The Voice of 5G & LTE for the Americas

Select a Language: Spanish | Portuguese

Paths converge for AI, Edge and 5G

By Chris Pearson, President, 5G Americas

For decades, technologists have been talking about the impact of how their various technologies will mix and match to create amazing new outcomes. How would the home personal computer and visual operating systems come together? What can you do by networking together large computers with telephone wiring? If you put together small amounts of capacitive electrical storage in a panel, could you control your computer monitor with a simple touch?

In 2021, today’s modern wireless cellular networks are well on their way to bringing together the firepower of three amazing technologies that are shaking up the Internet: artificial intelligence, edge computing, and 5G. In 5G Americas’ latest white paper “5G Edge Automation and Intelligence,” we explore how these unique technologies are converging to accelerate the automation and optimization of wireless networks – and what that means for specific industries.

But what’s so great about network automation and optimization? The latest Ericsson Mobility Report demonstrates that mobile network data traffic grew a staggering 44 percent between Q2 2020 and Q2 2021 – to a monthly total of 72 exabytes per month. Let’s put that into perspective. At that rate of growth, mobile networks would need to *double* their capacity every 1.63 years. Imagine if you had to build a system of roads or airports that would double its capacity roughly every year and a half or so. That would be one heck of a job!

Clearly, you need the tools to be able to stay ahead of the crushing load of data. Moreover, it’s important to understand that the data is not being generated in one place alone. As networks get bigger and more mobile, more data gets generated further and further away from the data center. Data processing must occur closer to where the data is being generated – at the network edge. In the grand scheme, networks will become more de-centralized.

This is where edge computing comes in. But where exactly is “the edge”?

In 5G Edge Automation and Intelligence, the edge is really a continuum of “edge zones” that are comprised of four key areas:

  • device edge: where signal and data processing on the device occurs
  • premise edge: where processing that occurs on site, such as in the home, car, or enterprise campus happens
  • access edge: involves processing at cell sites or points of presence
  • metro edge: which is at upstream aggregation centers like Internet service providers

What’s important to understand here is that each of these edge zones has different forms and requirements. Edge computing delivers rapid processing of data that’s typically best suited to its local environment and can close the loop on travel time between where the data is generated (near the user) and where it’s processed, so that bits of data no longer have to travel to the data center, but are instead processed locally. For instance, IoT devices can do autonomous and intelligent local computation based on their sensed environment. Self-driving cars can process data from the premise edge, as well as load training data to the metro edge.

Now with information being processed near the device, on-site, at the cell site, and at the Internet service provider’s data center, things start to get wickedly complicated. The management of all this data storage, transfer, and processing needs to be organized efficiently. When done appropriately, 5G edge automation will allow for an incredible number of management tasks such as distributed data collection, normalization, real-time processing, context discovery, situational awareness, network slicing, and dynamics.

However, with each new capability also comes with new challenges. 5G edge automation will need dedicated resource management, specific infrastructure, location awareness, and location-dependent capabilities, as well as an automation and control framework with controllers, security, and a healthy marketplace for applications. How are human beings supposed to manage this level of complexity when human brains are only so big?

The answer? As human beings with limited amount of brain resource at any dynamic point in time, we just can’t. However, we can develop and deploy innovative solutions with processing optimization that address this needed dynamic resource.

This is where artificial intelligence steps in, offering capabilities that allow for the most efficient management of this tsunami of information. Intelligence at the 5G edge allows for numerous automated functions like statistics collection, anomaly detection, and prediction of infrastructure and network overload.

But how does AI do this? First, it’s important to realize that every autonomic decision-making system must include elements of a triggering system, stimulus, context awareness, and an adaptive policy engine that tells the system what to do to achieve the business goals. It must also include the tools that allow it to respond.

Figure 1 Autonomic decision-making driven by external goals and context ("5G Edge Automation and Intelligence")

In a 5G radio access network, a lot of these activities are built into the RIC (Radio Intelligent Controller), which helps to orchestrate all of these decisioning activities. As edge nodes work together to share sensor data from each other, the collaborative AI/ML engines work in a dynamic environment parse out resources that improve the efficiency and productivity (and safety!) for different 5G edge applications.

Particularly in areas like intelligent transportation systems, smart factories, or smart energy/smart home situations, there will become an increasing reliance on the intelligence that AI/ML can bring to the table. For instance, according to Omdia, the global AI edge chipset forecast is forecast to grow to $51.9B by 2025 – that is a *lot* of intelligence that will be laid down in these networks over the next few years.

Creating an intelligent automation architecture is clearly going to need to a lot of different participants, from hardware vendors, platform companies, app developers, system integrators, cloud service providers, and of course – 5G network operators. For instance, take a look at the figure below – you’ll see ample opportunity for a variety of ecosystem players to participate in the development intelligent edge networks.

Figure 5 Intelligent Automation solution architecture. ("5G Edge Automation and Intelligence")

As you can see, not every smart 5G edge network is going to be the same. Different industries and use cases will drive different equipment and architectures. An autonomous industrial solution might need both high-speed bandwidth and very low latency, as well as address multi-tenancy to secure access to the factory floor. An intelligent transportation system might involve vehicle-to-infrastructure (V2I), as well as infrastructure-to-pedestrian (I2P) sensing and capabilities with incredibly sharp video resolution capabilities. Smart energy and smart home applications will need constant connections, as well as security and a high level of user control.

Figure 8 Distributed edge to meet the need of 5G applications. ("5G Edge Automation and Intelligence")

The distributed edge is in its early days of development. As our 5G wireless networks grow in size and capabilities, they become more far-flung challenging to manage. While hyperscale data centers will probably never go away, the future for 5G networks increasingly looks tied to the fortunes and strengths of automated intelligence at the edge.

There is much work ahead as we add key ingredients to the 5G recipe for success. I’ll continue to chronicle more changes in our rapidly changing industry as we go forward.

-Chris Pearson

See how edge computing and AI are changing 5G networks.

Edge computing and AI are forever impacting 5G network automation and optimization.

Share this post

Share on facebook
Share on twitter
Share on linkedin
Share on pinterest
Share on print
Share on email

Sign up to receive our announcements