What’s on the radar for the radar sector?

We’re hearing that machine learning is the next big thing in radar – what’s its importance?

“Machine learning is a technology that is very relevant to radar, in terms of not only adding new capability or functionality to a radar, but also in the wider manufacturing environment, or in the design lifecycle to optimise designs and minimise rework.

We are introducing AI into the way we build things in manufacturing, so we can improve the yield – so it’s not just about new features in a radar, it’s about how you go about manufacturing them. And we use that to optimise the selection and the design of the electronic components that go into a radar system. We’re investing a lot of money in machine learning, right across the board.”

In fact Machine Learning/AI in radar is one of the key themes being explored in this year’s conference. We’re having talks specific to the verification challenges; how do we ensure the programs are classifying the data correctly and what happens if it’s not? We will also have an entire track dedicated to automatic classification coming in from the technical papers, plus we’re running a machine learning/signal processing challenge at the conference, where teams will compete to accurately classify data.

Are there any challenges on the practicality of the machine learning side – is there enough training data, and if not, what is the bottleneck?

“Having substantial training data is always the problem. There are plenty of open-source datasets, classifying cars and buses etc, but military radars can be very different. There’s currently very little training data available, and it can be incredibly expensive to gather.

Access to enough data to train the machine learning algorithm is always going to be a problem. However, there’s ways around that in the sense that you can mix real data with some model data, or use techniques such as transfer learning. Then you come across the next hurdle. In order to be able to use the data for machine learning, it has to be properly labelled and indexed, so you know that you’ve got the right “truth” data associated with it.

So, having the right amount of data is a challenge, but also having correctly labelled and indexed data sets to train on is a challenge as well. Maybe a final thing is that some of these data sets can become highly classified, so how you handle and make them accessible can then be difficult.”

And on the other side of that question, are there safety implications of Machine Learning validation in radar?

“In terms of the safety, from our perspective, most of the radars we build are not what we call “safety involved”. They’re not flying the aircraft, but the pilot or operator is making decisions on the basis of the data they are given from the radar.

Our customers want to have confidence in what the radar’s telling them. If that includes an AI/ML component then we would call this “Trusted AI”. You may have a situation in which a component is learning as it’s being used – how do you have confidence that it’s still giving you a correct answer and hasn’t been confused or learned the “wrong” thing?

The topic of trusted AI and certification of products with AI in them is wider than radar, and still an immature field. I think what will end up happening is that you’ll end up doing some training off-stage, you’ll load an AI component into a radar and then it won’t be allowed to adapt thereafter. It’s something that we are observing, and I don’t think we’ve got to a complete closed answer on it yet.”

With the upcoming 6G communication technology it may be possible to apply the technology for joint communication and sensing. Is this something you are excited by?

“The majority of contemporary radar systems employ Active Electronically Scanned Antennas (AESA). Versions of this AESA technology is used in mobile communications, and increasingly there is a convergence in the underlying components and solutions adopted in both radar and communication domains. AESA systems have always been capable of acting as multi-function RF systems, combining traditional functions of detecting, tracking and mapping, with non-traditional radar functions including communications, electronic surveillance, and a whole bunch of different things.

If we look at 6G comms, they will likely use multi-channel AESA technology with agile waveform control, and be spread across a network of base stations, so in principle will be able to provide communication and sensing capabilities simultaneously. However, 6G will operate at very high RF frequencies (circa 100GHz and beyond) therefore will be relatively short range compared to airborne sensors operating in the 1-10GHz bands due to the increase atmospheric attenuation at these higher frequencies. This will be fine for local area surveillance but airborne systems will still be required to provide long range surveillance capability.”

What are the key technical challenges you’re likely to be facing over the next ten years or so?

“One of the key challenges is data. Modern radar systems are capable of generating huge volumes of data and with this comes the problem of how you process this and how you store it. How do you store, manage and process that data when you’re doing development and trials? It’s a challenge, but also an opportunity. We’re beginning to recognise that the data itself is a service that we can offer.

Another challenge for those who build military radars is the increasing need to have sovereign & secure supply chains for critical components and manufacturing processes. Examples include critical RF semiconductor devices which can reveal the underlying performance capabilities of a radar, as well as supply chain packaging technologies and processes that handle these components.

In order to protect the military capability delivered by these radar systems we are increasingly required to use UK supply chains which means that we need to work across industry and with UK government departments, UKRI & Catapults to foster and support UK SMEs and larger businesses that can provide these capabilities.”

Which new technologies are you really excited about?

“AI and machine learning is going to be a step change in the functions and capabilities we can offer.

Historically, radars would be lots of bespoke components all put together, and you end up with something that’s quite large, heavy, and expensive. Nowadays, the level of functionality that you can incorporated in commercial RF system-on-chip devices, that’s been driven by 5G and 6G mobile phones, means that you can get an incredible amount of functionality in a very small area.

That means we can have many more channels in a radar system, and you can squeeze more capability into a smaller space. That allows us to do lots of things e.g., MIMO (multiple input, multiple output) radar systems and multistatic operation. You can tie or chain radar systems together, and you can offer novel & differentiating functions. So, the emergence of these RF system-on-chip devices is something that we’re looking at right now. It’s going to provide a step change in the level of capability that you can get into radar systems.”

What motivated you to specialise in radar systems?

“I always had an interest in STEM subjects at school, more specifically maths, physics and computing.”

The interest in those subjects guided Iain towards a degree in Electrical and Electronic Engineering at Heriot-Watt University, where his curiosity in Mathematics confirmed that he was interested in the algorithmic side of engineering.

“That led on to a PhD, sponsored by Leonardo, in radar systems and looking at Space Time Adaptive Processing (STAP). It was a fairly seamless flow from my undergraduate interest in the algorithmic side of engineering, to my PhD at Edinburgh University. That was really how I gravitated towards radar systems. I’m now following that circle as CTO at Leonardo by sponsoring PhD students.”

As your PhD was sponsored by Leonardo and you’ve been with them for 23 years – how well do you think industry does in supporting and nurturing the next generation of engineers through their academic studies? What are the benefits for an organisation to sponsor PhD students with a view to recruiting them?

“Broadly, I think the industry invests significantly in supporting young people through their education, but that’s not to say we couldn’t be doing more. It’s very important to support the next generation of engineers, as ultimately, we’re only as good as the people we have in the company, and the kind of people we employ, so it’s necessary and vitally important that we continue this.”

Leonardo have a number of STEM ambassadors within the company that go out to local schools and show them how exciting a career in STEM can be. They host engagement days and Leonardo support a number of UK-wide initiatives as well as sponsoring dozens of PhD students right across the UK.

“But we need to do more. Right now, the UK has approximately 170,000 STEM related vacancies, so the competition for talent is fierce. I’m a living embodiment that engagement in the early stages is crucial. We’ve got between 20 and 30 PhD students across the UK that Leonardo sponsor and hopefully that number will only get bigger.”

Find out more