OFEME 2024 Programme

Next-generation wireless base stations and access points will utilize an extremely large number of antennas for transmission and reception. Dynamic Metasurface Antenna (DMA) is a brand new concept and a promising technology to achieve such massive arrays in a dynamically controllable and scalable manner, while reducing cost and power consumption, which involves radiating metamaterial elements as opposed to reflective element as in RIS. DMAs offer an efficient realization of massive antenna arrays for wireless communications with advanced analog signal processing capabilities for next-generation transceivers. They provide beam tailoring capabilities and facilitate processing of the transmitted and received signals in the analog domain in a flexible and dynamically configurable manner using simplified transceiver hardware with much less power and cost compared with conventional antenna arrays.

Although conceptually feasible, DMAs for communications in the high millimeter wave (around 60 GHz) and THz (above 100 GHz) bands are still in relatively early stages of research and require immense exploration. This set the stage to aim for experimental validation of DMAs for communication, imaging, sensing and ISAC avenues especially in indoor scenarios to unveiling their potential in future 6G transceivers at 60 GHz. Our designed DMA prototype is the first ever tested DMA at high mmWave frequency band of 60 GHz, which will serve as a catalyst to unfold the research avenues of this technology towards THz regime.

Method:
We developed the first prototype of a programmable Dynamic Metasurface Antenna (DMA) at the license-free 60 GHz mmWave band. A novel complementary electric inductive-capacitive (CELC) metamaterial element resonating at 60 GHz was designed, exhibiting left-handed metamaterial properties. Excited by a low-loss substrate integrated waveguide (SIW) and controlled by high-frequency PIN diodes, the CELC meta-element switches between radiating and non-radiating states to offer 1-bit digital control. Embedding 16 such elements into a one-dimensional DMA excited through a direct edge-feed, we designed an electronic beamsteering array, dynamically controlled by a high-speed FPGA. The PIN diodes’ biasing network is integrated using a 4-layer PCB for agile control over the radiation pattern.

Results and Conclusion:
Excellent and well controlled versatile beam synthesis (such as narrow beams, wide beams, and multiple beams) is achieved based on different digital coding combinations from FPGA. The prototypes of the single DMA element and 16-element 1-dimensional DMA array are fabricated and verified through practical measurements. High radiation efficiency and -10 dB impedance of 3 GHz (around 60 GHz) is achieved. The beam-switching agility is quantified and observed to be within 5 ns, indicating significant promise for real-time URLLC applications.

In conclusion, the first-ever proposed DMA prototype is a potential enabler to unfold a diverse range of next-generation mmWave wireless applications such as agile electronic beam-steering, adaptive beamforming, holographic imaging, mmWave industrial wireless communication, cognitive radars, as well as integrated sensing and communication (ISAC).

Co-Authors and Affiliation: Mostafa Elsyed, Jalil Ur Rehman Kazim, Qammer Abbasi, Muhammad Ali Imran, Masood Ur-Reman – University of Glasgow

The work developed a metasurface-antenna capable of independently steering its front-lobe, providing gain, whilst also independently steering a null in its back-lobe.
Independent front-lobe and back-lobe-null steering has various applications in Electronic Warfare (EW) and Electronic Surveillance (ES). As an example application, the metasurface-antenna may be part of an EW transmitter. The front-lobe may be directed towards the enemy whilst directing the back-lob-null towards friendly forces. Through software control, the front-lobe and back-lobe-null may be independently adjusted in response to the movement of the enemy and friendly forces.

The metasurface-antenna consists of a metasurface placed in front of a slot antenna, delivering gain through the Fabry-Perot effect, to focus the radiation of the slot antenna. A second metasurface is placed behind the slot antenna to generate a null in the antenna back-lobe. The metasurface-antenna is electrically small and therefore there is significant diffraction around the rear metasurface, limiting the depth of any null. A unique approach was therefore developed whereby emissions diffracted around the rear metasurface were cancelled by emissions passing through the metasurface.
Steering of the front-lobe and back-lobe-null was achieved through adjustment of passive components placed within the unit cells of the metasurface. This demonstrated the vision of steering through software control, though the implementation of the control circuitry was outside the scope of work.

Design of the metasurface-antenna was achieved through simulation using COMSOL Multiphysics.

Simulation of the metasurface-antenna achieved steering of deep nulls, < -20dBi, across approximately 180° of the rear metasurface whilst independently steering gain, approximately 8dB, in the forward direction, at +/-30°.

This work was conducted by PA Consulting and supported by the University of Sheffield.

Co-Authors and Affiliation: Nikhil Antony – PA Consulting

RF sensing almost exclusively uses architectures which are based on conventional electronic radio systems: antenna, amplifier, frequency conversion and demodulation. This architecture has been highly optimised, but may not be ‘optimal’ from a physics perspective. For instance, it may be possible to improve the SWaP, sensitivity or bandwidth of a system using newly emerging quantum sensing techniques.
Although these techniques can apparently sense electric, magnetic and RF fields with high sensitivity, it is unclear how to best exploit this technical advantage. For instance, although the new generation of magnetometers have very low noise figures, their applications tend to be limited by interference from external sources rather than noise inherent to the sensor.

This poster aims to prompt an open, and technology neutral discussion about how three types of quantum sensors can fit current and future EW activities. It presents outline order of magnitude sensitivity calculations as to the signal to noise (SNR) levels that might be achieved and compares them to conventional approaches. It will also discuss some other likely advantages such as size, frequency agility and resistance to deliberate overload conditions such as electronic attack or jamming.

We will consider (at least):
• Magnetometry using alkali metals (rubidium / caesium),
• Electrometry using highly excited vapours (Rydberg atom),
• RF sensing using synthetic diamonds (nitrogen vacancies)

Co-Authors and Affiliation: Edmund Owen, Martin Cooper – Cambridge Consultants

Future radar systems are set to rely on reconfigurability and multi-functionality, driving a shift from analogue to digital beamforming. However, fully digital beamformers pose challenges in terms of complexity, weight and cost. Therefore, hybrid architectures, comprising digital beamformers and analogue beamformers, are a viable choice for many radar and sensing applications. The hybrid array architecture can be streamlined by replacing analogue phase shifters with an analogue beam-former, such as a low-cost and passive Butler matrix. This enables sub-array beams to be steered in a discrete number of angular directions, allowing for more advanced beam steering algorithms to be included. The cost of a hybrid sub-array is estimated to be only 60% of a fully digital system equivalent.
Following these concepts, a novel architecture combining sub-arrays based on Butler matrices with fully digital beamformers has been developed. Since amplitude tapering requires greater excitation of central array elements, adding fully digital sub-arrays can offer significantly improved pattern and controllable sidelobe levels (SLLs), versus a traditional hybrid beamformer. In the developed architecture, excitation of digital sub-array elements is individually controlled, whereas in the Butler matrix-based sub-array all elements have equal excitation magnitude. However, amplitude tapering remains effective for lowering SLLs. For example, in a 16-element array with two digital sub-arrays and two Butler matrix-based sub-arrays, SLLs can be kept below –10 dB for a ±45° steering range. When scaling to larger arrays, a 160-element array (for example) with one-third hybrid sub-arrays and an adapted minimum variance beamforming, can have SLLs lowered to -30 dB with a half-power beamwidth of about 1°. This is a significant improvement versus SLLs of -21 dB for Taylor amplitude tapering alone. This shows how incorporating hybrid sub-arrays into fully digital arrays and applying appropriate beamforming algorithms enhances the performance while keeping the cost, complexity and weight manageable.

Co-Authors and Affiliation: Alexander Don, Ronnie Frith, Symon Podilchak – The University of Edinburgh

Retrodirective arrays (RDAs) utilize phase conjugation, which allows the received signal to be re-transmitted in the direction it was received, without prior knowledge of the signal’s origin. This automatic steering capability makes RDAs an attractive solution and has led to their use in a variety of applications, including mobile communications, wireless power transfer, and radar systems. Most of the work in the open literature predominantly involves analogue systems, which, while typically more responsive and less power-hungry than their digital counterparts, lack flexibility and are generally not reconfigurable without significant alterations to the system architecture. Additionally, most array designs rely on physical separation between TX and RX elements and/or frequency offsets to achieve sufficient isolation between TX and RX channels.

This work proposes a new full-duplex, dual-polarized, co-located digital RDA, which operates both with and without a programmable frequency offset. The supporting S-band array is compact and composed of co-located transmit and receive stacked patch elements with isolation values of 50 dB. Each element is fed using H-shaped slots and dual-differential integrated couplers, offering dual-polarized operation as required, while a synchronized software-defined radio transceiver network is used to apply complex weights and phase conjugation at baseband.

Moreover, this system demonstrates proof-of-concept by forming sum and difference patterns and also exhibits self-steering in response to an interrogating signal from the far field. Results show that the digital retrodirective antenna array (RDA) system can achieve a competitive -5 dB monostatic angular scan range of ±47°, requires no frequency offset to operate, and offers a 6% matching impedance bandwidth, all in a single compact form factor. To the best knowledge of the authors, no similar digital IBFD antenna RDA system has been designed, programmed, and tested.

Co-Authors and Affiliation: Zain Shafiq, Ronnie Frith, Symon K. Podilchak – The University of Edinburgh

Radio blackouts occur when a plasma sheath surrounds a platform travelling fast enough to ionise the atmosphere around it. The plasma frequency is dependent on factors such as platform trajectory, size and shape. Although hypersonic platforms fly at speeds lower than those experienced for Earth re-entry vehicles, they do move fast enough to produce a plasma layer. The challenge, therefore, is how can communications links for command and control or sensor data be established to these platforms through the plasma. The aim of this study was to review potential options, highlight the challenges and potential solutions, such as where the plasma density peaks and how can RF waves be directed to the platform surface. A Computational Fluid Dynamics (CFD) model was used to develop an initial assessment of the plasma environment surrounding a platform and basic link budgeted analyses were conducted to illustrate communications range. Ray-tracing was also used to investigate the propagation close-in to the platform. The findings indicated that although communications may be feasible based on technical solutions already used by re-entry platforms, modelling capabilities need to be further developed to include additional factors such as ablation products.

Co-Authors and Affiliation: Tajinder Chemar, QinetiQ, Steve McDowell, David Evans – Fluid Gravity Engineering

Rydberg atom receivers enable a fundamentally different mode of access to the electromagnetic spectrum – they operate over a wide bandwidth and with minimal spectral disturbance. Building a Rydberg system is non-trivial in that it requires high performance, long lead and scarce optical components, and optimally exploiting a system requires the intersection of optical and RF engineering.

PA has recently completed the build of such a system to operate in the GHz bands and now has a benchtop demonstration system in our Cambridge Technology Centre. This system is based on a small gas cell filled with Caesium atoms, pumped with a significant optical power; this system is designed as a test bed to explore a range of Rydberg use cases and is necessarily designed with a high degree of futureproofing. Our system has been tested with carrier frequencies of order 2-20GHz and is capable of simultaneous reception of multiple carriers; we have also designed a range of extensions to the system’s capability including increasing the range of operational frequencies and the potential data speed.

We discuss the lessons learned from building a Rydberg system, the opportunities and challenges associated with exploiting the technology, how it may be extended to a wider range of use cases, and the benefits of this technology.

Co-Authors and Affiliation: Paul Marsh – PA Consulting

Conventional intelligent metasurfaces, also known as reconfigurable intelligent surfaces, are composed of a large array of meta-atoms reconfigured using active components such as p-i-n diodes and varactors. Several intelligent metasurfaces shown in the literature utilize thousands of such diodes to control the incident beam and are being envisioned as a replacement for large dish antennas in space. This allows us to control the radio environment efficiently, but it comes at the cost of high-power consumption, maintenance, and fabrication costs. Such metasurfaces need consistent power to maintain the beam reconfiguration and hence are power-hungry.

This paper will present a novel concept of a 3D structural metasurface with mechanically morphing capability that can be used as a transmit- or reflect-array to manipulate the incident electromagnetic beam for applications in RF sensing and communications for both space and ground stations. The proposed simulated structural metasurfaces are presented as a sustainable replacement for conventional intelligent metasurfaces. They can be controlled using low-power actuators to deform the surface profile, which is then utilized as a lens (in case of transmitarrays) or reflectors (in case of reflectarrays) for beamforming and steering.

As an example of such structural metasurfaces for electromagnetics, we demonstrate a shape morphing metasurface with lensing behaviour. This shape morphing metasurface concept, when fed with a horn antenna at a fixed position, is shown to present a dynamic beam-steering capability of ±30° within simulation with the mechanical actuation. Furthermore, this one-dimensional lens designed with the help of a morphing metasurface is shown to offer a gain enhancement of 8 dB compared to a standard Ka-band horn antenna.

The ability to sense across a broad range of the electromagnetic spectrum is vital for defence and security operations. The goal of this Industrial Cooperative Awards in Science & Engineering (ICASE) PhD project is to develop nanoscale optical sensors based on a new class of two-dimensional semiconductors (2SEM). These will operate across a wide range of the spectrum for high-sensitivity, broad-band imaging. Leveraging a unique, state-of-the-art facility at the University of Nottingham for the EPItaxial growth and in-situ analysis of 2SEM (EPI2SEM) [1-2], we have developed novel material platforms for sensing applications. Here, we report on the 2SEM gallium selenide (GaSe), a new building block for nanoscale optoelectronics. We present the development of optical sensors based on GaSe on both sapphire [3] and epitaxial graphene/SiC substrates [4], which operate in the VIS-UV range. In addition, epitaxial GaSe offers an avenue for obtaining high quality wide bandgap gallium oxide (Ga2O3) via a post growth oxidation process [5]. The selective optical absorption of Ga2O3 in the deep-UV is exploited for photon sensing in the UV-C (200-280 nm) spectral range, offering a scalable route to deep-UV optoelectronics. Finally, we introduce mid-infrared photodetectors based on GaSe/graphene heterojunctions, which exploit a charge dipole and built-in potential at the GaSe/graphene interface for photosensing. These developments demonstrate a route to next generation scalable, high-performance optoelectronic devices for broadband sensing.

