Welcome to our Publications page!

In this paper, a MIMO antenna is presented consisting of 4 composite elements. Each antenna element consists of a U-shaped conducting structure and a perturbed barrel Dielectric Resonator Antenna (PB-DRA) structure. The former is loaded with a bow-tie patch, and it is parasitically excited through the U-shaped microstrip element. The additional bow-tie structure is designed for the gain enhancement in the 5G NR frequency range 1 (FR1) band and frequency range 2 (FR2). The PB-DRA has a dielectric constant of 8, and it is excited in higher-order mode to achieve resonance in FR2. The composite structure offers dual-band resonance with 3.85 GHz resonant frequency in the FR1 band and 26.65 GHz in the FR2 band, giving a large frequency ratio radiation characteristic. The proposed antenna offers impedance bandwidth of 2.02 GHz in the FR1 band, and 5.3 GHz in the FR2 band. Achieving multi-band resonance with a large frequency ratio is essential for exploiting the true benefit of 5G communication, as it enables operation across widely separated bands and supports multi-operator radio access networks (MORAN). The peak gains observed in the FR1 and FR2 bands are 8.23 dB and 13.14 dB, respectively. The proposed antenna is specifically designed to meet the requirements of the Open RAN compliant shareable 5G small cell radio unit specifications, by offering resonance in n77/n78 and n257/n261 bands, and end-fire radiation characteristics, ensuring sufficient coverage in the indoor and dense urban outdoor environment.

Open Radio Access Networks (O-RAN) offer a flexible RAN architecture for future 6G systems, yet their complexity and lack of real-world testbeds pose interoperability challenges, particularly with emerging software platforms and robotic systems. Here we present a real-world software-defined radio testbed based on an open-source 4G long-term evolution (LTE) system, integrated with the near-real-time (Near-RT) RAN Intelligent Controller (RIC) via standard O-RAN E2 interfaces. It enables connectivity with robotic end devices such as a haptic controller and robotic arm, demonstrating the activation of E2 functionality within a live RAN environment. The testbed enables haptic operation with sub-one-second latency and block error rate (BLER) under 12% for tasks such as dental inspection use cases. We also demonstrate replacement of software-defined radios (SDRs) with low-power mobile dongles, achieving comparable 10 Mbps throughput while cutting power consumption by 90%. This setup establishes a foundation for advancing research and integration in managing next-generation RANs.

Human activity recognition using WiFi sensing has become an important technology for smart environments, enabling device-free and privacy-preserving monitoring. However, existing approaches often face limitations in feature representation efficiency and generalisation, particularly when operating with constrained hardware setups.

This paper proposes Continuous Angle-of-Arrival and Time-of-Flight Maps, or CATM, a novel feature extraction framework that jointly captures the spatial and temporal dynamics of human activities using commercial WiFi devices. By combining smoothed channel state information with MUSIC-based signal processing, CATM constructs two-dimensional heatmaps that integrate angle-of-arrival and time-of-flight features. This enables a robust representation of both broad and fine-grained human movements.

Unlike conventional methods that rely on fragmented temporal or spectral features, CATM embeds continuous spatiotemporal patterns, simplifying downstream model learning. A lightweight Res-BiLSTM network trained on CATM features achieved 93.2% accuracy across eight activities and five users under three displacement settings, outperforming other single-domain methods.

Importantly, CATM also demonstrates strong transferability. When adapted to unseen device placements, the framework retained 80% accuracy using only 20% retraining data, reducing the need for extensive labelled datasets. Overall, the results show that CATM offers a more informative and transferable feature extraction method than traditional approaches based on CSI amplitude and Doppler information for human activity recognition.

Sign language is an essential form of communication for people who are deaf or have hearing impairments. However, a communication barrier remains between the deaf community and individuals who are not familiar with sign language. This paper explores the potential of Sign Language Recognition technology to help bridge this gap, focusing on phrase-level recognition using deep learning models.

A dataset of eight common phrases — Again Please, Bless You, Excuse Me, Good Afternoon, Good Morning, Hello, Welcome, Slow Please, and Stay Safe — was collected using the Xethru X4M03 Ultra-Wideband radar sensor. The dataset included 240 samples across eight classes. Two pre-trained deep learning models, VGG16 and GoogleNet, were applied to classify the phrases.

Model performance was evaluated using accuracy, precision, recall, and F1 score. The results showed that VGG16 outperformed GoogleNet, achieving an accuracy of 80%. Overall, the study demonstrates the feasibility of combining radar-based contactless sensing with deep learning to support sign language recognition and improve communication accessibility for deaf and hard-of-hearing communities.

Contactless vital signs detection has the potential to transform healthcare by enabling accurate, convenient, and non-invasive patient monitoring. By supporting continuous, real-time assessment of vital signs, this approach can help identify anomalies earlier and enable timely clinical intervention.

This paper presents a novel framework for contactless vital signs detection using continuous-wave radar and advanced signal processing techniques. The proposed method achieves high precision in capturing radar-based heart sound waveforms, with 1,261 samples compared against ground truth ECG signals. The heart sound-based approach delivers highly accurate pulse readings, outperforming previous benchmarks with a mean absolute percentage error of 0.0129 and a mean absolute error below one at 0.8712.

In addition, heart rates were derived from the heart sound waveforms and compared with both conventional radar-derived heart rates and ground truth ECG data. The analysis identifies areas where traditional radar-based methods show limitations. Overall, the proposed approach demonstrates minimal error and superior accuracy across different heart rate states, offering a promising step towards more reliable, non-invasive vital sign monitoring.

Curved reconfigurable intelligent surfaces (RISs) represent a promising frontier for next-generation wireless communications, enabling adaptive wavefront control on nonplanar platforms such as unmanned aerial vehicles and urban infrastructure. This work presents a systematic investigation of cylindrical RISs, progressing from idealized surface-impedance synthesis to practical implementations based on simple one-bit meta-atoms. Exact analytical and geometrical-optics-based models are first developed to explore fundamental design limits, followed by a semi-analytical formulation tailored to discrete, reconfigurable architectures. This model enables efficient beam synthesis using both evolutionary optimization and low-complexity strategies, including the minimum power distortionless response method, and is validated through full-wave simulations. Results confirm that one-bit RISs can achieve directive scattering with manageable sidelobe levels and minimal hardware complexity. These findings establish the viability of cylindrical RISs and open the door to their integration into dual-use wireless platforms for real-world communication scenarios.

Semantic communication (SemCom) shifts the focus from bit-level accuracy to the preservation of meaning, enabling more efficient and robust transmission. To improve wireless channel utilisation in SemCom, this paper proposes a channel assignment approach for polar code-based SemCom that allocates polarised channels according to semantic importance.

The proposed method combines eye-tracking data with semantic segmentation to define two metrics that capture the contribution and correlation of semantic entities within an image. Using these semantic metrics alongside polarised channel reliabilities, the paper formulates a constrained 0-1 optimisation problem for polarised channel assignment.

A priority-based algorithm is then developed to dynamically prioritise semantically important content during transmission. Simulation results show that the proposed method significantly outperforms traditional channel allocation policies, particularly under harsh channel conditions, by preserving critical visual information while reducing overall transmission redundancy.

Model checking and verification are essential for ensuring the correctness of software systems by systematically analysing whether a system satisfies the required specifications. When a Linear Temporal Logic (LTL) property is violated, diagnosing the resulting counterexample is a key step in debugging and refining the system, helping developers identify faults and ensure compliance with requirements.

Several existing approaches use causal reasoning to support the identification and explanation of counterexamples. However, these methods can be limited in their expressivity, their ability to capture complex dependencies in system behaviour, or their computational feasibility.

This paper proposes a revised definition of causes for LTL violations in counterexamples, enabling the identification of conjunctive causes. It also introduces a novel algorithm for computing these causes and evaluates its performance against a baseline causal computation algorithm using a benchmark of industrial specifications.

Across all comparisons, the proposed algorithm either subsumes the baseline causes, identifies stronger causes, or differs from the baseline in explainable ways. These results demonstrate its effectiveness in improving counterexample analysis and supporting more precise debugging of software systems.

Gait analysis is an important decision-making tool in the treatment and management of children with cerebral palsy (CP). CP refers to a group of congenital disorders caused by abnormal brain development during the fetal stage, affecting motor function and movement.

Each year, around 1,000 children are born with CP, while approximately 1,500 children between the ages of 4 and 12 are diagnosed with the condition. Improving posture through gait analysis is therefore an important area of focus, particularly through the use of smart insole systems.

In this work, a smart insole is designed using capacitive pressure sensors made with black carbon from a disposable, non-rechargeable alkaline 9V battery. The sensors demonstrate notable performance characteristics, including 99% linearity between applied weight and output capacitance, a minimum hysteresis loss of 2%, and a fast response time of 110 µs to reach 90% of the output value from 10%. They also recover to 10% within 161 µs once the input is removed.

A virtual reality effect is presented through a Python-based graphical user interface, which displays real-time insole system values using colour indicators corresponding to four different pressure levels. Overall, the proposed system has the potential to support improved biomechanical assessment and posture correction in children with cerebral palsy.

Generative artificial intelligence (AI), particularly large language models (LLMs), is expected to play a central role in sixth-generation (6G) wireless networks, enabling use cases such as autonomous troubleshooting, intelligent configuration, and real-time decision support.

Open Radio Access Network (O-RAN), with its modular and disaggregated architecture, provides a suitable platform for integrating these AI capabilities. However, deploying LLMs within O-RAN environments presents key challenges, including high inference latency, limited domain grounding, and resource-intensive retrieval workflows.

To address these challenges, this paper proposes a cache-augmented Retrieval-Augmented Generation (RAG) and Graph-RAG framework tailored specifically for O-RAN environments. The framework combines semantic caching at the User Plane Function (UPF) with centralised model orchestration to support low-latency, high-throughput, and contextually accurate responses.

The system is evaluated using the ORAN-Bench 13K benchmark, which includes 13,952 domain-specific queries derived from official specification documents. Results show that cache-enhanced RAG reduces average latency by 35.8%, increases cache utilisation to 76.38%, and achieves a cache hit rate of up to 42.5%, significantly outperforming baseline LLM and Graph-RAG configurations in responsiveness.

In terms of factual correctness, Graph-RAG achieves the highest score at 71.2%, followed by RAG at 70.5% and the baseline LLM at 64.8%. These results demonstrate that semantic caching can help overcome key performance bottlenecks in LLM-based reasoning for O-RAN, while supporting the development of scalable, AI-native telecommunications infrastructure.

In healthcare IoT networks, network anomalies can disrupt the flow of reliable data, potentially compromising healthcare data’s security and integrity. To address this challenge, several anomaly detection methods have been developed using artificial intelligence (AI) algorithms. However, finding an optimal AI model with the best tuning parameters for effective anomaly detection is a time-consuming and resource-intensive task. To address this issue, we propose an Automated AI (AutoAI) approach to optimise the tuning of hyperparameters in healthcare data anomaly detection. By leveraging the power of AutoAI, our goal is to streamline the anomaly detection process, making it more accurate and efficient. Our method is designed to adapt dynamically to the ever-changing nature of healthcare data, ensuring robustness against emerging anomalies. The proposed AutoAI method was validated in a realistic scenario, and the outcomes depict the superiority of the proposed approach as compared to existing schemes on various performance evaluation metrics.

This paper presents a bi-layer, double-F-shaped asymmetric transmissive chiral metasurface (ATCM) designed to achieve strong triple-band circular dichroism (CD) and asymmetric transmission (AT) in the microwave frequency range.

The proposed ATCM enables selective polarisation transmission under right-hand circularly polarised (RHCP) and left-hand circularly polarised (LHCP) waves incident from both forward and backward directions. This results in a significant CD effect at three distinct K-band frequencies: 18.788 GHz, 20.140 GHz, and 21.528 GHz, with maximum CD values of 0.75, 0.6, and 0.4, respectively.

The high-performance characteristics of the ATCM, including its strong polarisation selectivity and asymmetric transmission capabilities, make it a promising candidate for a wide range of applications, including analytical chemistry, imaging, sensing, spectroscopy, satellite communications, and the development of advanced microwave devices.

This work proposes a novel multi-band reflective metasurface that is capable of linear polarisation (LP), and circular polarisation (CP) conversion in Ku, K, Ka, and U Bands. The metasurface design involves a combination of ring and square elements strategically arranged and printed on a 0.76 mm thin-grounded Rogers RO3003 substrate. The metasurface achieves LP for y-polarised incident electromagnetic (EM) waves in 16.2–17.2 GHz, 23.0–25.4 GHz, 40.3–54.35 GHz frequency bands. The polarisation conversion ratio (PCR) for LP frequency ranges is a minimum 90% with a fractional bandwidth (FB) of 2.94%, 9.91%, and 26.9%, respectively. Moreover, metasurface achieves CP for y-polarised incident EM wave in 16.1–16.55 GHz, 17.5–22.15 GHz, 26.65–37.75 GHz, and 55.6–59.8 GHz frequency bands. In addition, the axial ratio (AR) for CP frequency ranges is less than 3 dB with a FB of 2.75%, 23.45%, 34.47%, and 7.27%. The device performance is considerably stable under oblique incidences up to 45 degrees. The metasurface unitcell is compact with a structural size of, and. The proposed prototype is fabricated, and the measured results are in good agreement with the simulated ones. Overall, the proposed metasurface exhibits promising performance characteristics and holds potential for multiple applications in satellite-based networks.

An ultrathin, single-layer planar terahertz metamaterial absorber (MMA) is numerically demonstrated for chemical sensing applications. The fractal MMA consists of a 2nd order Cayley tree-shaped resonator deposited on a polyimide substrate. The designed absorber is polarisation-insensitive and has six resonance frequencies at 3.376, 6.49, 6.92, 8.083, 8.227, and 9.36 THz. For the absorption peak at 3.376 THz, the proposed sensor has a thickness of 0.067 λ and its largest dimension is 0.38 λ. The electric field map is analysed to interpret the fundamental characteristics of the six resonances. The refractive index-based sensing is evaluated by placing an analyte on top of the absorber. The proposed sensor exhibits a good Quality factor (Q) value of 98.85, and its highest sensitivity is 1.35 THzIRIU. The designed MMA sensor is suitable for various chemical and bio sensing purposes.

In drone swarms, where multiple drones collaborate closely to achieve shared objectives within constrained spatial domains, the intricacies of these interrelated actions can lead to potential issues. Despite rigorous pre-deployment planning, the inherent probability of complications persists. These complications stem from onboard computational resources, hardware failures, and network communication disruptions. While the malfunction of an individual drone may seem inconsequential, it can escalate into a substantial predicament when it disrupts the seamless coordination of the entire swarm. Therefore, the need to proactively monitor drones for predictive failure analysis and the subsequent examination of failed drones to mitigate future occurrences becomes imperative. This paper introduces a comprehensive framework for systematically collecting and processing data within drone swarms. The framework gathers critical information about onboard characteristics and commu-nication metrics. These data points are subjected to advanced analysis using Complex Bayesian Networks to probabilistically uncover complex and hidden relationships between random features. The results demonstrate exceptional accuracy, with influences ranging from 99 % to 79 %, which ensures the reliability and effectiveness of the predictive capabilities in enhancing drone safety and network performance.

We propose a broadband metasurface consisting of a novel compact resonator to achieve linear polarisation conversion in reflection mode. This compact resonator is designed on Roger 5880 substrate, measuring 1.575 mm in thickness, and possessing a loss tangent of 0.004. This structure is also upheld by a metal ground. The described unit cell effectively sends back incoming waves by converting 90⁰ across a wide range of frequencies. This unit cell presents an efficiency exceeding 90 % for polarisation conversion in the following frequency regimes: 12.2–41.5 GHz for normal incident waves. At the same time, this design includes polarisation conversion with 90 % efficiency in broadband from 12.5 GHz to 30.6 GHz for angularly polarised waves up to. Behind this broadband polarisation transformation lies the concept of surface current distribution and a high impedance surface. Given the broad-spectrum coverage and high-efficiency polarisation conversion capabilities, our proposal holds significant potential for a wide range of applications.

Regular physical activity is essential for adult health and wellbeing. In response to the global challenge of physical inactivity, international targets aim to achieve a 10% relative reduction by 2025 and a 15% reduction by 2030, compared with 2010 levels.

As many adults spend 8 to 10 hours at work each day, the workplace offers a valuable opportunity to encourage physical activity. However, existing exercise monitoring systems are often camera-based, which can be limited by poor lighting conditions and raise privacy concerns, making them less suitable for workplace environments.

To address these challenges, this paper proposes a radar-based approach for monitoring employee fitness, with radar signals represented as spectrograms. The system focuses on five seated workplace exercise categories: lower back, glutes, arm stretches, neck stretches, and back shoulder exercises.

Deep learning models, including MobileNet, ResNet50, VGG16, and VGG19, are used to process the radar data and classify exercise patterns. Among these models, VGG16 achieves the strongest performance, with a classification accuracy of 100%. This demonstrates the potential of radar-based monitoring as an effective, privacy-preserving solution for workplace fitness applications.

High-clarity, precision imaging of biological and non-biological samples is essential across healthcare, education, and research. However, the small size and low contrast of many samples continue to present significant imaging challenges. Ultra-compact, multi-modal imaging systems with enhanced performance offer a promising way to overcome these limitations.

This work presents a highly efficient, ultra-compact imaging system that integrates multi-modal imaging capabilities using advanced metasurface-based optical design. By leveraging phase-gradient metasurfaces, the proposed optics combine bright-field imaging with a spiral phase configuration carrying a topological charge of +1, enabling edge-enhanced imaging and improved visualisation of low-contrast samples.

A tunable varifocal feature is also introduced to provide greater imaging control, allowing precise and convenient focus adjustment. The system uses an all-dielectric metasurface composed of subwavelength resonators to deliver the required multiplexed phase modulation with high efficiency.

For practical device demonstration, the lens performance is carefully evaluated through aberration and imaging analyses. With its multifunctional design, compact form, straightforward fabrication, and simple design approach, the proposed system offers significant potential for advanced optical imaging applications in healthcare and education.

Early detection of skin cancer is essential for improving survival rates and treatment outcomes. This study presents a highly sensitive, multi-resonance terahertz (THz) sensor designed to differentiate cancerous tissue from healthy tissue by analysing refractive index profiles.

The sensor features a T-shaped fractal gold resonator with three distinct resonances, each achieving absorption efficiencies above 99.5%. It also demonstrates a peak quality factor of 269.7, indicating sharp, well-defined resonances, and a maximum sensitivity of 432 GHz/RIU, enabling the accurate detection of subtle refractive index variations associated with cancerous cells.

Its advanced performance is supported by a compact metamaterial-based structure, which enhances both absorption and sensitivity. These features make the proposed sensor a promising platform for non-invasive skin cancer diagnostics.

