IMPROVED DETECTION OF HYPERSONIC

THREATS WITH RADAR USING IRREGULAR

WAVEFORMS AND ADVANCED PROCESSING
Pepijn Cox1, Keith Klein1, Mario Coutiño1, and Laura Anitori1.

Abstract

Hypersonic weapons can pose a significant threat to the international security, as the characteristics in terms of
speed, cruise altitude, and manoeuvrability differ significantly from other threats. This implies that the engagement
timelines become extremely short and, hence, the detection, tracking, classification, and identification should be
accomplished at large distances. To fulfil these tasks, current radar sensors will be pushed to or beyond their current
limits. A potential solution to overcome certain limitations of modern radar systems is by using novel waveforms and
advanced signal processing. Hence, the goal of this work is to demonstrate the potential of irregular waveforms and
advanced processing for the detection of hypersonic threats. It is shown that their combination can significantly
increase the detection performance and the measurement accuracy compared to multiple, medium pulse repetition
frequency waveforms with linear signal processing.

Keywords

Advanced radar signal processing, Irregular waveforms, Hypersonic threats, Radar systems.

1. INTRODUCTION
The introduction of hypersonic weapons to the battlefield will be disruptive. High velocities, manoeuvrability, and
relatively low cruising altitudes of these threats makes effective engagement difficult, as the first generation
hypersonic missiles have unfortunately shown in the Ukraine [1]. The threat characteristics of the hypersonic
weapons are significantly different than that of ballistic and cruise missiles and novel solutions are needed for
successful interception [2], [3]. The complexity of the problem suggests that a sensor network could be an effective
solution for a successful kill-chain.

That said, the required performance for the tasks assigned to each sensor within such a network might be pushed
to or beyond the current limits. The engagement timelines become extremely short due to the hypersonic speed,
cruise altitude, and threats’ manoeuvrability requiring that the detection, tracking, classification, and identification
are accomplished at large distances to have sufficient time for target engagement, where several tasks are
primarily fulfilled by radar systems.

Detection by radar systems depends on the system parameters and on the radar cross section (RCS) of the target,
where a small value implies that it is more difficulty to observe. Currently, there are several studies in e.g., EDA and
NATO to understand the phenomenology of hypersonic threats [3], [4]. Plasma effects surrounding the hypersonic
vehicle makes the to-be-expected RCS value highly uncertain and it is possibly subjected to significant variations.

Modern surveillance radar systems employ multiple, medium pulse repetition frequency (PRF) type of waveforms
with linear signal processing on receive for the detection of typical current targets. The waveforms commonly
consist of several bursts, each burst containing multiple identical pulses at a constant carrier frequency and
PRF. To obtain unambiguous estimates of the target’s range and velocity from the measurements, the concept of
staggered PRF waveforms is often used. This concept results in longer transmission times and processing losses,

1 Radar Technology Department, TNO, The Hague, The Netherlands {pepijn.cox;keith.klein;mario.coutinominguez;laura.

anitoro}@tno.nl

140
 EDA RESEARCH, TECHNOLOGY, AND INNOVATION PAPERS AWARD 2023

but the signal processing is rather straightforward and suited for real-time implementation. For the surveillance of
hypersonic threats, the staggered PRF waveforms and linear processing are pushed to their practical limits [5] in
terms of detection performance and in terms of unambiguously estimating both the range and the radial velocity
parameters of the target.

To summarize, the hypersonic threats pose challenges to individual radar systems in a network as: 1) the threat
needs to be detected at a large range leading to a low signal-to-noise-ratio scenario; and 2) the staggered PRF
waveforms with linear processing experience practical limitations leading to significant detection performance
degradation.

