Note: On 17 – 18 November, our DVN conference will take place in Stuttgart and for the first time, host special sessions on dual use and on “The Road to Type Approval: Mastering End-to-End AI Systems”.
In the runup to the DVN conference, we are presenting Key ADAS/AV players , and sharing their results, perspectives in this field. The first in this series was our report on a preview drive with Wayve’s latest vehicle; now comes the FAU interview, and further instalments will follow every fortnight until the conference begins
First Interview: Prof. Dr. Norman Franchi / FAU
Interview by DVN’s Dr Jürgen Dickmann and Eric Amiot
DVN: Hello! Will you please tell us about yourself and your institute?
Norman Franchi: At FAU, I head the Institute for Smart Electronics and Systems. With up to 80 employees, it is one of the largest institutes in the Department of Electrical Engineering. We develop chips and individual components, especially in the analog, RF, and mixed-signal domains, while also working on electronic system design and the corresponding applications – from radar, communications, and integrated sensing to smart cities, automotive, connected robotics, smart medical devices, and industrial automation.
For years, the city of Erlangen has been regarded as one of Germany’s most innovative locations and is part of the Nuremberg Metropolitan Region, where Erlangen, Fürth, and Nuremberg form a strong network.
We are active in safety scanning for airports, but also in medical technology and automotive. A good example is our 60-GHz radar for vital-sign measurement. It can be integrated into seatbacks or seats, for example, to detect stressful situations in a vehicle. We now use the same technology in clinical settings: at Erlangen University Hospital, beds are equipped with our radar technology, which records vital data and transmits it directly into the infrastructure. We are also looking at similar approaches for in-cabin applications in aircraft. Overall, we work with radar at 24, 44, 60, 77, 110 to 170 GHz – the D-band – as well as at 320 GHz, and we are currently developing further toward 400 GHz+. That makes us one of the leading research groups in Germany. Our focus is on RF integrated circuits and systems. Together with colleagues in RF engineering, we are also working on antennas, increasingly using 3D printing, so that at very high frequencies they can move as close as possible to the chip.
DVN: Why are these very high frequencies now interesting? D-band at 120 GHz, for example, or 320 GHz in the G-band: where do you want to use this?
Dr. Franchi: These high frequencies are particularly interesting wherever we need very high resolution at short and medium range. In automotive, it is therefore less about maximum long-range performance and more about a precise safety shell around the vehicle. This allows contours to be captured much better – in a sense, a cost-effective step toward lidar-like resolution. Typically, we address ranges of about 5 to 25 metres. With massive MIMO, narrow beams, and targeted focusing, larger distances can also be covered. In addition, the higher the frequency, the shorter the wavelength, and the more precisely I can measure and resolve. This is relevant for contour detection as well as for vital parameters.
DVN: Joint Communications and Sensing (JC&S) and V2X communications for connected mobility and smart cities sound far in the future, no?
Dr. Franchi: Let’s start with Vehicle-to-X. For us, the focus is mainly on direct vehicle-to-vehicle communication – without going through a cellular base station. In 3GPP terminology, this is PC5 or direct communication. It becomes particularly relevant in platooning: radar always reacts only after changes have occurred, whereas wireless communication can share intentions such as braking or accelerating at an early stage. That is precisely what makes simultaneous reaction possible. Our research shows that this does not require extremely low latencies across the board. The decisive factor is to think about communication and control together – our communication-control co-design.
DVN: Is the communication channel then just normal wireless communication, or is it already moving toward communicating radar?
Dr. Franchi: Initially, it is a classic radio channel, though a very demanding one. The direct connection between vehicles is much more difficult than the connection to a base station because of dynamics, the metallic environment, and similar antenna heights. That is why we focus strongly on antenna positioning – from the roof and bumper to the mirrors – and on channel measurements in real driving. In parallel, we are working on Joint Communication and Sensing, internationally usually called Integrated Sensing and Communications (ISAC). The idea is to stop looking at radar and communications separately. In automotive, what is particularly exciting for us is: radar first, communications second. If I already have radar in the vehicle and only need a few Mbps for certain applications, an integrated solution can become very attractive from an architecture perspective.
DVN: Over the past 15 years, ad hoc networks and platooning have received substantial funding, but they have often failed due to liability, infrastructure, and business models. Has this blockade become any smaller?
Dr. Franchi: Fundamentally, I would say no. What has changed is primarily the technological direction. For a long time, WLAN-based approaches and cellular were in opposition to each other. China has clearly committed to cellular here – and that has given the development direction. But the real difference lies in the infrastructure: China understood 5G very early as a strategic digitization infrastructure and expanded it massively. With that kind of coverage, you can talk about guarantees, availability, and connected vehicle applications in a completely different way. That is precisely the starting point we still lack in Europe.