Respiration rate is a critical parameter for assessing the physiological well-being of patients. Traditional methods typically rely on contact-based and visual devices, which pose challenges for long-term monitoring due to issues of comfort and privacy. This paper proposes a non-contact method for acquiring respiration rates using Wi-Fi Radio Frequency (RF) electromagnetic waves. We address the challenge of accurately estimating respiration rates by presenting a robust approach that leverages receiver antenna diversity to enhance the signal-to-noise ratio (SNR), thereby significantly improving estimation accuracy. Our method employs the maximal ratio combining technique to integrate the received signals effectively. Experimental results demonstrate that the proposed method achieves the overall Mean Absolute Error (MAE) of 0.7 bpm across all trials, compared to ground truth wearable respiration belts, indicating reliable respiration estimation using Wi-Fi system.

The multiband, multifunctional chiral metasurface with asymmetric transmission exhibits significant potential for diverse applications in modern communication systems, ranging from enhanced signal modulation and polarisation control to advanced beam steering and compact antenna design. This research presents a versatile and advanced chiral metasurface operating at multiple bands with diverse functionalities, including asymmetric transmission. The proposed metasurface effectively transforms an incoming Linearly Polarised (LP) wave into a Circularly Polarised (CP) wave. Additionally, it functions as a 90° polarisation rotator for the incident LP wave. The design starts with an element of a 2 × 2 supercell comprising a Square Split Ring Resonator (SSRR) and an I-shaped resonator. The right diagonal elements of a supercell undergo scaling down, giving rise to a rotational asymmetry. Chirality is introduced into the design, and cross-polarisation conversion is enhanced by rotating all four elements by 90° relative to each other. On the back side of the substrate, each element undergoes a 90° rotation compared to its counterpart on the front side, realising the asymmetric transmission feature. The incorporation of multiband and multifunctional features within a single supercell equips the subject chiral metasurface to be utilised in various engineering applications.

State-Of-the-art deep transfer learning methods depend on exhaustive, trial-and-error fine-tuning of pre-trained models—a process that is both computationally expensive and unreliable when data in the target domain are scarce. To overcome these limitations, we propose a domain-informed fine-tuning strategy built upon a novel Innately-Intelligent Neural Network (IINN) architecture. Unlike how state-of-the-art deep learning models are heuristically constructed, IINN constructs each layer in a domain-informed manner by directly mapping the mathematical operations of analytical equations (e.g., 3GPP propagation models) into its network architecture prior to any training. This “innate” design strategy inherently aligns each layer with specific physical parameters, making the model fully interpretable. As a result, we can pre-identify the exact layers associated with parameters that change between source and target domains and fine-tune only those—eliminating the need for iterative layer-by-layer retraining. This targeted fine-tuning approach reduces computational overhead and data requirements. We validated IINN on radio-propagation modelling for cellular networks, achieving faster adaptation and higher accuracy than the conventional fine-tuning approach. Experimental evaluations demonstrate that our proposed domain-aware transfer learning framework achieves up to 16.4% improvement in sector-based performance and approximately 10.3% gain in adapting to varying base station heights, with overall average gains in the 10–15% range over state-of-the-art DNN transfer learning approaches. The proposed framework offers a promising direction for data-efficient learning in next-generation wireless systems.

The effectiveness of software applications largely depends on the user experience (UX), since it has a direct impact on user engagement and satisfaction. Empathy mapping is an important design thinking technique that organizes user perceptions into distinct categories for better understanding. However, traditional empathy mapping methods rely entirely on interviews and manual analysis which are both time-consuming and costly, thereby limiting the scalability of UX design and research. To address these challenges, this study presents an automated process for empathy mapping by analyzing user-posted app reviews. This study uses the Bidirectional Encoder Representations from Transformers (BERT) model for sentiment analysis, classifying user reviews as either positive (gain points or desires) or negative (pain points or frustrations). Latent Dirichlet Allocation (LDA) is then used to apply topic modeling to pinpoint preferences and important themes. By concentrating on gains and pains, this method automates the traditional manual and costly process of design thinking and empathy mapping, making it more scalable and efficient through data-driven insights. In training, the proposed model with several versions of BERT model, the binary accuracy improved from 78.14 to 98.61%, with precision achieving 97.82%, F1 score of 98.62%, and recall up to 99.42%. The validation accuracy also increased from 87.40 to 92.58%, with an F1 score 92.59%, precision of 92.43%, and recall of 92.75%. These accurate results indicate that the proposed model may be used by user experience design teams, which will help them improve and streamline UX design while also assisting developers in promptly receiving user feedback.

We study the changes in morphology of the photoplethysmography (PPG) signals-acquired from a select group of South Asian origin-through a low-cost PPG sensor, and correlate it with healthy aging which allows us to reliably estimate the vascular age and chronological age of a healthy person as well as the age group he/she belongs to.Approach. Raw infrared PPG data is collected from the finger-tip of 173 apparently healthy subjects, aged 3-61 years, via a non-invasive low-cost MAX30102 PPG sensor. In addition, the following metadata is recorded for each subject: age, gender, height, weight, family history of cardiac disease, smoking history, vitals (heart rate and SpO2). The raw PPG data is conditioned and 62 features are then extracted based upon the first four PPG derivatives. Then, correlation-based feature-ranking is performed which retains 26 most important features. Finally, the feature set is fed to three machine learning classifiers, i.e. logistic regression, random forest, eXtreme Gradient Boosting (XGBoost), and two shallow neural networks: a feedforward neural network and a convolutional neural network.Main results. For the age group classification problem, the ensemble method XGboost stands out with an accuracy of 99% for both binary classification (3-20 years vs. 20+ years) and three-class classification (3-18 years, 18-23 years, 23+ years). For the vascular/chronological age prediction problem, the ensemble random forest method stands out with a mean absolute error of 6.97 years.Significance. The results demonstrate that PPG is indeed a promising (i.e. low-cost, non-invasive) biomarker to study the healthy aging phenomenon.

In the era of the Internet of Things (IoT), the transmission of medical reports in the form of scan images for collaborative diagnosis is vital for any telemedicine network. In this context, ensuring secure transmission and communication is necessary to protect medical data to maintain privacy. To address such privacy concerns and secure medical images against cyberattacks, this research presents a robust hybrid encryption framework that integrates quantum and classical cryptographic methods. The proposed framework not only secures medical data against cyber threats but also protects the secret security keys. Initially, a Quantum Key Distribution (QKD) is employed to generate a shared key, which is then used to secure the symmetric keys via One-Time Pad (OTP) encryption. Next, bit-planes are extracted from each colour component. The rows and columns of the extracted bit-planes are scrambled using random sequences which are generated by a 6D hyperchaotic Chen system and the Ikeda map. To further increase confusion in the original data, multiple-step pixel scrambling operations such as pixel shuffling, pixel value shuffling, and rotational and flipping operations are implemented. After the confusion phase, a combination of affine transformations with non-linear functions, Discrete Cosine Transform (DCT) with complex modulation, Discrete Wavelet Transform (DWT) with random phase modulation, bilinear transformation, and nonlinear polynomial mapping are employed to create diffusion in the scrambled components. These multiple encryption operations aim to maximise randomness in the final ciphertext image. Additionally, to reduce computational complexity, only the Most Significant Bit-Planes (MSBs) are encrypted, as they contain more than 94% of the plaintext information. Several experimental results and analyses are conducted to assess the proposed encryption framework, including entropy analysis, key sensitivity analysis, correlation analysis, lossless analysis, and histogram analysis. Furthermore, the framework is tested against various cyberattacks, such as brute-force attacks, clipping attacks, and noise attacks on the ciphertext images, to demonstrate its resilience against such threats.

Multiband implantable antennas are crucial components of biomedical implantable devices (BIDs), enabling the establishment of wireless communication links with external base stations. These types of antennas perform various functions such as data transmission, wireless power transfer, and control signalling. However, this scenario requires an external multiplexer in the BIDs to separate various frequency bands, imposing size constraints on the BIDs. This work proposes a self-duplexing circularly polarised implantable (SDCPI) antenna for wireless capsule endoscopy (WCE) application, having two separate ports, with port 1 providing a wideband response covering MICS (402 MHz), ISM (433 MHz) band, and port 2 covering ISM (915 MHz) band. The proposed SDCPI antenna achieved a very compact volume of π × (5.1)² × 0.127 = 10.3 mm³ by semi-circular slots on the radiating patch and shorting pins. The proposed capsule-integrated implantable antenna was thoroughly analysed through simulations in various parts of the digestive tract (stomach, small intestine, and colon) and was later fabricated and tested. The measurements were carried out in minced pork, and the results obtained showed close resemblance to the simulated results. The proposed SDCPI antenna offers a -10 dB impedance bandwidth and CP bandwidth of 151 and 273 MHz, and 10% and 15% at 402 and 915 MHz, respectively. Furthermore, it exhibits measured gain of -36 and − 25 dBi at 402 and 915 MHz, respectively. To evaluate the human safety of the proposed SDCPI antenna, specific absorption rate (SAR) at 402 and 915 MHz was estimated in the stomach, small intestine, and colon, and was found to be within the limits allowed by the IEEE standards. Additionally, the wireless communication link establishment capabilities of the proposed SDCPI antenna were gauged through link margin analysis. This analysis confirmed that at 402 MHz, the presented SDCPI antenna can establish reliable communication up to 25, 9.7, and 4.5 m when placed in the small intestine for bit rates of 1, 12, and 78 Mbps, respectively. Likewise, at 915 MHz, the suggested SDCPI antenna offers seamless communication up to 28, 14.7, and 5.3 m when placed in the small intestine for bit rates of 1, 12, and 78 Mbps, respectively. These results verify that, to the best of the authors’ knowledge, the proposed highly miniaturised SDCPI antenna is the first self-duplexing CP implantable antenna for WCE applications offering simultaneous transmission and reception without requiring an external multiplexer.

This research paper introduces a new design of metasurface for polarisation conversion applications, functioning as both a cross (half-wave plate) and circular (quarter-wave plate) polariser in reflection mode. Comprising unit cells on one side and a metal layer on the other, with a Roger 5880 substrate, the metasurface demonstrates its ability to reflect an incident or polarised wave as a polarised wave across multiple frequency bands: 9.72–10.00 GHz, 17.65–41.87 GHz, 45.67–45.80 GHz, and 49.66–49.84 GHz. The design achieves a noteworthy 24.82 GHz bandwidth with a 98.72 % fractional bandwidth for linear-to-linear conversion, demonstrating efficiency exceeding 90 %. Simultaneously, the metasurface converts the incident wave into a right-hand circularly polarised (RHCP) wave at frequencies ranging from 9.38 to 9.61 GHz, 45.9–46.1 GHz, and 49.96–50 GHz. It transforms the wave into a left-hand circularly polarised (LHCP) wave within the frequency band from 10.19 to 10.61 GHz, 15.60–16.82 GHz, and 45.45–45.6 GHz. The design also exhibits angular stability up to 45 degrees. Experimental validation using the fabricated prototype confirms the findings, showing good agreement with numerical results. This metasurface comes in handy for future communication, radar applications, and health applications. This metasurface is highly suitable for future communication systems, radar applications, and healthcare technologies.

A high-gain and low-side-lobe pencil beam 8×8 antenna array using substrate integrated waveguide (SIW) operating at 28 GHz millimetre wave (mmWave) spectrum is designed. The antenna array consists of a multilayer structure with a microstrip feed network to feed SIW with a slot. The feed network is implemented using Taylor distribution for low side lobe level in one plane, and the same distribution is also applied on series-fed patch antenna elements to get low side lobe level in both planes. The antenna array provides a bandwidth of 1.6 GHz with a maximum gain of 20.3 dBi at 28.5 GHz. It provides a side lobe level of less than -17.3 dB in both X-Z and Y-Z planes at 28.5 GHz. The simulated and measured results of the antenna array are in good agreement. The proposed antenna features low cost, high gain, and high efficiency, making it a potential candidate for 5G applications.

We propose a framework to design integrated communication and computing (ICC) receivers capable of simultaneously detecting data symbols and performing over-the-air computing (AirComp) in a manner that: a) is systematically generalizable to any nomographic function, b) scales to a massive number of user equipments (UEs) and edge devices (EDs), c) supports the computation of multiple independent functions (streams), and d) operates in a multi-access fashion whereby each transmitter can choose to transmit either data symbols, computing signals or both. For the sake of illustration, we design the proposed multi-stream and multi-access method under an uplink setting, where multiple single-antenna UEs/EDs simultaneously transmit data and computing signals to a single multiple-antenna base station (BS)/access point (AP). Under the communication functionality, the receiver aims to detect all independent communication symbols while treating the computing streams as aggregate interference, which it seeks to mitigate; and conversely, under the computing functionality, to minimise the distortion over the computing streams while minimising their mutual interference as well as the interference due to data symbols. To that end, the design leverages the Gaussian belief propagation (GaBP) framework, relying only on element-wise scalar operations coupled with closed-form combiners purpose-built for the AirComp operation, which allows for its use in massive settings, as demonstrated by simulation results incorporating up to 200 antennas and 300 UEs/EDs. The efficacy of the proposed method under different loading conditions is also evaluated, with the performance of the scheme shown to approach fundamental limiting bounds in the under/fully loaded cases.

Driver fatigue is a critical factor in road accidents, often resulting in severe consequences due to delayed reaction times and impaired decision-making. Traditional fatigue detection methods, such as camera-based systems, have significant challenges related to intrusiveness, privacy concerns, and reliability under varying environmental conditions, associated with them. This article introduces an innovative driver fatigue detection system, 3D-DFD, which leverages advanced 3-D millimetre-wave (mmWave) imaging radar and artificial intelligence algorithms for driver fatigue detection. By monitoring physiological and behavioural indicators, such as normal posture, yawning, nodding, and rapid blinking, using high-resolution 3-D radar imagery, we enable noninvasive and privacy-preserving detection. The integration of 3-D radar enhances spatial feature extraction, providing robust classification across a wide range of diverse detection scenarios. The system demonstrates an average accuracy of 93.16%, with precision rates of 92.5% for yawning, 94.2% for nodding, and 93.8% for rapid blinking based on testing with 19 volunteers across three different scenarios, showcasing its effectiveness and reliability. This research underscores the potential of 3-D mmWave radar technology in driver fatigue detection and lays a strong foundation for safer and more intelligent automotive systems.

In the rapidly evolving landscape of wireless networks, accurate and resilient propagation models are essential to achieve optimal performance and reliability. This paper presents a novel domain-aware framework for interpretable and resilient propagation models. The proposed approach represents an innovative architecture framework that is not only interpretable but can also deal with training data size scarcity. Bridges domain knowledge with machine learning. The proposed approach leverages a combination of domain expertise, analytical modeling, and customized neural networks to construct interpretable models that excel in both identical distribution and non-identical distribution test-train dataset scenarios. Through a comprehensive analysis, we demonstrate the proposed approach’s ability to adapt and refine models in response to real-world variations, ensuring consistent, high-quality performance. The proposed framework not only enhances our understanding of complex systems but also paves the way for the creation of digital twins for wireless networks. Furthermore, the root mean square error of the performance metric for the proposed approach is reported as 6.97 dB, further confirming its effectiveness in accurately predicting the results of wireless propagation.

Human–autonomy teaming in the Low Altitude Economy (LAE) requires operators to manage both ground and aerial autonomous agents under time pressure, spatial uncertainty, and cognitive load. This study investigates how visual and haptic feedback affect operator situational awareness (SA) in simulated collision avoidance tasks involving cars and drones. A high-fidelity virtual environment was built using Unreal Engine 4 and AirSim, with haptic cues delivered through a wearable bHaptics vest. Twenty-two participants performed within-subject trials across visual-only and visual–haptic conditions. Results showed that haptic feedback significantly enhanced SA, particularly in dimensions related to information acquisition and spare mental capacity. Improvements were more consistent in car-based tasks, while drone scenarios exhibited greater inter-individual variability. These findings demonstrate the potential of multimodal interfaces to support cognitive performance and reduce platform-related disparities in operator SA. This work provides empirical evidence for designing adaptive, perception-aware interfaces in safety-critical human–autonomy teaming systems.

Mobile network performance plays a crucial role in ensuring seamless connectivity for users, yet significant regional disparities persist across urban and rural areas. This study evaluates the coverage quality of four major mobile network operators (MNOs) across multiple counties around the River Severn Catchment Area, including performance metrics based on Acceptable Voice, Essential Data, Good Data, and Excellent Data, etc. However, one limitation is that only the best available network connection is measured, meaning that areas where a handset holds onto a 4G connection despite reverting to 2G for voice calls (due to VoLTE limitations) may not be fully accounted for. Furthermore, call setup times and failure rates, key indicators of real-world voice service reliability, are not explicitly captured in this study, although they remain essential factors for future research. The study uses coverage measurements and correlation matrices to highlight network uniformity, infrastructure sharing patterns, and independent deployment strategies. The findings indicate that urban centres exhibit strong inter-provider correlations, suggesting some degree of infrastructure sharing. Many operators argue that network density remains insufficient due to high contention levels and regulatory constraints, such as planning departments restricting new infrastructure deployment. Particularly with EE, which operates with a more independent deployment model. Vodafone consistently provides superior data coverage, whereas EE leads in essential connectivity, ensuring basic service availability. These findings underscore the need for targeted rural investments, infrastructure-sharing policies, and AI-driven network optimisation to enhance service equality. The study provides valuable insights for policymakers, telecom providers, and researchers to bridge the digital divide and improve nationwide network reliability.

Emergency communication is critical but challenging after natural disasters when the ground infrastructure is devastated. Unmanned aerial vehicles (UAVs) have enormous potential for agile relief coordination in such scenarios. However, effectively leveraging UAV fleets poses additional challenges in terms of security, privacy, and efficient collaboration across response agencies. This paper presents a robust blockchain-enabled framework to address these challenges by integrating a consortium blockchain model, smart contracts, and cryptographic techniques to securely coordinate UAV fleets for disaster response. Specifically, we make two key contributions: a consortium blockchain architecture for secure and private multi-agency coordination and an optimised consensus protocol balancing efficiency and fault tolerance using a delegated proof of stake practical Byzantine fault tolerance (DPoS-PBFT). Comprehensive simulations show the framework’s ability to enhance transparency, automation, scalability, and cyber-attack resilience for UAV coordination in post-disaster networks.

We present the first-ever digitally coded Dynamic Metasurface Antenna (DMA) designed for the 60 GHz millimetre-wave (mmWave) band.

The work begins with the design of a novel, fully addressable complementary electric inductive-capacitive (CELC) metamaterial element, embedded with PIN diodes to enable 1-bit control. A 16-element 1D DMA array is then developed to generate a range of beam patterns through digital coding, controlled by an external FPGA.

