Rakuten Mobile scientists build the future of telecommunications with autonomous networking

In April 2020, Rakuten Mobile made history by launching the world’s first fully virtualized cloud-native mobile network. The company’s ambitious undertakings have attracted global talent from all corners of the telecommunications industry and beyond, each with their own ideas on how to push this revolutionary network even further.

Dr. Pierre Imai joined Rakuten as Principal Scientist for the Rakuten Institute of Technology in 2016, and today heads up Rakuten Mobile’s recently-established Rakuten Mobile Innovation Studio. 

“Scientists are not the best futurists. We aren’t really the best people to look into the future. We’re just the ones who build it.”

Dr Pierre Imai, Rakuten Mobile Innovation Studio

The studio was also founded in part by Research Lead Paul Harvey, PhD, who received his doctorate in Computing Science from the University of Glasgow. Imai and Harvey are two of the driving forces behind the push for autonomy in Rakuten Mobile’s new network.

“An artificial engineer: That’s what we want to build,” Imai explains. “Basically, we want to get some intelligence into the network that is capable of managing it as if it were a real engineer. That is the long-term goal.”

A truly autonomous network is a network that evolves

Imai is not thrilled with the “autonomous” label that much of the telecommunications industry applies to its forays into modern tech. “Many of the things you see in networking lately are actually automation,” he argues. “They’re not true autonomy.”

True autonomy is the term Imai and Harvey use to describe their ambitions with the Rakuten Mobile network. “No one has actually defined exactly what autonomous networks mean,” Harvey notes. “The autonomy that we are seeking — this true autonomy — is a system that can operate all by itself.”

The network Harvey envisions will be able to diagnose and solve problems without human intervention. “The system should be able to understand: What’s wrong with me? Am I behaving in a way that isn’t normal, in a way that doesn’t make sense?” he continues. “And it should be able to self-heal: What do I do about it? How do I go from this situation where I think I’m in some bad situation to one in which I feel good again?”

Autonomy vs. automation

A crucial ingredient in achieving true autonomy is the concept of evolutionary computation. This is what sets autonomous systems apart from automated ones, Harvey argues. 

“The fundamental question we ask when trying to apply autonomy to our network is the question of what happens when things change,” he begins. “If your self-driving car is driving you on a road from A to B, that’s fine — but what happens if you crash and end up in the river? It can’t do anything. It wasn’t designed for this; no one thought about it ahead of time.”

Mobile networks, however, need to be able to adapt to this dynamic landscape of problems.

“You have all of these different technologies working together, all of these different services happening at the same time, and it never stops. It’s online 24 hours a day, 7 days a week,” he explains. “It’s always happening, and customers are using it in increasingly diverse ways. So you have this really changeable environment — one that’s completely dynamic and never stops. And in a situation like that, what happens when something changes becomes an even more fundamentally important question.”

Static intelligence that is programmed into a network may not be able to handle unexpected changes — that’s why the ability to evolve is so important. “Evolution is not intelligent in the sense that it has an ingrained understanding of what’s going on,” Imai explains. “Evolution doesn’t understand. It tries out stuff and sees whether it works.”

Like a smart vacuum cleaner, Imai suggests. “Your Roomba at home; when it’s running around trying to find where the dust is, it doesn’t know where it is at first. Over time, the system might improve and notice that most of the dust is concentrated over here, and then it’s more likely to show up in this area.”

To allow evolution to take place, an autonomous network needs room to experiment. Too much room, however, can be unproductive. 

“If you have total freedom to do anything, in many cases nothing totally new comes out of it. But if you bring in restrictions and try to break through them, that’s where true creativity happens,” Imai notes. “The more unbounded you are, the slower the evolution process is. But if you’re too restricted, it basically becomes automation. So you just have to give the system enough freedom so that it can evolve and come up with something new, but not so much freedom that it’s totally unguided.”

The team isn’t too concerned about accidentally creating Skynet, Harvey jokes, as while the network is autonomous, it is also controllable. “At the end of the day, it’s guided by the human,” he reassures. “The management of the company is going to say my network should be optimized for the quality of service that my customers experience, or it should be optimized for power. The human gives the intent or the purpose that drives the actions or the mechanisms implemented within the autonomous network. The autonomous network finds a way to make it happen. It’s there to represent the how, but the what comes from the human.”

Why does it need to be autonomous?

So why are Imai, Harvey and their crew at Rakuten Mobile going to all this effort? What does the end user get out of it all?

“A cheaper, better network,” Imai argues. “Because you will have a system that can adapt to the needs of the users. One that’s tailored for you and for the application that you’re using without having to massively overprovision the network.”

Demands on mobile networks are becoming more and more diverse. Imai and Harvey believe that human engineers won’t have the capacity for manual operation in the future.

“The network is changing all the time,” he says. “Think about the iPhone; when it first came out, basically nobody was playing games on their phone or watching videos. Then all of a sudden, the market changed. 10 years ago, nobody was envisioning that the future of telecom networks would be that phones are basically becoming computers… A computer 20 years ago was the domain of nerds like Paul and I, but nowadays basically everybody has a computer in their hand.”

“Subscriber numbers are exploding — not because people are making more phone calls, but because of the massive increase in the number and types of devices connected, like the Internet of Things,” Harvey adds. “So not only do you have an explosion in the number of people making demands of your network, but also in the types and ways in which they use it.”

“Right now you always need to have an engineer who is adapting the network to the situation,” Imai continues. “There are more and more applications, more and more devices… you can’t just throw a million engineers at it. However, an autonomous network can adapt to much of this — and come up with the best solution on its own.”

“The human gives the intent or the purpose that drives the actions or the mechanisms implemented within the autonomous network. The autonomous network finds a way to make it happen.”

Paul Harvey, PhD, Mobile Edge Computing Research Lead at Rakuten Mobile Innovation Studio

Harvey concludes: “Autonomy, by having software that can control and respond at a speed that is much faster than a human being is capable of, means not only that we can have better networks, but that we can still have networks that can function and operate more cost effectively.”

Harvey sees this push for autonomy not only as necessary, but as an opportunity, made possible by the fully virtualized foundation Rakuten Mobile has already built. 

“This is the place where it can happen,” he says. “All these other companies around the world are still working on achieving full virtualization, but we already have it. That’s what we’re starting with. We’re starting with something fresh and clean. So rather than just sitting with our feet up and saying OK, let’s just automate everything and we’re done — we’re saying, well what’s next?”

Building for future generations

Harvey is looking forward to watching how the technology plays out.

“As we increase the level of autonomy and start handing off different operations to these different systems, I think it’s going to be really interesting to see what our networks look like,” he poses. “What do the artificial engineers look like? How do they organize themselves? How do they communicate with each other? How do different networks from different operators start interoperating, and what do these different autonomous entities have to say to each other?”

Imai is eager to build a system that can adapt to future technologies — unlike many of the protocols in use today that were originally created for networks of the 1970s.

“Do we want a system that can actually adapt to our needs? Or do we want a system that binds us down to what has been engineered into it long before anybody knew what the future was going to be like?” he poses. “I’m going with the former: A system that can adapt to what our needs in the future will be — not what somebody envisioned on a drawing board 30 years ago, and not something that’s set in stone that you have to work around.”

As for what those future needs might be?

“Honestly, 20 years ago, we wouldn’t have expected to be where we are now,” Imai says. “Go back 70 years and people thought that by now we would all have colonies on the moon, on Mars, flying around in rocket ships, robots taking care of everything we do. And yet nobody mentioned the internet.”

“Scientists are not the best futurists,” he continues. “We aren’t really the best people to look into the future. We’re just the ones who build it.”