[1] EPI2SEM Animation, YouTube, rb.gy/iuaht8.
[2] EPI2SEM Interviews, YouTube, rb.gy/sm21ti.
[3] M. Shiffa et al., Small 2024, 20, 2305865.
[4] J. Bradford et al., Small 2024, 2404809.
[5] N. D. Cottam et al., ACS Appl. Nano Mater. 2024, 7 (15), 17553-17560

Co-Authors and Affiliation: N. D. Cottam, M. Shiffa, J. Bradford, T. S. Cheng, S. V. Novikov, C. J. Mellor, O. Makarovskiy, K. Rahman, J. N. O’Shea, P. H. Beton and A. Patanè – University of Nottingham. S. Lara-Avila, Chalmers -University of Technology. T. Ben and D. González – Universidad de Cádiz. J. Harknett, M. T. Greenaway – Loughborough University

This poster presents a compact wide bandwidth antenna with a predominantly end-fire main beam suitable for microwave and millimeter-wave applications. The novelty of this antenna is in its ability to maintain a relatively stable radiation pattern and direction over a large bandwidth, which offers better performance than the state of the art.

The planar microstrip-fed antenna comprises of two half circular rings of different diameters, one printed inside the other, with a partial ground plane on the underside. The antenna is printed on Rogers RT Duroid 5880 substrate and has dimensions 18 × 12 × 0.8 mm3, making it ideal for devices with very limited space.

The antenna is intended for operation between 24 and 48 GHz, although it has an operational bandwidth with return loss better than 12 dB from 11.5 to 62.5 GHz and gain that exceeds 6.5 dBi across the entire frequency band.
The double half circular rings and partial ground plane have been optimised to maintain a stable beam predominantly in the end-fire direction. Across the operating frequency range, the main beam maintains a direction of 65ᵒ ± 10ᵒ in azimuth plane and 85ᵒ ± 5ᵒ the elevation plane. The realised gain ranges between 7 dBi and 11.6 dBi. Although well matched, outside the operating frequency range, the main beam direction changes from being predominantly end-fire to being boresight and can split into two or more beams.

The prototype of the design has been fabricated and measured, and the results show good agreement with simulated results, indicating the proposed antenna is suitable for microwave and mmWave applications requiring a simple, compact, wide bandwidth, and end-fire radiation beam.

Co-Authors and Affiliation: William Whittow, Chinthana Panagamuwa – Wolfson School of Mechanical, Electrical and Manufacturing Engineering, Loughborough University

The field of quantum communications exploits quantum phenomena to both enhance classical communication methods and enable entirely new methods. As part of the Deployed Communications Evolving Against the Threat (DCEAT) project, Dstl has brought together a Subject Matter Expert Team consisting of experts from QinetiQ, PA Consulting, Thales, and L3Harris to carry out quantum communications work. Part of this work focuses on novel quantum receivers. These can provide significant advantage over conventional antennas and aid communications in an increasingly contested electromagnetic environment. This poster gives an introduction to two types of quantum receiver which are both based on the transmission of lasers through vapour cells: the optically pumped magnetometer (OPM) and the Rydberg receiver. It presents the context behind the use of these receivers, their working principles, and their advantages over conventional antennas. The work done by QinetiQ on OPMs and PA Consulting on Rydberg receivers is also described. OPMs are magnetic field sensors and are proposed as receivers for the novel method of magnetoinductive (MI) communications. This uses modulation of the magnetic field component over short ranges to transfer information. This is advantageous in environments such as underwater where electromagnetic waves are heavily attenuated. QinetiQ are investigating the performance of OPMs as MI receivers underwater. Using a water tank in the lab, they are exploring the effect of parameters such as transmitter frequency, water salinity, and temperature on the received signal. Rydberg receivers are electric field sensors that can be used in place of conventional receiving antennas for communications. Unlike conventional antennas, the size of the Rydberg sensor head is not related to the receiving frequency so it can be made compact while also being passive, widely-tuneable, and highly-sensitive. PA Consulting have developed a Rydberg receiver and conducted a literature review to identify methods of improving the instantaneous bandwidth.

The Defence Science and Technology Laboratory (Dstl) and RAL Space Chilbolton have established a 500m Radio Frequency (RF) terrestrial link transmitting on V&W (73-83GHz) bands in horizontal polarisation. The signal is received in both horizontal and vertical polarisations to allow for analysis of the atmospheric weather effects which impact the RF link.

Growing congestion at lower frequency bands has led to the need for new operating bands at higher frequencies. Understanding how extremely high frequency (EHF) links such as V&W band are impacted by tropospheric effects will influence where and how V&W links are utilised in the future, such as in ground terminals as high capacity anchors to alleviate bandwidth demands at lower frequencies for transmission of data.

Data was collected from January 2023 until March 2024. The RF dataset includes samples of phase and amplitude. Meteorological data is collected which includes but not limited to cumulative rain count, relative humidity and air temperature. A laser precipitation monitor dataset includes for precipitation size, speed and hydrometeor categorisation.

Initial correlations suggest moderate negative relationships between air temperature and relative humidity. Mild negative correlations between channel suppression and relative humidity indicates that attenuation can be present in very humid conditions through the introduction of partial attenuations. Similarly, a moderate negative correlation between channel suppression and category of particle. This indicates a relationship between smaller particles and increased channel suppression.

Future work is split across three categories. Development of the rain attenuation calculation will continue to improve accuracy and model fitting. Concurrent data collection of Ka/Q band AlphaSat data in 2024 will enable a frequency scaling exercise between the terrestrial and non-terrestrial experiments. Finally, an empirical model will be developed using relationships identified between hydrometeor production and meteorological parameters to enable early warning of suspected severity of signal attenuation.

Co-Authors and Affiliation: Tim Koetser – Dstl

Improving energy efficiency is a key requirement for future wireless communication systems, and antenna selection is considered a promising solution. Most research in this area has focused on developing antenna selection strategies, particularly for uniform linear arrays. In contrast, practical systems more commonly employ uniform planar arrays (UPAs). Moreover, an important factor—the interplay between selected antenna subsets and directivity—has been overlooked.

Aims: This work aims to address these gaps by exploring how different subsets of selected antennas affect directivity and by developing effective antenna selection strategies for UPAs with a rectangular array structure.

Methods: In this study, we incorporated array directivity into the antenna selection process. Note that the array directivity depends on antenna spacing, and different subsets of selected antennas result in varying inter-element spacings. Therefore, it is crucial to include directivity when selecting a subset of antennas. Furthermore, focusing on uniform rectangular arrays and we proposed four antenna selection strategies: 1) Column Selection, 2) Row Selection, 3) Full Aperture Greedy Selection, and 4) Hybrid Initialization Binary Particle Swarm Optimization (HBPSO).

Results: Our findings show that different subsets of selected antennas can result in significant directivity variations, influencing antenna selection outcomes compared to methods that do not consider directivity. Additionally, HBPSO, with careful particle initialization, achieved spectral efficiency close to exhaustive search and outperformed other strategies. Furthermore, incorporating directivity into the antenna selection process notably impacted the performance of these strategies.

Conclusion: This study highlights the practical importance of incorporating directivity into antenna selection strategies for UPAs and devise antenna selection strategies for uniform rectangular arrays. It also demonstrates the trade-off between directivity and antenna selection and underscores the need to incorporate directivity as a key factor when developing antenna selection strategies for UPAs.

Co-Authors and Affiliation: Geoffery Hilton, Angela Doufexi and Mark Beach – University of Bristol

This poster presents an investigation of a potential weight optimisation strategy for designing antennas. This strategy consists of adding and removing dielectric to specifically selected sections of an antennas substrate. These modifications will be performed in such a way as to maintain the antenna’s operation while simultaneously using less dielectric producing a net reduction in weight.

At their core, antennas are electrical components dedicated to controlling the emission and reception of electromagnetic (EM) radiation. Numerous types of antennas exist, all of which present notable trade-offs, but patch and microstrip antennas have been the centre of development over the last few decades and as such are some of the most widely used antennas for smaller electronic devices. A dielectric material is present in both of these antenna types in the form of a PCB substrate that has an antenna design created on its copper surface. The permittivity (ε_r) of the dielectric used has a large effect on the properties of the antenna with typical values ranging from 2 to 10 depending on specific application. However, in most cases not all the dielectric used is efficient at producing the desired antenna properties. In most cases this is irrelevant as dielectric is integrated into the PCB and removing it would serve no benefit. However, in our case when optimising for weight every gram of misplaced dielectric is a potential weight saving.

Utilising the unique properties of additive manufacturing techniques allows for exceptional precision when deciding how much dielectric to place at any given location. Additionally, utilising these techniques allows for the substrate to be easily re-optimised on any antenna design change. These weight reductions are of notable importance to certain defence applications such as small UAVs/drones. The weight reduction potential of this technique can reduce the weight of the smaller antennas used on such devices while maintaining antenna performance.

Co-Authors and Affiliation: James Flint, Chinthana Panagamuwa – Loughborough University

Multifunction phased array antennas for radar require calibration in the factory in order to align beamforming channel imbalances. This process can be expensive and time-consuming. Recent studies have suggested that employing measurements in parallel can offer many advantages over sequential individual channel measurements. In particular, the calibration speed for large arrays and arrays with digitally controlled transmit/receive beamforming may benefit the most. Orthogonal coding allows measurements on multiple channels to be made in parallel by making use of mathematically orthogonal signals and forming linear systems of equations to recover individual channel responses. This poster will investigate some simulation and benchtop measurement results on a simple phased array antenna evaluation board to show the potential benefits when using orthogonal coding with phased array antenna calibration in comparison to traditional techniques.

Co-Authors and Affiliation: Alexander Don, Dr Symon Podilchak, Dagna Wójciak – University of Edinburgh

This research investigates a novel approach to achieving beam-scanning capabilities using leaky coaxial (LCX) arrays. LCX cables are typically used for distributed wireless communication in confined spaces and are being explored for their potential use in antenna arrays due to their unique radiation properties along the cable length. In this research, we propose the use of a power divider network to introduce phase shifts between individual LCX cables, enabling beam steering by controlling the phase of the input signals. By adjusting these phase shifts, we aim to dynamically direct the radiated energy of the array without the need for complex mechanical movements or traditional phased array setups.

The array design is currently being modelled, and upcoming simulations will examine the effects of phase variation on the radiation pattern, directivity, and beam-steering capabilities. This setup is anticipated to offer an effective alternative to traditional phased arrays, leveraging the simplicity and flexibility of LCX cables while achieving dynamic beam-scanning characteristics. The use of LCX cables also introduces the potential for broader applications in challenging environments, where traditional antenna solutions may not be feasible.

Preliminary analysis suggests that controlled phase shifts through the power divider should allow for effective steering of the array’s beam, with promising applications in areas such as wireless communications and radar. Full wave simulations will focus on optimising the configuration to improve scanning performance and the bandwidth which is being utilised. The results from these simulations are expected to be encouraging based on previous simulation work and research into this area. This can pave the way for future experimental validation and system refinement.

Co-Authors and Affiliation: Babar Abbasi, Okan Yurduseven – Supervisory Team

With increasing congestion and demand for the electromagnetic spectrum, alternatives are required for future communication capabilities. Free space optical communications (FSOC) has the potential to achieve secure, reliable, and high-performance wireless connectivity. However, due to the perceived lack of resilience to natural weather conditions, such as rain and fog, FSOC has yet to reach mass market adoption. The development of quantum cascade lasers (QCLs) has enabled semiconductor based devices in the mid wave infrared (MWIR) region (8um-14um), such as QCLs using the AlInAs/GaInAs material systems and HgCdTe detectors. The anticipated properties of this wavelength range with regards to photon transmission in foggy weather conditions is highly beneficial for communications through the atmosphere, this is due to the free space transmission windows and scattering due to particle size. In the post we demonstrate a communications link using commercial-off-the-shelf components and equipment, which are not designed for high-speed communications, in the 10 um wavelength over distances of up to 290 mm through an artificial fog. Insignificant changes are observed to the communication performance, where data rates of 200 kbit s−1 are achieved over 3 modulation schemes (on-off keying, 4-level pulse position modulation, 4-level pulse amplitude modulation). Additionally there will be a discussion of future requirements for components in the MWIR region.

Co-Authors and Affiliation: I.M.E.Butler, J.Liang, E. Guler, W.O.Popoola

Seeking to make a step change in communications burden exploring systems without a transmitter. The Scatterbrains work aims to explore techniques that make a step change to the Size, Weight and Power (SWaP) requirement, thermal dissipation and detectability of a communications system by the application of novel backscatter techniques. The benefit will be a reduced weight and power burden to promote the Freedom of Access and Manoeuvre (FOAM) for dismounted personnel or uncrewed vehicles. The technique achieves the benefit by removing the need for an active transmitter at one end of the communications link, where the transmitter typically contributes significantly to the power consumption and hence battery and cooling requirements in the system.

The work explored use cases that could exploit the technology (such as dismounted personnel, UaV’s and tracking of high value assets) and demonstrated the principle operating over 230m range.
This work was funded by DASA (contract ACC6042061) completing in March 2024.

Co-Authors and Affiliation: Nikhil Antony – PA Consulting Group

Electromagnetic Warfare (EW) databases exist for the reprogramming of various platforms and provide a repository to contain the Electronic Order of Battle (EoB), where that EoB can become a support to a Situation Awareness (SA) picture. The emphasis for EW databases is commonly equipment reprogramming and, in many cases, it seeks to provide the best reprogramming value for that equipment. As such, they commonly wish to provide a single value, but to arrive at the singular unambiguous value, they require layers of threat Weapon System Analysis (WSA), Threat Vulnerability Analysis Countermeasure Design (TVACD) tactic development, EOB geolocation, networking, collection tasking, production, Test and Verification activities to provide an audit trail between the programming and the intelligence. That link between the evidence and the reprogramming may be weaker. The proposal here is for a Markup Language that allows the source data mapping of fields towards gaining a singular most likely encountered value; that is while accounting for all observational evidence, and all holistic rules from a body of knowledge, to aid that linkage in a complete and consistent way. That approach also provides an automated confidence weight value where that confidence is provided by a data convergence approach covering both the agreement in value and the outlier distance within the valid scope of the known body of knowledge. This approach represents a motivation for EW data management to reflect the source data ‘colourisation’ and its translation from intelligence sources, rather than just the fixed data structures for reprogramming the equipment data fields. As such, it represents a motivation not to compromise the representation of the source or the programmed equipment, in its structure. Furthermore, it perhaps represents a motivation toward knowledge graphs rather than the rigid database structures.

