AI Has Central Role in Helping Verizon Build Out its 5G Cell Tower Network 

For years, Verizon engineers have been using a laborious process of spreadsheets and manual data crunching to figure out the best locations for where to install the company’s huge national network of thousands of cellular towers and equipment.

But today, as Verizon is deploying many more thousands of its latest 5G towers and transmitters across the United States, the planned locations are not being left up to spreadsheets anymore. Instead, AI and complex modeling are being used to perform the analyses that are helping to company to strategically position its expanding network of 5G towers and transmitters to better serve its customers.

“There was a lot of work happening in spreadsheets, and there were a lot of people who were individually doing a lot of active planning” to position the towers in the past, Linda Avery, Verizon’s chief data and analytics officer, told EnterpriseAI. “We are [using AI to take] the place of … people who were doing their best to piece this together. It used to take weeks to analyze, but we are now able to do scenario analyses in minutes, where before it would take a great deal of time.”

One huge benefit is that those engineers are now being freed up to work on more essential projects in their jobs for Verizon, said Avery. “What this is allowing to happen is that we are really getting to the higher level capabilities that we want our engineers to be focusing on.”

Linda Avery, of Verizon

The use of AI for locating 5G towers and equipment is part of Verizon’s ongoing plans to integrate AI into the fabric of how the company delivers services across its consumer and business communications operations, said Avery.

To do that, the company has created and utilizes a digital twin of its entire network, giving it the ability to emulate and simulate the network under varying conditions, she said.

“It allows us to make rapid decisions,” said Avery. “When you look at what we are doing with 5G we are trying to create an optimized build-out plan. We are doing an optimized amount of capital planning. We are using network digital twin to allocate our resources and we put together a staffing plan for our forecasts.”

The digital twin also allows Verizon to rapidly apply different scenarios, such as “what if” analyses for new optimizations, OpEx or CapEx expenditures, project planning and many more variables, she said.

“We can basically do all these different scenarios in ways that we have never been able to do before,” said Avery.

“There has always been a real pioneering aspect of Verizon in this respect, so I don't think it was that much of a leap,” said Avery. “It all begins with the data, and there has been a lot of focus on having a top notch user network from a data perspective. It began with things like managing the network optimally and how to do predictive maintenance on the network and prevent catastrophic events. It's an evolution. I can't say there was this aha moment.”

How AI is Helping Verizon

To plan locations for future 5G towers and transmitters, Verizon engineers use AI to sort through the huge amounts of data gathered by the company to see where cellular usage levels are high, as well as analyzing the topography, landscape, user demographics, population density, the nature of nearby structures and more, said Avery. The goal is to optimize the positioning of new towers and equipment to meet the needs of the network today and tomorrow.

“We use drone capabilities quite often,” said Avery. “This is really a matter of continuing to build out and optimize. All of that information is being captured and we have an awareness of network traffic, but only at a very high level.”

Engineers then must evaluate many factors on their own, such as how cellular traffic loads change as users people move across an area and the cell towers hand off their calls from one tower to another.

“We have many models that come into play because we have many dimensions of the data coming in, and each of them needs to be represented by these models,” she said. “So, what you have here are models that are interacting with one another. That is how we can do the scenarios as quickly as we can.”

Verizon does all this work using its own homegrown software. The company has been using the Verizon Media Group grid network, which includes Yahoo’s network, to run its digital twin and AI workloads. That will change in the future as it moves this work into the cloud with new vendors following the coming sale of its Yahoo and AOL brands to private equity firm Apollo Global.

The use of AI for the project has been very successful, said Avery, who joined the company two years ago after the work was already under way.

“Yes, absolutely,” she said. “Even just in terms of capital planning, we have been able to save millions of dollars … that we get to reinvest in the network. We are seeing multi-million dollar results that we are applying to optimize the build.”

Ultimately, the 5G tower locating project was created to solve “a classic optimization problem,” said Avery. “It is right in the [heart] of what AI can do. We have been creating enormous acceleration around it and expanding its origins.”

As part of this goal, Verizon also created an internal AI center for excellence to help explore the possibilities of AI across its business operations.

Analysts on Verizon’s AI Strategy

Several industry analysts told EnterpriseAI that Verizon’s use of AI to do this work is part of a wise roadmap.

“Cell site placement has historically been a matter of designing a network using specialized CAD tools—not spreadsheets—that play with geographies, spectrum allocations, transmitter positioning and tilt and frequency reuse patterns,” said James Kobielus, senior research director for data communications and management at TDWI, a data analytics consultancy. That work has then been typically followed by testing of radio frequency signals for strength, interference and signal hand-off to validate the planned tower layouts and cellular engineering metrics, said Kobielus.

“This iterative process can take days or weeks to converge on an optimal network-wide cellular network design,” he said. “Using AI to design the network, predict its performance and even optimize how the testing is carried out can save cellular networks time, money and personnel.”

Avi Greengart, the founder and lead analyst at Techsponential, agreed.

“Determining where to put a cell tower may not be what people think of when you say, ‘artificial intelligence,’ but lots of AI use cases end up in definitional knots,” said Greengart.

Avi Greengart, analyst

While Verizon has developed “an impressive algorithm that weighs multiple variables to determine tower placement,” Greengart said that the algorithm itself does not appear to use machine learning to evaluate data from past tower locations to refine future placement. “So, I do not know if it is really AI or just sophisticated automation, but if it leads to better outcomes without manual entry in spreadsheets then it is all good.”

Another analyst, Jack E. Gold, the president and principal analyst at J.Gold Associates, said that cellular companies including Verizon have been working with AI for this kind of work for some time.

“It is probably not the high end, heavy duty AI that people think about needing supercomputers, but it is of great value, nonetheless, in optimizing their networks, which is very difficult to do manually,” said Gold. “They do this not only for tower placement, but for analyzing network traffic patterns and user needs, and modifying the network architecture – thanks to the use of virtualization – to optimize operations.”

Gold said he sees the use of AI growing for Verizon and other carriers in increasingly complex ways for many different operational requirements.

“And it gets even more important with the advent of 5G and the many classes of network services that need to be deployed and adjusted for specific customer and geographical needs,” said Gold.

Bill Menezes, an analyst with Gartner, said Verizon’s implementation of AI here “does seem like a repeatable and useful AI solution given the huge numbers of new 5G small cells that Verizon must deploy to have any competitive reach for its mmWave 5G Ultra Wideband deployment.”

Since mmWave transmissions are very sensitive to obstacles, identifying small cell sites that are not only optimal for performance and coverage but are also available and accessible is critical for the company to reach its TCO and ROI targets, said Menezes. “Assuming their AI has access to the right data and incorporates the huge number of variables – such as different foliage obstacles at different times of year and over time as trees grow, for example – relevant to all the new sites, it is conceivable Verizon has significantly cut both time and cost for their site planning.”