Recent developments of commercial high performance computing using graphical processing units increases the
processing capabilities far beyond the requirements of linear signal processing used in many radar systems. It
potentially allows for the use of waveforms consisting of irregular intervals and/or irregular modulations combined
with advanced, and possibly non-linear and iterative, signal processing techniques. This combination will [6], [7]:
a) not experience ambiguities in the range and the radial velocity measurements, b) significantly decrease the
processing losses currently experienced by linear processing, c) increase the accuracy of the range and velocity
measurements, and d) decrease the susceptibility against deceptive electronic counter measures. Figure 1
provides an artist’s impression of naval radar employing irregular waveforms and advanced processing using high
performance computing. The potential of these novel concepts for radar is currently also explored within a DARPA
program [8].

The goal of this work is to demonstrate the potential of irregular waveforms and advanced processing for the
detection of hypersonic threats. In particular, it shows that the combination of irregular waveforms and advanced
processing offers a potential solution for unambiguous range and velocity measurements while simultaneously
increasing the detection range. In addition, this work also discusses our developments in 1) decreasing the
computational load of the proposed advanced algorithms and 2) effective suppression of clutter when using
irregular waveforms.

The paper is organized as follows. Irregular waveforms and advanced processing concepts for radar is given in
Section 2. In Section 3, the performance of this solution for the hypersonic threat detection is given. The usage of
irregular waveforms and advanced processing in the civil domain is briefly discussed in Section 4 followed by the
conclusions in Section 5.

Figure 1. Artist’s impression of naval radar employing irregular waveforms and advanced processing against hypersonic threats.

2. IRREGULAR PULSED WAVEFORMS AND ADVANCED PROCESSING

The emerging threat of hypersonic weapons imposes the need to expand the maximum detection range and
the capabilities of unambiguously measuring the targets’ ranges and velocities. To understand the limitations of
systems employing staggered PRF waveforms with linear processing, we will discuss them briefly. Staggered PRF
waveforms consist of a number of bursts where the i-th burst contains multiple identical pulses with a constant PRFi
and a constant carrier frequency fi. The regularity in each burst leads to ambiguities in the range and the velocity

141
measurements of the target. As the required range and velocity spans far exceed the requirements for surveillance
of hypersonic threats, multiple bursts at different PRFs and carrier frequencies are transmitted and processed,
with varying unambiguous ranges and velocities, see blue and orange areas in Figure 2.(left). A target appears for
each burst at a different location, see blue crosses and orange plusses. Unfolding the measurements of each burst
beyond this ambiguous domain and then overlaying the measurements of the different bursts will reveal the true
location of the target, see green circle. In practice, the unfolding capabilities are limited [5], e.g., due to noise, blind
ranges, and clutter filtering, which significantly decreases the detection performance.

Introducing irregularities in the pulse interval and/or pulse modulation is a way to resolve the range and the velocity
ambiguities [7]. Absence of ambiguities removes the need for multiple incoherent bursts and the transmission time
could be used for a single coherent waveform that significantly increases the detection performance2. On the other
hand, processing irregular waveforms requires 1) advanced match filtering schemes to produce a high quality range
and velocity image without significant losses and 2) the target detection requires iterative optimization techniques
to handle the higher sidelobe levels of irregular waveforms. With increased sidelobes, strong targets can mask weak
targets. For example, an airliner or sea clutter in the beam might mask a hypersonic target. To mitigate both issue,
advanced processing plays a crucial role.

In this section, we highlight our recent developments [7], [9]-[14] in the design and processing of irregular
waveforms, show how the extra degrees of freedom can be exploited, and how clutter can be supressed in an
efficient way. In this section, we focus our discussion on irregular pulse intervals, but the techniques can also be
extended to waveforms with irregular pulse modulation.

Figure 2. Illustration of the unfolding of a staggered waveform employing two bursts (left) and (right) an illustration of amplitude over time
of an irregular pulse interval waveform.