Datasets are often complex and expensive, but if available, they can exercise model fitting in a machine-learnt solution, and when from a synthetic source they can be controlled in dimensionality. Synthetic sources can provide constraints to the dimensionality that stimulate, and exercise known ambiguities and discrimination criteria in a controlled way. In this research a radar emitter behaviour model is demonstrated from a behavioural description using an emitter markup language for a Machine Learning (ML) example. That emitter behaviour model is used in a physics engine to generate high bandwidth data within a dataset. Furthermore, the dataset is then used to train an emitter classification model and achieves 99.3% accuracy in emitter identification using image classification. The emitter model and dataset are available for further research, as well as for other methods such as Transferred Learning (TL), the Generative Adversarial Network (GAN) method, and AI Cyberattacks such as the Fast Sign Gradient Method (FSGM).

Decision making in a complex and fast moving EM environment can be challenging for both humans and autonomous systems. Many operational decisions can be expressed as combinatorial optimisation problems, which allows them to be expressed mathematically but can incur high computational overheads to solve. Annealing techniques have shown promise in solving these operational challenges, both on classical and quantum hardware. “digital annealers” are specialised hardware that bridge the gap between quantum and classical capability. We demonstrate that they currently offer high speed solution of intermediate to large combinatorial optimisation problems, for example the optimal allocation of tasks between fifty or more automated platforms; setting the benchmark for quantum systems.

Co-Authors and Affiliation: Gemma Shepherd, Rowan Curtis, Richard Claridge – Cranfield University

The Single Information Environment (SInfoE) has resulted from a series of interlinked Dstl research studies subsequently taken up by Futures Lab’s MDIS programme and demonstrated at Ex ACHERON. It is now at TRL7 with STRATCOM wishing to take its development further. It is a major success story for Dstl.

The SInfoE allows data to be found and accessed anywhere in the defence enterprise, regardless of the system or platform the data resides on. It is analogous to the business model used Google or Apple that allows any third party to contribute or extract data through a single interface. Like these commercial providers, the SInfoE contains information brokers to allow data to be found.

This approach has profound implications for future capabilities like AI, autonomy, Space, future sensing, Zodiac and RAS/UXV, because the original (SME?) manufacturer / developer of such tools can build an interface to the SInfoE and share data. On ACHERON, interface creation took between 2 days and 2 weeks, allowing MOD to access and exploit emerging technologies and new solutions faster and more cheaply than before, helping low TRL developments from Dstl and DASA to be pulled into service.

Traditionally attempts at integration focused on a system-to-system approach at the network level, and on data harmonisation. The SInfoE concentrates on data integration. At the heart of the SInfoE is a Common Protocol (SCP), effectively the Google or Apple sharing environment. This contains multiple open standards, DefStans and STANAGs; new ones can be easily added. At its edge are API Service Providers that translate data from the attached system into the SCP and visa versa, while also collecting metadata on the attached system. The information broker searches the metadata to identify what data is available and a connection between the requester and provider systems is created across the SCP.

Co-Authors and Affiliation: Steve Clark, Willow Capel – Nexor Ltd

Positioning, Navigation & Timing (PNT) techniques and technologies are considered within the context of the Electronic Attack (EA) landscape against a number of defined EA use cases covering a range of PNT requirement complexities.

A selection of positioning and navigation techniques are covered as well as a comprehensive overview of synchronisation and syntonisation techniques. Current areas of research and future synchronisation techniques are also presented.

A use case matrix is presented which correlates synchronisation techniques against the identified EA use cases and areas that require future work are outlined. It highlights that positioning solutions such as single point positioning Global Navigation Satellite System (GNSS), differential GNSS, and image-aided Inertial Navigation System (INS) are appropriate solutions for the majority of low and medium complexity positioning requirements where the accuracy required is between 1m – 10m. Some higher complexity positioning requirements (1cm – 1m) can be met by advanced techniques such as Precise Point Positioning (PPP), Real-Time Kinematic (RTK), and PPP-RTK and that further research is required for very high positioning techniques below 1cm. Similarly, the use case matrix indicates that the majority of EA use cases require a time synchronisation solution that can provide low nanosecond accuracies and as such, techniques such as Precision Time Protocol (PTP), White Rabbit and Two-Way Time and Frequency Transfer (TWTFT) are commonly the most appropriate solution. Low complexity synchronisation use cases can be met with coarse synchronisation techniques such as Timing-sync Protocol for Sensor Networks (TPSN), radio beacon signals, GNSS or Network Time Protocol (NTP). Use cases requiring very high accuracy time synchronisation such as collaborative coherent jamming may require a time synchronisation solution in line with techniques such as optical or quantum TWTFT, however these techniques are in their infancy and would require further research and development before deployment to EA applications.

Co-Authors and Affiliation: Calum Hunter – BAE Systems

The discrimination between false and real targets is an important challenge in monostatic SAR imaging. Specifically, a deceptive jammer has the capability to introduce false targets in the focused image, potentially hiding real targets and corrupting the scene. One possible solution to mitigate and probably overcome this issue is through the use of multistatic SAR sensors. Analysing images generated from different bistatic pairs, results to the false target appearing in a different position, due to the change in the receivers’ position, while the real targets will remain the same.
The purpose of this paper is to analyse the capabilities of different multistatic geometries to reject false targets from the imaging area.

Co-Authors and Affiliation: Christos Ilioudis, Malcolm Macdonald, Carmine Clemente – University of Strathclyde

Forsberg is committed to providing our customers with resilient PNT solutions. Our GNSS receivers and protected antennas already provide robust solutions for PNT in GNSS denied environments; however, the increasing prevalence of jamming, spoofing and meaconing has required the development of sophisticated alternative-navigation (Alt-Nav) techniques. Each Alt-Nav technique can contribute a useful component to the development of a trustworthy, fused, all-source location. Alt-Nav covers a wide set of innovative navigation techniques. Forsberg has for many years provided technology solutions for situational awareness, including an image based navigation and survey product called OPTOnav. This poster seeks to describe a new real-time video-based Alt-Nav technique developed from Forsberg’s existing OPTOnav product.

Project Orion is Forsberg’s internal project name for the upgrade of the OPTOnav product. The three phases of Orion are described:

1. The application of classical Data Science techniques to image classification.
2. The introduction of limited artificial intelligence from pre-trained models to automatically track multiple reference points in a real-time video from a moving vehicle to retain position lock.
3. Integration of a Machine Learning solution to provide an end-to-end alternative/backup (Alt-Nav) navigation system.
   The use of AI in an Alt-Nav product requires the developers to balance the needs of functionality against the requirement for explainability. Later versions of ORION lean towards a “black box” approach, potentially limiting use cases. The application of AI in Alt-Nav is use-case dependent, allowing for the use of AI more in lower-risk applications where autonomy is limited. Ethical considerations are another important factor. This poster seeks to explore these competing requirements of the AI used in the Alt-Nav context.

Co-Authors and Affiliation: Nathan Daniels – KTP Associate, Lancaster University

Measurement of variations in local gravity yields information about subterranean density, which can be used to infer the presence of underground infrastructure, such as tunnels and pipes. However, environmental noise such as vibration limits the speed of performing surveys, making gravity measurements impractical for a range of use cases.
Quantum sensors that use atom interferometry to perform gravity gradiometry are emerging as a promising alternative to classical methods, due to their ability to suppress the effects of vibration and other systematics. These sensors use a single reference laser beam to prepare quantum superpositions of atoms at two different heights. The use of a common beam makes external noise sources identical, meaning these can be suppressed on performing a differential measurement.

The translation of atom interferometry sensors from the laboratory to real-world scenarios is an active area of research The Quantum Sensing group at the University of Birmingham has developed a portable atom interferometry gravity gradient sensor.

This sensor was used to detect a multi-utility tunnel beneath a road, showcasing the potential of quantum sensing for infrastructure monitoring. The sensor has also been active in trials on ships, where it has potential to act as an alternative navigation device via gravity gradient map matching. We are now investigating further trials in areas such as hazard detection, navigation and environmental sciences.

Whilst the gravity gradient provides significant information about the underground, as with all single-mode remote sensing some ambiguity will remain. We are now investigating multi-modal sensing and data fusion, combining the gravity gradient output with data streams such as magnetometry and ground penetrating radar. Our objective is to provide more actionable information to end users, for example to increase productivity in infrastructure and road maintenance and detect natural hazards such as erosion features.

Co-Authors and Affiliation: David Sedlak, Adam Seedat, Andrew Hinton, Michael Holynski – The Quantum Sensing Group, University of Birmingham

Micro-Doppler analysis of SAR signals has a broad range of applications including target vibrometry, structural health monitoring, and maritime surveillance. Location of sources of micro-Doppler signals in SAR images is however not straight- forward, where typical artefacts such as ‘ghost’ targets may be unnoticeable if, for example, a vibration is of low amplitude and frequency. This issue is exacerbated by background noise and clutter. In this work, a method for locating micro- Doppler signals using scale invariant coherent change detection between subaperture images is developed and tested on experimental SAR data with vibrating targets. The results indicate that this method is promising, with micro-motion targets being successfully identified.

Co-Authors and Affiliation: Aleksanteri Vattulainen – University of Strathclyde

A variety of techniques exist to schedule the use of a disparate mix of sensors to extract information from the environment. However, only a modest amount of research has been conducted on how to schedule computation and communication to produce useful information
and, in particular, determine what subset of a disparate mix of distributed processing is applied.

For example, in a network of sensors receiving data, future middleware must consider the long-term effects of sending data to a central server for an accurate but time-consuming analysis, versus processing the data locally near the sensor for a quicker, though less precise, result. The key challenge in this context is ensuring we can reason about what tasks we might calculate elsewhere and in the future within the network, without actually performing the calculation
itself: we need statistical models for the future distributed processing tasks.

Efficient distributed computing with future inference is important for improving resource use, decision making, scalability, accuracy, adaptability, and supports emerging technologies. To investigate this problem, we emulate an array of sensors within the Common Open Research
Environment (CORE), utilising Extendable Mobile Ad-hoc Network Emulator (EMANE) to handle network structure and interactions. Our contribution is in implementing a constraint programming structure, leveraging Google’s CP-SAT solver, to optimise criteria within the novel
use-case. Currently this produces results that allow us to measure performance of the network
based on latency of jobs and the accuracy to which they are computed.

The aim of this work is to investigate the effects that setting different constraints on the network will have, and whether this can be used to optimise a range of distinct criteria. Future work will involve exploring more sophisticated optimisation techniques such as using statistical methods
to incorporate a level of inference within the network.

This work focuses on efficient computation for wide-band RF array sensing. As established in previous research, wide-band array sensing offers several advantages, including improved resolution and reduced hardware costs. However, these benefits come with increased computational complexity, as signals across a wide spectrum of frequencies must be processed jointly. In this work, we explore wide-band RF imaging to extract range, angle, and frequency-selective responses from the observed scene, leading to an inverse problem typically of very large scale. While recent studies have examined various approaches to wide-band angle sensing, there is still a lack in addressing the computational challenges of extracting these three-dimensional parameters.

To enhance computational efficiency, we propose a sensing scheme that integrates the fast Fourier transform (FFT) into the inverse process and incorporates FFT into several steps of modern proximal algorithms. To ensure compatibility with current commercial sensing platforms, multiple narrow-band frequency modulated continuous wave (FMCW) signals are used for probing. The inverse problem is formulated by constructing a multi-dimensional FFT computation based on the greatest common divisor (GCD) of the different narrow-band frequencies. We investigate both first-order and second-order proximal algorithms, demonstrating that both gradient and Hessian computations can be efficiently performed using FFT, making these algorithms highly efficient for implementation.

Numerical simulations are presented to demonstrate significant improvements in computational efficiency and imaging effectiveness. Results show that the computation time for gradient evaluation and Hessian inversion can be reduced by 100 and 10,000 times, respectively, leading to overall processing time reductions of orders of magnitude, depending on the specific proximal algorithm employed. Additionally, the simulations indicate that the proposed sensing scheme can generate reliable RF images across various SNR levels, including as low as -10 dB.

Co-Authors and Affiliation: Dr. Wei Dai – Department of Electrical and Electronic Engineering, Imperial College London

Electromagnetic Warfare (EW) databases exist for the reprogramming of various platforms and provide a repository to contain the Electronic Order of Battle (EoB), where that EoB can become a support to a Situation Awareness (SA) picture. The emphasis for EW databases is commonly equipment reprogramming and, in many cases, it seeks to provide the best reprogramming value for that equipment. As such, they commonly wish to provide a single value, but to arrive at the singular unambiguous value, they require layers of threat Weapon System Analysis (WSA), Threat Vulnerability Analysis Countermeasure Design (TVACD) tactic development, EOB geolocation, networking, collection tasking, production, Test and Verification activities to provide an audit trail between the programming and the intelligence. That link between the evidence and the reprogramming may be weaker. The proposal here is for a Markup Language that allows the source data mapping of fields towards gaining a singular most likely encountered value; that is while accounting for all observational evidence, and all holistic rules from a body of knowledge, to aid that linkage in a complete and consistent way. That approach also provides an automated confidence weight value where that confidence is provided by a data convergence approach covering both the agreement in value and the outlier distance within the valid scope of the known body of knowledge. This approach represents a motivation for EW data management to reflect the source data ‘colourisation’ and its translation from intelligence sources, rather than just the fixed data structures for reprogramming the equipment data fields. As such, it represents a motivation not to compromise the representation of the source or the programmed equipment, in its structure. Furthermore, it perhaps represents a motivation toward knowledge graphs rather than the rigid database structures.