DVN: So is there a real danger that the Western focus on highly complex, self-contained individual vehicles will be technologically and economically devalued by the Chinese path of systemic connectivity?
Dr. Franchi: Yes, that danger exists. In my view, China is at least two steps ahead in smart cities: first in infrastructure coverage, and second in building urban sensor and monitoring infrastructure. As a result, large data spaces already exist there that can be condensed into a digital twin of the city. Europe still lacks this scaling. At the same time, with data protection and trustworthy technology, we would actually have a potential asset – we have just been playing it more as a brake than as a strategic advantage so far.
DVN: When do you think China could really implement such a smart-city model across the board?
Dr. Franchi: I am reluctant to name exact years, but I consider it realistic that China will move into broader implementation starting around 2028 and that between 2031 and 2032 the first model segments will emerge in which such a system largely works. The conditions there are simply more favourable than in many European brownfield cities because the infrastructure is newer and more plannable.
DVN: In the end, isn’t smart city mainly a robotaxi topic?
Dr. Franchi: No, robotaxis are only one building block. Smart city also includes energy supply, environmental monitoring, traffic management, and cooperation between the city and authorities. Especially in complex urban spaces, we will see that purely autonomous vehicles are not the sole solution. They need guidance from infrastructure – for example through cameras, sensors, and connected intersections. This connection between vehicle, communication, and urban infrastructure is the real core of the smart city.
DVN: And who will ultimately provide the vehicle in such a model, the state or an operator?
Dr. Franchi: Who will provide the vehicle later is still open, in my view. What is clear, however, is that from around 2028 we will see very affordable and at the same time technologically well-equipped Chinese vehicles in Europe. The key point is not only the price, but market penetration through connectivity, software, and data. That is exactly where a new technological dependency can arise for Europe if we do not invest consistently enough in future fields such as smart cities, AI, and data centres.
DVN: Let’s move to sensor technology. Where is the added value of THz radar compared with today’s systems?
Dr. Franchi: I see a bridge to robotics here. With humanoid and collaborative robots, the question is how the machine interacts with humans while the human is still directly involved in the process. Today, that is done almost exclusively through cameras and microphones; lidar is often too expensive. In the close range of five to ten meters, a robot needs a robust technology for pose, gesture, and intention recognition. And that is where radar comes in.
DVN: Are cameras alone not sufficient for that?
Dr. Franchi: No. Video data is extremely compute-intensive and requires high GPU performance, which increases energy consumption. In addition, optical systems reach their limits in metallic or highly reflective environments. Radar offers a leaner, more efficient alternative here. By using high frequencies with low transmit power, we create a local ‘radar shell’. This limits the reflection space and makes the system’s decision-making much more precise.
DVN: You mentioned five metres’ range around the vehicle or robot. Do you really see these THz applications coming outdoors?
Dr. Franchi: I believe the development will take place first in the vehicle interior and only then move to the exterior. As frequency increases, resolution increases massively, which allows us to capture gesture patterns much more finely than current systems. In the automotive sector, this enables contactless interaction that goes beyond conventional haptic controls.
DVN: What challenges arise in signal processing at these high data rates?
Dr. Franchi: The sheer volume of IQ data at wide bandwidths overwhelms classical processor architectures. We rely here on specialized RF frontends and, over the long term, are researching more energy-efficient architectures, possibly in the direction of neuromorphic computing, to convert IQ data into parallel spike representations.
DVN: Let’s talk about AI and approval readiness. How can a robust safety case for autonomous driving be justified to regulators when both the software stack (end-to-end learning) and the validation tools are increasingly based on AI black boxes? Can this ‘AI ping-pong’ between the driving function and generative simulation ever achieve the necessary traceability for public approval in the civilian sector?
Dr. Franchi: To be honest, I have not yet formed a robust opinion on that.My impression from automotive practice is this: the assess ability of such systems is still heavily shaped by manual processes. From my time in the industry, I know how radar and communications software is developed, tested, and validated. With AI-based systems, exactly this evidence process becomes significantly more difficult, both with regard to functional safety and with regard to the question of how to assess the quality of the system in a robust way at all.
What I can say, however, is this: we are working exactly on the simulation side of this problem, including with our startup FiveD. FiveD comes from the radar environment here at the Chair of High-Frequency Engineering. Our aim is to build a platform together; Rohde & Schwarz has now also joined. There, we are building a simulator that transfers what has so far been rather static ray tracing into very dynamic environments. In addition, multiple radars can be integrated into a scenario simultaneously.
DVN: Why not full-wave simulation? It is more realistic, especially when I see what, for example, the military can do with it.