2.1. WAVEFORM WITH IRREGULAR INTERVALS

To suppress periodic ambiguities in both range and velocity, irregularity can be introduced in the pulse interval
(PI), e.g., by selecting the pulse interval of sequential transmitted pulses randomly on a uniform interval between
a minimum and maximum value, see Figure 2.(right). In Figure 3, two ambiguity functions are shown for a) a linear
frequency modulated (LFM) waveform with a regular PRF of 2 kHz at 1.2 GHz, and b) an LFM waveform with random
PIs. Clearly, in Figure 3.(left), the range ambiguities at multiples of 75 km and the velocity ambiguities at multiples
of 250 m/s are visible for the regular PRF waveform as peaks with almost equal amplitude to the main peak in the
origin. The sidelobes, i.e., contributions in regions outside the main peak, are significantly lower for the regular PRF
waveform compared to the irregular PI waveform, in both the range and radial velocity domain. However, removing
the periodic ambiguities of the regular PRF waveform by using random PI waveforms come at the cost of increased
sidelobes.

The ambiguity function represents the signal contribution of a single object at zero range and with zero radial

2 The detection performance will increase under the assumption that the target remains coherent during the coherent
integration time.

142
 EDA RESEARCH, TECHNOLOGY, AND INNOVATION PAPERS AWARD 2023

velocity. A complex radar scene is a superposition of amplitude scaled ambiguity functions of all targets and clutter,
where the peak at (0,0) is moved to the range and velocity of the targets and clutter. Detecting these peaks allows
to find objects in the scene. However, the strong sidelobes of slow moving clutter nearby can mask a fast moving
hypersonic far away.

Figure 3. Ambiguity function of (left) a regular LFM waveform with PRF of 2 kHz at 1.2 GHz and (right) LFM waveform with random PIs on
[0.25, 0.75] ms interval.

The sidelobes of the waveform can be modified and lowered in particular regions by optimizing the PI for particular
scenarios. For example, lowering sidelobes of a clutter in the range-velocity region where the hypersonic target
should be initially detected. See [9]-[11] for our current developments in waveform optimization.

2.3. CLUTTER SUPPRESSION

The detection of hypersonic threats at a large distance by radar systems close to the earth’s surface implies small
grazing angles, i.e., observations close to the horizon. This will introduce strong surface clutter into the observation.
However, efficient mitigation of clutter for irregular waveforms is an open topic in the literature.

To mitigate the clutter in the received signal for irregular waveforms, we propose a modified filtering approach.
Similar to standard moving target indication (MTI), our irregular MTI method exploits the fact that most clutter is found
at close range and has a small velocity component [12]. When the irregular waveform has pulses with equivalent
modulation, a simple and computationally efficient clutter filter can be implemented, see [12] for the details. The
effectiveness of our irregular MTI method is highlighted in the next section.

The irregular MTI clutter filter is inadequate for filtering of waveforms with irregular modulation. We recently
introduced a clutter filtering technique at a slight increased computational cost [13] to handle waveforms with both
irregular pulse intervals and irregular modulation.

2.4. TARGET ESTIMATION WITH IRREGULAR WAVEFORMS

Irregular waveforms require advanced processing to 1) avoid large matched filtering losses due to the Doppler
sensitivity of the waveform and 2) to handle the high sidelobe levels.

To generate an image of the scene similar to Figure 3 used for target detection, matched filtering (MF) is applied.
Linear processing usually employs a one dimensional (1D) range MF and a separate 1D Doppler filter bank under the
assumption that range and velocity processing can be performed independently. However, for irregular waveforms,
independent processing leads to significant MF losses. To mitigate the matched filtering losses, in this paper, a two
dimensional (2D) MF is applied, where a 2D MF accounts for all velocity shifts of interest for each pulse delay [14].
However, the 2D MF significantly increases the computational load compared to independent linear 1D processing.

143
Recently, we proposed an approximate 2D MF based on subpulse processing [12] that reduces the computational
overhead and that can be efficiently computed by hardware accelerated FFTs.

To mitigate the high sidelobe levels of irregular waveforms, iterative processing techniques can be applied. As the
radar only detects a couple of objects in the scene, the solution to our detection problem is sparse and, therefore,
sparse optimization is applied in this paper. Many solvers exist to solve a sparse optimization problem, e.g., see [15],
in our example the non-linear, iterative orthogonal matching pursuit (OMP) [14] is applied in Section 3.