This work addresses the problem of gridless direction-of-arrival (DoA) estimation from the superposition of unknown source signals over a wide frequency range. Jointly processing signals across multiple frequencies can improve the accuracy of DoA estimation, but it significantly increases the dimensionality of the associated optimization problem, thereby substantially raising the computational complexity. As more frequencies are considered, the computational demands may render wideband DoA estimation impractical.

This work focuses on improving the computational efficiency of gridless wideband DoA estimation. To achieve this, we employ the Hankel matrix lifting technique, formulating the DoA estimation as a nonconvex low-rank Hankel matrix recovery problem. Specifically, we construct a block-wise Hankel matrix with a carefully chosen block size tailored to the problem. A key feature of our formulation is that the optimization matrix variable exhibits both low-rank and Hankel structures, enabling efficient computations of matrix-vector products, truncated singular value decompositions (SVD), and gradient and Newton direction evaluations. Our analysis shows that the computational complexity per iteration of our approach is $\mathcal{O}(KMF^{2} \log MF)$, where $K$, $M$, and $F$ represent the number of sources, the number of receiving antennas, and the number of frequencies, respectively. This represents a substantial improvement over other benchmark methods, which typically have a complexity of $\mathcal{O}(K^2 MF^3)$ or higher. By leveraging a modern second-order nonconvex optimization technique, our approach not only achieves a low computational cost per iteration but also exhibits a fast overall convergence rate.

Numerical experiments demonstrate the substantial gains in computational efficiency and empirical DoA estimation accuracy. The results show that our method reduces the overall runtime by several orders of magnitude while also achieving superior DoA estimation accuracy across a wide range of SNRs compared to benchmark methods.

Co-Authors and Affiliation: Wei Dai – Imperial College London

Inverse Synthetic Aperture Radar (ISAR) is a crucial radar imaging technique that relies on the relative motion between the radar and the target to generate high-resolution images. However, traditional ISAR methods are highly sensitive to inaccuracies in the estimation of rotation parameters, such as roll, pitch, and yaw. Errors in these estimates can lead to significant image degradation: underestimation of the rotation rate results in motion blur, while overestimation causes image distortions due to excessive motion compensation. In this paper, we introduce a novel approach for imaging dynamic rotating scenes using a multistatic ISAR setup, where we estimate motion parameters through Differential Semblance Optimization (DSO)—a technique originally designed for seismic imaging. By minimizing discrepancies between images formed from multiple transmitter-receiver pairs, our method enhances the accuracy of target rotation estimation, resulting in well-focused ISAR images. We demonstrate the efficacy of this approach through a series of numerical experiments, highlighting its robustness in both noise-free and noisy environments, and discuss its potential for improving ISAR imaging in complex scenarios.

Co-Authors and Affiliation: Francis Watson William (Bill) – Lionheart University of Manchester

Rate-Splitting Multiple Access (RSMA) is a powerful signal processing technique for suppressing interference in multi-antenna wireless communications systems, which outperforms Space Division Multiple Access (SDMA) – the 5G state-of-the-art.

SDMA suppresses interference by designing beamformers to minimize the signal leakage towards undesired users. This approach becomes unsustainable with increasing user density, as the signal leakage scales with interference. RSMA circumvents this problem by having each user partially decode and subtract the interference before decoding its desired signal. As an analogy, RSMA is akin to listening and tuning out background chatter at a party, while SDMA resembles seeking a quiet corner for a conversation. Hence, RSMA achieves higher data rates in the high interference regime than SDMA.

RSMA’s superior interference suppression capability enables new applications that are critical to defense, such as (i) multi-group multicast (MGM), where a message is meant for a group of users (e.g., tactical communications support), as opposed to unicast where each user’s message is unique. The one-to-many communications in MGM means that SDMA cannot effectively suppress interference for large group sizes, which RSMA is better equipped to handle; and (ii) Integrated Sensing and Communications (ISAC), where the limited available resources (spectrum and power) need to be judiciously used for both radar sensing and wireless communications. By additionally exploiting the fact that communications signals can be reused for sensing, RSMA offers a larger communications-sensing performance frontier than SDMA.

Building on OFEME 2023, where we highlighted RSMA’s superior unicast performance over SDMA using our in-house prototype, we will present experimental results demonstrating RSMA’s superiority over SDMA for MGM and ISAC this year. Through these, we hope to draw attention to RSMA’s considerable potential for defense applications.

Co-Authors and Affiliation: Sundar Aditya, Bruno Clerckx – Imperial College London

This poster presents a new remote monitoring method for extracting vibrational information from very high-resolution SAR data. This is demonstrated on the use case of structural health monitoring (SHM), where the use of vibrometry is a key technique for the maintenance of bridges and other infrastructure, conventional monitoring methods requiring in-situ sensor placement which can be costly and inconvenient. Apart from SHM, the remote monitoring technique described is also expected to be useful for a broad range of applications, such as the surveillance and classification of ships using vibrational feature maps. The approach involved using a novel method to form the signal of interest with a modified back-projection algorithm (BPA). The micro-Doppler extraction was then achieved using a combination of a spectrogram and cadence-frequency analysis, which was applied to SAR data of a bridge. The extracted vibrations were then validated against newly acquired, synchronous, in-situ ground truth measurements. The vibrational frequency estimates showed a good match to the ground truth, where for over half of these the difference was less than a frequency spectrum bin width. The validation of these results with accurate ground truth measurements is a key novelty in the field. It is expected that these promising results can be improved further, potentially using alternative micro-Doppler analysis techniques on the signal of interest formed using the BPA, where this method of signal formation imposes no restrictions on further analysis techniques.

Co-Authors and Affiliation: Finlay Rollo, Sebastian Diaz Riofrio, Enrico Tubaldi, Christos Ilioudis, Carmine Clemente – University of Strathclyde, Alessandro Lotti, Daniel Tonelli, Daniele Zonta – University of Trento, Pietro Milillo – University of Houston

With the demand for reliable access to the electromagnetic environment (EM) set to continue to grow across the civilian and military sectors (including for sensing and communication), it is becoming increasingly important to be able to robustly characterise and identify radio frequency (RF) emitters in congested and contested environments. This can support various intelligence, surveillance and reconnaissance applications (e.g. revealing malicious or unauthorised transmissions and early threat detection), cueing electronic warfare (EW) countermeasures to disrupt operations of adversaries and many others. In this poster, we present a generic and versatile machine learning (ML) framework that can profile RF emitters (i.e. learn their RF fingerprints RFFs) and simultaneously carry out one or more of the typical downstream tasks such as specific emitter identification/classification (SEI/C), emitter data association (EDA) for aiding geolocation and emitter data clustering from unlabelled data. It is applied directly to in-phase and quadrature (IQ) data of the received transmission(s), for example from a general purpose software defined radio (SDR). This approach has been developed and evaluated on several use cases such as: a) spaceborne SEI and EDA for geolocation using real RF captures from SDRs on-board LEO satellites aimed at improving maritime situational awareness; and b) recognising individual Digital Mobile Radio (DMR) handsets in urban settings from single or multiple, commercial off-the-shelf, SDRs for threat detection-tracking. Here, we show selected results from the latter two applications to demonstrate the efficacy of the introduced ML framework. We will also highlight the key enablers, challenges and future prospects of this promising RF fingerprinting technology, for instance training data requirements as per the downstream task (e.g. SEI and EDA), and its applicability to other RF emitters such as radars, EW countermeasures (e.g. jammers), beacons or transponders in maritime surveillance and others.

Co-Authors and Affiliation: Bashar Ahmad, Sam Slade – Cranfield University

Situational awareness provided by robust navigation solutions is mission-critical for military and emergency services. Global Navigation Satellite Systems (GNSS) provide the most common positioning solutions. However, GNSS positioning solutions are degraded or unavailable in many operational environments, necessitating other positioning navigation and timing (PNT) technologies. No single PNT technology, including GNSS, is sufficient in all operational environments. Therefore, a multisensor integrated system using a system of subsystems is required.

This PhD project focuses on dismounted soldier navigation on subsystems derived from inertial and magnetic measurement units (IMMU). While inertial navigation is well-researched for this platform type, deriving position solutions with map matching using magnetometers is under-researched, particularly the effect on the performance of different environments. The magnetic-derived positioning subsystem does not require new sensors but complements the commonly used IMMU system with no additional hardware.

Research from this project proves that the characteristics of magnetic flux density measurements vary by environment type, including various indoor environments, identifying specific characteristics for each environment. Furthermore, magnetometer measurements were independent of sensor height on the user in specific, but not all, environment types. In some, but not all, indoor environments, foot-level sensors have a more significant measured MFD variation than sensors placed higher on the body.

Unpredictable temporal variations in magnetic flux density can severely disrupt magnetic map matching. A significant source of these is electrical power cables; research on how underfloor cabling affected magnetometer measurements has been conducted. The variables examined include the type of magnetometer, electrical power transmitted through the cable, the cable type, and the distance from and along the cable. Because some electrical power is alternating current (AC), its magnetic field can be separated using frequency-domain filtering; however, this requires high-bandwidth magnetometers. This knowledge can be used to improve the magnetic-derived positioning subsystem, improving the overall multisensor integrated positioning system.

This research, conducted at University College London (UCL), is funded by the Engineering & Physical Sciences Research Council (EPSRC) and the Defence Science and Technology Laboratory (Dstl) as part of the ICASE scheme.
Co-Authors and Affiliation: Paul D. Groves, David R. Selviah – University College London, Owen J. Griffiths, Robin J. Handley – Dstl

Geolocation of an unknown transmitter is a great tactical advantage in modern electronic warfare. This study aimed to evaluate the potential of using passive UAV reflectors to produce position estimates of an unknown transmitter. A model is developed to conduct inverse path calculations of reflected transmitter signals that are purposefully directed towards a friendly ground based receiver. This has the potential advantage of Beyond Visual Line Of Sight (BvLOS) electronic surveillance at a ground based receiver, without requiring the UAV reflector to process the incoming signal. The simple nature of the UAV reflector has the potential to provide a rapid cost effective capability using a single UAV with minimal modification. The capability could be enhanced with multiple UAVs reflecting signals to the ground based receiver, allowing for additional measurements of transmitter location and redundancy. To evaluate the technique, a preliminary system has been developed, with ray tracing, to test its effectiveness and potential challenges. The challenges encountered include reflector position errors, and reflected power at the receiver site.

Co-Authors and Affiliation: Anil Shukla – QinetiQ

The EM-CAT or the Electromagnetic Activities Planning tool brings together multiple information feeds to inform, understand and control the EM battlespace.

It can be used to understand and process live Cyber & Electromagnetic Activity data streams, scenario planning, war gaming manoeuvres and replaying real world EM activities to understand the effectiveness of a given tactic or strategy.

EM-CAT is also a control interface for deployed real world CEMA assets, allowing the detection and identification of real world signals of interest. Once a signal of interest is identified the EM-CAT can be used to decide a response to a particular signal or manoeuvre by blue or red forces, e.g. to trigger an electronic attack, to degrade the capabilities of red forces or deceive them. In particular the EMCAT is able to repurpose remote SDR sensors to dynamically update them from sensor to effect and vice-versa according to the operational requirement.

Whatever mode the EM-CAT is in an operator is able to model a 3D picture of RF propagation with terrain data to understand how an EM emission, such as a communications system or an electronic attack will propagate and effect blue units or red forces.

The EM-CAT includes a database of EM equipment to enable an operator to decode and refine the Electronic Order of Battle (EOB) both for blue and red forces. EM-CAT utilises standard C2 protocols providing integration with the wider CEMA enterprise. This allows it to operate at both the tactical and strategic level, configuring and tasking both local and remote sensors and effectors.

Co-Authors and Affiliation: BAE Systems, Dstl

Traditional PNT using GPS or other GNSS has known vulnerabilities in the battlefield, from intentional jamming and spoofing to unintentional interference and limitations due to the environment. Alternative Positioning, Navigation and Timing (APNT) is a range of techniques, including RF, Visual and high-accuracy inertial devices. RF APNT to date has exploited terrestrial broadcast signals (“signals of opportunity”), such as TV, Radio and Cellular transmissions, leveraging time-of-arrival (TOA), carrier phase (CP) and angle of arrival observations for derivation of a PNT solution. Terrestrial APNT has limitations, however, including variation of density and distribution of transmitters and varying transmission standards. Importantly, though, for deep sea and littoral operation, terrestrial APNT is limited by coverage to the extent where it is not useable.

There is a growing number of low earth orbit (LEO) satellite networks provisioning voice, messaging and data services to subscribers. These networks provide global coverage, including Iridium, Orbcomm, Globalstar and Starlink. Transmissions can be readily received by simple SDR or similar receivers. Doppler-based PNT techniques can be exploited.

A demonstrator was put together using several in-house software-defined radios (SDR), COTS antennas and frequency reference and back-end MATLAB software to capture and process transmissions from several networks, the Ring Alert channel on Iridium (L band), Orbcomm (VHF) and Starlink (Ku band). Initial results were obtained with a single satellite, following on with results from two satellites. These results indicate better than 1km positioning accuracy with minimal processing.

This low TRL work has demonstrated that Doppler-based PNT is achievable using LEO satellite transmissions. Inclusion of observations from several networks provides a 3D positioning capability using simple COTS receiver hardware and processing. Long term plans include development of TOA PNT using wideband transmissions (e.g. Starlink) to supplement the Doppler-based approach, and resilience of the processing developed to date.

Co-Authors and Affiliation: David Mcilwraith, Rory Jenkins – BAE Systems

Radio Frequency (RF) sensing involves monitoring RF signals affected by human presence to detect movements or gestures using Channel State Information (CSI). Movements affect the signals, creating identifiable patterns. These patterns can be recognised by AI algorithms to classify human movements.

RF sensing’s civilian uses extend to defence, such as monitoring wounded soldiers’ movements or vital signs, offering a cost-effective alternative to wearables. Gesture recognition can aid in silent military operations with autonomous robots, and surveillance systems can detect intruders in secure areas.

