Programa del congreso

Energy efficiency has emerged as a cornerstone in the development of b5G/6G networks, particularly within the framework of Open Radio Access Networks (O-RAN). This paper presents a monitoring and prediction system designed to reduce the energy consumption of critical O-RAN components, namely the Central Unit (O-CU) and Distributed Unit (O-DU). The proposed approach leverages the MAPIE algorithm to generate robust energy consumption forecasts with theoretically grounded prediction intervals, allowing for real-time network resource optimization without compromising service quality. Experimental validation was conducted in a disaggregated commercial O-RAN environment, emulating realistic operational conditions, supporting the identification of consumption patterns and enabling proactive network management. The proposed system serves as a foundation for intelligent, energy-aware management strategies in future mobile networks, paving the way for automated decision-making in Service Management and Orchestration (SMO) systems.

The fifth generation of networks arrived to provide not only enhanced network capabilities but also to establish unprecedented security and privacy levels aimed at protecting end-user security and privacy. While standardization bodies have defined the required security mechanisms, there is a lack of systematic approaches for evaluating the security measures applied by 5G networks and their resiliency under fuzzing attacks. In fact, security compliance certification of any given interface is only required when there are enough offerings from the testing providers, according to TS 33.117 [2]. Currently, this approach leaves multiple critical surfaces out of the testing scope. To address this gap, we propose a complete framework built on open-source software to assess the robustness of 5G base stations in real over-the-air environments, enabling a comprehensive evaluation of security and privacy protections provided to end users.

Brain-Computer Interfaces (BCIs), are revolutionizing human-machine interactions, providing remarkable advancements across medical, industrial, educational, entertainment, and security domains. Despite BCIs' great potential, they simultaneously raise critical ethical concerns regarding user data privacy and their Neurorights, which aim to safeguard individuals against the risks of emerging neurotechnologies, especially as they become increasingly available on the market. Utilizing a comprehensive literature review and implementing a passive attack on two commercial BCI devices, the study reveals critical vulnerabilities during all the phases of the system. Our findings underscore the urgent need for robust cybersecurity frameworks, including Privacy-by-Design strategies. To address this problem, we designed a scoring framework based on the Bluetooth standard, that guides manufacturers and helps regulators, guaranteeing the safe adoption of BCIs. The proposed solution aims to enhance the security of the pairing process on BCI devices, thereby ensuring user neural data integrity and protection.

Garantizar la confianza en las redes 6G es crucial para mantener la seguridad, la fiabilidad y la idoneidad de muchas aplicaciones. La evaluación de la confianza requiere validación, integridad y protección contra manipulaciones. Este artículo explora cómo puede aprovecharse la tecnología blockchain para la evaluación de la confianza a prueba de manipulaciones en las redes 6G. Analizamos las ventajas de utilizar blockchain para una evaluación de la confianza transparente y descentralizada, destacando su resistencia a los ataques y su capacidad para garantizar la integridad de los datos. Además, abordamos los retos de aplicar blockchain en diversas aplicaciones 6G, como la utilización de recursos, la respuesta en tiempo real y la escalabilidad. El artículo también hace hincapié en el papel de los contratos inteligentes en la automatización de la gestión de la confianza y el almacenamiento seguro de la información relacionada con la confianza dentro de la blockchain, ofreciendo un enfoque innovador para mejorar la seguridad y la rendición de cuentas en las redes de próxima generación.

Traducción realizada con la versión gratuita del traductor DeepL.com

Trust and security are paramount in the emerging landscape of 6G networks, particularly as these systems become increasingly decentralized and heterogeneous. Accurate node reputation prediction is essential for maintaining robust network functionality in such environments, enabling the identification of trustworthy nodes and the mitigation of malicious behaviors. In this work, we present a novel reputation evaluation framework tailored for 6G scenarios, combining Graph Neural Networks (GNNs) for representation learning with Support Vector Machines (SVMs) for effective reputation classification. This hybrid approach captures both the structural and behavioral dynamics of network nodes, offering resilience against common threats, such as Sybil attacks and collusion. Our method addresses critical challenges in reputation prediction, including dynamic node behavior, data sparsity, and scalability across massive device deployments. Experimental results demonstrate that our GNN–SVM pipeline significantly enhances reputation prediction performance, offering a scalable and intelligent trust management solution for the next generation of 6G networks.

