Earlier this month, Infovista hosted a two-day summit to discuss the latest and hottest topics in radio access network engineering.
With hundreds of registrants from around the world and a healthy turnout on both days, we covered a huge amount of ground, taking in conversations on:
- Artificial intelligence in RAN planning
- The latest trends in this space
- What’s new in planning and testing
- Various 5G case studies focusing on best planning practice among customers and partners
I was delighted to take part in several sessions in this new annual event focusing entirely on RAN, including a heavily attended one on AI and the latest Planet evolutions. We didn’t have time to address all of the questions in that session, so in this blog post, I’ll focus on the most interesting unanswered questions from that, and provide a quick overview of some of the others – all with links to where you can watch any that you missed.
AI and the latest Planet evolutions
In this, the most popular session on both days, we invited attendees to discover all of the innovative functionalities around AI and crowdsourcing that we’d introduced to our Planet 7.5 release, and understand how RF engineers can leverage them to improve their planning and optimization processes.
From the ground up, AIM is a neural network-based propagation engine. It ‘looks’ at all the relevant data points to develop a ‘site picture’ – the environmental geodata at the cell location, cell information such as the heights, antenna and power etc – then, based on its understanding relative to its machine learning (ML)-calibration ‘knowledge’ or enhanced tuning calibration, it finds the best possible combinations of model parameter to fit the current radio design scenario.
One question focused on what we mean by ‘ML-calibration’.
In short, AIM taps into Infovista’s vast historical database of environmental calibration, accessing the best geodata available. With this, we can build an extensive working library of data, millions upon millions of data-points, which the neural net can compare with models of real-world data-points. This means that, in effect, we’re training the AI to understand the world it perceives using ML technology.
But the big question was on what savings can be made by using AIM. Is it time? Money? Operational reductions? Well, it’s yes to all of those.
Compared to the traditional drive test approach, our AI model significantly reduces tuning costs by up to 70%. AIM also reduces the number of models required. For example, one of our customers traditionally required 300+. This dropped to around 30, vastly reducing the complexity of their work and bringing them huge operational and time savings.
But these savings are all underpinned by the increased accuracy of our predictions, with significant reductions in the standard deviation of predicted vs real-world testing data. Thanks to improved accuracy in network plans, we’ve calculated that CAPEX savings are likely to be in the region of 10 to 20%.
For more on AIM, take a look at our recent blog post, Using artificial intelligence in RAN planning to improve accuracy and save time.
Evolution and trends in the wireless market
Closely aligned with the above, our CTO, Yann Le Helloco, led a discussion in this session on the latest wireless industry insights and our vision on network lifecycle automation (NLA). Presenting the evolution of the Infovista’s RAN products towards automation, he focused on network planning and optimization solutions.
For more on the background to network lifecycle automation, see our blog post, How network lifecycle automation could boost telecoms in the 5G era, which links to a recently published report we commissioned.
Today’s CSPs face many strategic challenges in automating and optimizing their network lifecycle for 5G, including those of complexity, efficiency roadblocks, budget constraints and customer churn. NLA offers a huge competitive edge to carriers as 5G is rolled out. So how are companies preparing?
Infovista commissioned Forrester Consulting to interview 104 senior wireless strategy decision-makers around the world to investigate their understanding of NLA, and how they plan to adopt it.
Download the report, Network Lifecycle Automation: the telecoms industry view.
Data-driven 5G planning: the Ericsson story
In this session, Ericsson, a long-time partner of Infovista, shared what their customers have learned so far about 5G and new challenges encountered for 5G network design and optimization. Participants learned how this infrastructure provider is using a large variety of data, including crowdsource, as well as AI and ML, in their planning and design practices to increase speed, scale and accuracy for 5G deployments.
Better planning with umlaut crowd data
This session provided attendees with a chance to learn about umlaut’s unique crowdsourcing technology, which collects data about real-world experiences wherever and whenever customers use their smartphones, and how Planet users will be able to easily leverage this type of data to improve RAN planning and optimization activities.
Also featuring umlaut, take a look at our recent blog post, A case study in managing multiple network benchmarking test projects.
Cloud-native planning: the Rakuten story
In this session, our panel discussed how Rakuten – featured in this recent press release – delivers the first carrier-grade, cloud-based platform for radio network planning, optimisation and modelling powered by Infovista Planet microservices technology. The solution aims to support the rapid roll-out of their cloud-native 5G mobile network across Japan as the world’s first fully virtualized New Generation network.
Improving planning and testing workflows
This session covered a number of bases, including: increased efficiency for 5G site testing; integrated workflows from planning to testing; and the Planet and TEMS single site verification solution. It concluded with an all-important discussion on the outcome of these initiative: time savings and increased accuracy in planning.
Take a look at our ongoing series of blog posts on drive testing, which starts with How technical disruptions are changing the way 5G NR testing must be done.
More to come
We’ll continue to publish new material related to the sessions presented at the event – there were three more – so readers should check out our resources library for upcoming info. But for now, take a look at those you missed or want to re-watch from the six above.
See you this time next year for our second Infovista RAN Summit – but before then, feel free to book an in-person or virtual meeting at next week’s MWC Barcelona, which we discuss in this blog post.