Movements affect channel frequency response, observable in RF communication with Orthogonal Frequency-Division Multiplexing (OFDM). Utilizing existing Wi-Fi for CSI sensing is practical, but jamming Wi-Fi signals, common at 2.4 GHz and 5 GHz, can degrade these systems. Inexpensive Software Defined Radios (SDRs) can jam these frequencies by transmitting noise.
Bluetooth, also operating at 2.4 GHz with Wi-Fi, uses Frequency Hopping Spread Spectrum (FHSS) to mitigate congestion, jumping between frequencies to avoid interference. While FHSS aids communication in congested areas, SDRs can jam a wide bandwidth covering Bluetooth channels.

This work implements an RF sensing system using FHSS to avoid jamming by switching between WLAN channels 6 (2.437 GHz) and 116 (5.580 GHz). This enables protection from noise jamming on a single spectrum. The challenge lies in training AI algorithms to interpret CSI patterns with a changing centre frequency. This paper seeks to achieve a binary classification of sitting and standing movements, testing the proposed FHSS efficacy under jamming conditions. Experimental results are presented to demonstrate the feasibility of robust RF sensing systems with anti-jamming capabilities.

Co-Authors and Affiliation: Akram Alomainy – Queen Mary University of London, Oluwakayode Onireti, Shuja Ansari – University of Glasgow

In this work we detail an ongoing project for developing a convolutional neural network model tasked with automated processing of pulse compressed maritime radar signals to detect and classify targets. We discuss the design and performance of the current model. The training data contains proprietary IQ radar signals collected by BAE Systems maritime radar division. Based on the nature and the quantity of the training data, we devised bespoke modelling and training strategies. The presented model is capable of accurately categorizing a target’s speed based collected radar signals. Moreover, we used state-of-the-art (SOTA) explainability models for determining how the black-box model comes to its decisions in categorising the input signals. We will also describe our strategy for addressing the potential difficulties with generalising the model performance to out-of-distribution datasets. Specifically, we will describe how SOTA adaptive and continuous learning strategies are used to ensure the model can be used in new scenes, without the need for extensive retraining.

Co-Authors and Affiliation: Matthew Stewart, Chris Willis, Phill Clark – BAE Systems

Joint Radar and Communications (JRC) has gained significant attention due to spectrum scarcity, emerging 6th Generation (6G) cellular technologies and increase in number of wireless devices. This has created an urgent need to combine radar and communication functions. In disaster scenarios where conventional cellular infrastructure is damaged or overwhelmed, JRC functions can be crucial in supporting rescue operations. JRC offers solutions to spectrum congestion and unlocks new benefits beyond what independent sensing and communication could achieve. Currently, most of the research in this domain revolves around beamforming techniques, maximizing sensing and communication performance through constrained optimization, maximizing energy efficiency, formulating theoretical performance limits and channel estimation. However, given the numerous applications of JRC systems such as in military and commercial scenarios, to the best of our knowledge, no work has yet addressed the detection of JRC based systems. This entails inferring the presence of JRC behaviour or determining whether a particular transmitter operates as a JRC system. This is crucial in detecting and monitoring JRC enabled adversaries and enforcing regulatory measures on JRC systems for appropriate resource utilization. In our work, we consider the problem of detecting the presence of a cooperative based JRC system that monitors radar targets and communicates to users simultaneously. We use a number of randomly and spatially distributed isotropic sensors that detect the beamformed radar and communication signals over an extended observation period. We develop metrics, filters and a detection scheme that makes use of various observed signal parameters (such as power and angle of arrival) to determine whether the received signals originate from a JRC system or not. Our results show that our technique is able to detect JRC behaviour accurately even at low SNRs with few sensors, and we demonstrate the effect of varying the various system parameters on the overall detection accuracy.

Co-Authors and Affiliation: Dr Ahmet Burak Ozyurt, Dr John Thompson – Institute for Imaging, Data and Communications, The University of Edinburgh

YOLO is an machine learning based object detector that can be trained to detect specific objects in images and videos.

By using training YOLO on the time-frequency characteristics of radar signals and then applying it on rolling spectrograms generated from IQ recordings, YOLO can provide automatic modulation classification of pulses and can also localise them in the spectrogram, thus providing useful parameters such as the bandwidth, time-of-arrival and pulse width which are key to ESM processing further down the chain.

Unlike time-domain methods, YOLO can easily manage multiple signals simulatenosuly.

Co-Authors and Affiliation: Matthew Ritchie, University College London

Rate-Splitting Multiple Access (RSMA) is a powerful signal processing technique for suppressing interference in multi-antenna wireless communications systems, which outperforms Space Division Multiple Access (SDMA) – the 5G state-of-the-art.

SDMA suppresses interference by designing beamformers to minimize the signal leakage towards undesired users. This approach becomes unsustainable with increasing user density, as the signal leakage scales with interference. RSMA circumvents this problem by having each user partially decode and subtract the interference before decoding its desired signal. As an analogy, RSMA is akin to listening and tuning out background chatter at a party, while SDMA resembles seeking a quiet corner for a conversation. Hence, RSMA achieves higher data rates in the high interference regime than SDMA.

RSMA’s superior interference suppression capability enables new applications that are critical to defense, such as (i) multi-group multicast (MGM), where a message is meant for a group of users (e.g., tactical communications support), as opposed to unicast where each user’s message is unique. The one-to-many communications in MGM means that SDMA cannot effectively suppress interference for large group sizes, which RSMA is better equipped to handle; and (ii) Integrated Sensing and Communications (ISAC), where the limited available resources (spectrum and power) need to be judiciously used for both radar sensing and wireless communications. By exploiting the fact that communications signals can be reused for sensing, RSMA offers a larger communications-sensing performance frontier than SDMA.

Building on OFEME 2023, where we highlighted RSMA’s superior unicast performance (higher data rates) over SDMA using our in-house prototype, we will present experimental results demonstrating RSMA’s superiority over SDMA for MGM (higher data rates and more fairness) and ISAC this year. Through these, we hope to draw attention to RSMA’s considerable potential for defense applications.

Co-Authors and Affiliation: Sundar Aditya, Bruno Clerckx – Imperial College London

The MWIR (mid-wave infrared) spans the 2-10 micron wavelength range of the electromagnetic spectrum. The MWIR includes the atmospheric windows at 3-5 and 8-10 micron wavelength, with potential for atmospheric communications, direct chemical sensing, counter-measure applications in defence and collision-avoidance.

In 2021, we reported the world’s first ever continuous wave MWIR fibre laser operating > 4 micron wavelength. Such fibre lasers promise far higher beam quality than any other MWIR lasers viz.: OPOs (optical parametric oscillators); ICLs (interband cascade lasers); QCLs (quantum cascade lasers) and gas lasers like CO2 and CO. In addition, fibre lasers exhibit large energy storage which uniquely enables high peak power pulses useful for generating MWIR wavelengths via nonlinear processes.

The PhD Programme builds on this first, MWIR >4 micron lasing fibre, extending the technology to include investigation of pulsed behaviour, further rare earth ion emitters across the MWIR and applying Bragg grating cavities, with demonstration of fibre laser sensing of chemical moieties of interest. The chalcogenide glass fibres, on which the brand new fibre lasers are based, are low loss, stable at ambient temperature, can be co-doped for potential tuneable emission across 2-10 micron wavelengths and pumped with low-cost near-infrared laser diodes. The chalcogenide glass low phonon energy and high refractive index means low non-radiative decay rates for efficient light emission and high cross-sections for short devices.

AIM & OBJECTIVES
This PhD project will produce breakthrough fibre lasers with pulsed operation and explore laser operation in the 4 to 7 m wavelength region. Objectives:

Explore other lanthanide ion dopants to cover MWIR

Fabricate double clad fibre for efficient pumping

Write Bragg gratings for internal cavity

Apply 5 micron acousto-optic modulator to cavity for pulsed operation

This paper proposes a technique for near-field source localization using an extremely large-scale antenna array (ELAA) while considering direction-dependent mutual coupling (MC). The three-dimensional (3D) multiple signal classification (MUSIC) algorithm – which is a high-resolution algorithm – can be used to estimate the near-field source location in the presence of MC. However, this method suffers from a high computational load due to the requirement to perform 3D-based search steps. We address this issue by using one sub-array of ELAA which holds far-field assumptions to estimate the direction of arrival (DOA). The DOA estimation relies on a one-dimensional (1D) MUSIC-based algorithm, drastically reducing the computational complexity. Then, the MC coefficients are calculated using the estimated DOAs. Finally, another 1D MUSIC step is used to estimate the ranges corresponding to the estimated DOA and MC coefficients. Simulation results verify the superiority of the proposed near-field source localization in the presence of MC in terms of computational complexity while offering comparable accuracy with computationally expensive multi-dimensional search-based techniques. By taking into account electromagnetic (EM)-informed phenomena, such as the ELAA MC response, the proposed technique also offers a path towards practical scenarios involving wireless channel characterization, a key requirement for communications and sensing in EM-congested environments.

Co-Authors and Affiliation: Zohreh Ebadi, Amir Masoud Molaei, Muhammad Ali Babar Abbasi, Simon Cotton, Okan Yurduseven – Queen’s University Belfast, Anvar Tukmanov – BT Group

Optical fibres are a convenient transportation medium for light within optical sensing. Single-mode fibres, which maintain spatial coherence, are the common fibre type when fibre coupling within an interferometric optical sensor. However, multimode fibres carrying many spatial modes, can increase the étendue of a system. Having an increased étendue enables the multimode fibre to be used as a more effective collection optic than a single-mode fibre with a limited étendue. We have demonstrated that multimode fibres can be used in interferometric sensing as a collection optic. Using a multimode fibre, with a 910 μm core and 0.22 NA, as the only collection optic, to collect backscattered light from a vibrating surface, we have measured signals across the bandwidth of the sensor, between a range of frequencies from 212 Hz to 2199 Hz, at volumes between 51.3 dB and 112.4 dB, with a signal to noise ratio up to 45.0 dB. By applying an off-axis holography technique we can recover the full complex amplitude of the transmitted speckle pattern in a single frame. Applying a Gaussian mask in the Fourier plane to remove the DC-components and recover the phase and amplitude of each frame, we measure the phase shifts driven by a sound source positioned behind a target surface. We collect the global change in phase, from the frame before, and apply a Butterworth bandpass filter, set between 70 Hz and 2398.5 Hz, the minimum frequency the sound source can produce and the Nyquist limit of the system, respectively. The Mach-Zehnder interferometric sensor, applied as a laser vibrometer has also recovered complex sounds such as human speech, with the sound source being directly from a person, positioned over 1 metre from the vibrating surface.

Co-Authors and Affiliation: Osian Wolley, Hao-Wei Hu, Simon Peter Mekhail, Miles Padgett – University of Glasgow

Synthetic aperture radar (SAR) has been a mainstay of the remote sensing field for many years, with a wide range of applications across both civilian and military contexts. SAR imagery is challenging for a human operator to analyse and requires specialist knowledge of the sensing modality to interpret reliably. Historically, the lack of openly available datasets of comparable size and quality to those available for optical imagery has severely hampered work on open automation problems such as automatic target recognition, image understanding and inverse modelling.

This poster presents an overview of new modelling techniques based on the SARCASTIC v2.0 engine. SARCASTIC is capable of supporting full-complex phase-sensitive simulations at high fidelity and providing diagnostic outputs whilst maintaining near-realtime performance on complex scenes. This enables the application of inverse modelling techniques to iteratively interrogate a collection and refine hypotheses about the targets under investigation. This capability is currently being used to develop novel ATR processing for hyperfine-resolution systems in defence applications.

By way of demonstration, the initial modelling results for a large structural target are shown and compared to real data captured by an UMBRA SAR satellite sensor. The outputs from this proof of concept demonstrate that the expected modelling properties of the approach are valid and that the SARCASTIC outputs can be processed through the appropriate tooling to generate direct comparisons between the real and simulated data. Future work will focus on refining the inverse modelling approach to exploit more of the information available in the real capture and improving the fidelity of the generated results.

Co-Authors and Affiliation: Dr Matthew Ritchie, Prof Hugh Griffiths

The number of satellite launches is growing exponentially; commercialisation of various space-based imaging services and the number of space-capable nations continues to rise; and on-orbit collisions between active satellites and other objects is an ever present and growing risk. In tandem, the amount of observational and technical data available on Earth satellites is growing, delivering a more complex but potentially more veracious view of the situation.

These two factors – increasing domain congestion and an overload of data, presents a significant challenge to orbital analysts attempting to maintain space domain awareness. However, it is shown that Machine Learning (ML) techniques can potentially offer assistance to analysts managing these burdens: ML can leverage high volumes of data to provide timesaving, wider-reaching, or novel capabilities, and release human effort to focus on complex or higher-priority issues.

This work demonstrates the feasibility and effectiveness of machine learning models for satellite characterisation. Dstl have developed models that successfully predict the operational status of satellites using both optical and RADAR data, (with an accuracy of around 86% for LEO, and 92% for GEO). Notably, the accuracy of the ML assessment only dropped by a few percent when only presented with a single track (observed overhead pass) rather than the entire dataset, demonstrating the utility of even a single light-curve collection when using these characterisation techniques. We also developed machine learning models to identify the bus type of a satellite from the same data (with an accuracy of >98%).

We found classical ML models (e.g. Random Forests) performed just as well as Deep Learning approaches: our LSTM (long short-term memory) models struggled to retain their high performance when generalising. The full range of characterisation tasks tackled by analysts is significantly broader than those investigated here, and considerable scope remains for further research.
© Dstl Crown Copyright 2024

There is growing interest in the development of secure quantum sensing schemes, both for their potential applications in insecure and noisy environments and the fact they provide a new example of hybridising quantum technologies, giving us a better understanding of the advantages of doing so. Promising results have already been achieved for single sensors [1,2] and attention is now turning to developing networks of quantum sensors. In such cases, entanglement is known to give a metrological advantage when we are interested in measuring a function of the parameters at each sensor [3]. However, it is also known that such an approach becomes exponentially inefficient (with the number of sensors) in detecting an eavesdropper [4]. One way around this is to use a mixture of entangled and separable states [5]. Another approach is to not use entanglement at all. This preserves the security and has SWaP advantages. Here we demonstrate preliminary results showing that the precision of the parameter estimation can be significantly enhanced in an entanglement-free scheme by using the James-Stein estimator to process the data.

