Skip to content

Spotlight on Young Researchers: Computational radars & the quest for high quality data from low-complexity measurements

Computational sensing is everywhere – for example, radar. The use of radar is still expanding – it is for example, essential for self-driving cars. Radars only work thanks to a delicate balance between the sensor and the software. Researchers are working to improve how radars work and how they can be used, for example, by studying the acquisition and reconstruction of radar signals for different applications.

Computational sensing leverages the interplay between sensors and their associated algorithms to generate higher-quality results. Imaging applications such as cameras are well-known examples of the power of computational sensing. Through dedicated filters that consider the lens, and other parameters of the camera, better-quality images are generated (i.e., computed).

Other examples of computational systems are MRI machines, LIDARs, microscopes and, where most of my research has been focused on up to now, RADARS. These applications are ubiquitous and thus impact our daily lives in many aspects.

For cameras, computational Sensing works through dedicated algorithms (e.g., filters) that consider the sensors properties (for examples the lens and other parameters of the camera), and so generates – computes – better quality images. MRI machines, LIDARS, microscopes and radars are other known examples that rely on computational sensing.

Radar systems: From live traffic to the mapping of earthquakes

Radar systems allow the estimation of specific information about objects, such as live maps showing traffic. They also help monitor the environment: thanks to the processing technology known as ‘Synthetic Aperture Radar’, high-quality radar images of the earth can be captured from space and used for flood monitoring and other geo-sensing applications, such as the mapping of earthquakes.

Radar is also traditionally used for tracking objects in the sky but has also become a vital tool for self-driving cars: new advancements mean smaller sensors can now have made it possible to use radar for contactless vital sign monitoring. Regardless of what radars are deployed for, they share a delicate balance between the radio frequency system– the sensor – and the software – the algorithm.

Combining hardware (sensor) and software (algorithm) is always challenging, as each side has to keep up with the other. This complex but rewarding research is also inherently interdisciplinary, which requires the input of a broad range of expertise and scientists.
Thomas Feuillen Research Associate in the field of Computational Sensing (SnT, University of Luxembourg)

Ever-evolving systems

As computational sensing leverages the interplay between the system and the processing, advancing one of the two can help advance the other.

“In this sense, these systems are ever-evolving. One aspect that is a challenge for many applications is the efficient acquisition of data.”

Researchers are exploring using easy or cheap acquisition platform that acquire low-complexity measurements to generate high-quality data, thanks to dedicated algorithms. For example, this would entail measuring signals in a binary fashion (with only 1s and 0s/ black or white) but still recovering the information of interest.

Novel type of computational radar prototype developed

My work focuses on the study of the acquisition and reconstruction of radar signals for different applications. My PhD, for example, studied how really coarse acquisition using cheap 1-bit ADC can be slightly modified (by adding a small random component) to allow for higher quality reconstruction matching normal and costly setups.
Thomas Feuillen Research Associate in the field of Computational Sensing (SnT, University of Luxembourg)
“More recently, my focus has shifted to a new type of non-linear acquisition that tries to remove limitations on the dynamic range of signals, called Unlimited Sampling. My work focuses on bringing these new mathematical concepts to fully practical demonstrators to highlight their feasibility and efficiency by addressing real-world problems.”
“Since starting here in the SPARC research team, and in collaboration with the team of Ayush Bhandari in Imperial College London in the UK, we realised the world’s first computational radar prototype that leverages this Unlimited Sampling theory to remove common limitations on dynamic range often encountered in radar systems.”

Dr Thomas Feuillen is a Research Associate in the SPARC group, headed by prof. Bhavani Shankar MRR, at the SnT at the University of Luxembourg. He is currently working on the FNR funded CORE project SENCOM, and secured his own CORE project in the 2023 call, in which he will be Junior Principal Investigator (PI).


Describing his research in one sentence

“Translating and studying complex mathematical concepts into innovative radar demonstrators where the acquisition and the processing work together.”

Why he chose this research field

“I like to do research where the theory that is developed is then used on real systems. For this, radar is the perfect candidate as its associated mathematical model is somewhat pure and can then be easily combined with new concepts and algorithms. I chose Luxembourg and specifically the SPARC research group in SnT because it is the perfect balance between theoretical and applied research in a dynamic country.”

What he loves about research

“I like to understand the underlying mathematical model of phenomena and see their effects on the world. Most of signal processing theory and applications are not known to the public, but without its fundamental results, our modern world wouldn’t exist as we know it. Spectral analysis through the Fourier transform, for example, is the basis that underpins telecommunications, modern photography, music, and for me, radar systems.”

Mentors with an impact

“During my PhD, my supervisor Laurent Jacques was the first one to guide me on the path to this discipline by helping me combine my practical knowledge of radar systems with new and challenging ways of acquiring data. I also had the opportunity to work for and alongside Dr Petros Boufounos, who heads the computational sensing research teams in the Mitsubishi Electric Research Lab in Boston USA, which also cemented my idea to continue my work in this area.”

Why Luxembourg?

“The research ecosystem in Luxembourg is, compared to other countries in Europe, continuously growing, and new opportunities for collaboration and private-public partnership are always appearing. This, combined with the numerous opportunities given to young researchers at the University of Luxembourg, made it a no-brainer for me to join. What I also deeply enjoy about working at SnT (or any other research entity in Luxembourg for that matter), is that it is truly multicultural with no clear majority language or culture. This, for me, fosters a real open research and learning environment in which I love working.”

Where he sees himself in 5 years

“As a PI on new exciting projects on computational sensing. Although I like working with radar sensors, I would appreciate branching out to other modalities and new applications like biomedical imaging, e.g. I find the domain of vital sign monitoring of infants using radar very inspiring.”

Related contacts

Emily Iversen

Digital Communication Manager