Nokia Bell Labs’ Peter Vetter talks 6G research

November 22, 2025

2 Views 0

SaveSavedRemoved 0

At the 2025 NYU Brooklyn 6G Summit, we chatted with Peter Vetter, Nokia Bell Labs President of Core Research about the technologies heading for 6G and about how researchers work at the lab.

Brooklyn, NY, November 6, 2025 — For 100 years, Bell Labs research has changed the world. Originally AT&T, then Lucent and now Nokia Bell Labs the facility was the birthplace of the transistor, information theory, and Unix, among many others.

Today, researchers at Nokia Bell Labs continue to break new ground as they look into ways to improve wireless and wired communications from the radio to the core network. Leading a team of some 450 researchers is Peter Vetter, Nokia Bell Labs President of Core Research. EE World spoke with Mr. Vetter at the 2025 Brooklyn 6G Summit on the campus of the NYU Tandon School of Engineering. Research today focuses on integrated sensing and communication, and of course, AI and how it could fit into 6G networks.

An edited transcript follows the video.

EE World: Martin Rowe, Senior Technical Editor with the E world. I am here in Brooklyn, New York, at the Brooklyn 6G Summit. This is the NYU Tandon School of Engineering, and I have the pleasure of meeting with Peter Vetter, who is the President of Core Research at Nokia Bell Labs. Peter, thank you so much for taking the time to meet with me. I just want to have a few questions for you while we have a little time.

The first thing is, Bell Labs is famous. Everybody knows Bell Labs. What are some of the projects that you’re working on now, or that your people are working on now that you think will relate to 6G and I realize that there may be some things that you’re looking at that are probably beyond 6G. What kind of projects are working on now at Bell Labs related to 6G?

Vetter: There are three main reasons why you want a new generation. One is higher capacity. Every generation, the capacity goes up by a factor of ten. We work on technologies that can improve capacity. Typically, a new generation is a new technology that you want to leverage and what is that new technology? AI. So, we work on what AI means for 6G and future applications that can leverage 6G. A third is typically there’s also a new capability, and for 6G that new capability that we’re investigating is integrated sensing and communication. Beyond that, there’s also work on making the network secure, including quantum security. Everybody knows that a future quantum computer can hack security code that is deemed safe today. So how do you deal with that in the 6G era when we expect a quantum computer to become a reality?

EE World: You’ve given me several questions already, but let’s go back to capacity. What sort of things do you see coming along with capacity? For example, new radio, some new modulation, perhaps, or it doesn’t even have to be in the radio. What if it’s in the network, beyond the radio, beyond the tower?

Vetter: What we see is every generation, capacity goes up by a factor of ten, sometimes a bit slower, sometimes it’s faster. It depends on suddenly a new service type comes up. The expectation is that AI, the use of AI, and the massive scale deployment of sensors will drive capacity mainly in the uplink. So how do you support that? You need more spectrum, new spectrum, and you want to support that from existing cell sites. Lesson learned in that we thought of millimeter wave that it has limited reach, every 300 meters, you would have a small cell. But the reality is, it didn’t happen. It’s not commercially viable to deploy from existing cell sites. We look now for 6G frequency ranges just above the current mid-band, 7 GHz, which is about double of the three and a half that is used today.

To maintain approximately the same capacity, or nearly the same capacity, you need larger antenna arrays. In massive MIMO, you form beams, but you can find the beams and the radio energy more focused on the end devices.

“You want capacity to go up by 10x, but not the power consumption.”

Now the challenge with that higher capacity is you want capacity to go up by 10x, but not the power consumption. We want the power consumption to stay at least flat, possibly even in absolute numbers, reduced by 50% so the energy efficiency of our systems needs to go up by a factor of 20 or more. So the research is then also, what can we do is clever tricks to achieve that capacity, in massive MIMO, using hybrid beamforming, using such things as meta services, clever compute architectures, clever algorithms to get to the higher capacity, but without the power consumption and the cost going up?

EE World: Speaking of power consumption, do you look at things at the circuit board level or more at the network level?

Vetter: We look at all aspects. The strength of Bell Labs, multi-disciplinary research to solve the problems, but at the same time, we, of course, also leverage what’s already out there, because the type of ASIC powering are problems that are being solved by others in the industry, so that we focus on those things where we can make the difference in network systems. We look at what we can do at the physical layer, the processor architecture, novel concepts, also inside the processor architecture, like computing memory and aspects like that.