Co-Authors and Affiliation: P. Yin, S.W.Moore and J.A.Dunningham – AVS Quantum Science, T.J.Proctor, P.A.Knott, Z.Huang, C. Macchiavello, L.Maccone

Currently the choice of attack techniques, used to deploy an effect against a target, are based on knowledge of the platform and system which signals are emitted from.

However, given limited or inaccurate information about the target system, it may not be possible to accurately identify it. In this work, we investigate an approach to circumvent the need to explicitly identify the system using Knowledge Graphs.

Knowledge Graphs allow information to be stored as entities, relationships, and features in a graph structure. Using domain knowledge a simple but representative knowledge graph is constructed, consisting of signal characteristics which may be observed, systems which they may be observed from, and attack techniques which may be applicable to these systems. In addition, the knowledge graph is extended to incorporate signatures, where previously observed systems can be matched according to their observed signal characteristics and position.

A rules-based path tracing algorithm is then developed in order to demonstrate the matching process on this knowledge graph. It is shown that the algorithm, given a set of observed signal characteristics and a desired effect (deny, degrade, deceive), is able to make valid technique recommendations, without knowing the system which has been observed. The algorithm developed is fully explainable and interpretable, with technique recommendations being traceable back to the observed characteristics which led to them (and the systems identified in-between). Further, the algorithm is expanded to incorporate confidence metrics, which allow uncertain observations to be factored for in the recommendation process. This further allows valid technique matching even if one of the signal characteristics is observed incorrectly.

Finally, the limitations of the current approach are investigated and identified, and recommendations for further work to overcome these, expand upon the current capabilities, and incorporate new information, are provided.

Co-Authors and Affiliation: Mohammad Beit-Sadi, Lee Mackmurdie

Learning the behaviour of the target of an electronic attack for optimising the method of attack is modelled as a multi-armed bandit problem. An Active Inference-inspired model is proposed to learn the target’s behaviour and select the optimal attack strategy. The proposed model is compared to random and sweep jamming. The percentage effectiveness of the proposed is highly dependent on the standard deviation of the probability distribution used by the target to select the channels used. The proposed model achieved over 60% jamming effectiveness for a target selecting communication channels at random.

Co-Authors and Affiliation: Dr. Shuja Ansari, Dr. Oluwakayode Onireti – University of Glasgow

This study demonstrates a proof of principle that Reinforcement Learning (RL) can be used in a scheme to manage the power output of radio transmissions with the intent of maximising battery life. A secondary benefit is that minimising the transmit power may also lead to a reduction of interception probability.

In this study the RL agent is rewarded based on a target receiving a specified signal strength relating to a minimum quality of service. The agent receives less reward if it fails to hit, or exceeds, this target signal strength and maximises its reward as it approaches this target signal strength. The agent then learns to transmit with just enough power to achieve the objective received power, thus preserving battery life.

This study explored two different scenarios, in the first, a single transmitter moves along a path in Canary Wharf transmitting to a fixed receiver, with various obstructions in the transmission path. In the second, a transmitter moving along a path in Cannock Chase transmits alternatively to two receivers. One that is in close proximity to the transmitter and another that is very distant.

Matlab provides a signal propagation environment which is used as the synthetic environment in which the reinforcement learning agents can be developed. Matlab is also used to specify and train the agents which streamlines the development path.

This work proves the concept of using reinforcement learning in order to moderate transmit power and thus preserve battery life. However, although a proof of concept is demonstrated more work is needed to understand and optimise the reinforcement learning agents, how these agents would be deployed operationally and how they would make use of real measurement data during deployment as this may differ from the synthetic performance used to develop the agent.

Co-Authors and Affiliation: Anil Shukla – QinetiQ

The rapid evolution of wireless communication systems requires transceiver architectures to dynamically adapt to varying frequency conditions, optimizing spectrum utilization and transmission diversity. Power amplifiers (PAs) are critical in these systems but face challenges with efficiency, linearity, broadband operation, and thermal management.

Aims:
This research aims to:

1. Develop High-Efficiency, Linear PAs: Investigating novel materials and circuit topologies to improve efficiency and linearity across a wide frequency range.
2. Explore Advanced Cooling Techniques: Managing thermal output to enhance reliability in high-power environments.
3. Design Broadband PAs: Creating PAs that support broadband operation to enable dynamic spectrum sharing.

Methods:
The research focuses on developing multistage Doherty PAs under various biasing conditions, exploring innovative materials, and using simulations to assess performance in terms of efficiency, thermal management, and broadband capabilities.

Expected Outcomes:
Although in its early stages, the project is expected to result in designs that address current limitations, improving PA efficiency, thermal management, and broadband operation. These advancements will contribute to more flexible, scalable transceiver architectures, enhancing spectrum utilization and transmission diversity in next-generation wireless systems.

As human factors professionals, we are always keen to be involved earlier in the process. We have clear impact in design. But what about concept stages and even earlier. When we are asked to consider ‘Generation After Next’ how do we deal with that? We’re designing for a generation that isn’t born yet, considering the impact of technology and systems that may only be an idea, if that. It has always been a challenge for human factors to be applied in the early stages of design, let alone the concept stages. Development of applying human factors so early is certainly a step in the right direction. Current PhD research is tackling this challenge – how do you predict the future, understand the decisions that will be made, the context and environment they’ll be made in and how we support them.

Co-Authors and Affiliation: Dr Ella-Mae Hubbard

The advent of 5G technology has brought significant advancements in communication capabilities, particularly through the innovative concept of Integrated Access and Backhaul (IAB). This poster explores the potential of 5G IAB in defence applications, focusing on its ability to enhance communication networks in challenging environments.

5G IAB offers a flexible and efficient solution by combining access and backhaul links, enabling rapid deployment of communication infrastructure with minimal reliance on wired backhaul. This capability is particularly advantageous in defence scenarios where traditional infrastructure may be compromised or unavailable. The poster delves into specific defence use cases, including battlefield communications, secure data transmission in remote locations, and support for autonomous systems and drones.

Through a detailed analysis of these use cases, the poster highlights how 5G IAB can improve situational awareness, decision-making, and operational efficiency in defence operations. Additionally, it discusses the challenges associated with implementing 5G IAB in military environments, such as ensuring cybersecurity, maintaining network resilience under hostile conditions, and managing spectrum resources.

The findings presented in this poster emphasize the transformative potential of 5G IAB in modern defence strategies, providing a robust framework for enhancing communication capabilities in the field. As defence forces worldwide look to leverage emerging technologies, 5G IAB stands out as a key enabler of future military operations.

We present state-of-the art modelling and design environment for a range of Electromagnetic Applications – from EMC to Photonics. Originally developed at UoN and currently being sold by Dassault as CST software package, the time-domain numerical method Transmission Line Modelling (TLM) method has been re-defined to operate on an unstructured Delaunay mesh. This gives us distinct advantages compared to its cartesian equivalent i.e. reduced numerical noise as a result of smooth discretisation. Other advantages of the method are: a) Broadband time domain method not restricted to any type of materials/excitations; b) Unconditionally stable algorithm; c) not restricted to any type of geometry; d) multiscale capability, e) massively parallelised; f) fully integrated with industrial scale CAD and process control requirements.

However, it is not just the TLM methodology we want to highlight. Our innovation in generating clean geometries and innovative 3D Delaunay meshing capabilities are fundamental to our ability to consider industrial scale problems.

Our talk/poster would briefly highlight main advances in the methodology and the holistic approach we take to EM design and validation. This will be followed by some illustrative examples that highlight our state-of-the art capability. Some typical examples include:

• Complex carbon fibre composite structural joint (for which we received the BAE Chairman’s award for “Predicting the Effects of Lightning);
• Installed antenna modelling – antenna on an airplane wing or inside the airborne radome;
• Impact of radome lightning diverter strips on antenna performance.

We conclude that we combine the latest advances in geometry generation, meshing and electromagnetic simulations. This makes us uniquely placed to perform electromagnetic simulations on industrial scale problems. We believe that our capability is unique in the world.

Co-Authors and Affiliation: Phillip Sewell – University of Nottingham

Satellite-based synthetic aperture radar (SAR) is widely used for imaging the earth surface. However, existing SAR satellites are large instruments designed to satisfy multiple earth-observation applications. The high costs of developing and launching large (>1000 kg) satellites means these are typically operated individually (e.g., TerraSAR-X) or in small constellations (e.g., RADARSAT Constellation Mission, which uses three satellites). Accordingly, the revisit time for existing SAR systems limits these platforms for real-time earth observation, e.g., vessel detection or following a natural disaster.

The growth in nano-satellite technology and the reduced development and launch costs of these platforms have led to the idea of SAR satellite constellations, which would enable near real-time imaging. However, the size and mass restrictions of nano-satellites place significant limitations on a SAR instrument. In particular, the maximum power available on a nano-satellite is typically under 100 W, and the small form-factor limits the size of the antenna aperture. The reduced power and antenna aperture will significantly reduce the quality of the SAR image (i.e., increased noise levels) and limit the spatial resolution capabilities. Previous research has identified several use-cases, namely vessel detection and earth deformation monitoring using persistent scatterer interferometry, where nano-satellite SAR can provide meaningful results in the presence of increased noise.

This poster will present some of our recent theoretical, simulation, and experimental work around developing SAR systems suitable for nano-satellites, including: the impact of hardware impairments present in commercial-off-the-shelf (COTS) radio frequency components (e.g., mixers and amplifiers), the design of deployable antennas (and impact of antenna vibrations), and the use of simultaneous-transmit-and-receive (STAR) technology to reduce peak power consumption.

Co-Authors and Affiliation: Prof. Mark Beach, Dr Geoff Hilton – University of Bristol

We present a versatile and flexible environment for the development and evaluation of adversarial Reinforcement Learning (RL) algorithms for Cognitive Electronic Warfare (CEW) concepts. The focus here is on a cognitive jammer versus a cognitive or multi-function radar with Electronic Counter-Counter Measures (ECCMs) capability. The former continuously assesses its impact on the latter’s performance and actions as well as changes in the Electromagnetic Environment (EMA). The CEW then dynamically adapt its jamming strategy to maximise the efficacy of its Electronic Attack (EA). To this end, we pose the CEW problem as an adversarial learning task, this is radar (i.e. red) agent versus jammer (i.e. blue) agent. This is formulated in the context of a practical scenario where both agents have realistic observation(s) from their counterparts. Results from simulated data and applying state-of-the-art adversarial deep reinforcement learning algorithms demonstrate the potential of the introduced Adversarial Multi-agent Deep RL (Adv-MADRL) framework for novel CEW concepts. For example, the introduced Ad-MADRL approach enables the CEW system to learn robust jamming strategies to conceal a key (own) asset from being detected by the adversary radar. Thus, it can enable disruptive closed-loop EA and EA mission-planning capabilities, such that the jammer can autonomously establish the “optimal” Course Of Action (COA) and learn to quickly respond to the victim system’s ECCMs from limited noisy observations of radar parameter-actions and the EM environment. This can also substantially reduce the burden on the human operator and facilitate timely and possibly automated decision-making (e.g. by providing best COAs) in complex and congested environments.

Co-Authors and Affiliation: Alex Hiles, Daniel Harrold, Sam Slade – BAE Systems Digital Intelligence

We present an overview of a unified machine learning (ML) framework for tasking swarms of UXVs (e.g. uncrewed air systems, UAS or drones, ground vehicles UGVs and underwater vehicles UUVs, etc.) such that they can dynamically self-organise, in a decentralised way, and autonomously navigate the scene. This is in order for the swarm to collectively achieve a particular goal, such as improving situational awareness in ISR contexts across all physical domains (including for threat detection-tracking and enabling the full targeting cycle), maintaining data connectivity for beyond-line-of-sight (BLOS) communications, deception, distributed RF sensing and more. The introduced, learning-algorithm agnostic, framework is underpinned by a principled Adversarial Multi-Agent Deep Reinforcement Learning (Adv-MADRL) that incorporates Graph Machine Learning (GML) techniques. This ensures that: a) the learnt blue (i.e. swarm) policies are robust and generalise across many possible scenarios and against (red) opponents of varying capabilities (i.e. overcoming susceptibilities to overfitting or exploitability typically associated with MADRL methods); b) UXVs can maintain local connectivity and efficiently share information (e.g. with multi-hop communication). UXVs thus need not communicate directly with a remote command node or operator to accomplish their objective(s), which can be difficult in the contested environments (including electromagnetic environments) of the contemporary battlespace. After outlining the various challenges and enablers of the proposed ML framework for swarming UXVs, we will present results from a recently completed DASA-funded project focused on underwater surveillance, namely to detect high value threats detection (e.g. submarines), using a swarm of low-cost, small, UUVs.

Co-Authors and Affiliation: Mohammad Beit-sadi, Bashar Ahmad – BAE Systems Digital Intelligence

We typically consider physics, electronics and maths in designing systems; additional inspiration can be found in nature, a two-billion-year-old battlefield. This inspiration can take the form of how nature conducts particular operations or strategies, but realised with conventional technology, or to leverage biology in order to deliver effects.

Due to the breadth of possible areas of inspiration provided by the natural world we have split the biomimetics space into three thematic areas: biology-based technologies; biology-inspired technologies; and biology-inspired strategies. Each thematic area has potential to enable new capabilities in systems for applications in the EM domain and beyond. Our research revealed interesting and initially unexpected advantages of using biology as inspiration for designing systems: directly applying biology to EM activities benefits in some instances from the inherent properties of biological systems (e.g. self-healing, low power consumption, covertness, and EM resilience); drawing inspiration from biological systems can enhance performance, create new design principles or enable new capabilities, whilst retaining the core performance of established manmade systems; and strategies taken from nature have the potential to enable capability enhancement across a range of use cases.