The proposed 60 GHz DMA prototype supports customised beamforming and signal processing, making it highly relevant for beyond 5G and 6G applications, including ISAC, holographic communications, sensing and imaging, and mmWave-enabled smart industries.

In this study, a 1-bit phase reconfigurable unit cell (UC) is proposed for an X-band intelligent reflective surface (IRS). The UC is designed by integrating a PIN diode into a slotted copper patch. The maximum phase shift occurs at the design frequency of 9.5 GHz and the phase difference varies within 180∘±20∘ from 9 to 10 GHz in the X-band. A 16 ×16 IRS is designed and simulated using the proposed UC to verify gain and beam steering performance. The IRS has a gain of 20.5dBi at θ=0∘. At θ=±40∘ the gain is 18.6 dBi which validates good beam steering performance.

Large language model (LLM)-based agents are emerging as a key enabler of autonomous, zero-touch operations in 6G AI-RAN systems. Many current frameworks position these agents as decision-makers, capable of optimising network configurations, orchestrating resources, and interacting with users and connected applications.

However, these systems face significant challenges. Intrinsic limitations—such as hallucinations and misalignment with human values—combined with external threats like jailbreaks and prompt injections, introduce serious risks to network safety, reliability, and privacy. These issues remain largely underexplored in existing research.

This paper addresses that gap by reviewing the latest guardrail techniques for 6G AI-RAN. It categorises safeguards at both the model and agent levels and maps them to common agent application patterns within 6G networks. In doing so, it provides a practical foundation for designing more trustworthy, agent-driven decision-making systems.

The emergence of large language models (LLMs) and agentic systems is enabling autonomous 6G networks with advanced intelligence, including self-configuration, self-optimization, and self-healing. However, the current implementation of individual intelligence tasks necessitates isolated knowledge retrieval pipelines, resulting in redundant data flows and inconsistent interpretations. Inspired by the service model unification effort in Open-RAN (to support interoperability and vendor diversity), we propose KP-A: a unified Network Knowledge Plane specifically designed for Agentic network intelligence. By decoupling network knowledge acquisition and management from intelligence logic, KP-A streamlines development and reduces maintenance complexity for intelligence engineers. By offering an intuitive and consistent knowledge interface, KP-A also enhances interoperability for the network intelligence agents. We demonstrate KP-A in two representative intelligence tasks: live network knowledge Q&A and edge AI service orchestration. All implementation artifacts have been open-sourced to support reproducibility and future standardization efforts.

Integrated sensing and communication (ISAC) enables the efficient sharing of wireless resources to support emerging applications, but it also gives rise to new sensing-based security vulnerabilities. Here, potential communication security threats whereby confidential messages intended for legitimate users are intercepted, but also unauthorized receivers (Eves) can passively exploit target echoes to infer sensing parameters without users being aware. Despite these risks, the joint protection of sensing and communication security in ISAC systems remains unexplored. To address this challenge, this paper proposes a two-layer dual-secure ISAC framework that simultaneously protects sensing and communication against passive sensing Eves and communication Eves, without requiring their channel state information (CSI). Specifically, transmit beamformers are jointly designed to inject artificial noise (AN) to introduce interference to communication Eves, while deliberately distorting the reference signal available to sensing Eves to impair their sensing capability. Furthermore, the proposed design generates artificial ghosts (AGs) with fake angle-range-velocity profiles observable by all receivers. Legitimate receivers can suppress these AGs, whereas sensing Eves cannot, thereby significantly reducing their probability of correctly detecting the true targets. Numerical results demonstrate that the proposed framework effectively enhances both communication and sensing security, while preserving the performance of communication users and legitimate sensing receivers.

This work highlights the key aspects of developing a digital twin (DT) for photovoltaic (PV) systems, with a focus on digital models that enable the safe and efficient use of autonomous platforms. In this context, autonomous systems refer to unmanned air systems (UAS) and ground robots.

The integration of autonomous systems introduces the need for a different approach to modelling principles in order to enhance navigation coordination and real-time image transfer. As part of ongoing research, this study investigates aspects of 3D modelling and database implementation.

We present findings derived from terrestrial laser scanner (TLS) and structure-from-motion (SfM) 3D models, alongside their interaction with the database structure for an experimental PV farm.

As part of the project outcomes, we clarify the distinguishing factors between digital twins for PV systems and advanced robotics systems. The potential applications, research gaps and critical lessons learned are also highlighted to support a deeper understanding of the field.

This report aims to provide foundational documentation for future research or industrial implementation with a similar focus. It is envisaged that current and future solar photovoltaic research in this area will build on these insights, highlighting the realities and benefits of DT technology while enhancing the management, control and deeper integration of renewable energy systems.

However, traditional model-based phase-shift optimization is highly sensitive to imperfect CSI and becomes computationally prohibitive for large UPA-based RIS, while existing model-free solutions relying on single-agent DRL struggle with the exponentially growing action space. This paper presents a scalable multi-agent deep Q-network (MADQN)–based RIS controller designed for large-scale UAV–RIS systems under realistic channel dynamics. An end-to-end channel inference architecture is first introduced to mitigate CSI imperfection and reconstruct stable channel representations under UAV mobility. A multi-objective formulation is then developed to jointly optimize sum rate, energy consumption, and control latency, which is transformed into a multi-agent Markov decision process (MMDP) compatible with quantized RIS hardware. Building on this formulation, a dual-agent RIS controller is proposed, in which row and column agents cooperatively determine the quantized phase configuration of a large UPA RIS. Extensive simulations demonstrate that the proposed framework significantly outperforms benchmark schemes, showing acceptable robustness against varying Rician factor SNRs, UAV densities, and RIS sizes. These results confirm that the proposed MADQN-based controller is a promising and practical solution for scalable RIS control in large-scale multi-UAV communication systems.

The rapid development of Intelligent Transportation Systems (ITS) and the Internet of Vehicles (IoV) has revolutionized transportation networks by enabling real-time communication between vehicles, road infrastructure, and cloud systems. One such advancement is the vehicular reverse offloading system, which allows Road Side Units (RSUs) to offload tasks to vehicles on the road. This paradigm makes full use of the dynamic computational resources available on vehicles and helps reduce the overall carbon emissions of IoV systems. In this paper, we establish a multi-hop reverse offloading vehicular edge computing model, enabling RSUs to utilize dynamic computational resources beyond their communication range. Considering the potential variations in power supply sources for RSUs, we further construct a carbon intensity adaptive carbon emission optimization model for RSUs and optimize the system’s overall carbon emissions through deep reinforcement learning(DRL). Through extensive simulations, we demonstrate that our DRL-based approach significantly reduces carbon emissions compared to traditional task-offloading methods.

To advance towards carbon-neutrality and improve the limited performance of conventional passive wireless communications, in this paper, we investigate the integration of noise modulation with zero-energy reconfigurable intelligent surfaces (RISs). In particular, the RIS reconfigurable elements (REs) are divided into two groups: one for beamforming the desired signals in reflection mode and another for harvesting energy from interference signals in an absorption mode, providing the power required for RIS operation. Since the harvested energy is a random variable, a random number of REs can beamform the signals, while the remainder randomly reflects them. We present a closed-form solution and a search algorithm for REs allocation, jointly optimizing both the energy harvesting (EH) and communication performance. Considering the repetition coding technique and discrete phase shifts, we derive analytical expressions for the energy constrained success rate, bit error rate, optimal threshold, mutual information, and energy efficiency. Numerical and simulation results confirm the effectiveness of the algorithm and expressions, demonstrating the superiority of the proposed integration over conventional noise-modulation systems. It is shown that by properly allocating the REs, both the EH and communication performance can be improved in low to moderate interference scenarios, while the latter is restricted in the high-interference regime.

We consider analog over-the-air federated learning, where devices harvest energy from in-band and out-band radio frequency signals, with the former also causing co-channel interference (CCI). To mitigate the aggregation error, we propose an effective denoising policy that does not require channel state information (CSI). We also propose an adaptive scheduling algorithm that dynamically adjusts the number of local training epochs based on available energy, enhancing device participation and learning performance while reducing energy consumption. Simulation results and convergence analysis confirm the robust performance of the algorithm compared to conventional methods. It is shown that the performance of the proposed denoising method is comparable to that of conventional CSI-based methods. It is observed that high-power CCI severely degrades the learning performance, which can be mitigated by increasing the number of active devices, achievable via the adaptive algorithm.

The screening tools for respiratory diseases typically involve spirometry (for asthma and COPD), CT scans (for interstitial lung disease), chest X-rays (for pneumonia and tuberculosis), and sputum analysis (for tuberculosis).

This work examines a diagnostic approach whereby a subject’s chest is radio-exposed to non-ionising 6G/WiFi multi-carrier radio signals at a frequency of 5.23 GHz. The fact that each respiratory disease modulates the amplitude, frequency, and phase of each radio frequency differently allows us to screen for five respiratory diseases: asthma, chronic obstructive pulmonary disease, interstitial lung disease, pneumonia, and tuberculosis. We collect a new dataset (OFDM-Breathe) from 220 individuals in a hospital setting, including 190 patients and 30 healthy controls. The dataset contains over 26,000 s of radio signal recordings across 64 frequencies. Several machine learning and deep learning models are evaluated to classify disease type based on the discriminatory signatures of radio signals.

Coordinated controlling a large UAV swarm requires significant spectrum resources due to the need for bandwidth allocation per UAV, posing a challenge in resource-limited environments. Over-the-air (OTA) control has emerged as a spectrum-efficient approach, leveraging electromagnetic superposition to form control signals at a base station (BS). However, existing OTA controllers lack sufficient optimization variables to meet UAV swarm control objectives and fail to integrate control with other BS functions like sensing. This work proposes an integrated sensing and OTA control framework (ISAC-OTA) for UAV swarm. The BS performs OTA signal construction (uplink) and dispatch (downlink) while simultaneously sensing objects. Two uplink post-processing methods are developed: a control-centric approach generating closed-form control signals via a feedback-looped OTA control problem, and a sensing-centric method mitigating transmission-induced interference for accurate object sensing. For the downlink, a non-convex problem is formulated and solved to minimize control signal dispatch (transmission) error while maintaining a minimum sensing signal-to-interference-plus-noise ratio (SINR). Simulation results show that the proposed ISAC-OTA controller achieves control performance comparable to the ideal optimal control algorithm while maintaining high sensing accuracy, despite OTA transmission interference. Moreover, it eliminates the need for per-UAV bandwidth allocation, showcasing a spectrum-efficient method for cooperative control in future wireless systems.

This letter presents a deterministic diffusion Q-learning scheme for resource allocation in Open Radio Access Network (O-RAN) digital twin systems. The proposed approach jointly optimizes resource allocation across all user equipments (UEs) by formulating Q-values that reflect the dynamics of the O-RAN environment. By incorporating diffusion policies into Q-learning, the method enhances exploration and enables smoother policy optimization through the diffusion process, resulting in more robust and efficient learning.

To improve computational efficiency, a deterministic diffusion mechanism is introduced to eliminate redundant iteration steps in the forward–reverse diffusion process. Compared with traditional stochastic diffusion models, the proposed method significantly reduces computational complexity while supporting customizable optimization objectives. Simulation results demonstrate that the proposed approach improves total throughput by 8.2% compared to proportional fairness, while achieving comparable performance to classical Q-learning using 7.9% fewer resources. Furthermore, the model can be readily adapted to diverse O-RAN scenarios with minimal fine-tuning, leveraging the inherent flexibility of the diffusion-based framework.

Dynamic and uncertain environments pose major challenges for multi-agent autonomous systems,particularly in achieving robust simultaneous localization and mapping (SLAM) and efficientknowledge sharing across robots. Conventional data-driven methods often overlook underlyingcausal structures, resulting in spurious correlations and limited generalization. To address this, wepresent CASK—an edge-assisted causal knowledge aggregation framework that fuses structuredcausal inference with data-driven learning to improve adaptive decision-making. A key feature is atime-based normalization mechanism that ensures mapping consistency across varying operationalspeeds, enabling speed-independent transfer of spatial knowledge between heterogeneous agents.We validate CASK through simulations and real-world experiments using autonomous groundvehicles, a class of mobile robots. Results show substantial gains over state-of-the-art methods: up to20% higher success at low speeds, 40% at high speeds, 50% lower trajectory deviation, and 45%fewer re-planning steps. Thesefindings demonstrate how causal inference combined with mobileedge computing enables scalable, reliable, and generalizable autonomy in multi-agent systems.

Generative AI (GenAI) is expected to play a pivotal role in enabling autonomous optimization in future wireless networks. Within the ORAN architecture, Large Language Models (LLMs) can be specialized to generate xApps and rApps by leveraging specifications and API definitions from the RAN Intelligent Controller (RIC) platform. However, fine-tuning base LLMs for telecom-specific tasks remains expensive and resource-intensive. Retrieval-Augmented Generation (RAG) offers a practical alternative through in-context learning, enabling domain adaptation without full retraining. While traditional RAG systems rely on vector-based retrieval, emerging variants such as GraphRAG and Hybrid GraphRAG incorporate knowledge graphs or dual retrieval strategies to support multi-hop reasoning and improve factual grounding. Despite their promise, these methods lack systematic, metric-driven evaluations, particularly in high-stakes domains such as ORAN. In this study, we conduct a comparative evaluation of Vector RAG, GraphRAG, and Hybrid GraphRAG using ORAN specifications. We assess performance across varying question complexities using established generation metrics: faithfulness, answer relevance, context relevance, and factual correctness. Results show that both GraphRAG and Hybrid GraphRAG outperform traditional RAG. Hybrid GraphRAG improves factual correctness by 8%, while GraphRAG improves context relevance by 11%.

We propose a resource allocation optimisation framework that integrates the O-RAN architecture with out-band Integrated Access and Backhaul (IAB). Using mixed-integer linear programming (MILP), we coordinate multiple control functions across O-RAN to achieve global optimality. The MILP model jointly optimises: (i) radio-resource allocations among users (power, numerology, and resource block) at the DU level to satisfy diverse data rate and latency requirements in frequency-selective channels; (ii) carrier bandwidth partitioning among numerology based Bandwidth Parts (BWPs) at the CU control-plane level as a longer-term policy; and (iii) wireless backhaul routing over mmWave links from IAB Donors to IAB Nodes at the CU user-plane level. Our framework leverages mmWave frequency bands for backhaul despite their path-loss and penetration-loss challenges. The main objectives are to minimise downlink transmit power and co-channel interference (CCI) while maximising the number of simultaneously served users. Results show that integrating IAB reduces average total transmit power by 35%. We also compare flexible and non-flexible numerology allocation, with IAB and flexible numerology demonstrating significant improvements in both CCI and downlink energy efficiency. For faster implementation, we develop two hybrid genetic algorithms that converge rapidly, delivering near-optimal solutions far more quickly than the MILP model. This makes them well-suited to real-time resource allocation and dynamic network management in O-RAN with out-of-band IAB.

6G’s AI native vision of embedding advance intelligence in the network while bringing it closer to the user requires a systematic evaluation of Generative AI (GenAI) models on edge devices. Rapidly emerging solutions based on Open RAN (ORAN) and Network-in-a-Box strongly advocate the use of low-cost, off-the-shelf components for simpler and efficient deployment, e.g., in provisioning rural connectivity. In this context, conceptual architecture, hardware testbeds and precise performance quantification of Large Language Models (LLMs) on off-the-shelf edge devices remains largely unexplored. This research investigates computationally demanding LLM inference on a single commodity Raspberry Pi serving as an edge testbed for ORAN. We investigate various LLMs, including small, medium and large models, on a Raspberry Pi 5 Cluster using a lightweight Kubernetes distribution (K3s) with modular prompting implementation. We study its feasibility and limitations by analyzing throughput, latency, accuracy and efficiency. Our findings indicate that CPU-only deployment of lightweight models, such as Yi, Phi, and Llama3, can effectively support edge applications, achieving a generation throughput of 5 to 12 tokens per second with less than 50\% CPU and RAM usage. We conclude that GenAI on the edge offers localized inference in remote or bandwidth-constrained environments in 6G networks without reliance on cloud infrastructure.

The rise of large language models (LLMs) has transformed natural language processing (NLP) and generative artificial intelligence (AI) applications. However, deploying these transformer-based models in resource-constrained environments poses a significant challenge due to their high computational and memory demands. To address this, we introduce in this article generative AI (GenAI) on the Edge, a comprehensive benchmarking dataset designed to evaluate the performance of LLMs deployed on edge devices. Leveraging a distributed testbed of Raspberry Pi 5 devices orchestrated with lightweight Kubernetes (K3s), the dataset captures a broad range of performance metrics essential for assessing the feasibility of local inference in constrained environments. These metrics include detailed measurements of throughput, inference latency, memory utilization, and computational efficiency, along with granular timing data for key stages of the inference pipeline—sample, prefill, and decode phases. We systematically evaluate LLMs of varying sizes under real-world deployment scenarios, with a particular emphasis on CPU-based edge platforms. By conducting multiple runs of conversation-based evaluations, GenAI on the Edge provides actionable insights into the tradeoffs between performance and resource efficiency, enabling better decision-making for LLM deployment in edge environments. IEEE SOCIETY/COUNCIL Communications Society (ComSoc) DATA TYPE/LOCATION Structured Text Data (CSV); Leeds, U.K. DATA DOI/PID 10.21227/7d08-8655

This paper envisions 6G as a self-evolving telecom ecosystem, where AI-driven intelligence enables dynamic adaptation beyond static connectivity. We explore the key enablers of autonomous communication systems, spanning reconfigurable infrastructure, adaptive middleware, and intelligent network functions, alongside multi-agent collaboration for distributed decision-making. We explore how these methodologies align with emerging industrial IoT frameworks, ensuring seamless integration within digital manufacturing processes. Our findings emphasize the potential for improved real-time decision-making, optimizing efficiency, and reducing latency in networked control systems. The discussion addresses ethical challenges, research directions, and standardization efforts, concluding with a technology stack roadmap to guide future developments. By leveraging state-of-the-art 6G network management techniques, this research contributes to the next generation of intelligent automation solutions, bridging the gap between theoretical advancements and real-world industrial applications.

Generative multiagent systems are rapidly emerging as transformative tools for scalable automation and adaptive decisionmaking in telecommunications. Despite their promise, these systems introduce novel risks that remain underexplored, particularly when agents operate asynchronously across layered architectures. This paper investigates key safety pathways in telecomfocused Generative MultiAgent Systems (GMAS), emphasizing risks of miscoordination and semantic drift shaped by persona diversity. We propose a modular safety evaluation framework that integrates agentlevel checks on code quality and compliance with systemlevel safety metrics. Using controlled simulations across 32 persona sets, five questions, and multiple iterative runs, we demonstrate progressive improvements in analyzer penalties and AllocatorCoder consistency, alongside persistent vulnerabilities such as policy drift and variability under specific persona combinations. Our findings provide the first domaingrounded evidence that persona design, coding style, and planning orientation directly influence the stability and safety of telecom GMAS, highlighting both promising mitigation strategies and open risks for future deployment.