3. SIMULATION STUDY OF THE HYPERSONIC THREAT DETECTION

In this section, the performance of the proposed irregular waveform and advanced, iterative 2D signal processing
is compared to that of a typical staggered, medium PRF waveform with linear 1D processing for the detection of a
hypersonic threat.

3.1. HYPERSONIC SURVEILLANCE SCENARIO DESCRIPTION

We consider the case of a single hypersonic threat modelled with a Swerling case I fluctuation model. The radial
velocity can range between Mach 2 and Mach 18 and the threat is placed at a range between 300 km and 700 km
from the radar system. The simulation includes sea surface clutter with sea state 3 modelled by the NRL model [16].
The clutter horizon is 12 km and the velocity spectrum is taken as Gaussian with 0.72 m/s standard deviation. For
the sea clutter generation, it is assumed that the radar beam is fixed at an elevation of 3°. The target range and
velocity are uniformly drawn from the above given brackets and only one target is present in every Monte Carlo run.
For every simulation of the 3000 Monte Carlo runs, a new realization of the thermal noise, of the clutter, and of the
target is generated on which the staggered PRF waveform with 1D processing and the irregular PI waveform with 2D
processing are evaluated.

The staggered medium-PRF waveform consists of four bursts where the carrier frequency, number of pulses, and
PRF are changed per burst. The pairs are chosen (1.23 GHz, 13, 1600 Hz), (1.20 GHz, 15, 1800 Hz), (1.18 GHz, 17, 2000 Hz),
(1.21 GHz, 19, 2300 Hz), respectively. All pulses have an LFM modulation with a pulse length of 50 µs. For processing, a
three-pulse MTI filter is used to mitigate clutter, a Hamming window for pulse compression is applied and for Doppler
filtering a Hanning window is used. For the detector, the individual bursts are processed with the cell-averaging
constant false alarm detector and the individual burst detections are then combined using an 2-out-of-4 detector.

The irregular waveform consists of irregular PIs with equal modulation for each pulse within the burst. The waveform
is composed of 64 pulses with an LFM modulation, a pulse duration of 50 µs at 1.2 GHz centre frequency. Each PI is
chosen randomly between [0.25, 0.75] ms and the initial phase of each pulse is chosen randomly. For processing of
the irregular waveform, the irregular MTI clutter filtering is applied and the detection is performed by OMP combined
with 2D matched filtering. Note that both waveforms have equal energy on target, i.e., equal number of pulses and
pulse length, and they have both a dwell time of 32 ms.

The detection threshold is chosen for a probability of false alarms of 10-8. To account for migration artifacts, a
detection is considered to be correct when the range and radial velocity are estimated within ±150 m and ±75 m/s,
respectively, of the true target range and radial velocity.

3.2. SIMULATION RESULTS

The simulation results are presented and discussed in this subsection. The percentage of detected targets for the
staggered medium-PRF waveform with linear processing is 11.8% compared to 73.5% for the irregular PI waveform
with 2D matched filtering for 3000 Monte Carlo runs. The detection performance of the staggered medium-PRF
waveform with linear processing is significantly less than the irregular PI waveforms with advanced processing.
The performance degradation is partially caused by the three-pulse MTI clutter filter. Yet, a significant part of the
performance degradation is due to the practical limitations of the staggered medium-PRF waveform with linear
processing [5]. This part of the performance degradation may be solved by increasing transmit power, i.e., deploying

144
 EDA RESEARCH, TECHNOLOGY, AND INNOVATION PAPERS AWARD 2023

a larger-sized system, however, the irregular waveforms with advanced processing may also offer a solution.

Figure 4 and Figure 5 show the detection histograms in range and velocity, respectively, for the staggered
medium-PRF waveform and irregular PI waveform. The significant increase in detected targets can be noted even
at large ranges. The detection histogram for the velocity is non-uniform for the staggered medium-PRF waveform,
due to the practical limitations of linear processing. For the irregular waveforms with advanced processing, the
detection histogram for velocity is uniform, as the 2D matched filter compensates for the velocity of the target.