Dr. Franchi: What has already been integrated there needs to be looked at in detail in each case, because such a startup naturally develops very dynamically in the early phase. Fundamentally, both automotive and defense partners contribute their concrete use cases: Which environment should be modeled, which objects, which interactions? The goal is to represent the most realistic human-environment models possible. In that respect, FiveD would certainly be an exciting conversation partner for you.
At its core, FiveD provides a software tool that can reproduce any number of indoor and outdoor scenarios in order to better understand the behavior of radar and communication – for example reflections, bounces, resolution, and interference. There, we work primarily on coherent radar, meaning on operating multiple radars within a vehicle, or prospectively even across vehicles, coherently and evaluating the data jointly. That is exactly what should also be represented in the simulator. For security scanners, for example, this approach is particularly valuable: rather than measuring an extensive range of configurations – clothing, posture, hidden objects – entirely in the real world, they are first generated in simulation, partly also with the help of generative AI. In effect, this creates digital twins, because real learning phases are extremely time-consuming.
At the same time, this also shows how difficult the final step remains: you can simulate and train a great deal, but the regulatory proof is not automatically solved by that. It reminds me of earlier projects, for example, involving taillights with countless software variants. There, the real problem was not the component itself, but the explosively growing state space. You had to decide which test cases were truly critical and which had to be prioritized. This is exactly where AI can help cluster states and filter out relevant cases. In the end, it nevertheless remains a matter of probability. And if the state spaces eventually become too large, it even becomes interesting to think about quantum computers or quantum sensing, because they can deal with electromagnetic fields physically much more precisely. That is not a final answer, but from my point of view the direction is clear.
DVN: Shifting now to dual use, where do you see the most concrete potentials between civilian mobility and defense technology?
Dr. Franchi: I see dual use in many places, also beyond the topics we have already touched on today. A ‘security valley’ will emerge here in the coming months. We already have a very strong base here, e.g., with Diehl Defence, Saab, Rheinmetall, and others. In the radar field, new startups are being added, other companies are currently expanding their focus from space into defense, and others are also building up their presence in the region. In parallel, we are also talking about a large computing data center intended specifically for this context.
This region has historically had a strong foothold in the security environment. Medav is an early example of a company here that worked in the field of audio and speech analysis for security agencies and was later acquired by Saab. A lot is being built on that foundation today. I am helping to build this Security Valley together with partners from across Bavaria, as well as key players from Germany and the European Union.
That is why, in November last year, we launched the Forum for Security and Innovation. There, we make our research and our startups in this field visible, deliberately in a closed format. In terms of content, we bring two worlds together: the traditional defense sector and the civilian field of critical infrastructure. That is exactly where I currently see the largest dual-use lever.
Much of what we develop in the drone sector, for example, is motivated by defense, but can be used just as well to protect chemical plants, energy grids, or other critical infrastructure. What is interesting is that the requirements of these infrastructure operators have often been formulated more clearly for years than those in the defense sector; they simply did not have the same lobby for a long time. We are now deliberately bringing the two together. Technologically, we position ourselves with radar, communication, and resilience – that is, with the question of how to make communication networks and overall systems robust.
One example is our Chaos 6G Lab, which is funded by the Federal Ministry of the Interior. There, we investigate vulnerabilities in telecommunication networks in a similarly systematic way as the Chaos Computer Club does for IT systems. All these activities are now feeding into the establishment of the Security Valley. The domains are very different, but much of what is being conceived today in a defense context is also highly relevant for automotive, medical technology, or automation.
DVN: Is that the same sensor technology?
Dr. Franchi: Partly yes, partly no. For example, if a sensor on a drone has to analyze very large data volumes, it needs enormous bandwidths – not just a few megahertz, but up to many gigahertz. That is why we contribute our expertise in mixed-signal chip design, ADCs, and special RF frontends. Some of these sensors operate with very high carrier frequencies and very wide bands, which quickly overwhelms classical processor architectures. Today, many tasks are still solved with GPUs, as in automotive. In the long term, however, we will probably need more energy-efficient, more strongly parallelized architectures – possibly moving toward neuromorphic computing. That raises the question of how analog data can be converted into digital parallel spike representations in the first place. That is exactly what we are currently researching.
DVN: We have planned a Dual Use session, Smart Defence, at our upcoming automotive conference.
Dr. Franchi: In my view, that makes a great deal of sense. A growing number of automotive companies are now getting involved in the defense sector, highlighting just how important the concept of dual-use has become. In defense, requirements are initially higher in many areas, but once a technology has been developed, the question immediately arises of how it can be transferred to automotive or other civilian applications. It would be economically unreasonable to think about such developments only for one side.
DVN: Norman, thank you for this insightful view into your world of joint communication and sensing, and for your multifaceted insights into smart cities and dual use in defense. I hope we can welcome you at the DVN conference.
Further links to Norman Franchi and his activities:
