimization still happens mostly within its microcosm. Operators typically try to improve RAN KPIs on the implicit assumption that this gives the best user experience the network can support.

With the combination of 5G, AI, advanced analytics, and automation, a new approach becomes available: to fine-tune the RAN to optimize the user experience – rather than to optimize the RAN performance as an end to itself.

The novelty in this approach is that end-to-end network performance data is combined with the user experience measurements and predictions to create the training set for the AI platform. The goal is to find the RAN configuration that maximizes the user experience and to identify the RAN updates and expansions needed to meet subscribers’ evolving demand cost-effectively – i.e., providing the user experience subscribers expect and are paying for, without overprovisioning the network.

The shift to maximizing user experience is crucial because the effectiveness of AI-based optimization depends not only on how good the learning algorithms and training are but also on what the target is. Operators can choose a variety of optimization targets – e.g., minimize capex, maximize throughput, or manage congestion – to direct optimization along a different path. While it may be relatively easy to meet these targets, they are too narrow to ensure the best user experience – which is arguably the primary function of the RAN. For example, the ability to manage congestion is vital to providing a good user experience, but avoiding congestion alone is not a guarantee for a good user experience.

How can operators do this? The evolution to 5G brings challenges and additional complexity, but operators have to deal with them to unlock the opportunities that 5G offers. RF planning and RAN optimization still play a critical role in planning, operating and expanding the RAN. What must be added to the mix is the data on end-to-end network performance and user experience to optimize the user experience.

Of course, operators are already using estimates of user experience. But these typically are averages, insufficient to capture the distribution of user experience – i.e., how it varies among subscribers in different locations, with different devices, and using applications with different requirements, as well as how capacity and density of demand (or congestion) affects it. And this is the granularity that operators need to apply user experience data to optimize RAN performance.

Analyzing 5G data with AI allows operators to go beyond the average user experience and estimate the range and mix of user experiences that subscribers will encounter across the network footprint.

Is collecting and using a much larger set of data going to add to the cost of network planning and operations? This is where automation plays a key role. It is not a cost issue. It would be impractical to manage such a high volume of data manually. Automation makes it possible and affordable to process that data in real-time, increase the accuracy of user experience estimates and predictions, and use them to make RAN planning, testing, and monitoring more precise.

In the planning phase, the ability to predict the distribution of user experience would, for instance, allow operators to avoid the deployment of unnecessary infrastructure where expected demand does not require it – or delay deployment until it is going to be needed. It can also provide specific advice on how to best dimension and expand the network – i.e., which spectrum bands to use, which architecture is best suited (e.g., macro or micro or small cells), and whether to deploy outdoor or indoor infrastructure.

After deployment, real-time monitoring of the distribution of user experience can improve the ongoing optimization of the RAN, keep the operator informed of performance issues, and provide an assessment of capacity and coverage limitations.

With 5G, AI and automation, the optimization of user experience comes full circle. User experience is no longer a network metric to be passively measured. It provides the ongoing, real-time input to optimize the RAN to do its job: delivering the best user experience.