Example use cases include understanding how predator-prey behaviour could be used to create operational strategies, and similarly how immune system behaviours could increase operational resilience; both of these examples can be expressed as algorithms, running on conventional hardware, with complex emergent behaviours resulting from a simple series of rules. We can also consider how biology creates and uses resources that we need operationally, such as energy and modulation; this can inspire technologies both based on the architecture of biological systems, and/or the biology itself.

Prototyping biology-inspired and/or biology-based solutions is accessible in the near term and does not necessarily require high performance bio-science capability.

Co-Authors and Affiliation: Millie Rose, Richard Claridge – PA Consulting

Bright Corvus is maturing the technology and understanding of the military benefits of Distributed RF Sense and Effect. It has progressed UAS based Multi Static Radar, Electronic Attack, Coherent Electronic Surveillance, Foliage Penetrating Radar and TDOA ES to TRL4/5 using low SWAP-C Software Defined Radios.  The work is domain and host platform agnostic.  The project contains a strong Position and Timing element to help enable the path towards coherency.  Funded by Spending Review 2020 work ceases Mar 25.  The project team are keen to find exploitation partners who might fund or encourage progression of one or more of the techniques to TRL6 over the next 2-3 years. 

The EM environment is crucial to the Defence sector’s operations, yet it is becoming increasingly congested and complex. To maintain effective performance, respond to evolving threats and deliver improved effects, there is a need for the Defence sector to adapt new technologies which bring about disruptive capabilities. PA Consulting’s recent work with Dstl has identified a broad range of novel technologies and devices, spanning photonic, spintronic, solid-state transduction and analogue compute technologies, as well as identifying innovative approaches to RF transmission. Many of the 150 disruptive technologies had the potential of enabling Generation After Next (GAN) capabilities in the EM environment. We identified potential academic and/or industrial partners for selected technologies and articulated potential development approaches to reach tangible milestones. Our research highlighted that many GAN technologies act as enablers for adjacent technologies, creating systems that offer a significant step change in capability over existing technologies. Whilst some technologies will be mature through market forces, several technologies will require the Defence sector to take the lead in their development for future EM and EW systems.

Co-Authors and Affiliation: PA Consulting

Prevalent wireless connectivity becomes indispensable in the modern society, which demands high-level security in wireless communication systems. This requirement is paramount for defence applications in congested EM environments. Wireless data transfer, however, is particularly vulnerable due to their broadcast nature.

Current wireless systems are protected by mathematical-based encryption techniques, which are exclusively applied in higher communication layers. However, for some mobile applications the battery-powered communication nodes have very limited computation capability and have limited (or no) access to secret key management infrastructure. These constraints render the conventional encryption method less secure, or even infeasible. More critically, the current cryptographic encryption can be readily cracked with the emerging quantum computing that will be available in the near future.

In this context, focus is now on physical-layer wireless security. Unlike classical cryptographic methods that aim to conceal information from eavesdroppers (using mathematical methods) in the digital domain, physical-layer wireless security ‘contaminates’ or ‘obliterates’ unintended information leakage in the analogue radio frequency waveform domain. This is achieved by exploiting randomness as a ‘security key’, which is inherently present, e.g., wireless multipath channels and radio frequency fingerprints (RFF) in frontend hardware, and that is uniquely shared among legitimate communication parties. Physical-layer wireless secure communication techniques contaminate the analogue waveform such that the useful information signals are masked for eavesdroppers located away from the legitimate recipients. Consequently, the capability to extract information away from the legitimate receivers is fundamentally limited and absolute wireless security can be achieved. This approach is, thus, resilient to quantum computing.

In this poster, we would like to showcase our research progress on physical layer security techniques including; exploitation of power amplifier nonlinear memory effects as RFF for secure device authentication; autonomous wireless link security via directional modulation implemented in advanced RF hardware such as Fourier Rotman lens and retrodirective arrays.

Co-Authors and Affiliation: George Goussetis – Heriot-Watt University

We develop a two-dimensional wireless channel model for wave propagation through scattering clusters from first electromagnetic (EM) principles. The mathematical model is devised through the many-body scattering EM theory and is: i) physics based, as it includes the electromagnetic interactions among objects; ii) unified, as it is valid in both near- and far-field; iii) universal, as it can work at arbitrary frequency ranges, also modelling objects of different shape and materials. The dipole approximation is employed to devise a general mathematical framework for EM fields scattered by objects of arbitrary shape, as represented by clusters of dielectric cylinders supporting only one scattering mode each.

We apply the proposed model to analyse the trade-off between communication performance and localisation accuracy of MIMO communications for the sixth generation (6G) wireless communication systems.

Our contribution will propose an electromagnetic definition of channel transfer matrix, explore the synergies with – as well as advance – modern channel models for integrated sensing and communication (ISAC). We will discuss our approach and its main advantages with respect to existing channel models present in the literature, as well as provisioned contributions to the standardisation activity undergoing in both ETSI and 3GPP.

Numerical results are presented for both exterior and interior boundary value problems. The former includes the scattering by multiple large asymmetric objects; the latter includes outdoor-to-indoor transmission and coupling between rooms. The case study provides a validation of the transfer impedance via commercial full-wave EM simulation software.

Provisions are made for the fundamental generation of multipath fading statistics from the channel model, hinting that the integration of random processes with electromagnetic scattering will drive the next generation of statistical channel models of interest in 3GPP.

Co-Authors and Affiliation: Srikumar Sandeep, Ahmed Elzanaty, Mohsen Khalily, Morteza Kheirkhah, Rahim Tafazolli – University of Surrey

Adaptive beamforming offers a promising approach to enhance signal transmission by dynamically adjusting beam direction based on real-time environmental data. This concept introduces a digital twin designed to accurately represent its surroundings, and compute best signal paths, enabling the optimization of phased array antennas’ beam direction for optimal signal strength. The methodology focuses on determining the most effective beam direction by evaluating various signal paths, including direct line-of-sight, scattering, reflection, and diffraction. By integrating these diverse paths, the system can select the optimal beam direction to ensure robust signal transmission.

The proposed methodology leverages several technologies. Blender’s BLOSM Add-on constructs a precise environment model, incorporating custom radio materials for various surfaces. Nvidia Sionna, an open source library for link level simulations, facilitates the simulation for propagation between transmitters and receivers offering detailed insights into signal behaviour within complex environments. Data is collected by placing 218 transmitters and receivers at 5-meter intervals across all three spatial dimensions in Nvidia Sionna, creating a comprehensive dataset that includes, interaction distances, materials, channel coefficients, delays, and path types.

A neural network is trained on this dataset to predict channel coefficients and time delays, enabling the identification of the most effective signal paths. Nvidia OptiX, a general purpose ray-tracing tool, supports accurate ray generation and can compute interaction properties, Providing the inputs for the neural network. Using this information, the best beam direction can be formed for optimal signal strength.

The anticipated outcome is a digital twin capable of significantly improving signal transmission in complex environments. By utilizing advanced modelling, neural networks, and path optimization techniques, this approach aims to maintain robust communication links even in challenging scenarios, ultimately leading to more effective and reliable communication systems.

Co-Authors and Affiliation: Oluwakayode Onireti, University of Glasgow, Shuja Ansari – University of Glasgow

Flexible radio platforms are advantageous because they can have multiple personalities and functions. This multi-modal operation allows for different radio functions to be implemented through a single antenna or aperture. In this work we show how a wideband adaptable radio platform can be leveraged to operate in a wideband / high resolution RADAR sensor mode and then rapidly switch 1 to a high bandwidth OFDM communications mode to despatch stored sensor data or establish a communications link in a burst before returning to its main RADAR functional mode of operation. This approach not only reduces the number of antennas, it also provides for reduced RF hardware and lower probability of intercept for the communications link.

This multi-modal demonstration was carried out using a PCIe version of the wideband adaptable radio platform which is a high-performance radio developed in-house at Slipstream Engineering Design. At the heart of the platform lies the ASTRO AMD RFSoC module, a Direct RF on-the-edge processor module for accelerated digital beamforming, which features 8 channels of 14-bit data converters capable of 2 GHz of instantaneous bandwidth. The converters digitally generate and receive RF signals up to 6 GHz in an ultra-low SWaP FMC+ form factor that contains all required clocking and synchronisation peripherals and allows a user to configure these directly from software using a bespoke code-based board support package. Use of the PCIe platform allows for easy control and configuration directly from a PC.

Measured spectrograms of generated waveforms hopping between different modes of operation are presented.

Co-Authors and Affiliation: Edward Pipe – Slipstream Engineering Design Ltd

Optical devices can be used for the generation and processing of radio frequency (RF) signals, with applications in radar systems and telecommunications networks – this field is called microwave photonics (MWP). When the optical devices are integrated into photonic integrated circuits, the devices provide not only the wide bandwidth, low loss, and electromagnetic interference immunity offered by photonics in general but also reduced footprint, cost, and sensitivity to external perturbations due to the nature of integrated systems. So far, most photonic circuits for integrated MWP have been application-specific, meaning the functionality of the circuit cannot be changed after fabrication. Recently, different versions of reconfigurable MWP signal processors have been demonstrated, but common for all of these is that they are volatile, meaning they require a constant supply of power to keep the configuration. Furthermore, the footprint of the building blocks of the programmable mesh at the heart of the signal processor is currently so large that it limits the performance of the processor. This work aims to develop a programmable photonic mesh for optical signal processing and other MWP applications. In this study, we have fabricated a programmable photonic mesh unit cell consisting of 2×2 tunable couplers. The coupling between the input and output ports of the tunable coupler is determined by the refractive index of the non-volatile, low-loss phase-change material Sb2Se3, which is tuned by sending short heat pulses to the material from integrated microheaters. We show that the mesh unit cell can route optical signals, act as a delay line, and synthesise an optical ring resonator, which in turn can be used in optical signal processing. Our work paves the way for non-volatile and small-footprint programmable photonic meshes for reconfigurable MWP applications.

Co-Authors and Affiliation: Wen Zhou, Harish Bhaskaran – University of Oxford; Xi’an Jiaotong University, Bowei Dong – University of Oxford, Institute of Microelectronics, A*STAR Singapore

The Standing Wave Fibre Interference Trap (SWIFT) is a levitated optomechanical trap employing fibre optics and a planar geometry to create a standing wave optical field in which individual nanoparticles can be trapped [1]; phase coherent light, sent through three single-mode fibres, converges on a common central point. This geometry offers is compact, stable, and offers fast and accurate 2D positioning, including active feedback cooling, via relative
phase control of the optical fields. Our device uses a silicon wafer with diced trenches, created in collaboration with the Optical Engineering Group (Southampton) [2], to provide
accurate alignment of the fibres, which are glued in place, over a millimeter-sized through hole. In contrast with other work [3], the hexagonal arrangement offers a standing-wave
without counter-propagating light so that scatter from the nanoparticle can be more easily separated from the strong laser field used for trapping, enabling homodyne detection with a single optical wavelength. I will report on the challenges and overall progress of realising this
device.

[1] Backaction suppression in levitated optomechanics (2024), Rafal Gajewski :
https://levitation.wales/theses/2024_Rafal_Gajewski.pdf
[2] Optoelectronics Research Centre, Southampton : https://www.orc.soton.ac.uk/
[3] Vacuum levitation and motion control on chip, Melo et al. : https://arxiv.org/abs/2311.14016
Co-Authors and Affiliation: Dr. James Bateman – Swansea University

A component of work completed under the Diamond Flow construct involved a modification/generalization of the “Array Pattern Explorer” (APE) code. Originally used to model the performance of array antennas from a given immersed element pattern and slew angle, the code is improved with the added capability to estimate the power flux density (PFD) incident over a target of given size, speed, distance, and altitude.

Three dynamic scenarios are studied, each for the domains of land, sea, and air involving different targets, speeds, and distances. We consider a generic 32 x 32 element triangular pitched rectangular array with electronic scanning capability, and each element is supplied with a normalized power of 1W, enabling easily scalable estimates of the incident PFD distribution over the targets given area. Additional information is also determined such as beam characteristics, maximum slew speed, and duration within the steerable zone.

This work provides a useful tool for investigations involving planar array designs and the assessment of their capabilities in a dynamic setting. It enables study into the potential impacts and trade-offs to performance, with the potential to highlight where improvements are needed. Scenarios in which electronic warfare (EW) is employed are becoming more dynamic, and it is becoming increasingly more important for electronic support (ES) and attack (EA) capabilities to work collaboratively in a constantly changing landscape.

Co-Authors and Affiliation: Liam Scott – BAE Systems Digital Intelligence

This research project advances the principles of Electromagnetic Information Theory (EIT) and deep physical layer modelling, integrating these concepts into radar detection systems. The study addresses the growing complexity and sophistication of electronic attacks, which increasingly challenge traditional radar systems.

Starting this year, the project will run through 2027 and is divided into several key tasks. Initially, the research successfully explored radar systems’ Probability of Detection (PoD) under various EW scenarios. This foundational work examined radar and jammer characteristics, demonstrating how factors like frequency, bandwidth, and waveform influence radar detection, especially against sophisticated jamming tactics. Accurate modelling of the radar equation in contested electromagnetic (EM) environments has been established as essential for predicting the signal-to-noise ratio (SNR) at the radar receiver, a critical determinant of radar performance under interference. The ongoing analysis of PoD is expected to account for varying environmental conditions and system parameters, to provide insights into the likelihood of successful target identification and the risk of false alarms. Advanced radar techniques, such as enhanced waveforms and pulse compression, are planned to be evaluated to maintain detection efficacy in noisy environments.

Building on the foundation of the first task, the second task involves applying EIT to co-model communication and interference channels within the radar’s operating environment. This task involves the evaluation of channel capacity, degrees of freedom, and the influence of multiple dimensions (time, frequency, and space), leading to providing a robust theoretical foundation for enhancing radar and communication system performance in contested EM environments.