Figure 4. The detection histograms in range with (left) the staggered medium-PRF waveform with linear processing and (right) irregular PI
waveform with advanced processing.

Figure 5. The detection histograms in velocity with (left) the staggered medium-PRF waveform with linear processing and (right) irregular
PI waveform with advanced processing.

The bias of the range and velocity estimates of the target using the staggered medium-PRF waveform with linear
processing are 8.56 m and -0.14 m/s, respectively, compared to -1.42 m and 0.02 m/s for the irregular waveform
with advanced processing. The standard deviation of the estimates are 14.21 m and 7.08 m/s for the staggered
medium-PRF waveform and 2.32 m and 1.10 m/s for irregular waveform. Clearly, our proposed irregular waveform
and advanced processing improves the accuracy of the estimates roughly by a factor 6. Improved accuracy of the
estimates improves radar tracker initialization and it increases the track accuracy.

The irregular waveforms and advanced processing significantly improves the detection range that can also be
traded-off for shorter transmission times and/or usage of less transmit power. The shorter transmission time means
that the freed radar time budget can be used for other tasks, e.g., more simultaneous active tracks. Decreasing the
transmission power would allow for smaller-sized systems or improved energy/covertness profile of the sensor.

145
4. IRREGULAR WAVEFORMS AND ADVANCED PROCESSING IN THE CIVIL DOMAIN
The strength of irregular waveforms, or non-uniform sampling, combined with advanced processing has not gone
unnoticed in the civil domain. In particular, advanced processing for imaging in medical and acoustic sensors and
in radioastronomy have been matured and deployed in products. Within the medical imaging field, the usage of
advanced processing significantly reduces the acquisition time, i.e., patient in the machine, while simultaneously
sharpening the image. For example, Philips, General Electric and Siemens exploit it for MRI, CT, PET, and X-rays
scanners [17], [18], [19]. Another field that highly benefits of sharpened images using advanced processing is in the
acoustic domain. In particular, the oil and gas industry uses, e.g., seismic imaging, to geologically map the Earth’s
crust for natural resources [20]. Within radioastronomy, example of advanced processing for sharpened imaging
and calibration can be found in the LOFAR and SKA telescopes [21], [22].

5. CONCLUSIONS
This paper has demonstrated the potential of waveforms consisting of irregular pulse intervals combined
with advanced processing for the detection of hypersonic threats. In particular, we have shown that the novel
waveforms and associated iterative 2D processing offer a potential solution to overcome the practical limitations
of staggered PRF waveforms and linear 1D processing. Particularly for the hypersonic threat detection, this novel
combination can significantly increase the detection performance in terms of the detection range and the accuracy
of the estimated range and velocity measurements of the target. It has also been shown that the irregular MTI clutter
filtering technique can sufficiently suppress the simulated sea clutter to be able to detect the hypersonic threats at
large distances. Hence, the presented approach has the potential to significantly improve the detection range of the
radar system and, simultaneously, it improves the track quality that can be constructed from these measurements.
These advantages can contribute to early warning detection and increased time to engage hypersonic threats
using radar systems.

Current efforts are focussed on thoroughly analysing the impact the usage of irregular waveforms and advanced
signal processing within the complete radar processing chain and to improve robustness of these methods.
In addition, the focus is on efficient implementations tailored to (specific) processing platforms in terms of
computational load and memory capabilities. Moreover, designing waveforms in dynamic environments particularly
for the hypersonic threats is an ongoing research topic.

ACKNOWLEDGEMENTS
The results in this paper are financed by the Ministry of Defence in The Netherland covered partially by the Nationaal
Technologie Project MANDRAKE and the V1908 Radar Programma.

