Elisa Polystar and partners collaborate in ground-breaking Catalyst Innovation Project – autonomous CEI for zero-touch 5G networks, Phase III
TMF Catalyst Innovation projects are designed to bring together relevant stakeholders into collaborative groups to explore a concept or idea – and to deliver a working proof of concept that delivers value to the communications industry. Recently, Elisa Polystar worked with Elisa, Telefonica, NTT Data, Nokia, and Optare Systems to explore the autonomous customer experience management in the context of 5G – which brings a host of new challenges that must be addressed by operators.
Complexity complicates customer experience
5G is based on a cloud-native, virtualized architecture that (with full 5G SA) offers new levels of performance – but also an entirely new level of complexity when compared to previous generations of mobile network technology. This has implications for ensuring that the right experiences are delivered to network users and connected devices.
One way of measuring service performance is to create a Customer Experience Index (CEI) – a measure that’s based on aggregated feedback from active customers. When processed by AI and ML, such data can provide insights that help operators to optimize performance. However, it is largely retrospective – and actions can only be taken after the fact.
What’s needed is a way to provide proxies for a CEI that can allow network issues to be identified and resolved before they impact service performance and network quality – and customer experience.
In parallel, the industry has also recognized that 5G networks need to operate autonomously, to ensure they deliver the agility and performance required by new applications and services. So, can autonomous operations also be combined with measures that enable automated problem detection, analysis, and remediation – as well as fault prediction – to deliver customer-centric performance and new CEI measures generated from the network directly?
It is this question that the Catalyst was designed to explore, through four selected use cases:
- CEI prediction
- Persona models and problem management
- 5G network digital twin
- CEI network predictive and automated 5GSA resynchs
Each was based on considerations included in TR284E – the TM Forum’s technical report on closed-loop automation and digital twins.
CEI prediction
In contrast to leveraging real customer inputs to generate a CEI, accurate prediction of experience based on network information would enable a predictive and proactive approach. The key innovation is to enable different CEIs to be created for different services. This is essential, because the users of 5G may not be people, but could be industrial processes, autonomous vehicles, and so on – each having both a different performance expectation and, hence, a different perspective in terms of CEI.
With AI and ML processing, data captured from the network (via the NWDAF), is fed into different processing functions to generate the relevant CEI values. In practice, this was shown to enable CEI prediction with accuracy of more than 90%. The CEI values captured could then be used to predict deviations from the expected range, perform root-cause analysis, and trigger remedial actions so that experience can be maintained before any impairment in service performance occurred.
Persona models
Understanding customers and services requires their classification into generic types – or personas. By creating personas to represent users based on consumption, service usage and other factors, the team hoped to boost the CEI models generated – enriching them with data regarding network complexity and experience.
With such models, the specific service set offered to a group of similar users can be management more effectively – leading to autonomous operations and automated performance management, through the activation of corrective measures to manage the requisite CEI models.
5G network digital twin
The group created a digital twin for 5G networks – a complete virtual representation of the network, built on real-time data from the live network. By using the CEI data from the first use case, an accurate, up to date representation of real experiences and network performance can be obtained.
There are three aims. First, to expose network capabilities, so that developers can model new services with real network data and experiment with new performance demands, particularly for critical communications services, across different verticals. Second, to enhance coverage planning and strategic investment decisions and, third, to model new scenarios based on the CEI data.
CEI network predictive and automated 5GSA resyncs
Problems with cellular base stations and the synchronization between them can have serious detrimental effects on mobile performance and customer experience. They can also lead to total loss of performance in a given location. Early identification of such issues, both through detection and prediction will minimize their negative impact and enable the correct remedial steps to be taken – up to and including a complete reset of the base station in question, which for obvious reasons has been a task traditionally undertaken at night.
In this case, Elisa Polystar and Elisa combined to capture and filter alarms from base stations and, depending on the predicted severity of the issue, take the appropriate corrective actions. The results indicated that the new algorithms reduced the number of users affected by any issues by 30 – 40% and, in some cases, by as much as 90%.
The way ahead
The full results of this catalyst are available on the TMF’s website, together with reports and supporting documentation. They clearly show that automation can also deliver enhanced customer experience and accelerate network maintenance activities – promoting subscriber retention, reducing costs – and supporting innovative yet increasingly complex new services. In particular, the use of personas to build new CEI profiles shows that a more granular approach to service assurance and performance can be delivered, tailored to an increasingly diverse range of services, with predictive maintenance optimized for each new use case.