Then we look at the algorithmic level, at such things as antenna muting, power back-off if you need lower power. We look at all these aspects, and then also in the network itself. It’s not only the power consumption of the radio. It’s also the power consumption of the fiber network that interconnects these radios with the data centers and connects the data centers with each other and inside the data centers in those parts of the network. Due to the rise of AI, we expect, even in one decade, an increase of 100x in capacity need. Again, nothing to sneeze at, right? How do you solve that? Don’t let power consumption go up in the same way. These are all challenges that we’re looking at in research.

EE World:You mentioned AI, and even here at this conference, if there are three of the most important things that I have heard, four of them are AI. How are you and Bell Labs using AI in terms of your research and in terms of researching AI itself?

Vetter: So there’s most multiple facets. Again, we look at how can AI be used for the network, and the most obvious one is the autonomous service generation, that you automatically set up services. But it goes beyond that. We see what we call the self-evolving network, where network functions are auto generated. I mean software generation in the application field is already a table steak. We are evaluating, researching, what that means for a network, self-evolving network functions, and how they are finding each other in an agentic, dynamic architecture.

We also look at the physical layer, what AI can mean for a radio, for instance, where the base station and the handset learn the channel. Do what we call channel estimation, automatically find the proper beam setting, find the proper equalization, new waveforms, and we’ve done actually already. A few years ago was the world’s first demonstration of such an AI radio access network. Now this is becoming part of the 6G standard discussions. Then, it is also research on AI and future applications, how we can use AI to well. Long before ChatGPT became a thing, we had our own research on Nokia large language models where we trained the system specifically for network questions. We trained with specific standards and specifications of products. We’re now going into a phase where we do large world models, where AI becomes physical, where AI enriches digital twin models of the physical world to understand what is happening in the physical world and then help with automating actions in industrial environments or even in your private home environment.

EE World: What about AI in the radio access network? Are you looking at that as well?

Vetter: So we’re looking at it in different ways. AI for the RAN is using AI processing for the AI radio air interface. For that, you need processors like GPUs, very energy efficient processors, and you’re actually seeing a shift of the type of processors that are being embedded in the radio access network. We are also looking at, can you leverage these GPUs for other applications, since they’re already there? In addition, how do you make that easy so that other parties can use automated smart contracts? Use those capabilities wherever that GPU programming capability is embedded, is it in the RAN or in the core, or maybe also other computes?

EE World: 6G seems to be moving, in a way, I’ll say, away from the consumer smartphone, not leaving it behind or anything, but the growth seems to be “what else can we do with it?” You mentioned something about sensing, and that seems to be one of the themes that we’ve seen today here in Brooklyn. Can you tell us a little bit about what kind of work you’re doing regarding using the cellular network as a sensing device, as a sensing system?

Vetter: We’re not reinventing the radar, but it’s a principle of of radar, using the typical 3GPP waveforms and signals. We are at an interesting time to indeed consider base stations as a potential sensor, because these base stations are large-scale antenna arrays, massive MIMO, which are used to create beams to your handset. You can then create beams to sweep an environment and get a sense of angle of arrival, where is an object, at which angle? The larger the number of antenna elements, the more accurate you can do that. The other interesting thing is that you have a very high bandwidth, 100 MHz, 200 MHz. So that means you can detect very fast signal changes. The back reflected signal you can measure with a fairly high accuracy the time of arrival of that back reflected signal. You get a good range estimation. We’ve shown that you can actually localize people using a base station, massive MIMO array with an accuracy of a meter.

EE World: Do you see these? This ability to, as you said, detect a person or an object? What kind of use cases do you see for that?

Vetter: One use case is detecting people is helping to see around the corner for, for instance, safety so you detect and you can say to a driverless car or to an active driver, “send an alert, there is somebody around the corner, slow down.”

It gets, admittedly, a bit tricky if there’s multiple people, what our current systems cannot discern are individuals because there’s too much clutter. But an interesting use case that we get a lot of interest for is detection of drones. That’s because in the sky, there’s no clutter. There are not many drones flying but in present day, these drones fly literally under the radar system of avionic radars. Those base stations on rooftops in urban environments, they’re perfectly positioned to pick up and see these, these drones, and we have actually a proof of concept that have shown that at a distance of 150 meters and more, you can detect a small drone that is invisible with a camera — less than a pixel — but it picked up the radius back reflection of that drone. So this sensing is going to work.

EE World: I’ll say at the same time the network is also doing the communications. How is that, I’ll say, multiplexed? How does that work?

“We need to make sure that the quality of experience of the communication is not jeopardized by that sensing capability.”

Vetter: We need to make sure that the quality of experience of the communication is not jeopardized by that sensing capability. We have come up with clever algorithms that do the multiplexing of sensing an object of interest and at the same time providing the connectivity. These algorithms to optimize that we have in place. At the moment, most of our research is using a single antenna system, but you can actually then also do what is called bistatic use, a different antenna system, one that is used for communication, and then the other picks up that signal. Through central processing, you get even a better understanding of what is happening and what you’re observing.

EE World: Interesting. I saw, a demonstration here that talked about using time-domain multiplexing where every so number of (I’m not sure if I’m getting this), whether it’s packets or symbols or there was some time slicing where once every six time slots the network would switch over to doing the sensing. It would not do it every sixth or seventh time slot. But sometimes it would skip one or move to another one. Is that something that you’re working on here?

Vetter: I know that’s something that’s being done here at NYU so I’ll generalize it. It’s indeed how do you multiplex
the communication slots in one direction while in multi-user massive MIMO, we can even imagine multiple beams at the same time and another beam that is following a device of interest for sensing. So this kind of algorithms we have, and you described one example of an algorithm where you do time-division multiplexing, but with massive MIMO, we could go also in a direction where you do space multiplexing.

EE World: Let me switch a little bit and talk about the workings of Bell Labs. Now you’ve got many projects going on.How many do you have going on at the current time?

Vetter: It depends on what you call a project, right? I mean, we go in a typical life cycle. We start on average, ten years ahead of the cycle. And these early exploration projects are typically one or two people that work out an idea. They have identified the problem, we always start with, “what is the hard problem the real world needs, and what are possible ideas to solve for it that do it differently and better than how it’s done today?”

These are the feasibility studies, where we do a bit of simulation and maybe a small experimental breadboard. When we are onto a path that is promising, we invest more people, then the projects get larger research projects of five to ten people, even where we connect the dots of some concepts and make more of the sum of the parts. Then we get into a phase where, okay, now we get it onto a roadmap. So why I’m saying this? I give you a rough idea of the number of people that work on a similar topic. I’d roughly say we have between 20 and 30 different research projects going on [in Bell Labs Core Research]. Then the feasibility studies. There’s even more if you count these small conceptual studies.

EE World: How do you decide which people should work on a particular project, and given that they might be working on something else as well? How do you arrange that? How do you put a team together?

Vetter: So in Bell Labs, we have very much a bottoms up research culture where we set the large strategic direction. We expose our researchers to what the problems are.

We formulate a vision, but then let researchers themselves come up with ideas, new concepts, and we make sure that they find each other through readouts, through coffee-corner talks, through virtual coffee corner talks online so that they themselves identify “I have this problem. Can you help me with this?” And that they find each other, and then, as research managers, we give them the room to okay, we sometimes nudge them a bit to collaborate, but we maintain that because that’s where the energy comes from, right? “Hey, this is a cool problem. I want to work with you on this.”

EE World: Okay, one last question. It relates to this, and you’ve answered it to some extent. But how do you manage an organization? First of all, how many people do you manage? Let’s start with that.

Vetter: That is the easier question. I’m managing around 450 people [in Bell Labs Core Research] at the moment.

EE World: That was my next question. You’ve got some very smart people there. How do you do that?

“The freedom of research is very important to the creative idea generation.”

Vetter: You don’t. The freedom of research is very important to the creative idea generation. If you start managing researchers too much, you get what you ask for, but you don’t get the spark of something groundbreaking. It is more create the environment. where researchers can flourish. Make sure they understand the overall vision, the direction that they get exposure to the problems, but don’t dictate what they need to work on.

EE World: Peter Vetter, president of core research at Bell Labs, thank you so much for taking the time to meet with EE World. For EE World, I’m Martin Rowe, thank you for watching.

Filed Under: 5G, 5G antennas, 6G, AI/ML, Featured, MIMO, mmWave, Radar, Sub-6 Ghz RF, Video
Tagged With: 6G, BellLabs, communications, sensing