The increasing adoption of Extended Reality (XR) applications demands advanced traffic modelling to optimise next-generation wireless networks. Existing models often rely on fixed data rates (e.g., 50 Mbps) and frame rates (e.g., 60 Hz), which are insufficient for modern VR systems. This paper introduces an enhanced traffic modelling framework for highly interactive 6-degree-of-freedom (6DoF) VR gaming, incorporating video, audio, and control streams with support for higher frame rates (up to 120 Hz) and data rates (up to 251.9 Mbps). Built on empirical data from a high-fidelity VR setup, the model uses statistical analysis to fit data rate and inter-arrival times to probability distributions. The generalized logistic distribution is identified as the best fit for data rates, while the generalized extreme value distribution accurately represents inter-arrival times. Model accuracy is validated using Kullback-Leibler divergence. The findings offer critical insights for traffic generation, network planning, and resource allocation, supporting scalable and immersive XRexperiences in future wireless networks.

Open Radio Access Network (O-RAN) architecture provides an intrinsic capability to exploit key performance monitoring (KPM) within Radio Intelligence Controller (RIC) to derive network optimisation through xApps. These xApps can leverage KPM knowledge to dynamically switch on/off the associated RUs where such a function is supported over the E2 interface. Several existing studies employ artificial intelligence (AI)/Machine Learning (ML) based approaches to realise such dynamic sleeping for increased energy efficiency (EE). Nevertheless, most of these approaches rely upon offloading user equipment (UE) to carve out a sleeping opportunity. Such an approach inherently creates load imbalance across the network. Such load imbalance may impact the throughput performance of offloaded UEs as they might be allocated a lower number of physical resource blocks (PRBs). Maintaining the same PRB allocation while addressing the EE at the network level is a challenging task. To that end, in this article, we present a comprehensive ML-based framework for joint optimisation of load balancing and EE for ORAN deployments. We formulate the problem as a multi-class classification system that predictively evaluates potential RU configurations before optimising the EE, mapping network conditions to three load balance categories (Well Balanced, Moderately Balanced, Imbalanced). Our multi-threshold approach (Conservative, Moderate, Aggressive) accommodates different operational priorities between energy savings and performance assurance. Experimental evaluation using 4.26 million real network measurements from simulations demonstrates that our Random Forest model achieves 98.3% F1-macro performance, representing 195% improvement over traditional baseline strategies.

6G wireless networks aim to exploit semantic awareness to optimize radio resources. By optimizing the transmission through the lens of the desired goal, the energy consumption of transmissions can also be reduced, and the latency can be improved. To that end, this paper investigates a paradigm in which the capabilities of generative AI (GenAI) on the edge are harnessed for network optimization. In particular, we investigate an Unmanned Aerial Vehicle (UAV) handover framework that takes advantage of GenAI and semantic communication to maintain reliable connectivity. To that end, we propose a framework in which a lightweight MobileBERT language model, fine-tuned using Low-Rank Adaptation (LoRA), is deployed on the UAV. This model processes multi-attribute flight and radio measurements and performs multi-label classification to determine appropriate handover action. Concurrently, the model identifies an appropriate set of contextual “Reason Tags” that elucidate the decision’s rationale. Our model, evaluated on a rule-based synthetic dataset of UAV handover scenarios, demonstrates the model’s high efficacy in learning these rules, achieving high accuracy in predicting the primary handover decision. The model also shows strong performance in identifying supporting reasons, with an F1 micro-score of approximately 0.9 for reason tags.

This paper presents a novel framework for optimising energy consumption in ORAN networks using machine learning (ML) models integrated with realistic mobility and spatial data aggregation techniques. The proposed approach leverages real-time Key Performance Metrics (KPMs) to dynamically manage the power states of Radio Units (RUs), ensuring energy efficiency while maintaining network performance. A dense urban simulation environment with realistic mobility patterns, based on a Poisson Point Process and Dijkstra’s Algorithm, models user movement and traffic dynamics. To address the challenges of large-scale dataset management, an H3 spatial indexing system aggregates data into hexagonal grids, reducing data size by 74% without sacrificing spatial accuracy. Five ML-based classifiers, including ensemble and regression-based methods, were trained and evaluated using the aggregated dataset based on actual data from the city of Leeds. The results demonstrate high accuracy for optimal power plans, with models achieving up to 97.8% accuracy. Network performance metrics, including throughput and energy efficiency, highlight significant improvements over a Full Power Baseline (FPB), with energy consumption reduced by up to 33.88% using the proposed models. These findings underscore the potential of ML-driven approaches to optimise energy usage in ORAN networks, providing a scalable and effective solution for sustainable network operations.

Using Unmanned Aerial Vehicles (UAVs) in Search and rescue operations (SAR) to navigate challenging terrain while maintaining reliable communication with the cellular network is a promising approach. This paper suggests a novel technique employing a reinforcement learning multi Q-learning algorithm to optimize UAV connectivity in such scenarios. We introduce a Strategic Planning Agent for efficient path planning and collision awareness and a Real-time Adaptive Agent to maintain optimal connection with the cellular base station. The agents trained in a simulated environment using multi Q-learning, encouraging them to learn from experience and adjust their decision-making to diverse terrain complexities and communication scenarios. Evaluation results reveal the significance of the approach, highlighting successful navigation in environments with varying obstacle densities and the ability to perform optimal connectivity using different frequency bands. This work paves the way for enhanced UAV autonomy and enhanced communication reliability in search and rescue operations.

The rapid growth of unmanned aerial vehicle (UAV) applications—including last-mile delivery, medical support, and agricultural monitoring—has underscored the need for reliable beyond-line-of-sight (BLoS) operations supported by wide-area connectivity. Although 5G network densification is well suited to meet these requirements, conventional cellular networks are not optimised for serving aerial users. This paper presents a novel approach to enhancing delivery-UAV connectivity with ground base stations using low-complexity online reinforcement learning and multiple Q-learning algorithms. By training multiple Q-learning models, the proposed system predicts optimal UAV trajectories that minimise flight path length while maintaining continuous connectivity with ground infrastructure. The results demonstrate a practical and scalable solution for reliable UAV communications in real-world deployment scenarios.

The rapid proliferation of consumer UAVs, or drones, is reshaping the wireless communication landscape. These agile, autonomous devices find new life as UE in cellular networks. This paper explores their integration, emphasizing the myriad applications, standardization efforts, challenges, and research community solutions. Key areas of investigation include the complexities of 3D deployment, channel modelling, and energy efficiency. Moreover, we highlight the open questions and research opportunities these flying UEs present. The evolving landscape of UAV integration into cellular networks promises transformative enhancements for next-generation communications, addressing challenges while fostering innovation across industries. The paper encapsulates the essential aspects of UAV integration within the cellular ecosystem, offering a concise yet comprehensive overview of this dynamic field, where UAVs as UEs redefine wireless communication with promise and complexity.

The disaggregated, distributed and virtualised implementation of radio access networks allows for dynamic resource allocation. These attributes can be realised by virtue of the Open Radio Access Networks (O-RAN) architecture. In this article, we tackle the issue of dynamic resource allocation using a data-driven approach by employing Machine Learning (ML). We present an xApp-based implementation for the proposed ML algorithm. The core aim of this work is to optimise resource allocation and fulfil Service Level Specifications (SLS). This is accomplished by dynamically adjusting the allocation of Physical Resource Blocks (PRBs) based on traffic demand and Quality of Service (QoS) requirements. The proposed ML model effectively selects the best allocation policy for each base station and enhances the performance of scheduler functionality in O-RAN – Distributed Unit (O-DU). We show that an xApp implementing the Random Forest Classifier can yield high (85\%) performance accuracy for optimal policy selection. This can be attained using the O-RAN instance state input parameters over a short training duration.

The vision of AI-RAN convergence, as advocated by the AI-RAN Alliance, aims to unlock a unified 6G platform capable of seamlessly supporting AI and RAN workloads over shared infrastructure. However, the architectural framework and intelligent resource orchestration strategies necessary to realize this vision remain largely unexplored. In this paper, we propose a Converged AI-and-ORAN Architectural (CAORA) framework based on O-RAN specifications, enabling the dynamic coexistence of real-time RAN and computationally intensive AI workloads. We design custom xApps within the Near-Real-Time RAN Intelligent Controller (NRT-RIC) to monitor RAN KPIs and expose radio analytics to an End-to-End (E2E) orchestrator via the recently introduced Y1 interface. The orchestrator incorporates workload forecasting and anomaly detection modules, augmenting a Soft Actor-Critic (SAC) reinforcement learning agent that proactively manages resource allocation, including Multi-Instance GPU (MIG) partitioning. Using real-world 5G traffic traces from Barcelona, our trace-driven simulations demonstrate that CAORA achieves near 99\% fulfillment of RAN demands, supports dynamic AI workloads, and maximizes infrastructure utilization even under highly dynamic conditions. Our results reveal that predictive orchestration significantly improves system adaptability, resource efficiency, and service continuity, offering a viable blueprint for future AI-and-RAN converged 6G systems.

Federated Learning (FL) enables distributed model training on edge devices while preserving data privacy. However, FL deployments in wireless networks face significant challenges, including communication overhead, unreliable connectivity, and high energy consumption, particularly in dynamic environments. This paper proposes EcoFL, an integrated FL framework that leverages the Open Radio Access Network (ORAN) architecture with multiple Radio Access Technologies (RATs) to enhance communication efficiency and ensure robust FL operations. EcoFL implements a two-stage optimisation approach: an RL-based rApp for dynamic RAT selection that balances energy efficiency with network performance, and a CNN-based xApp for near real-time resource allocation with adaptive policies. This coordinated approach significantly enhances communication resilience under fluctuating network conditions. Experimental results demonstrate competitive FL model performance with 19\% lower power consumption compared to baseline approaches, highlighting substantial potential for scalable, energy-efficient collaborative learning applications.

Next-generation wireless networks require intelligent traffic prediction to enable autonomous resource management and handle diverse, dynamic service demands. The Open Radio Access Network (O-RAN) framework provides a promising foundation for embedding machine learning intelligence through its disaggregated architecture and programmable interfaces. This work applies a Neural Architecture Search (NAS)-based framework that dynamically selects and orchestrates efficient Long Short-Term Memory (LSTM) architectures for traffic prediction in O-RAN environments. Our approach leverages the O-RAN paradigm by separating architecture optimisation (via non-RT RIC rApps) from real-time inference (via near-RT RIC xApps), enabling adaptive model deployment based on traffic conditions and resource constraints. Experimental evaluation across six LSTM architectures demonstrates that lightweight models achieve R2≈0.91—0.93with high efficiency for regular traffic, while complex models reach near-perfect accuracy (R2=0.989—0.996) during critical scenarios. Our NAS-based orchestration achieves a 70-75\% reduction in computational complexity compared to static high-performance models, while maintaining high prediction accuracy when required, thereby enabling scalable deployment in real-world edge environments.

This work proposes an integrated approach for optimising Federated Learning (FL) communication in dynamic and heterogeneous network environments. Leveraging the modular flexibility of the Open Radio Access Network (ORAN) architecture and multiple Radio Access Technologies (RATs), we aim to enhance data transmission efficiency and mitigate client-server communication constraints within the FL framework. Our system employs a two-stage optimisation strategy using ORAN’s rApps and xApps. In the first stage, Reinforcement Learning (RL) based rApp is used to dynamically select each user’s optimal Radio Access Technology (RAT), balancing energy efficiency with network performance. In the second stage, a model-based xApp facilitates near-real-time resource allocation optimisation through predefined policies to achieve optimal network performance. The dynamic RAT selection and resource allocation capabilities enabled by ORAN and multi-RAT contribute to robust communication resilience in dynamic network environments. Our approach demonstrates competitive performance with low power consumption compared to other state-of-the-art models, showcasing its potential for real-time applications demanding both accuracy and efficiency. This robust and comprehensive framework, enabling clients to utilise available resources effectively, highlights the potential for scalable, collaborative learning applications prioritising energy efficiency and network performance.

The deployment of AI agents within legacy Radio Access Network (RAN) infrastructure poses significant safety and reliability challenges for future 6G networks. This paper presents a novel Edge AI framework for autonomous network optimisation in Open RAN environments, addressing these challenges through three core innovations: (1) a persona-based multi-tools architecture enabling distributed, context-aware decision-making; (2) proactive anomaly detection agent powered by traffic predictive tool; and (3) a safety, aligned reward mechanism that balances performance with operational stability. Integrated into the RAN Intelligent Controller (RIC), our framework leverages multimodal data fusion, including network KPIs, a traffic prediction model, and external information sources, to anticipate and respond to dynamic network conditions. Extensive evaluation using realistic 5G scenarios demonstrates that the edge framework achieves zero network outages under high-stress conditions, compared to 8.4% for traditional fixed-power networks and 3.3% for large language model (LLM) agent-based approaches, while maintaining near real-time responsiveness and consistent QoS. These results establish that, when equipped with the right tools and contextual awareness, AI agents can be safely and effectively deployed in critical network infrastructure, laying the framework for intelligent and autonomous 5G and beyond network operations.

6G open radio access networks (O-RAN) promises to open data interfaces to enable plug-and-play service apps, many of which are consumer and business-facing. Opening up 6G access lowers the barrier to innovation but raises the challenge of required communication specifications that are not fully known to all service designers. As such, business innovators must either be familiar with 6G standards, or consult with experts. Enabling consistent, unbiased, rapid, and low-cost requirement assessment and specification generation is crucial to the O-RAN innovation ecosystem. Here, we discuss our initiative to bridge service specification gaps between network service providers and business innovators leveraging large language models (LLMs). We first review the state-of-the-art and motivation in 6G plug-and-play services, capabilities, potential use cases, and LLMs. We identify an ample innovation space for hybrid use cases that may require diverse and variational wireless functionalities across its operating time. We show that the network specification can be automated, and present the first automatic retrieval-augmented network service specification framework for 6G use cases. To enable public acceptance and feedback, a website interface is published for the research and industrial communities to experiment with the framework. We hope this review highlights the need for emerging foundation models for this area and motivates

Data-driven proactive network optimisation is critical for 5G advanced and 6G, allowing operators to dynamically allocate cellular spectrum reuse in anticipating for demand surges. Current approaches to traffic prediction are largely temporal correlation based. We know causal inference of key factors can help to improve prediction accuracy for spike traffic events and identify pathways to improve services. Current causal inference identify stationary independent variables, but real environments have open challenges: (i) dynamic and heterogeneous causal maps, (ii) cascade partially observable variables, and/or (iii) have coupled / confounding relationships. Currently there is no research that dynamically configures the causal relationship according to emerging real-time data and shares inference outcomes across the data sharing and computing continuum of Open-RAN (ORAN) architecture. Here, we use both real cellular network traffic and social event triggers to perform nonlinear causal inference as an rApp: Predictability Improvement (PI), Conditional Mutual Information (CMI), and Convergent Cross Map (CCM). This causal knowledge is then shared across the ORAN to be embedded in traffic prediction xApps: hard causal embedding to Recurrent Neural Network (RNN) and soft causal feature embedding to a Gaussian Processes (GP). The results show a significant accuracy improvement (93-99%) over baseline non-causal correlated prediction (76-94%) and blind multi-variate approaches (87-95%). This work paves the way to causal proactive network optimisation.

Cascade stability of load balancing is critical for ensuring high efficiency service delivery and preventing undesirable handovers. In energy efficient networks that employ diverse sleep mode operations, handing over traffic to neighbouring cells’ expanded coverage must be done with minimal side effects. Current research is largely concerned with designing distributed and centralized efficient load balancing policies that are locally stable. There is a major research gap in identifying largescale cascade stability for networks with heterogeneous load balancing policies arising from diverse plug-and-play sleep mode policies in ORAN, which will cause heterogeneity in the network stability behaviour. Here, we investigate whether cells arbitrarily connected for load balancing and having an arbitrary number undergoing sleep mode can: (i) synchronize to a desirable load-balancing state, and (ii) maintain stability. For the first time, we establish the criterion for stability and prove its validity for any general load dynamics and random network topology. Whilst its general form allows all load balancing and sleep mode dynamics to be incorporated, we propose an ORAN architecture where the network service management and orchestration (SMO) must monitor new load balancing policies to ensure overall network cascade stability.

Future telecommunications will increasingly integrate AI capabilities into network infrastructures to deliver seamless and harmonized services closer to end-users. However, this progress also raises significant trust and safety concerns. The machine learning systems orchestrating these advanced services will widely rely on deep reinforcement learning (DRL) to process multi-modal requirements datasets and make semantically modulated decisions, introducing three major challenges: (1) First, we acknowledge that most explainable AI research is stakeholder-agnostic while, in reality, the explanations must cater for diverse telecommunications stakeholders, including network service providers, legal authorities, and end users, each with unique goals and operational practices; (2) Second, DRL lacks prior models or established frameworks to guide the creation of meaningful long-term explanations of the agent’s behaviour in a goal-oriented RL task, and we introduce state-of-the-art approaches such as reward machine and sub-goal automata that can be universally represented and easily manipulated by logic programs and verifiably learned by inductive logic programming of answer set programs; (3) Third, most explainability approaches focus on correlation rather than causation, and we emphasise that understanding causal learning can further enhance 6G network optimisation. Together, in our judgement they form crucial enabling technologies for trustworthy services in 6G. This review offers a timely resource for academic researchers and industry practitioners by highlighting the methodological advancements needed for explainable DRL (X-DRL) in 6G. It identifies key stakeholder groups, maps their needs to X-DRL solutions, and presents case studies showcasing practical applications. By identifying and analysing these challenges in the context of 6G case studies, this work aims to inform future research, transform industry practices, and highlight unresolved gaps in this rapidly evolving field.

With the advent of 6G, Open Radio Access Network (O-RAN) architectures are evolving to support intelligent, adaptive, and automated network orchestration. This paper proposes a novel Edge AI and Network Service Orchestration framework that leverages Large Language Model (LLM) agents deployed as O-RAN rApps. The proposed LLM-agent-powered system enables interactive and intuitive orchestration by translating the user’s use case description into deployable AI services and corresponding network configurations. The LLM agent automates multiple tasks, including AI model selection from repositories (e.g., Hugging Face), service deployment, network adaptation, and real-time monitoring via xApps. We implement a prototype using open-source O-RAN projects (OpenAirInterface and FlexRIC) to demonstrate the feasibility and functionality of our framework. Our demonstration showcases the end-to-end flow of AI service orchestration, from user interaction to network adaptation, ensuring Quality of Service (QoS) compliance. This work highlights the potential of integrating LLM-driven automation into 6G O-RAN ecosystems, paving the way for more accessible and efficient edge AI ecosystems.