REFERENCES

[1] H. Astier, “Ukraine war: Russia fires hypersonic missiles in new barrage,” on BBC News, Mar. 2023 [Online]. https://
www.bbc.com/news/world-europe-64903202. [Accessed: 13 March 2023].
[2] NATO AVT-ST-008, “Assessment of the Status and Challenges Posed by Hypersonic Operational Threats”, TR-AVT-
ST-008, 2020.
[3] NATO AVT-359, “Impact of Hypersonic Operational Threats on Military Operations and Technical High Level
Requirements Phase 1”, STO-TR-AVT-359-Part-I, 2022.
[4] European Defence Agency, “Hypersonic Threat Detection and Countermeasures (Hypotenuse)”, tender reference
nr. 21.RTI.OP.092, Jul. 2021. https://etendering.ted.europa.eu/cft/cft-display.html?cftId=8915. [Accessed 13 March
2023].
[5] C. Alabaster, E. Hughes, and J. Matthew, “Medium PRF radar PRF selection using evolutionary algorithms,” in IEEE
Trans. on Aerospace and Electronic Systems, vol. 39, no. 3, 990–1001, 2003.
[6] A. De Maio, Y. Eldar and A. Haimovich, Compressed Sensing in Radar Signal Processing, Cambridge University
Press, 2019.

146
 EDA RESEARCH, TECHNOLOGY, AND INNOVATION PAPERS AWARD 2023

[7] W. van Rossum and L. Anitori, "Simultaneous Resolution of Range-Doppler Ambiguities using Agile Pulse Intervals
with Sparse Signal Processing," in IEEE Radar Conf., Florence, Italy, Sept. 2020.
[8] Defence advanced research projects agency (DARPA), “Beyond Linear Processing”, notice id: HR001123S0008,
Oct. 2022. https://sam.gov/opp/818c44d90f884834bf01e2e1382956ac/view. [Accessed 13 March 2023].
[9] L. Anitori and E. Joachim, "Waveform Design for Sparse Signal Processing in Radar." in Proc. of the IEEE Radar
Conf., New York, NY, USA, 2021.
[10] M. Coutino and F. Uysal, “Reinforcement Learning for Radar Waveform Optimization” in Proc. of the 2023 IEEE
Radar Conf., San Antonio, TX, USA, May 2023.
[11] L. de Martín and W. van Rossum, “Optimization of Pulse Intervals for Unambiguous Doppler Recovery with
Oversampled Dictionary,” in Proc. of the 2022 IEEE Radar Conf., New York, NY, USA, Mar. 2022.
[12] K. Klein, M. Coutino, R. Struiksma, P. Cox, L. Anitori, “Efficient Processing of Irregular PRF Waveforms: Clutter
Suppression and Approximate 2D Matched Filtering” in Proc. of the 2023 IEEE Radar Conf., San Antonio, TX, USA, May
2023.
[13] P. Cox, M. Coutino, and W. van Rossum, “Kernel Design Meets Clutter Cancellation for Irregular Waveforms,” in
Proc. of the 2023 IEEE Radar Conf., San Antonio, TX, USA, May 2023.
[14] R. Struiksma, F. Uysal and W. van Rossum, "2D matched filtering with time-stretching; Application to Orthogonal
Matching Pursuit (OMP)," in Proc. of the 18th European Radar Conf., London, United Kingdom, April 2022.
[15] M. A. Hadi, S. Alshebeili, K. Jamil, and F. E. A. El-Samie, “Compressive sensing applied to radar systems: an
overview,” in Signal, Image and Video Processing, vol. 9, no. 1, pp. 25–39, 2015.
[16] V. Gregers-Hansen and R. Mital, "An improved empirical model for radar sea clutter reflectivity," IEEE Trans. on
Aerospace and Electronic Systems, vol. 48, no. 4, pp. 3512-3524, 2012.
[17] Philips Healthcare, “Snellere scans in alle lichaamszones met Compressed SENSE” [Online]. https://www.philips.
nl/healthcare/artikelen/medisch-perspectief/compressed-sense. [Accessed 13 March 2023].
[18] General Electric Care, “HyperSense”, [Online]. https://www.gehealthcare.com/-/jssmedia/files/us/non-gated/
mri/hypersense-booklet.pdf?rev=-1&hash=FAE260A3F4CA0A82E236DEDA5CAD39A6. [Accessed 13 March 2023].
[19] Siemens, “Compressed Sensing: Beyond Speed”, [Online]. https://www.siemens-healthineers.com/nl/magnetic-
resonance-imaging/clinical-specialities/compressed-sensing. [Accessed 13 March 2023].
[20] Georgia Tech, “SINBAD,” Seismic Laboratory for Imaging and Modeling, [Online]. https://slim.gatech.edu/projects/
sinbadseismic-imaging-next-generation-basis-unctions-decomposition. [Accessed 13 March 2023].
[21] H. Garsden et al., “LOFAR sparse image reconstruction,” Astronomy & Astrophysics, vol. 575, pp. 18, 2015.
[22] ASTRON, “Calibration and imaging,” [Online]. https://www.astron.nl/research-and-innovation/calibration-and-
imaging/. [Accessed 13 March 2023].