The vision for 6G is built on the need for predictability, reliability, and time-sensitiveness (i.e. deterministic networking), yet achieving deterministic communication across diverse network domains remains challenging. PREDICT-6G addresses this gap by pioneering an AI-driven, multi-domain framework that integrates deterministic networking principles across heterogeneous infrastructures. By combining a Multi-Domain Data Plane (MDP) with an AI-driven Multi-Stakeholder Inter-Domain Control Plane (AICP), the project enables seamless orchestration of end-to-end deterministic services, ensuring stringent latency, reliability, and synchronization across 3GPP (B5G), TSN, DetNet, and Wi-Fi technologies.

This paper reviews the key highlights of PREDICT-6G and introduces the results obtained.

Next-generation wireless networks will integrate sensing and communication, enabling intelligent and perception-driven connectivity. The MultiX approach introduces the MultiX Fusion Perceptive 6G-RAN (MP6R), a framework that leverages multi-band, multi-static, multi-sensor, and multi-technology capabilities to enhance situational awareness and network adaptability. To address network heterogeneity, MultiX proposes a distributed architecture comprising the MultiX Perception System (MPS) for embedded sensing, the MP6R Controller (MP6RC) for network orchestration, and the Data Access and Security Hub (DASH) for secure multi-sensor data processing. As a proof of concept, we present the Multi-Layer Digital Twin for Industrial Manufacturing, demonstrating how contextual data optimizes decision-making and real-time network adaptation. By integrating sensing across the RAN, MultiX transforms wireless networks into active perception platforms, supporting applications such as autonomous systems, smart manufacturing, and environment-aware wireless optimization,paving the way for future 6G networks.

This work presents a perspective on addressing the underutilization of computing resources in FPGA SoC devices deployed in 5G radio and edge computing infrastructure. The initial step in this approach involves developing a resource management layer capable of dynamically migrating and scaling functions within these devices in response to contextual events. This layer serves as the foundation for designing a hierarchical, data-driven micro-orchestrator responsible for managing the lifecycle of functions in FPGA SoC devices. In this paper, the proposed resource management layer is utilized to reconfigure a function based on events identified by a computer vision edge application.

Vehicular communications are expected to become one of the services that will benefit the most from 5G networks. The low latency and unprecedented high data rates 5G is expected to provide are a key enabler for new vehicular services. However, there are still different topics hindering its global adoption. For instance, certain cellular coverage holes may prevent having continuous V2X services as tele-operated driving. Similarly, situations such as traffic congestions can drain radio resource availability. In this context, the 6GTWINROAD-MNO project aims to investigate how to design the control plane of future 6G V2X network slices. Particularly, in this paper we merge the network exposure APIs defined in the 5G Core and the Open-Radio Access Network (O-RAN) architecture to provide a system that offers advanced exposure functions for future V2X services. In this context, we present two different exposure functions: predictive quality of service (QoS) for tele-operated driving and a digital twin for cooperative perception.

This paper provides an overview of the TRAINER-6G project, targeted to a special session devoted to spreading awareness within the URSI community on current research projects in 6G conducted by national academia and research centers. Since the project has just recently started, the paper describes its starting hypothesis and objectives. Besides, the link of the envisaged research with previous results is highlighted.

El aumento de las enfermedades óseas debido al envejecimiento poblacional requiere la búsqueda de nuevas técnicas de implantes, como son los andamios óseos. En esta contribución se propone el diseño de andamios óseos autosensores que funcionan como sensores pasivos de microondas sin necesidad de electrónica adicional, a la vez que mantienen sus características estructurales y favorecen la regeneración ósea. Para ello, se diseña un resonador dieléctrico (RD) poroso que se puede fabricar mediante técnicas de fabricación aditiva (impresión 3D), resultando en una estructura personalizable y de bajo coste. Se propone una prueba de concepto mediante la que se observa que la simulación de la regeneración ósea produce un cambio en la frecuencia de resonancia y ancho de banda del RD poroso. Estos resultados experimentales demuestran la validez de esta técnica para conseguir una monitorización no invasiva, continua, en tiempo real y barata mediante radiación no ionizante para realizar el seguimiento del proceso de regeneración ósea.