Mixed-precision computing, a widely applied technique in AI, offers a larger trade-off space between accuracy and efficiency. The recent purposed Mixed-Precision Over-the-Air Federated Learning (MP-OTA-FL) enables clients to operate at appropriate precision levels based on their heterogeneous hardware, taking advantages of the larger trade-off space while covering the quantization overheads in the mixed-precision modulation scheme for the OTA aggregation process. A key to further exploring the potential of the MP-OTA-FL framework is the optimization of client precision levels. The choice of precision level hinges on multifaceted factors including hardware capability, potential client contribution, and user satisfaction, among which factors can be difficult to define or quantify. In this paper, we propose a RAG-based User Profiling for precision planning framework that integrates retrieval-augmented LLMs and dynamic client profiling to optimize satisfaction and contributions. This includes a hybrid interface for gathering device/user insights and an RAG database storing historical quantization decisions with feedback. Experiments show that our method boosts satisfaction, energy savings, and global model accuracy in MP-OTA-FL systems.

Load balancing in open radio access networks (O-RAN) is critical for ensuring efficient resource utilization, and the user’s experience by evenly distributing network traffic load. Current research mainly focuses on designing load-balancing algorithms to allocate resources while overlooking the cascade stability of load balancing, which is critical to prevent endless handover. The main challenge to analyse the cascade stability lies in the difficulty of establishing an accurate mathematical model to describe the process of load balancing due to its nonlinearity and high-dimensionality. In our previous theoretical work, a simplified general dynamic function was used to analyze the stability. However, it is elusive whether this function is close to the reality of the load balance process. To solve this problem, 1) a data-driven method is proposed to identify the dynamic model of the load balancing process according to the real-time traffic load data collected from the radio units (RUs); 2) the stability condition of load balancing process is established for the identified dynamics model. Based on the identified dynamics model and the stability condition, the RAN Intelligent Controller (RIC) can control RUs to achieve a desired load-balancing state while ensuring cascade stability.

Efficient orchestration of AI services in 6G AI-RAN requires well-structured, ready-to-deploy AI service repositories combined with orchestration methods adaptive to diverse runtime contexts across radio access, edge, and cloud layers. Current literature lacks comprehensive frameworks for constructing such repositories and generally overlooks key practical orchestration factors. This paper systematically identifies and categorizes critical attributes influencing AI service orchestration in 6G networks and introduces an open-source, LLM-assisted toolchain that automates service packaging, deployment, and runtime profiling. We validate the proposed toolchain through the Cranfield AI Service repository case study, demonstrating significant automation benefits, reduced manual coding efforts, and the necessity of infrastructure-specific profiling, paving the way for more practical orchestration frameworks.

Holographic multiple-input multiple-output (MIMO) enables electrically large continuous apertures, overcoming the physical scaling limits of conventional MIMO architectures with half-wavelength spacing. Their near-field operating regime requires channel models that jointly capture line-of-sight (LoS) and non-line-of-sight (NLoS) components in a physically consistent manner. Existing studies typically treat these components separately or rely on environment-specific multipath models. In this work, we develop a unified LoS+NLoS channel representation for holographic lines that integrates spatial-sampling-based and expansion-based formulations. Building on this model, we extend the wavenumber-division multiplexing (WDM) framework, originally introduced for purely LoS channels, to the LoS+NLoS scenario. Applying WDM to the NLoS component yields its angular-domain representation, enabling direct characterization through the power spectral factor and power spectral density. We further derive closed-form characterizations for isotropic and non-isotropic scattering, with the former recovering Jakes’ isotropic model. Lastly, we evaluate the resulting degrees of freedom and ergodic capacity, showing that incorporating the NLoS component substantially improves the performance relative to the purely LoS case.

Reconfigurable intelligent surfaces (RIS) can both help and hinder the physical layer secret key generation (PL-SKG) of communications systems. Whilst a legitimate RIS can yield beneficial impacts, including increased channel randomness to enhance PL-SKG, a malicious RIS can poison legitimate channels and crack almost all existing PL-SKGs. In this work, we propose an adversarial learning framework that addresses Man-in-the-middle RIS (MITM-RIS) eavesdropping which can exist between legitimate parties, namely Alice and Bob. First, the theoretical mutual information gap between legitimate pairs and MITM-RIS is deduced. From this, Alice and Bob leverage adversarial learning to learn a common feature space that assures no mutual information overlap with MITM-RIS. Next, to explain the trained legitimate common feature generator, we aid signal processing interpretation of black-box neural networks using a symbolic explainable AI (xAI) representation. These symbolic terms of dominant neurons aid the engineering of feature designs and the validation of the learned common feature space. Simulation results show that our proposed adversarial learning- and symbolic-based PL-SKGs can achieve high key agreement rates between legitimate users, and is further resistant to an MITM-RIS Eve with the full knowledge of legitimate feature generation (NNs or formulas). This therefore paves the way to secure wireless communications with untrusted reflective devices in future 6G.

The research introduces an innovative metasurface design to provide multi-band capabilities for both cross-polarization and circular polarization. The metasurface consists of unit cells based on a ring-shaped configuration with a star inside. It is designed on a substrate made of Roger 5880. The thickness of this substrate is 1.575 mm, and the loss tangent is 0.009. The design acts like an efficient cross-polarizer, attaining a fractional bandwidth of 66% within the frequency bands spanning 23.24-34.64 GHz and 36.68-41.12 GHz. Additionally, this design functions as a circular polarizer in the frequency bands of 11.76-13 GHz, 19.43-21.6 GHz, 35.94-36.58 GHz, and 41.28-42.28 GHz.

The unprecedented advancement of Artificial Intelligence (AI) has positioned Explainable AI (XAI) as a critical enabler in addressing the complexities of next-generation wireless communications. With the evolution of the 6G networks, characterized by ultra-low latency, massive data rates, and intricate network structures, the need for transparency, interpretability, and fairness in AI-driven decision-making has become more urgent than ever. This survey provides a comprehensive review of the current state and future potential of XAI in communications, with a focus on network slicing, a fundamental technology for resource management in 6G. By systematically categorizing XAI methodologies–ranging from modelagnostic to model-specific approaches, and from pre-model to post-model strategies–this paper identifies their unique advantages, limitations, and applications in wireless communications. Moreover, the survey emphasizes the role of XAI in network slicing for vehicular network, highlighting its ability to enhance transparency and reliability in scenarios requiring real-time decision-making and high-stakes operational environments. Real-world use cases are examined to illustrate how XAI-driven systems can improve resource allocation, facilitate fault diagnosis, and meet regulatory requirements for ethical AI deployment. By addressing the inherent challenges of applying XAI in complex, dynamic networks, this survey offers critical insights into the convergence of XAI and 6G technologies. Future research directions, including scalability, real-time applicability, and interdisciplinary integration, are discussed, establishing a foundation for advancing transparent and trustworthy AI in 6G communications systems.

Cellular communication standards have been established to ensure connectivity across most urban environments, complemented by deployment hardware and facilities tailored for city life. At the same time, numerous initiatives seek to broaden connectivity to rural and developing areas. However, with nearly half the global population still offline, there is an urgent need to drive research toward enhancing connectivity in areas and conditions that deviate from the norm. This article delves into innovative communication solutions not only for hard-to-reach and extreme environments but also introduces “hard-to-serve” areas as a crucial, yet underexplored, category within the broader spectrum of connectivity challenges. We explore the latest advancements in communication systems designed for environments subject to extreme temperatures, harsh weather, excessive dust, or even disasters such as fires. Our exploration spans the entire communication stack, covering communications on isolated islands, sparsely populated regions, mountainous terrains, and even underwater and underground settings. We highlight system architectures, hardware, materials, algorithms, and other pivotal technologies that promise to connect these challenging areas. Through case studies, we explore the application of 5G for innovative research, long range (LoRa) for audio messages and emails, LoRa wireless connections, free-space optics, communications in underwater and underground scenarios, delay-tolerant networks, satellite links, and the strategic use of shared spectrum and TV white space (TVWS) to improve mobile connectivity in secluded and remote regions. These studies also touch on prevalent challenges such as power outages, regulatory gaps, technological availability, and human resource constraints, where we introduce the concept of peri-urban hard-to-serve areas where populations might struggle with affordability or lack the skills for traditional connectivity solutions. This article provides an exhaustive summary of our research, showcasing how 6G and future networks will play a crucial role in delivering connectivity to areas that are hard-to-reach, hard-to-serve, or subject to extreme conditions (ECs).

The recent advances in quantum information processing, sensing, and communications are surveyed with the objective of identifying the associated knowledge gaps and formulating a roadmap for their future evolution. Since the operation of quantum systems is prone to the deleterious effects of decoherence, which manifests itself in terms of bit-flips, phase-flips, or both, the pivotal subject of quantum error mitigation is reviewed both in the presence and absence of quantum coding. The state of the art, knowledge gaps, and future evolution of quantum machine learning (QML) are also discussed, followed by a discourse on quantum radar systems and briefly hypothesizing about the feasibility of integrated sensing and communications (ISAC) in the quantum domain (QD). Finally, we conclude with a set of promising future research ideas in the field of ultimately secure quantum communications with the objective of harnessing ideas from the classical communications field.

The recent advances in quantum information processing, sensing, and communications are surveyed with the objective of identifying the associated knowledge gaps and formulating a roadmap for their future evolution. Since the operation of quantum systems is prone to the deleterious effects of decoherence, which manifests itself in terms of bit-flips, phase-flips, or both, the pivotal subject of quantum error mitigation is reviewed both in the presence and absence of quantum coding. The state of the art, knowledge gaps, and future evolution of quantum machine learning (QML) are also discussed, followed by a discourse on quantum radar systems and briefly hypothesizing about the feasibility of integrated sensing and communications (ISAC) in the quantum domain (QD). Finally, we conclude with a set of promising future research ideas in the field of ultimately secure quantum communications with the objective of harnessing ideas from the classical communications field.

Global health challenges related to cardiovascular and pulmonary diseases drive the demand for more effective, non-contact methods to monitor vital signs such as heart rate (HR) and respiration rate (RR). Traditional monitoring techniques, while reliable, often rely on physical contact, limiting their utility in remote or long-term care scenarios. In this context, wireless monitoring systems leveraging radio frequency (RF) technologies, such as radar and Wi-Fi, have emerged as promising solutions for non-invasive and contactless HR and RR tracking. This paper delves into the principles and performance of state-of-the-art RF-based monitoring systems. Significant advances in signal processing techniques and machine learning (ML) models have markedly improved the accuracy and reliability of vital signs detection. These innovations underscore the potential of RF technologies to redefine healthcare monitoring. However, the adoption of RF-based systems is not without challenges. Key issues include optimizing signal processing for multiperson monitoring, mitigating interference, ensuring data security and privacy, and promoting open access to datasets. In addition, integrating edge computing and advanced ML models to enhance system intelligence and accuracy remains a critical area of research. This review provides a comprehensive understanding of the operational mechanisms underlying RF-based monitoring systems and proposes improvements to overcome current barriers, thereby advancing the field of non-invasive vital signs monitoring and broadening its application in healthcare.

Human activity recognition (HAR) via Wi-Fi sensing has emerged as a pivotal technology for smart environments, offering device-free, privacy-preserving monitoring. However, existing methods often face limitations in feature representation efficiency and generalization under constrained hardware setups. In this work, we propose continuous angle-of-arrival and time-of-flight maps (CATM), a novel and easily learnable feature extraction framework that jointly encodes spatial and temporal dynamics of human activities using commercial Wi-Fi devices. By integrating smoothed channel state information (CSI) with MUSIC-based signal processing, CATM constructs 2-D heatmaps that unify angle-of-arrival (AoA) and time-of-flight (ToF) features, enabling robust representation of both coarse and fine-grained movements. Unlike conventional approaches relying on fragmented temporal or spectral features, CATM inherently embeds continuous spatiotemporal patterns that simplify downstream model learning. Our lightweight Res-BiLSTM network, trained on CATM features, achieves 93.2% accuracy across eight activities and five users, under three displacement ways, and outperforms other single-domain methods. Crucially, CATM demonstrates exceptional transferability: when adapted to unseen device placements, the framework retains 80% accuracy with only 20% retraining data, significantly reducing dependence on extensive labeled datasets. These results demonstrate that CATM provides a more informative feature extraction method compared to traditional approaches relying on CSI amplitude and Doppler information for HAR.

Open Radio Access Network (O-RAN) introduces xApps to enable flexible, near real-time control of RAN functions. However, managing multiple xApps dynamically, under rapidly changing network conditions, becomes significantly more challenging in IoT-driven environments, where massive numbers of low-power devices may simultaneously access the network, triggering phenomena such as wake-up storms. In this work, we showcase and compare Digital Twin-enhanced Multi-Agent Reinforcement Learning (MARL) strategies for intelligent xApp management in IoT-enabled O-RAN systems. We explore how Digital Twins can simulate large-scale, event-driven IoT device behaviors, enabling proactive optimization of xApp scheduling while reducing communication and synchronization overhead. We further examine advanced MARL approaches, such as Hierarchical Agent Policy Optimization (HAPPO), to address scalability and coordination challenges under ultra-dense, low-latency, and energy-constrained LPWAN protocols like NB-IoT and LTE-M. We define two system-level KPIs to evaluate the performance of RL schemes: Conflict Avoidance Efficiency and Coordination QoS Improvement. The proposed MARL-based approach significantly outperforms conventional single-agent RL methods in both coordination effectiveness and overall QoS delivery, while maintaining a reasonable model complexity suitable for practical deployment.

The open nature of wireless communications renders unmanned aerial vehicle (UAV) communications vulnerable to impersonation attacks, under which malicious UAVs can impersonate authorized ones with stolen digital certificates. Traditional fingerprint-based UAV authentication approaches rely on a single modality of sensory data gathered from a single layer of the network model, resulting in unreliable authentication experiences, particularly when UAVs are mobile and in an open-world environment. To transcend these limitations, this paper proposes SecureLink, a UAV authentication system that is among the first to employ cross-layer information for enhancing the efficiency and reliability of UAV authentication. Instead of using single modalities, SecureLink fuses physical-layer radio frequency (RF) fingerprints and application-layer micro-electromechanical system (MEMS) fingerprints into reliable UAV identifiers via multimodal fusion. SecureLink first aligns fingerprints from channel state information measurements and telemetry data, such as feedback readings of onboard accelerometers, gyroscopes, and barometers. Then, an attention-based neural network is devised for in-depth feature fusion. Next, the fused features are trained by a multi-similarity loss and fed into a one-class support vector machine for open-world authentication. We extensively implement our SecureLink using three different types of UAVs and evaluate it in different environments. With only six additional data frames, SecureLink achieves a closed-world accuracy of 98.61% and an open-world accuracy of 97.54% with two impersonating UAVs, outperforming the existing approaches in authentication robustness and communication overheads. Finally, our datasets collected from these experiments are available on GitHub: this https URL\_data.

Integrated sensing and communications (ISAC) is essential for the sixth-generation (6 G) wireless systems. In contrast to most existing works that assume either the far-field or near-field model for both communications and sensing, this study addresses a new scenario where multiple targets and communication users exist in a mixed near- and far-field scenario. Specifically, two optimization problems are formulated for the sensing-centric (S-C) and communication-centric (C-C) scenarios. In the S-C scenario, the minimum signal-to-clutter-plus-noise ratio (SCNR) of each target is maximized by designing the transmit beamforming at the dual-functional base station (BS), subject to the signal-to-interference-plus-noise ratio (SINR) requirement at each CU and the transmit power constraint. In the C-C scenario, the minimum SINR at each CU is maximized, under the constraints of the SCNR requirement for each target and the transmit power. The two problems are solved using the Dinkelbach-based successive convex approximation (SCA) method. Numerical results highlight the tradeoff between mixed-field communications and sensing, as well as the performance difference in S-C and C-C scenarios due to the mixed field. They also examine how antenna size and transmit power affect sensing performance in mixed-field ISAC systems. In addition, our proposed design outperforms the benchmark in cases where the fairness profile optimization (FPO) is applied to both communications and sensing.

We present a deep learning (DL) approach for identifying random body movements (RBM) to enhance radio frequency (RF) sensing applications. The proposed method leverages a capsule neural network architecture to automate feature extraction, eliminating the need for manual feature engineering. This approach demonstrates robust performance by achieving an average RBM detection accuracy of 92% across diverse environments. The method enhances the accuracy and reliability of RF-based systems by mitigating RBM-induced interference, making it highly valuable for wireless sensing applications such as vital signs monitoring, facial recognition, and gesture detection.

Wavenumber-division multiplexing (WDM) was introduced as a counterpart of orthogonal frequency-division multiplexing in the spatial-frequency domain for line-of-sight holographic multiple-input multiple-output (MIMO) systems. In this paper, we extend WDM to holographic MIMO channels with non-line-of-sight (NLoS) propagation. We show that applying WDM to the NLoS channel yields the corresponding angular-domain representation, which we characterize through the power spectral factor and power spectral density. We further obtain a closed-form characterization for the case of isotropic scattering, recovering Jakes’ isotropic model. The analysis is complemented by numerical results evaluating the degrees of freedom and ergodic capacity under both isotropic and non-isotropic scattering.

Holographic multiple-input multiple-output (MIMO) enables electrically large continuous apertures, overcoming the physical scaling limits of conventional MIMO architectures with half-wavelength spacing. Their near-field operating regime requires channel models that jointly capture line-of-sight (LoS) and non-line-of-sight (NLoS) components in a physically consistent manner. Existing studies typically treat these components separately or rely on environment-specific multipath models. In this work, we develop a unified LoS+NLoS channel representation for holographic lines that integrates spatial-sampling-based and expansion-based formulations. Building on this model, we extend the wavenumber-division multiplexing (WDM) framework, originally introduced for purely LoS channels, to the LoS+NLoS scenario. Applying WDM to the NLoS component yields its angular-domain representation, enabling direct characterization through the power spectral factor and power spectral density. We further derive closed-form characterizations for isotropic and non-isotropic scattering, with the former recovering Jakes’ isotropic model. Lastly, we evaluate the resulting degrees of freedom and ergodic capacity, showing that incorporating the NLoS component substantially improves the performance relative to the purely LoS case.

A novel Hilbert-curve (HC) based microwave nondestructive testing (NDT) array sensor is presented for carbon fiber-reinforced polymer (CFRP) composites. The proposed sensor is designed on a cost-effective Rogers RT5880 by using second order Hilbert-curve. The transmission coefficient (S21) response and electric field distribution are examined. The result works as a bandstop filter when cracks are present and bandpass filter when cracks are absent within the sensing band for the sample-under-test (SUT). The HC-based unit sensor uses a fringing field sensing capacitor (FFSC) as a “sensing pad” and interacts directly with the SUT to detect surface cracks. The designed sensor can detect the minimum crack size of 0.5 mm while operating in the frequency range of 0.25−2.75 GHz. The designed sensors have potential applications for detecting surface cracks in valuable infrastructure including aircraft, bridges, pipelines and strengthening their safety and integrity.