Finally, in the third parallel running task, the project is expected to integrate deep learning techniques with physical layer modelling to tackle challenges such as motion blur in radar imaging. The proposed models leverage large datasets generated from detailed EIT simulations to train deep-learning algorithms, aiming to improve the accuracy and speed of radar detection in real-time EW scenarios.

Co-Authors and Affiliation: Oleksandr Malyuskin, Hien Quoc Ngo, Oluwakayode Onireti, Shuja Ansari, Okan Yurduseven, Babar Abbasi

In a real-world maritime CEMA environment, phase noise directly impacts the clarity, accuracy, and reliability of intercepted signals. For maritime operators, the quality of signal intelligence can be the difference between mission success and failure, especially in dense, contested, and complex electromagnetic environments.

High phase noise spreads signal energy, masking weaker signals and making it challenging to distinguish between closely spaced frequencies. It introduces errors in DF calculations, leading to imprecise location tracking of targets, which is critical when navigating open seas or contested zones. Increased phase noise reduces the receiver’s ability to detect low-level signals, particularly in environments with high signal density, such as near busy ports or active conflict zones.

CommsAudit’s new low phase noise technology will dramatically enhance the performance of our RF systems by reducing phase noise to levels previously unachievable in traditional receiver designs. This breakthrough provides significant real-world benefits, particularly for users operating in maritime CEMA applications.

For operators, the impact of improved phase noise directly affects their ability to make informed decisions. Enhanced signal clarity reduces the cognitive load on operators by providing clear and actionable data, reducing the risk of misinterpretation. Lower phase noise reduces the likelihood of false positives, ensuring that operators’ attention is focused on genuine threats rather than noise-induced artifacts.

CommsAudit’s commitment to innovation in phase noise reduction translates into a substantial enhancement of receiver performance for next generation product lines, particularly within the Spectra series. For maritime CEMA operators, this means a transformative improvement in signal detection, direction finding, operational range, and overall situational awareness. By deploying systems with our patented low phase noise technology, maritime forces can navigate complex electromagnetic environments with greater accuracy, confidence, and mission effectiveness, securing a decisive advantage in any CEMA scenario.

Co-Authors and Affiliation: Charles Vafiadis

Datasets are often complex and expensive, but if available, they can exercise model fitting in a machine-learnt solution, and when from a synthetic source they can be controlled in dimensionality. Synthetic sources can provide constraints to the dimensionality that stimulate, and exercise known ambiguities and discrimination criteria in a controlled way. In this research a radar emitter behaviour model is demonstrated from a behavioural description using an emitter mark-up language for a Machine Learning (ML) example. That emitter behaviour model is used in a physics engine to generate high bandwidth data within a dataset. Furthermore, the dataset is then used to train an emitter classification model and achieves 99.3% accuracy in emitter identification using image classification. The emitter model and dataset are available for further research, as well as for other methods such as Transferred Learning (TL), the Generative Adversarial Network (GAN) method, and AI Cyberattacks such as the Fast Sign Gradient Method (FSGM).

The aim of this work is to investigate how reinforcement learning (RL) techniques could improve electromagnetic warfare (EW) capabilities. EW is a contest against an adversary for utilisation of the electromagnetic environment (EME). Such contest aims to minimise the quality of the adversary’s utilisation of the EME while maximising friendly utilisation. Effective EW systems can perceive the EME, actively adapt to it, and counter coordinated adversarial systems in the increasingly congested and dynamic EME. To deal with these challenges, the concept of Cognitive Electronic Warfare (CEW) is being developed as a field of study. CEW enhances EW capabilities via efficient decision-making, handling of large data, situational awareness, visualisation , and generation of appropriate responses to threats. This work is interested in applying CEW capabilities to electromagnetic attacks, where maintaining a jamming effect depends on real-time adaption to physical obstacles and electromagnetic interferences.

To apply CEW techniques to a scenario, the state and action spaces must be modelled. This is typically done using Markov Decision Processes (MDP), game theory (for multiple players/agents), or a Multi-Armed Bandits framework. A significant challenge, at this stage, is training an agent with imperfect observation of states (for example, the jamming agent is incapable of knowing the target’s frequency-hopping strategy). Nonetheless, for single-agent jamming decision-making, common RL algorithms used were found to be Deep Q-learning Networks (DQN), Double DQN, and Asynchronous Advantage Actor-Critic. Many-to-many confrontations are also explored in the literature, for which a hierarchical RL framework was recently proposed. The poster will explore potential directions for furthering these approaching to improve the reliability of a jammer; specifically, using RL to optimise resource management, jamming style decision-making, and jamming waveform generation in response to dynamic adverse effects.

This project aims to explore the feasibility of using off-the-shelf 60GHz Wi-Fi routers for simultaneous sensing and secured communication. The 60GHz frequency band is known for its large bandwidth, making it ideal for high-speed wireless communication. Our approach involves developing a custom kernel for commercial WiFi routers, enabling them to use the beacon beam for both sensing and device discovery. The sensing antennas detect the beacon wave to identify the presence of a person before establishing a secure connection. Depending on the resource utilization by the WiFi communication, the system can switch between using communication beams or specialized beams formed by the transmitter for calibration and sensing tasks.

Additionally, the system will employ a fingerprinting algorithm to quickly adapt to dynamic environments and identify various objects in the environment in real-time. Although experimentation is ongoing, available literature supports our anticipation that the system will achieve high accuracy in detecting individuals while maintaining robust communication performance with minimal latency and high throughput. This innovative approach could have significant implications for secured military deployments, where the ability to simultaneously sense and communicate is crucial. Due to low penetration power and beam-forming capabilities of the 60 GHz signal through objects, the system will provide a secured and contained network that will be snooping-free from outside the designated zone.

Co-Authors and Affiliation: Mathini Sellathurai – Heriot-Watt University

The electromagnetic environment is more congested than ever, and with the growing reliance upon network connected devices, the need for accurate and timely location information is paramount. This research presents a novel approach, which we call swarm enabled spatial intelligence (SESI), for locating GNSS emissions and disruptions within the local environment. The GNSS Combined Vehicular Measurements open dataset was gathered around Bristol, featuring computer vision, inertial and lidar measurements, L1 band radio measurements on three coherent channels, and a survey grade GPS measurement. This dataset allows analysis of GPS propagation within the local environment, treating each separate antenna as a separate platform in a swarm moving in formation, or beamforming as a combined antenna array. The results are presented from each dataset, including natural and urban canyons, and the transition into an underground carpark. This approach allows the validation of the use of the spatial intelligence model to predict the influence of the local environment on radio propagation, and the timely localisation and identification of disruptions. This proof of concept is a foundation for future demonstrations of SESI with autonomous robots (rovers, drones) working as a swarm.

Co-Authors and Affiliation: Edmund Hunt, Andrew Austin – University of Bristol

Shifting towards millimetre-wave (mm-wave) frequency offers wider frequency bands, enabling emerging high data rates and high-mobility wireless applications, such as integrated sensing and communications, including vehicle-to-vehicle communications. However, designing transceivers for this band presents challenges in signal processing and radio frequency (RF) aspects. Research at the University of Sheffield, under the academic research in next-generation information networks (AR-NGIN) project, aims to develop wireless techniques for mm-wave frequencies, focusing on two topics waveform and RF design. In waveform design, the combination of mm-wave frequencies and high mobility introduces significant Doppler shifts, which degrade the performance of the commonly used orthogonal frequency division multiplexing (OFDM) modulation by causing inter-carrier interference (ICI). Orthogonal time frequency space (OTFS) modulation has been proposed as an alternative waveform, transforming time-varying channels into nearly non-fading ones in the delay-Doppler domain. The error performance of OTFS and OFDM in high-mobility mm-wave systems is investigated with various low-density parity-check (LDPC) coding rates. Simulation results show that OTFS outperforms OFDM with low-order modulations and weak coding rates, highlighting the need for further research to evaluate these methods for future mm-wave systems. From an RF perspective, software-defined radio (SDR) technology, with analogue-to-digital converters (ADCs) placed as close to the antennas as possible, provides flexibility in digital signal processing for filtering, down-conversion, and baseband processing, thereby reducing complexity. The primary challenge in direct mm-wave sampling is the high-power consumption of ADCs at Nyquist rates. However, since the data resides within the carrier envelope, the Nyquist sampling rate can be reduced to the order of the signal bandwidth, which is much lower than the carrier frequency. While previous studies have experimentally modelled the error vector magnitude (EVM) as a function of sub-sampling rate and design parameters, the theoretical impact of sub-sampling on EVM remains unclear and is investigated here.

Co-Authors and Affiliation: Mahmoud Mojarrad Kiasaraei, Amir Dayan

The number of satellites in orbit is increasing rapidly, bringing increased likelihood of collisions, damage, and more space debris. The monitoring and classification of space assets, and characterisation and identification of non-cooperative targets (including debris) is termed “space domain awareness”, (SDA), and is becoming essential. SDA has traditionally been obtained from ground-based radar and optical telescopes, but recent interest focuses on development of in-orbit capabilities for close monitoring of space residents. This novel approach has several advantages over ground-based SDA: ability to use higher frequencies (not attenuated by atmosphere) allowing for finer resolution and higher sensitivity to small features and surface texture; shorter imaging distance; and ability to view targets from aspect angles not visible from the ground. This approach uses inverse synthetic aperture radar (ISAR) at high frequencies (sub-THz band) which has advantages over optical systems: imagery is independent of target illumination, resolution is independent of range, and target motion is a prerequisite for image formation rather than a hindrance.

This research explores two approaches to the concept. First, an analysis of the power budget for the proposed system: expected signal-to-noise ratios, how radar cross-section of targets varies with aspect angle and frequency, and how changes in scattering mechanisms at high frequencies affect imagery and signal return.

Second, investigating image segmentation and classification techniques. The chosen frequency band allows centimetre-level resolution, producing images close to optical quality. This allows adaptation of existing optical image recognition techniques to ISAR imagery for classification and recognition. The presented process is based on contrast within imagery and machine learning – particularly, this work utilises support vector machines (SVM).

Current research focuses on feature tracking within image sequences (using techniques such as the SIFT algorithm), analysing effects of polarisation on high-frequency ISAR imagery, and developing a multi-modal classifier to combine multiple classification methods.

Co-Authors and Affiliation: Marina Gashinova, Mikhail Cherniakov, Gruffudd Jones – University of Birmingham

As airborne sensing technology advances, autonomous control of sensor platforms, such as drones, is becoming a reality. These platforms can communicate with each other, forming dynamic networks of mobile sensors. This capability requires innovative sensor management approaches to ensure real-time adaptability in rapidly changing and contested environments. This project addresses the optimization of airborne radar sensor networks, focusing on the adaptive control of sensor movement, frequency selection, and beam direction to improve target detection and tracking. Key research areas of focus include the application of Bayesian inference for uncertainty management and multi-target tracking, with a focus on scenarios involving an unknown number of targets. Game theory and information theory-based methods are then employed for sensor management. These methods offer robust, explainable solutions that directly interface with modern multi-target multi-sensor tracking techniques to inform optimal sensor decision-making. Real-world experiment data is currently being used in this project to study and  validate sensor management strategies, providing insight into how algorithms perform in live environments. The research also considers practical challenges such as platform size and weight restrictions and sensor positioning restrictions present in the available flight trial data. Multitarget tracking has also been performed on the experimental data to facilitate this. By investigating these strategies, the project aims to enhance the operational effectiveness of sensor networks, laying the groundwork for autonomous, resilient systems capable of thriving in complex electromagnetic environments.

Co-Authors and Affiliation: University of Edinburgh

Resilient, assured timing is a critical capability for defence and security, underpinning position, navigation and multiple other capabilities in defence, having mission-critical applications across all domains, on multiple platforms in all areas of sensing, situational awareness, intelligence surveillance and reconnaissance, Electronic counter-measures, radar, secure communications & networks and data fusion, amongst others, as well as UK civilian sector and infrastructure. While GPS may be utilised for these applications and can be synchronized to within half a nanosecond of Universal Coordinated Time (UTC), it is vulnerable to loss, degradation and spoofing. Hence there is a critical need to develop distributed resilient timing (DRT) systems. Two-Way Satellite Time & Frequency Transfer (TWSTFT) is a long-established method of time transfer used to sync clocks to within 1 ns of UTC which has yet to be fully exploited by defence and may utilise existing communications infrastructure. This work assesses the level of precision achievable using TWSTFT, the factors affecting delay instabilities and the requirements necessary to its implementation as part of a DRT network.

In recent years, spoofing has emerged as a powerful method of attacking GNSS. The aim of these attacks is to deceive a receiver via transmission of “fake” or “spoofed” GNSS signals; once these inauthentic signals are tracked by the receiver, incorrect timing and positioning information can be provided. Unlike traditional jamming attacks which tend to be high–power (and easily detectable), spoofing attacks can be tuned to match the expected power of the true GNSS signal and thus detection/mitigation is vastly more difficult.

Through DASA funding, Chelton has developed a novel approach to anti-spoofing, using phase array-based spatial discrimination to detect spoofed signals, and null-steering to mitigate the spoofing attack and provide a ‘clean’ signal the GNSS receivers. Chelton’s innovative technology requires no complex platform integration and is ideally suited for rapid capability upgrades and retrofit onto existing platforms. 

The technology can be implemented onto pre-existing GNSS anti-jam systems via a software update, removing the need for additional hardware. Chelton proved this by implementing the anti-spoof technology onto our DACU4c Anti-Jam hardware to produce a demonstrator system. 

In DASA sponsored trials, Chelton’s demonstrator system was used to detect and completely mitigate a sophisticated spoofing attack that would otherwise have tricked advanced GNSS receiver systems. This performance was repeated in all spoofer scenarios tested, proving the value of a spatial discrimination approach. 

The conclusion of the project is that phased-array anti-jam hardware can be software upgraded to provide a powerful anti-spoof capability that can be used either as a stand-alone defence or alongside receiver based methods. As anti-jam products increasingly become standard fit on military platforms, “free” capability upgrades such as a spoofer protection should be taken into account when selecting Anti-Jam products.