This paper presents a study related to the implementation of realistic breast models for the training and optimization of non-invasive microwave-based breast tumour detection systems. The use of these synthetic models, with similar properties to the biological tissues, is essential in order to evaluate the performance of tumour detection systems in controlled laboratory environments, prior to testing with patients in clinical settings.

In this paper, mixtures of commonly used materials have been used in the implementation of the synthetic tissues. From these mixtures, complex and realistic breast models with two and three layers have been implemented. The developed breast models have been analysed with a microwave medical imaging system. The signals obtained have been processed with complex techniques to detect both the breast tissues and, more importantly, to detect and locate the tumour tissue. The results obtained are promising, predicting good properties of these systems in the detection of breast tumours, especially in the screening phase.

We present a numerical framework to simulate the nonlinear electromagnetic response of biological cells subjected to microwave excitation. Building on recent experimental studies that detect third-order intermodulation products from living cells, our model captures the electrodynamic behavior of membranes via time-dependent conductivity and permittivity changes associated with electroporation. The governing equations are formulated in the time domain, based on a full-wave Maxwell model for TE polarization, and discretized using a semi-implicit scheme coupled with a high-order fast direct solver. Validation against reference solutions in both damped and lossless configurations demonstrates high accuracy and numerical stability. We further illustrate the nonlinear current response of a single cell placed between parallel electrodes under time-varying voltage stimulation. The results highlight the potential of this computational approach to predict nonlinear microwave interactions in cellular environments and to complement experimental microfluidic platforms for future sensing and biomedical applications.

Este trabajo presenta un novedoso sensor de microondas en guía de ondas rectangular no uniforme para la caracterización de las propiedades dieléctricas de materiales. Este describe el principio teórico de funcionamiento del sensor. Las dimensiones de la estructura se han optimizado para caracterizar las propiedades dieléctricas de aceites vegetales. La respuesta del sensor se ha validado midiendo un sustrato sólido con propiedades dieléctricas similares a las de estos aceites. La permitividad relativa del material que rellena la estructura se obtiene mediante la fase del coeficiente de reflexión, que se mide y compara con el análisis teórico y las simulaciones.

This work presents an improved low-cost system for measuring the complex permittivity of liquids and semi-solids using an open-ended coaxial probe. The new system replaces the previous SMA connector with a 0.047" semi-rigid coaxial cable assembly to extend the frequency range up to 20 GHz. The performance of the proposed system is evaluated by comparing measurements of various liquids acetone, methanol, ethanol, propan-1-ol, propan-2-ol, Triton, sunflower oil, saline water, and a 50% water-Triton mixture with those obtained using a commercial open-ended coaxial probe. The results demonstrate the accuracy and reliability of the proposed system, showing excellent agreement with commercial measurements, especially in the frequency range of 500 MHz to 20 GHz. This study highlights the potential of the proposed system as a cost-effective alternative for dielectric characterization, making it suitable for use in laboratories with limited resources.

Accurate sleep stage classification plays a significant role in sleep analysis and sleep disorder diagnosis. Traditional manual annotation of polysomnography (PSG) recordings is labour-intensive and time-consuming, prompting the adoption of automated sleep scoring methods. Although many approaches have been developed, challenges remain in terms of the temporal dependence of PSG signals and computational efficiency. To address these limitations, we propose a novel sleep stage classification model with strong clinical feasibility that balances performance and computational cost. Our method utilises two hierarchical components to process multichannel PSG recordings. First, handcrafted features are extracted in both time and frequency domains in accordance with the American Academy of Sleep Medicine (AASM) guidelines. By including the neighbouring epochs, it preserves temporal correlations. Second, an extreme gradient boosting (XGBoost) classifier is employed for final classification results. Experimental evaluations demonstrate that the proposed model outperforms state-of-the-art approaches in classification performance and computational efficiency, highlighting its potential for real-world clinical implementation.