This paper presents TelcoGPT, a specialised question-answering (Q&A) and code retrieval system for telecommunications that combines retrieval-augmented generation (RAG) with domain-specific optimizations. TelcoGPT introduces three key enhancements: (1) a HybridEmbedding method-ology integrating text-embedding models with telecom-specific filtering mechanisms; (2) an advanced document processing pipeline with adaptive chunking and technical term density scoring; and (3) a dual-path query engine optimized for both question-answering and code retrieval tasks. Evaluation on the RedPajama-Data-1T arxiv subset demonstrates that hybrid embedding approach achieves mean reciprocal rank (MRR) of 0.89 and hit rate (HR) of 0.94 with optimal configuration (thresh-old=0.8, chunk size=12K, k=15), outperforming single-embedding approaches by 5-7%. The hybrid RAG implementation increases MRR by 8.5% (0.82 to 0.89) and HR by 10.6% (0.85 to 0.94). TelcoGPT achieves 95% accuracy in domain-specific Q&A tasks versus 87% for base models, while maintaining higher technical term density scores (0.90 vs 0.81). For code retrieval, our system demonstrates 93% execution success rate with comprehensive error handling, surpassing baseline approaches by 6-8%. Comparative analysis with GPT-3.5, GPT-4, and LLAMA-2/3 shows significant improvements in context relevance (0.92 vs 0.84), information accuracy (0.95 vs 0.89), faithfulness (0.84 to 0.92), and relevancy (0.83 to 0.93), demonstrating the effectiveness of our architecture for telecommunications applications.

A scalable reconfigurable intelligent surface (RIS) architecture with both centralized and distributed control capability is presented. The RIS is implemented using a one-bit phase reconfigurable unit cell (UC) designed for X-band operation. The UC is realized by embedding a PIN diode into a slotted copper patch, achieving a maximum phase shift at 9.5 GHz with a phase difference of across the 9–10 GHz band. The proposed RIS enables real-time reconfiguration, with each element controlled through an FPGA achieving an update time below 0.1 ms. In addition, multiple control interfaces are supported, including LAN, USB, and WiFi. The measurement results of a single RIS tile prototype validate beam steering from to and gain performance, demonstrating a peak gain of 20.2 dBi at broadside and 18.2 dBi at . To demonstrate scalability, a four-tile prototype of (each tile consisting of UCs) is fabricated. Each tile can be independently programmed via its dedicated interface, or alternatively, a single interface module (WiFi, LAN, or USB) can distribute ON/OFF configurations across all tiles, thereby enabling flexible and scalable control of the RIS.

In this article, we propose an event-triggered H∞ filtering algorithm that uses hysteresis-quantized measurements to achieve accurate state estimation in networked mass-switching autonomous marine vehicles (AMVs). The method tackles several key challenges commonly seen in networked AMV filtering, including limited communication bandwidth, restricted energy resources, and signal transmission delays.

We begin by establishing a dynamic model that captures the parameter variations caused by mass changes in AMVs. To mitigate bandwidth constraints and minimise signal chattering, we introduce a hysteresis-based quantization scheme. We further design an event-triggered mechanism that is both energy-efficient and capable of preventing excessively long intervals without triggering, thereby reducing the frequency of data transmission.

To enable rigorous stability analysis, a new augmented Lyapunov–Krasovskii functional is constructed, and Wirtinger-based inequalities are applied to manage the time-delay-dependent integral terms. Building on this foundation, we develop a co-design strategy that jointly determines both the filter and the event-triggering mechanism.

Simulation results show that the proposed method provides accurate state estimation for networked mass-switching AMVs while reducing the transmission rate by 75%, and it significantly outperforms conventional logarithmic quantizers in suppressing chattering.

Human Activity Recognition (HAR) plays a crucial role in intelligent healthcare, smart environments, and elderly monitoring. Traditional deep learning-based HAR methods often function as black-box models, limiting their interpretability. The recently proposed Kolmogorov-Arnold Network (KAN) utilizes explicit, mathematically defined basis functions, which clarify its operation and enhance interpretability. However, these methods still face challenges, such as slow training speed, high computational costs and suboptimal performance. Here we propose the Daily Activity Recognition with Optimized Wavelet-based KAN (DARKAN), a lightweight architecture that leverages wavelet decomposition to boost performance, and simplifies KAN structure to lower model parameters and computational complexity. Specifically, low- and high-frequency inertial measurement unit (IMU) signals are extracted by a wavelet transform, while time-domain features are incorporated to enrich feature representation. Subsequently, the B-Spline is replaced by the wavelet function as the activation function in KAN (wav-KAN), and the network depth of wav-KAN is reduced to two layers. Finally, the optimized wav-KAN is utilized to classify daily activities by fusing the extracted time-frequency features. Extensive experiments on three open-source datasets demonstrate that DARKAN outperforms stat-eof-the-art methods, achieving 98.82%, 97.11% and 98.57% in classification accuracy respectively while reducing the number of model parameters by 1.45× and FLOPs by 4×.

The open nature of wireless communications renders unmanned aerial vehicle (UAV) communications vulnerable to impersonation attacks, under which malicious UAVs can impersonate authorized ones with stolen digital certificates. Traditional fingerprint-based UAV authentication approaches rely on a single modality of sensory data gathered from a single layer of the network model, resulting in unreliable authentication experiences, particularly when UAVs are mobile and in an openworld environment. To transcend these limitations, this paper proposes SecureLink, a UAV authentication system that is among the first to employ cross-layer information for enhancing the efficiency and reliability of UAV authentication. Instead of using single modalities, SecureLink fuses physical-layer radio frequency (RF) fingerprints and application-layer micro-electromechanical system (MEMS) fingerprints into reliable UAV identifiers via multimodal fusion. SecureLink first aligns fingerprints from channel state information measurements and telemetry data, such as feedback readings of onboard accelerometers, gyroscopes, and barometers. Then, an attention-based neural network is devised for in-depth feature fusion. Next, the fused features are trained by a multi-similarity loss and fed into a one-class support vector machine for open-world authentication. We extensively implement our SecureLink using three different types of UAVs and evaluate it in different environments. With only six additional data frames, SecureLink achieves a closed-world accuracy of 98.61% and an open-world accuracy of 97.54% with two impersonating UAVs, outperforming the existing approaches in authentication robustness and communication overheads. Finally, our datasets collected from these experiments are available on GitHub: https://github.com/PhyGroup/SecureLink_data.

Graph Neural Networks (GNNs) have demonstrated remarkable capabilities in handling graph data. Typically, GNNs recursively aggregate node information, including node features and local topological information, through a message-passing scheme. However, most existing GNNs are highly sensitive to neighborhood aggregation, and irrelevant information in the graph topology can lead to inefficient or even invalid node embeddings. To overcome these challenges, we propose a novel Space Gravity-based Graph Neural Network (Gravity-GNN) guided by Deep Reinforcement Learning (DRL). In particular, we introduce a novel similarity measure called ”node gravity”, inspired by the gravitational force between particles in space, to compare nodes within graph data. Furthermore, we employ DRL technology to learn and select the most suitable number of adjacent nodes for each node. Our experimental results on various real-world datasets demonstrate that Gravity-GNN outperforms state-of-the-art methods regarding node classification accuracy, while exhibiting greater robustness against disturbances.

In Vehicular Edge Computing (VEC), vehicles of fload computational tasks to Roadside Units (RSUs) equipped with edge servers to achieve efficient processing. Considering that vehicles switch connections between RSUs during highspeed movement, obtaining the state information of other RSUs is crucial for achieving global collaborative decision-making. However, frequent sharing of RSUs’ state data during the training of scheduling models may result in privacy leakage risks. To address this issue, we federally train a joint scheduling model for task offloading and resource allocation without the need for state sharing among RSUs. We prove that the proposed task offloading problem influenced by resource allocation is a strict multi-node non-cooperative potential game problem, and use the potential function as the reward function for MultiAgent Deep Deterministic Policy Gradient (MADDPG). Finally, we propose the Fed-MADDPG algorithm to find the equilibrium point of task offloading and apply the gradient descent method and the Lagrange multiplier method to maximize the average task completion rate among RSUs under constraints, ensuring the framework has optimal computational and transmission performance. We conduct simulation experiments using realworld datasets, and the results show that this method has superior performance compared to previous approaches.

In an era of rapidly increasing bandwidth demand and revolutionary wireless technologies, efficient spectrum management is essential. This paper proposes a novel multi-tier tokenisation approach for dynamic spectrum management. Using the concept of heterogeneous tokenisation of spectrum bands, we develop a decentralised blockchain-based framework that enables spectrum sharing among users.

The spectrum space is represented using multi-planes. The first plane consists of unique spectrum bands, which are converted into NFTs for long-term allocations. The second plane subdivides these NFT spectrum bands into smaller units for short-term usage by retail users through fungible tokens. These fungible tokens are dynamically traded and mapped using particle swarm optimisation (PSO) to manage demand and supply.

The paper presents formal models of the system entities, along with algorithms for multi-tier token creation, dynamic token trading, and demand–supply mapping using PSO. To enhance privacy, a zero-knowledge proof (ZKP)-based approach is employed for user authentication.

The proposed framework provides a secure, transparent, and scalable solution for spectrum management, addressing the limitations of traditional centralised methods. Simulation results demonstrate the effectiveness of the framework in supporting dynamic spectrum access while ensuring privacy and scalability.

6G applications rely on data-intensive AI models for network optimization. These demand a scalable and energyefficient framework to handle massive device networks with stringent latency requirements which current solutions struggle to support. Although reinforcement learning (RL) and split learning have matured to provide commercial solutions elsewhere. Current solutions in 6G have not used them systematically to achieve the sustainability goals. In this paper, we propose a three-layer framework that minimizes energy consumption of the communication system capable of handling large number of devices. The proposed solution uses RL agents at the edge layer to mathematically model the system and communicate to fog layer for aggregation. The aggregated feature maps are further communicated to cloud layer for global model training. We use split learning for communication and training, the learning at each device are communicated for global model creation effectively. Each edge device improves the overall RL model where system matures quickly consuming minimal energy. The proposed framework’s efficacy has been tested extensively for accuracy and scalability, in terms of energy consumption, latency and memory utilizations. The simulation results validate the claims of maturity in models across edge, fog and cloud levels.

Sensor-enabled distributed energy resources (DERs) provide various advantages, including a lower carbon footprint, yet effective management of millions of DERs is still an issue. Virtual power plants (VPP) can integrate several DERs into a unified operational digital twin to enable real-time monitoring, analysis, and control. VPP may utilize advanced solutions to improve operational efficiency by combining substantial measurement data from DERs. However, effectively managing the quantity and complexity of data flows, whether streaming data or high-impact low-frequency data, is essential in maintaining the performance of DERs at sustained levels. The vast amounts of diverse data generated from various DERs pose significant challenges for storage, processing, and resource management. This article proposes a comprehensive framework that employs a distributed big data cluster to ensure scalable and reliable data storage and utilizes a robust message broker system for efficient data queuing. In addition, we present innovative load-balancing strategies within the VPP digital twin system. A decision tree algorithm is implemented to calculate the forthcoming workload collected by various deployed sensors at various DERs. The required resources are identified per workload, and the numbers are forwarded to the orchestrator. The orchestrator scales up and down resources based on resource utilization suggested by the decision tree algorithm when the resources or nodes are insufficient to handle the sensor data, optimizing the utilization of computing resources. The framework also features a failure detection component that performs root cause analysis to provide actionable insights for system optimization. Experimental results show that this framework ensures high efficiency, reliability, and real-time operational capability in VPP digital twin by addressing critical challenges in data storage, streaming data analysis, and load balancing.

As the scale of data continues to grow exponentially, managing resource allocation and energy consumption in big data systems becomes increasingly complex and critical. Moreover, with big data systems, energy efficiency is more important daily. In cloud environments, it can be the determining factor between reduced costs and lowered environmental damage. This paper presents a deep learning-based framework for accurately predicting instant energy consumption in real-time and detecting anomalies of different sizes in big data clusters. We use SmartMonit to gather task execution and real-time infrastructure data. A Feedforward Neural Network (FNN) predicts energy consumption from CPU utilisation, memory usage, and task profiling research. The system will track any deviation from predicted consumption with root cause analysis (RCA) if there are significant anomalies. We also integrate an Autoencoder to identify straggler tasks and inefficient resource utilisation. Userdefined functions are next applied to examine these anomalies and try to detect the underlying reasons, like distributed data processing, locality of computation exploitation, or resource waste. Given the scale and heterogeneity of big data workloads, the system’s ability to dynamically adjust and optimise resource usage is essential for handling complex processing tasks. The experimental results prove that the proposed system effectively enhances resource allocation and decreases wasted energy.

With the widespread adoption of healthcare Internet of Things (IoT) devices, the need for effective intrusion and anomaly detection has become pivotal in ensuring network security. However, the optimization of machine learning (ML) and deep learning (DL) models for these detection tasks frequently necessitates extensive computational resources, adversely affecting both temporal and energy efficiency. This article introduces an automated ML (AutoML) framework specifically tailored to enhance anomaly detection models, with a strategic focus on energy efficiency throughout the optimization process. The process commences with data preprocessing, followed by feature selection employing a combination of recursive feature elimination (RFE) and SHapley Additive exPlanations (SHAP) to identify important features for anomaly detection. Subsequently, a baseline multilayer perceptron (MLP) neural network model is trained, and hyperparameter optimization is executed in a constrained search space to mitigate energy consumption. The framework produces optimized models, which are assessed based on accuracy and energy consumption at various checkpoints, with the models demonstrating inferior performance systematically excluded based on predefined accuracy or energy consumption objectives. Experimental outcomes reveal that the pipeline effectively balances detection performance with energy consumption, with certain cases showing minimal accuracy losses (less than 1%) accompanied by substantial energy savings (over 60%), presenting a sustainable and resource-efficient approach to anomaly detection in IoT systems.

Federated learning is a promising approach that enables collaborative machine learning (ML) in distributed environments, such as the Internet of Medical Things (IoMT) while preserving consumer privacy. It allows multiple consumers to collaboratively train a model using their own data, sharing only the locally trained model rather than the raw data. Most existing federated learning systems assume a high level of trust in participating nodes, which is unrealistic in real-world consumer-centric scenarios. Involving untrusted nodes can compromise the integrity of the training process and result in potential data breaches. To address these challenges, this paper presents REWARDCHAIN, a novel federated learning framework that leverages blockchain technology to ensure trust and accountability among untrusted IoMT consumers. By recording all model updates and client contributions on an immutable blockchain ledger, REWARDCHAIN allows auditing of the entire training process and attributing any malicious behaviour to specific nodes. Moreover, we design an incentive mechanism that evaluates contributions based on data quality and participant reputation. This system motivates participants to contribute high-quality data through a reputation-constrained reward allocation. Our evaluations show that REWARDCHAIN effectively balances trust, security, and model performance, facilitating a more secure and effective federated learning ecosystem.

In network security, the issue of imbalanced data, where certain classes are under-represented, poses a significant challenge for intrusion detection systems (IDS). Machine learning algorithms often prioritise the majority class, resulting in poor detection of minority classes and reduced overall model accuracy. Accurately identifying these rare intrusions is essential for predicting and mitigating emerging threats. This paper presents an advanced IDS designed for Open Radio Access Networks (O-RAN) to address class imbalance by employing Synthetic Minority Over-sampling Technique (SMOTE) variants and Generative Adversarial Networks (GANs). The system was evaluated on the CICEVSE2024 dataset and a real-world O-RAN dataset. Results show that SMOTE improved the accuracy of weaker classifiers such as Naive Bayes by up to 10% (from 53.9% to 64.0%), with corresponding increases in recall and F1-score. Ensemble methods like Random Forest and XGBoost maintained high accuracy ( 81%–89%) and benefited from balanced recall when synthetic data was applied. However, GAN-generated data (CTGAN) showed little to no improvement over baseline models, and in some cases, SMOTE reduced accuracy for classifiers such as Logistic Regression and Random Forest on the O-RAN dataset (by up to 30%). These results highlight that while SMOTE variants can significantly enhance minority class detection, especially for weaker classifiers, the utility of GANs remains limited in this context. Future work should therefore focus on improving GAN-based data quality, exploring hybrid deep learning approaches, and extending IDS to real-time and multi-class scenarios.

Formal verification is crucial for ensuring the robustness of security protocols against adversarial attacks. The Needham-Schroeder protocol, a foundational authentication mechanism, has been extensively studied, including its integration with Physical Layer Security (PLS) techniques such as watermarking and jamming. Recent research has used ProVerif to verify these mechanisms in terms of secrecy. However, the ProVerif-based approach limits the ability to improve understanding of security beyond verification results. To overcome these limitations, we re-model the same protocol using an Isabelle formalism that generates sound animation, enabling interactive and automated formal verification of security protocols. Our modelling and verification framework is generic and highly configurable, supporting both cryptography and PLS. For the same protocol, we have conducted a comprehensive analysis (secrecy and authenticity in four different eavesdropper locations under both passive and active attacks) using our new web interface. Our findings not only successfully reproduce and reinforce previous results on secrecy but also reveal an uncommon but expected outcome: authenticity is preserved across all examined scenarios, even in cases where secrecy is compromised. We have proposed a PLS-based Diffie-Hellman protocol that integrates watermarking and jamming, and our analysis shows that it is secure for deriving a session key with required authentication. These highlight the advantages of our novel approach, demonstrating its robustness in formally verifying security properties beyond conventional methods.

Recently proposed over-the-air (OTA) controllers enable spectrum-efficient data fusion and control of networked dynamical systems such as swarm of drones and smart-grids. By properly tuning wireless channels, control signals can be formed via electromagnetic superposition of nodes’ states transmitted on the same frequency, and received at the controller. However, the challenge of current OTA controllers lies in the insufficient optimization variables to compensate wireless channels for OTA control objectives. In this work, we leverage the massive channel tuning elements of a reconfigurable intelligent surface (RIS) to design a RIS optimized OTA controller for networked dynamical systems. We first formulate the RIS-assisted OTA control objective function and a non-convex minimax eigenvalue optimization problem. Then, an quadratic upper-bound of the maximum eigenvalue is found to approximate the original optimization problem into a convex form. A smart-grid case-study is explored to validate our proposed RIS-OTA controller. Simulation results illustrate better control performance of our proposed RIS-OTA controller than existing transmitter power allocation based OTA controller, showing the promising potentials of RIS-optimized OTA controller in practical networked dynamical systems.