147
Authors

Pepijn B. Cox has received his B.Sc. degree in Mechanical Engineering (cum laude)
in 2010 and his M.Sc. degree in Systems and Control Engineering (cum laude) in 2013,
both at the Delft University of Technology (TUDelft), The Netherlands. In 2018, he
obtained his Ph.D. degree in the Control Systems group at the Eindhoven University
of Technology (TUe), The Netherlands. In 2018, he was a postdoctoral researcher in
the Control Systems group at the TUe.

Since 2019, he works as a signal processing expert at the Radar Technology group
at TNO the Netherlands. Pepijn Cox’s main research interests are in multistatic and
netted radar systems, compressive sensing, numerical optimization, run-time
verification as well as in linear parameter-varying and nonlinear system modelling
and identification.

Keith T.J Klein received the M.Sc. degree in electrical engineering from Delft
University of Technology, Delft, The Netherlands, in 2020. Since 2021, he has been
with the Department of Radar Technology, Netherlands Organisation for Applied
Scientific Research, The Hague, Netherlands. His current research interests include
radar signal processing and radar system concepts.

Mario Coutiño received the M.Sc. and the Ph.D. degree (summa cum laude) in
electrical engineering from the Delft University of Technology, Delft, The Netherlands,
in July 2016 and April 2021, respectively. Since 2020, he has been a Signal Processing
and Machine Learning Researcher with the Radar Technology Department, TNO, The
Netherlands. He has held temporally positions with Thales Nederland, during 2015,
and Bang & Olufsen, during 2016 working on inverse problems. His research interests
include array signal processing, signal processing on networks, optimization, inverse
problems, machine learning and radar technology. He was the recipient of the Best
Student Paper Award at CAMSAP 2017, the CONACYT excellence scholarship and
was a Visiting Researcher with RIKEN AIP and the Digital Technological Center at the
University of Minnesota, in 2018 and 2019, respectively, working on graph methods
and theoretical foundations of graph neural networks.

Dr. Laura Anitori received her Master of Science degree (cum laude) in
Telecommunication Engineering from the University of Pisa, Italy, in 2005 and her Ph.D.
degree (cum laude) in Electrical Engineering from the Technical University of Delft,
The Netherlands, in 2013. Since 2007 she works at the Radar Technology department
of TNO, The Netherlands, where she is senior scientist and program manager of the
Defense funded radar research program. She is an IEEE senior member, member of
the Board of Governors of IEEE AESS, and chair of the IEEE AESS Radar System Panel.
She is government expert representative for The Netherlands within the European
Defence Agency Radar Captech, and Dutch national representative in the NATO
Sensors and Electronics Technology (SET) Panel. Her significant contributions to
NATO were recognized with the SET Early Career Award in 2018, the 2019 SET Panel
Excellence Award, and the NATO Scientific Excellence Award in 2022. She serves
on several technical program committees and student competition committees at
international scientific conferences.

148