Metasurfaces, composed of subwavelength-scale meta-atoms, offer extensive capabilities in electromagnetic (EM) wave manipulation. In this study, we present a dual-functional transmissive chiral metasurface (TCMS) designed to achieve dual-band circular dichroism (CD) at 12.62 GHz and 16.305 GHz within the Ku-band, with CD values of 0.5 and 0.4, respectively, alongside asymmetric transmission (AT). These functionalities are realized simultaneously for right-hand circularly polarized (RHCP) and left-hand circularly polarized (LHCP) incident waves. Simulation results show that the dual-band CD effect is driven by the structural asymmetry of the chiral metasurface. Compared to single-function metasurfaces, this TCMS offers enhanced performance and dual-functionality while maintaining compactness. We believe that this dual-band, dual-functional TCMS, with its simplified geometry, holds significant promise for applications in analytical chemistry, imaging, sensing, spectroscopy, satellite communication, and the development of advanced communication components that enable efficient polarization manipulation.

Circular dichroism (CD) in electromagnetic (EM) waves offers significant potential for advancing various scientific and technological fields. This study presents a unique approach using a C-shaped transmissive chiral metasurface, engineered to achieve strong CD within the X-band frequency range. The metasurface demonstrates outstanding selectivity for left-hand circularly polarized (LHCP) and right-hand circularly polarized (RHCP) incident waves, featuring enhanced bandwidth and a strong CD of 68.3%. With a thickness of 0.8 mm and built on an FR-4 substrate, the metasurface is optimized for high selectivity toward LHCP and RHCP waves, representing a significant advancement in the study of transmissive metasurfaces. By seamlessly combining theoretical knowledge with practical application, this research not only extends the boundaries of metasurface design but also paves the way for diverse applications across various scientific disciplines, including satellite communication, stealth technology, and antenna design, etc.

This paper presents a novel dual-mode chiral metasurface (CM) designed to achieve strong circular dichroism (CD) in both transmission and reflection mode within the Ku-band. The proposed dual-mode CM demonstrates CD i.e., an efficient conversion of linearly polarized (LP) electromagnetic (EM) waves into circularly polarized (CP) waves, both within a broader spectrum as well as at single frequencies in both transmission and reflection mode, exhibiting asymmetric transmission (AT) response. This is achieved through a judiciously designed unit cell structure, which eliminates the requirement for intricate supercell configurations or active circuitry. The metasurface comprises a circular ring structure embedded with intelligently placed angle-induced slot and a metallic strip, fabricated using the cost-effective FR-4 substrate. The structure on the front side of the dual-mode CM is replicated on the back side of the substrate with a 90° rotation to achieve a chiral configuration. Under the forward-propagating y-polarized incident wave, the dual-mode CM demonstrates the capability to convert LP wave into right-handed circularly polarized (RHCP) wave at 14.064 GHz. Additionally, in transmission mode, it converts LP waves to left-handed circularly polarized (LHCP) wave over a wider frequency range of 16.60 – 17.03 GHz with AT response. In reflection mode, the dual-mode CM converts LP wave into RHCP wave at 12.048 GHz when subject to x-polarized incident wave propagating in the backward direction.

This paper discusses a miniaturized multi-wideband tablet antenna having a compact size of 20 × 9.6 × 1.8 mm³. The antenna is modeled on FR-4 substrate with thickness of 0.8 mm. The antenna covers LTE 700/2300/2500/3500, GPS 1575.42 MHz, GSM 850/1800/1900, UMTS 2100, Bluetooth 2.4 GHz and 5.2/5.8 GHz WLAN bands. The wideband operation is achieved by a novel inverted-F type structure which is capacitively coupled with two λ/4 microstrip resonators. Impedance matching at lower bands is further improved by a high-pass matching circuit integrated with the antenna feed. The proposed antenna has reasonable gain over all covered bands, low specific absorption rate (SAR) and high radiation efficiency from 80-90% over all LTE and WLAN bands.

An optically transparent and flexible metasurface is proposed for the V-band applications. The metasurface uses a silver nanowire (AgNW) deposit on a transparent PEN substrate to achieve optical transparency of greater than 85 %. The proposed metasurface exhibits cross polarization conversion in reflection mode from 54.2 GHz to 70.7 GHz with an angularly stable field of view of up to 40°. The metasurface is also capable of dual-band linear-to-circular polarization conversion from 50 GHz to 52.7 GHz and 73.8 GHz to 80.8 GHz. Moreover, a L-blt phase resolution in reflection mode is achieved by 90° rotation of unit cell for cross polarization reflectarray applications. The proposed metasurface is simulated on CST Microwave Studio and features low insertion loss, high return loss, and a wideband polarization conversion ratio (PCR) of better than 90 %, making it a desirable candidate for V-band cross polarization applications.

In this study, a 1-bit phase reconfigurable unit cell (UC) is proposed for an X-band intelligent reflective surface (IRS). The UC is designed by integrating a PIN diode into a slotted copper patch. The maximum phase shift occurs at the design frequency of 9.5 GHz and the phase difference varies within 180∘±20∘ from 9 to 10 GHz in the X-band. A 16 ×16 IRS is designed and simulated using the proposed UC to verify gain and beam steering performance. The IRS has a gain of 20.5dBi at θ=0∘. At θ=±40∘ the gain is 18.6 dBi which validates good beam steering performance.

Circular dichroism (CD) in electromagnetic (EM) waves offers significant potential for advancing various scientific and technological fields. This study presents a unique approach using a C-shaped transmissive chiral metasurface, engineered to achieve strong CD within the X-band frequency range. The metasurface demonstrates outstanding selectivity for left-hand circularly polarized (LHCP) and right-hand circularly polarized (RHCP) incident waves, featuring enhanced bandwidth and a strong CD of 68.3%. With a thickness of 0.8 mm and built on an FR-4 substrate, the metasurface is optimized for high selectivity toward LHCP and RHCP waves, representing a significant advancement in the study of transmissive metasurfaces. By seamlessly combining theoretical knowledge with practical application, this research not only extends the boundaries of metasurface design but also paves the way for diverse applications across various scientific disciplines, including satellite communication, stealth technology, and antenna design, etc.

This paper presents a broadband circularly polar-ized (CP) antenna array for millimeter-wave (mmWave) applications. The overall size of the 1×4 antenna array is 28×23mm2 and covers the 28 GHz to 38 GHz bands. Each element of the antenna array consists of a wide square slot (WSS) with a T-shaped stub extended in the ground plane. The impedance matching and the CP bandwidth is improved by introducing asymmetric fractals in the ground plane. The antenna array elements are fed by a simple 4-way power divider network. The simulated and measured impedance bandwidth of the array is 49% (26.6 GHz to 42.9 GHz) and 49% (26.6 GHz to 42.4 GHz) respectively. The simulated and measured 3dB axial ratio (AR) bandwidth is 29.5% (34 GHz to 44 GHz) and 28.5% (35 GHz to 42GHz) respectively. The peak gain of 8.5-11.2 dBi is measured throughout the bandwidth. The proposed antenna array is suitable for commercial use in future 5G communication systems.

This paper presents the RF performance analysis of various designs of RF – MEMS shunt capacitive switches in coplanar configuration. The switches consist of slotted bridges of a variety of shapes such as circular, hexagonal, and rectangular. The design consisting of the rectangular slotted bridge exhibits the best performance with a maximum insertion loss of 0.08 dB, return loss of greater than 20 dB and isolation of more than 30 dB. The design has an operational frequency range from DC-40 GHz. The designs are modelled and simulated using High Frequency Structure Simulator (HFSS).

Human activity recognition (HAR) in indoor environments is essential for healthcare, elder care, and assisted living applications, especially in complex scenarios involving non-line-of-sight (NLoS) and long-range conditions. Traditional single-sensor HAR systems often struggle with accuracy and reliability in such environments. This study introduces the RFiDAR system, a novel fusion approach that combines radio frequency identification (RFID) and Radar technologies with an LSTM-variational autoencoder (LSTM-VAE) model to enhance HAR accuracy and reliability. The RFiDAR system applied data-level, feature-level, and decision-level fusion techniques to integrate temporal patterns from RFID and Radar data, facilitating the recognition of five distinct activities at varying distances. Results indicate that fusion methods, particularly feature-level fusion, significantly improve classification accuracy. For instance, feature-level fusion achieves up to 98.8% accuracy at 2 meters and 97.9% at 3 meters, outperforming single-sensor models by 5.3% and 6.2%, respectively. The RFiDAR system demonstrates superior performance in complex scenarios, offering reliable and cost-effective solutions for autonomous, long-term monitoring. The proposed approach has potential applications in healthcare and accessible IoT environments and demonstrates the innovative impact of multi-sensor fusion in developing safe, flexible, and inclusive technology.

This paper presents RadSpecFusion, a novel dynamic attention-based fusion architecture for multi-radar human activity recognition (HAR). Our method learns activity-specific importance weights for each radar modality (24 GHz, 77 GHz, and Xethru sensors). Unlike existing concatenation or averaging approaches, our method dynamically adapts radar contributions based on motion characteristics. This addresses cross-frequency generalization challenges, where transfer learning methods achieve only 11%–34% accuracy. Using the CI4R dataset with spectrograms from 11 activities, our approach achieves 99.21% accuracy, representing a 15.8% improvement over existing fusion methods (83.4%). This demonstrates that different radar frequencies capture complementary information about human motion. Ablation studies show that while the three-radar system optimizes performance, dual-radar combinations achieve comparable accuracy (24GHz+77GHz: 96.1%, 24GHz+Xethru: 95.8%, 77GHz+Xethru: 97.2%), enabling flexible deployment for resource-constrained applications. The attention mechanism reveals interpretable patterns: 77 GHz radar receives higher weights for fine movements (superior Doppler resolution), while 24 GHz dominates gross body movements (better range resolution). The system maintains 71.4% accuracy at 10 dB SNR, demonstrating environmental robustness. This research establishes a new paradigm for multimodal radar fusion, moving from cross-frequency transfer learning to adaptive fusion with implications for healthcare monitoring, smart environments, and security applications.

This paper addresses the problem of uniform random generation of vectors of values with a fixed sum, subject to upper and lower constraints on the individual component values. Solutions to this problem are used extensively in the generation of tasksets, specifically task utilization values, in support of the performance assessment of schedulability tests for real-time systems. This paper introduces a general-purpose solution in the form of an Inverse Volume Ratio Sampling method that is applicable provided that it is possible to determine the ratio of the volume below a given hyperplane to the total volume of the valid region in n-dimensional space, as demarcated by the constraints and the fixed sum. An efficient approach is derived for volume calculation using numerical convolution, thus instantiating the ConvolutionalFixedSum algorithm, which provides a user-specified level of precision, while scaling at O(n^3log(n)). A stringent uniformity test is developed, called the slices test, which is able to fully explore the extent of the valid region in each of the n dimensions. The slices test reveals that while the outputs of UUnifast and ConvolutionalFixedSum form uniform distributions, in some cases the outputs of prior state-of-the-art algorithms do not.

This paper provides a comprehensive understanding of data and information flow in Autonomous Vehicles (AVs) with a particular focus on authorisation and validation challenges across different data processing stages. It examines the stage-specific key challenges and reviews existing solutions based on the data flow stages: collection, transmission, processing, actuation and storage. The findings highlight critical gaps in current approaches and suggest future research directions to enhance AV data authorisation and validation.

This work develops a novel ontology-based framework for achieving personal data sovereignty, overcomes data migration challenges through a federation-based sovereignty approach, and provides more fine-grained control over personal data than current solutions. The implementation demonstrates the feasibility of personal data sovereignty in practice and validates practical deployment without extensive infrastructure modifications. 

This paper proposes a novel integration of noise modulation with an autonomous RIS, enabling zero-energy transmission by exploiting thermal noise for information encoding and interference for powering the RIS. A resource allocation strategy optimizes both energy harvesting and communication performance. Analytical and simulation results demonstrate its strong potential for sustainable 6G IoT.

In this letter, we experimentally present a customized hardware framework to implement contactless millimeter-wave (mmWave) vital sign sensing in the 60 GHz license-free industrial, scientific, and medical (ISM) band using Dynamic Metasurface Antenna (DMA). The proof-of-concept experiment highlights how mmWave RF sensing challenges can be addressed through the pattern diversity of a DMA-assisted setup, eliminating the need for commercial Wi-Fi routers or reflecting surfaces. The measured results demonstrate that mmWave signals can be precisely steered toward the human body, and analysis of the reflected signals enables accurate estimation of breathing and heart rates. The reflected signal strength exceeds ambient noise by over 30 dB, supporting reliable vital sign sensing over the mmWave channel. We experimentally evaluate breathing and heart rates using the magnitude signature of measured channel state information in a realistic indoor environment. Experimental results demonstrate low mean absolute errors for both respiration and heart rate compared to ground truth sensor data. The proposed DMA-assisted mmWave testbed unlocks the potential for the practical deployment of various mmWave sensing applications, such as fall detection, gait recognition, human localization and spatial tracking, and realtime vital sign monitoring in dynamic environments, as well as integrated sensing and communication.

The adoption of Generative Artificial Intelligence (GenAI) in Radio Access Networks (RAN) presents new opportunities for automation and intelligence across network operations. GenAI-powered agents, leveraging Large Language Models (LLMs), can enhance planning, execution, and decision-making for orchestration and real-time optimisation of 6G networks. Standardizing the implementation of the Agentic architecture for RAN is now essential to establish a unified framework for RANOps and AgentOps. One of the key challenges is to develop a blueprint that incorporates best practices for memory integration, tool generation, multi-agent orchestration, and performance benchmarking. This study highlights key areas requiring standardization, including agent tool specifications, RAN-specific LLM fine-tuning, validation frameworks, and AI-friendly documentation. We propose a dedicated research initiative on GenAI-for-RAN and GenAI-on-RAN to address these gaps and advance AI-driven network automation.

Reconfigurable intelligent Surfaces (RIS) and half-duplex decoded and forwarded (DF) relays can collaborate to optimize wireless signal propagation in communication systems. Users typically have different rate demands and are clustered into groups in practice based on their requirements, where the former results in the trade-off between maximizing the rate and satisfying fine-grained rate demands, while the latter causes a trade-off between inter-group competition and intra-group cooperation when maximizing the sum rate. However, traditional approaches often overlook the joint optimization encompassing both of these trade-offs, disregarding potential optimal solutions and leaving some users even consistently at low date rates. To address this issue, we propose a novel joint optimization model for a RIS- and DF-assisted multiple-input single-output (MISO) system where a base station (BS) is with multiple antennas transmits data by multiple RISs and DF relays to serve grouped users with fine-grained rate demands. We design a new loss function to not only optimize the sum rate of all groups but also adjust the satisfaction ratio of fine-grained rate demands by modifying the penalty parameter. We further propose a two-phase graph neural network (GNN) based approach that inputs channel state information (CSI) to simultaneously and autonomously learn efficient phase shifts, beamforming, and relay selection. The experimental results demonstrate that the proposed method significantly improves system performance.

To advance towards carbon-neutrality and improve the limited {performance} of conventional passive wireless communications, in this paper, we investigate the integration of noise modulation with zero-energy reconfigurable intelligent surfaces (RISs). In particular, the RIS reconfigurable elements (REs) are divided into two groups: one for beamforming the desired signals in reflection mode and another for harvesting energy from interference signals in an absorption mode, providing the power required for RIS operation. Since the harvested energy is a random variable, a random number of REs can beamform the signals, while the remainder blindly reflects them. We present a closed-form solution and a search algorithm for REs allocation, jointly optimizing both the energy harvesting (EH) and communication performance.

This work presents the design, fabrication, and over-the-air (OTA) characterisation of a high-performance millimetre-wave antenna array operating in the 5G FR-2 band. We propose a 1×8 linear series-fed patch array with optimised performance for both point-to-point and point-to-multipoint communication scenarios. Designed at 28 GHz, the antenna demonstrates high gain (~14 dBi), wide impedance bandwidth, and low sidelobe levels, making it suitable for dense urban deployments. The performance of the prototype was experimentally verified using a Compact Antenna Test Range (CATR), enabling precise beam pattern characterisation under realistic propagation conditions. The study also highlights the impact of array geometry on beam directionality and discusses the practical implications of integrating such antennas in 5G base stations and backhaul links.

Leaky wave antennas (LWAs) are a type of directional antenna known for their ability to scan beams by simply changing the operating frequency. This makes them attractive for some applications like frequency scanning radars and sensing. However, traditional leaky-wave antennas face a significant limitation known as beam squint, where the direction of the radiated beam shifts with changes in the operating frequency. This frequency-dependent behaviour is undesirable in many wireless applications where bandwidth is limited and wideband mmWave circuits are expensive and complex to implement. and poses a major challenge in achieving consistent beam steering at a fixed frequency.

These limitations have sparked interest in more advanced alternatives, such as programmable metasurface antennas. To overcome these limitations, the team introduced a metasurface antenna solution that employs a leaky-wave radiation mechanism through specially engineered materials, enabling electronic beam steering at a fixed frequency, without relying on bulky or power hungry phase shifters. This intelligent antenna leverages reconfigurable meta-elements, allowing it to generate multiple beam patterns at a fixed operating frequency through the digital coding of its constituent elements. Such a fixed frequency beamsteering mechanism is not readily available in conventional LWAs.

Tested at the 60 GHz mmWave band, the proposed antenna demonstrates superior beamforming capabilities at a fixed operating frequency compared to conventional slot-based LWAs. It supports a wide range of applications, including point-to-point and multi-point high-speed wireless communication, short-range radar systems, next-generation 6G networks, and advanced healthcare sensing technologies.

We present RFSensingGPT, an integrated framework for radio frequency (RF) sensing that combines technical question-answering, code retrieval, and spectrogram analysis through retrieval-augmented generation (RAG). Our framework addresses the fundamental challenge of applying large language models to RF sensing applications, where specialized domain knowledge is underrepresented in general training corpora.

Jinsheng Yuan at Cranfield University gave a talk on how mixed computation precision for over-the-air federated learning, titled “Mixed-Precision Over-The-Air Federated Learning via Approximated Computing,” at the IEEE Wireless Communications and Networking Conference (WCNC) in March 2025. Here, the work maps the tradeoff between efficiency and federated accuracy whilst enabling modulation space super-position over the air for different resolution constellations.

At the IEEE International Conference on Machine Learning and Autonomous Systems (March 2025), Prof. Weisi Guo presented a talk titled “Contextualized Autonomous Drone Navigation using LLMs Deployed in Edge-Cloud Computing.” His work demonstrates how large language models (LLMs) can be deployed within 6G Open RAN networks to autonomously generate mission code for drones. This innovation highlights the potential of future 6G infrastructure as a command-and-control centre for aviation.

This white paper discusses the role of large-scale AI in the telecommunications industry, with a specific focus on the potential of generative AI to revolutionize network functions and user experiences, especially in the context of 6G systems. It highlights the development and deployment of Large Telecom Models (LTMs), which are tailored AI models designed to address the complex challenges faced by modern telecom networks. The paper covers a wide range of topics, from the architecture and deployment strategies of LTMs to their applications in network management, resource allocation, and optimization. It also explores the regulatory, ethical, and standardization considerations for LTMs, offering insights into their future integration into telecom infrastructure. The goal is to provide a comprehensive roadmap for the adoption of LTMs to enhance scalability, performance, and user-centric innovation in telecom networks.

This paper addresses how the increasing demand for machine learning (ML) technologies has led to a significant rise in energy consumption and environmental impact, particularly within the context of distributed computing environments like the Edge-Fog-Cloud Continuum. This paper addresses the critical challenge of optimizing ML processes not only for performance but also for sustainability by introducing a novel Green Automated Machine Learning (AutoML) framework. The proposed framework integrates energy-aware task allocation into the AutoML pipeline, strategically distributing computational tasks across edge, fog, and cloud layers to minimize energy usage and carbon emissions without compromising model accuracy. To achieve this, the framework incorporates real-time monitoring and dynamic task allocation based on energy consumption, latency, and carbon footprint, utilizing a hierarchical approach that leverages the unique strengths of each layer within the continuum. The framework is supported by mathematical models that quantify energy consumption, communication latency, and environmental impact, offering a comprehensive metric for evaluating the sustainability of ML deployments. Through extensive simulations and real-world experiments, the framework demonstrates substantial improvements in energy efficiency and significant reductions in environmental footprint compared to traditional AutoML approaches. This research contributes to the advancement of sustainable AI by providing a practical solution for deploying ML models in a manner that balances performance with environmental responsibility.

In this paper, we propose an improved Transmission Control Protocol (TCP) algorithm called HYBIC, building upon existing CUBIC and HYBLA algorithms. This HYBIC algorithm is designed for improving capacity utilisation and transmission rate of heterogeneous networks such as the Integrated Satellite-Terrestrial Networks (ISTN) and long delay High Throughput Satellites (HTS) networks on Geostationary Earth Orbit that are characterised with high bandwidth-delay product path. Results analysed indicated that better performance is achieved using the proposed HYBIC algorithm. Considering the results, HYBIC achieved better performance in terms of window growth of 23×10 3 segments, transmission rate of 3Gbps, and capacity utilisation of 60 % compared with both CUBIC and HYBLA. However, the proposed HYBIC inherits the features of Round-Trip Time fairness, scalability, and friendliness of HYBLA and CUBIC algorithms.

In this work, we developed and proposed a real testbed with Integrated Satellite-Terrestrial Network (ISTN) scenario. This topology was used to measure the actual parameters that were used as the Smart Grid (SG) Quality of Service (QoS) metrics. Performance was evaluated with reference to the QoS requirements of SG applications. The emergence of new and evolving technologies, such as smarter energy utilities enabled by advanced communications technologies, necessitated evolutionary enhancements of both satellite and terrestrial communications systems. These would help improve energy efficiency and sustainability through the effective acquisition and analysis of data from energy systems. Hybrid communication technologies can be used to collect and share energy data efficiently and ubiquitously from generation stations to consumption sites. Thus, a topology using Inmarsat-4 satellites is presented that connects a robust, portable, and energy-efficient Broadband Global Area Network ground terminal. Performance evaluation was performed using measured parameters from the real ISTN topology testbed against key SG applications. The latency and bandwidth requirements for the 80-90% key SG applications were found to be within the QoS requirements range.

This work is part of Pillar 3 – Building Trustworthy Systems and examines the impact of customising the endorsement policy (EP) in Hyperledger Fabric (HLF) for data sharing in Autonomous Vehicles (AVs). In HLF, the EP can be tailored for each application, specifying the required approvals from the endorsing peers of participating organisations. Customisation is often essential to ensure that the transaction validation process aligns with the application’s specific security and business requirements. In our AV data-sharing platform, we implemented both default and customised EP configurations to assess their effectiveness.

This paper introduces the Search-Voxel Ellipse Normalisation technique designed to enhance the accuracy and reliability of respiratory rate monitoring using MIMO FMCW radar in environments impacted by metal-induced multipath distortions. The key contributions include the implementation of an ellipse normalisation method that effectively reduces noise and phase distortions, and the introduction of a voxel selection policy that integrates the Least Squares Method with Dynamic Time Warping and K-Nearest Neighbours to exclude low SNR voxel and refine signal quality. Additionally, the adoption of a Gaussian estimator offers a robust mechanism for continuous and precise respiratory rate estimation. Together, these innovations significantly enhance the efficacy of radar-based health monitoring systems, especially in complex indoor environments.

The aviation and air mobility sectors are undergoing rapid transformation, driven by 6G advancements and the need for secure and reliable airborne communication systems. The integration of Unmanned Aerial Systems (UAS) into commercial and logistical operations is redefining air mobility. The Carrier Sense Multiple Access/Collision Avoidance (CSMA/CA) protocol enhances UAS communication safety by minimizing data transmission collisions and ensuring efficient communication. Combining CSMA/CA with Sixth Generation (6G) technology further strengthens UAS operations’ reliability and effectiveness. MATLAB simulations provide insights into signal quality and latency as a function of distance, critical for analysing safety communications. Additionally, understanding handover procedures, especially vertical handovers in UAS operations, is vital for public safety and the development of advanced Unmanned Traffic Management (UTM) and Air Traffic Management (ATM) systems.

This new paper ‘Mixed-Precision Federated Learning via Multi-Precision Over-The-Air Aggregation’ demonstrates that Over-the-Air Federated Learning (OTA-FL) offers numerous advantages in challenging environments, including ultra-high radio resource efficiency and enhanced privacy preservation. It achieves this by combining client updates through the superposition of modulated waveforms in the propagation channel. Traditional research in OTA-FL assumes that the FL clients have the same computation bit resolutions (e.g., 4-bit, 8-bit 16-bit…etc.), and differential approximate computing (AxC) bit resolution updates pose a problem when reflected into the modulation constellation space superposition (e.g., QPSK, 8-PSK, 16-QAM). Our paper is due to appear in 2025 IEEE Wireless Communications and Networking Conference (WCNC) develops 2 innovations: (1) to create a common modulation basis for multi-resolution OTA-FL, and (2) to create a performance vs. energy efficiency trade-off mapping by adapting the client side bit-resolution. Our work paves the way for Auto-ML: the automatic orchestration of resources in distributed 6G and IoT computing architectures.

The work presented in this paper addresses the risk of misconfiguration introduced by logical inconsistencies when trying to balance QoS, for example between service availability and power efficiency, during the adoption of Open Radio Access Networks (O-RAN) and the deployment of AI/ML-driven applications (xApps) to enhance and simplify network management. This work proposes an approach to use probabilistic model checkers (PRISM and Storm) to quantitatively compute optimal thresholds between energy efficiency and service availability for a defined policy in a given scenario with three radio cells (RCs) and nine user equipments (UEs) while UEs are dynamically switched on and off under uncertainty. As a result, the work can prevent logical inconsistencies in xApp development and safeguard AI-driven applications to ensure network decisions are logically consistent, enhancing reliability and preventing misconfigurations.

The work presented in this paper addresses the accessibility issue of formal verification to security protocol designers by introducing animation as a formal way to verify protocols and the soundness issue of animation. This paper proposes an innovative and iterative workflow to allow designers to carry our verification in the early stage of design with a guarantee that a problem detected during animation must be a bug in the design. Most importantly, animators are automatically generated from security protocol models. As a result, designers can find bugs earlier and deliver secure protocols efficiently without a loss of guarantee.

The integration of machine learning (ML) algorithms with edge sensor systems has fundamentally transformed numerous industries. This convergence empowers real-time data processing, analysis, and decision-making at the network’s periphery. This paper investigates the latest advancements in this domain by examining two key communities: Sensys-ML and TinyML. While Sensys-ML concentrates on optimizing ML for sensor systems, TinyML prioritizes deploying ML models on resource-constrained devices. Through a critical analysis of these communities’ contributions and interactions, this work aims to provide a comprehensive overview of cutting-edge methodologies, persistent challenges, and promising future directions for ML at the edge within sensor systems. By tracing the trajectory of advancements in this field, we offer a critical reflection on the broader research landscape and its scope. Additionally, we identify emerging research areas as reflected in prominent forums and underscore persisting knowledge gaps that call for further investigation.

In the rapidly expanding landscape of Internet of Things (IoT) device manufacturing and deployment, concerns about security have become prominent. This demonstration involves practical attacks on a thread-mesh network within a controlled environment, exploiting vulnerabilities in various components of the Thread network stack. Our attack vectors successfully identified nearby Thread networks and devices by gathering 2-byte Personal Area Network ID (PAN ID) and device frequency information, serving as reconnaissance for potential additional attacks. The focus was on investigating susceptibility to replay attacks and packet injection into thread-mesh networks. Although the experiment attempted to capture thread packets to emulate an authorised sender, the cryptographic encryption and sequence numbers employed for integrity checks resulted in packet rejection by the network. Despite this, our successful injection of packets highlights the potential for battery depletion attacks.

Lip-reading has become a critical research challenge, aiming to recognize speech by analysing lip movements. Traditional methods rely on cameras and wearable devices, but these face limitations like occlusion, lighting conditions, privacy concerns, and user discomfort. In the COVID-19 era, with face masks being the norm, such systems are ineffective for hearing aids. To overcome these issues, this paper introduces an RFID-based smart mask for lip-reading, enabling effective speech recognition even under face masks. The system collects data using RF sensing, capturing three classes: vowels (A, E, I, O, U), consonants (F, G, M, S), and words (Fish, Goat, Meal, Moon, Snake). Machine learning models classify the data, achieving high accuracy across individual classes. On the combined dataset, the Random Forest model achieves 80% accuracy, demonstrating the system’s potential to enhance accessibility and speech recognition for individuals with hearing impairments.

The paper titled “Terahertz Photoconductive Antenna Based on Tai-Chi Totem Structure for Cancer Detection” introduces an innovative design for a terahertz (THz) photoconductive antenna (PCA) utilising a Tai-Chi totem structure. This design aims to enhance the efficiency and sensitivity of THz radiation sources, which are crucial for non-invasive and accurate cancer detection. By leveraging the unique properties of the Tai-Chi totem configuration, the PCA achieves improved performance in generating and detecting THz waves, offering significant potential in medical diagnostics. This research was presented at the 2024 IEEE International Conference on Communications Workshops (ICC Workshops).

The paper titled “A Fractal Dual-Band Polarization Diversity Antenna Array for 5G Communications” introduces a compact, dual-polarized antenna array tailored for the 28 GHz and 38 GHz frequency bands, which are pivotal for 5G applications. With dimensions of 32×20 mm², the antenna array achieves broad impedance bandwidths of 2.7 GHz (26.6–29.3 GHz) and 5.2 GHz (35–40.2 GHz). To facilitate polarization diversity, the design incorporates two orthogonal feed lines, enabling the antenna to support multiple polarization states. This innovation holds significant potential for enhancing the performance and versatility of 5G communication systems.

The paper titled “RIS-Enabled Near-Field Localisation with EMI” explores the utilisation of Reconfigurable Intelligent Surfaces (RIS) to enhance near-field localisation accuracy in environments with Electromagnetic Interference (EMI). By dynamically controlling electromagnetic signal characteristics such as scattering, reflection, and refraction, the study demonstrates how RIS technology can mitigate EMI effects, leading to improved precision in position estimation. This research holds significant potential for applications in wireless communication systems, indoor navigation, and IoT networks. The findings were presented at the 2024 IEEE International Conference on Communications Workshops (ICC Workshops).

The paper titled “Ultrahigh Sensitive Terahertz Metasurface with 2D MoS₂ for Refractive Index Biosensing” introduces a highly sensitive terahertz (THz) metasurface integrated with two-dimensional molybdenum disulfide (2D MoS₂) for advanced refractive index biosensing applications. By leveraging the unique properties of 2D MoS₂, the metasurface achieves an enhanced quality (Q) factor, leading to improved sensitivity in detecting minute changes in the refractive index. This innovation holds significant potential for applications in biosensing, medical diagnostics, and environmental monitoring. The research was presented at the 2024 IEEE International Conference on Communications Workshops (ICC Workshops).

The paper titled “Design of Intelligent Reflective Surface Unit Cell for 5G mmWave Communication Enhancement” presents a novel approach to improving 5G millimeter-wave (mmWave) communications through the development of an intelligent reflective surface (IRS) unit cell. The proposed design utilises a PIN diode switch to achieve a 180° phase shift in reflection mode between its ON and OFF states. The geometry of the unit cell’s top layer is optimised to support transverse electric (TE) polarized reflection with an angular stability of 30°, ensuring consistent performance across a range of incident angles. This research was presented at the 2024 IEEE International Conference on Communications Workshops (ICC Workshops).

The paper titled “RF-Based Respiration Disorders Sensing and Classification Using USRP” explores the application of Radio Frequency (RF) sensing technologies for real-time, contactless monitoring of respiratory disorders. Utilising Universal Software Radio Peripherals (USRP), the study demonstrates how RF signals can detect and classify various respiratory conditions, offering a non-invasive alternative to traditional monitoring methods. This research was presented at the 2024 IEEE International Conference on Communications Workshops (ICC Workshops).

The paper titled “DRL-based Joint Resource Scheduling of eMBB and URLLC in O-RAN” explores the application of Deep Reinforcement Learning (DRL) to optimise resource allocation in Open Radio Access Networks (O-RAN). By jointly scheduling Enhanced Mobile Broadband (eMBB) and Ultra-Reliable Low-Latency Communications (URLLC) services, the study aims to enhance network efficiency and meet diverse Quality of Service (QoS) requirements. This research was presented at the 2024 IEEE International Conference on Communications Workshops (ICC Workshops).

The paper “DRL-based Joint Resource Scheduling of eMBB and URLLC in O-RAN” explores the application of Deep Reinforcement Learning (DRL) to tackle the challenge of resource scheduling in Open Radio Access Networks (O-RAN). By jointly optimising Enhanced Mobile Broadband (eMBB) and Ultra-Reliable Low-Latency Communications (URLLC) traffic, the proposed DRL framework ensures efficient allocation of network resources while meeting the diverse Quality of Service (QoS) requirements of these services. This innovative approach paves the way for scalable and adaptive resource management in 5G and beyond, enhancing network performance and reliability. Presented at the 2024 IEEE International Conference on Communications Workshops, the research highlights significant advancements in AI-driven solutions for modern wireless networks.

Efficient utilisation of scarce resources, particularly radio spectrum and power, is a critical challenge in the development of future Internet of Things (IoT) networks. The paper “Throughput Optimisation in Ambient Backscatter-Based Cognitive IoT Networks” addresses this issue by proposing an innovative approach that combines ambient energy harvesting with backscatter communication in energy-constrained cognitive IoT networks. In this scheme, secondary network nodes effectively utilise the primary network’s resources. Depending on the availability of spectrum and energy, these nodes can switch between energy harvesting and backscatter communication modes, thereby optimising throughput and enhancing overall network efficiency.

In their study, “Performing QoS Measurements in UAV Field Trials with 5G-Advanced Communications at Various Altitudes,” presented at the IET 6G and Future Networks Conference (IET 6G 2024) in London, UK, the authors conduct field trials to assess Quality of Service (QoS) parameters of 5G networks in air mobility scenarios. They investigate latency, throughput, and handover attempts at altitudes of 50, 72, 88, and 110 meters, providing insights into the performance of UAV-to-ground connections across different heights. 

In “Moving Target Defence in 6G UAV Networks,” presented at the IEEE/AIAA 43rd Digital Avionics Systems Conference (DASC) in San Diego, Oct 2024, the authors explore innovative methods to enhance security in UAV networks using 6G technology. By leveraging predictive cyber-attack analysis, network monitoring, and management functions, the study demonstrates how to proactively secure communication data links. This work, recognized with the Best Session Award, highlights the potential of 6G-enabled computing to fortify UAV network defenses.  

Exploring the seamless integration of satellite and terrestrial networks, this publication addresses key challenges in 6G communications, including latency and dynamic user traffic. By leveraging a deep Q-learning-based algorithm for optimizing satellite trajectories, the research offers innovative solutions to enhance network reliability and performance. 

“This paper introduces a multi-band ultrathin reflective metasurface that supports polarization conversion across Ku, K, and Ka bands. The metasurface offers both linear and circular polarization options, making it a versatile solution for advanced communication and sensing applications.”

“This study delves into the use of reconfigurable intelligent surfaces (RIS) for precise indoor localization in ‘coded environments.’ By leveraging data-driven approaches, this research enhances indoor navigation, with potential applications in smart buildings and automated environments.”

“This paper presents the concept of ‘intelligent wireless walls,’ designed for contactless in-home monitoring. The technology enables continuous health and safety monitoring by transforming walls into sensor networks, offering potential applications in elder care and home security through seamless, unobtrusive monitoring.”

Papers before 2023 were not funded by CHEDDAR but are here as they influenced the design of the hub.

“In this innovative study, the authors explore the use of remote RF sensing to achieve lip reading even when individuals are wearing face masks. This research pushes the boundaries of RF technology, presenting methods that could be transformative for communication in healthcare and other masked environments.”

Papers before 2023 were not funded by CHEDDAR but are here as they influenced the design of the hub.

“This paper introduces a comprehensive multimodal dataset aimed at advancing contactless lip reading and acoustic analysis. By combining visual and audio modalities, the dataset provides valuable resources for developing robust algorithms that can accurately interpret speech in challenging environments, enhancing applications in fields like accessibility and remote communication.”

“In this paper, the authors explore data-driven approaches to indoor localization using reconfigurable intelligent surfaces (RIS). This ‘coded environment’ approach enhances accuracy in indoor navigation, making it a promising solution for applications such as smart buildings and automated logistics.”

“This study investigates the use of reconfigurable intelligent surfaces (RIS) to improve near-field localization accuracy. By implementing a practical phase shift model, the research demonstrates how RIS technology can enable precise localization, with potential applications in indoor navigation and IoT.”

“This research explores the development of a 60 GHz programmable dynamic metasurface antenna (DMA), showcasing its potential for next-generation communication and imaging systems. The paper provides a detailed journey from concept to prototype, emphasizing DMA’s role in enhancing wireless communication and imaging precision.”

“This paper introduces a wideband series-fed patch antenna array optimized for 5G Vehicle-to-Everything (V2X) communications. The array’s high gain and low sidelobe characteristics, coupled with its linearly and circularly polarized configurations, make it ideal for improving connectivity and reliability in 5G-enabled transportation systems.”

“This study presents a novel sub-terahertz (THz) planar antenna array designed to enhance sensing and imaging applications. With a focus on achieving high performance in THz frequencies, the research highlights significant advancements in antenna design, providing applications in fields such as medical imaging and environmental monitoring.”