En este trabajo se han analizado exhaustivamente las tolerancias de fabricación de la tecnología CLAF-SIW para aplicarlas al filtro conseguido en [1]. El filtro consiste en un BPF de cavidades apiladas en el rango de frecuencias milimétricas, la tecnología se adapta muy bien a este tipo de filtros y en ese rango de frecuencias. Aún así, hay que tener en cuenta las tolerancias en la fabricación. Se ha conseguido determinar los posibles errores en simulación, sobre la celda unidad EBG de la tecnología, sobre las dimensiones de los iris y cómo afectan en su conjunto al filtro. Posteriormente, se han analizado las tolerancias en medida, para conseguir mitigarlas y obtener la respuesta ideal del filtro.

En este trabajo se presenta un nuevo convertidor de modo TE10–TE20 para guías de ondas rectangulares (RW). La estructura propuesta se concibe como un único dispositivo sólido que se inserta en una sección recta vacía de una RW para obtener la conversión de modo deseada. El convertidor de modo diseñado se fabrica mediante técnicas de fabricación aditiva y dos materiales dieléctricos con diferentes permitividades. El trabajo incluye consideraciones teóricas para las dimensiones del diseño propuesto. Para validar la novedosa topología sugerida, se diseñó un convertidor de modo TE10–TE20 para su inserción en una sección recta vacía de 70 mm de longitud de WR-137. Los resultados simulados muestran que la pureza de modo es superior al 98 % para un rango de frecuencias entre 9.5 GHz y 11.6 GHz, mientras que la eficiencia de conversión es superior al 85 % para un rango de frecuencias entre 10.1 GHz y 11.7 GHz. El convertidor de modo fabricado se caracterizó en una configuración back-to-back utilizando transiciones de WR-90 a WR-137. Las medidas concuerdan satisfactoriamente con las simulaciones electromagnéticas. Además, se comprobó el comportamiento del prototipo fabricado midiendo el diagrama de radiación de una apertura WR-137 alimentada con el modo TE20 generado por el convertidor de modo fabricado.

This paper presents initial research on differentiable solvers for electromagnetics based on Automatic Differentiation (AD), aimed at efficiently integrating forward solvers into gradient-based optimization workflows and neural network training. Specifically, we explore differentiable programming techniques applied to two electromagnetic problems: parameter estimation in a simple RC circuit governed by ordinary differential equations (ODEs), and reflection coefficient computation in multilayer electromagnetic wave propagation. We systematically compare FiniteDiff.jl and two AD libraries ---ForwardDiff.jl and Zygote.jl--- in the Julia programming environment, evaluating their performance and precision in computing derivatives and gradients. Results highlight Zygote's superior efficiency and scalability, particularly advantageous in large-scale inverse problems involving numerous parameters. The proposed approach demonstrates significant potential for addressing complex inverse electromagnetic problems through differentiable programming paradigms.

This paper presents a systematic assessment of four machine learning (ML) models used for the simultaneous extraction of complex permittivity (𝜺) and permeability (𝝁) in magnetodielectric composites. Results are analyzed by focusing on metrics and techniques to improve models explainability and identify the main characteristics involved in the predictions. The models have been trained with 2554 samples obtained from full-wave simulations in CST microwave studio. The best model has been chosen using MSE and R<sup>2 metrics. An analysis of explainability has been performed to enhance understanding of the results and to establish feature importance in the model predictions.

This article describes the design procedure of a “filtenna” (filtering antenna). The “filtenna” is conceived as a third order filter consisting of two open ring resonators which are coupled to a resonant radiating patch through an aperture. The radiating patch behaves as the third resonator of the filter. The “filtenna” is used as the building block of a dual-polarization “diplectenna” (diplexing antenna). In particular, four pairs of coupled open ring resonators are connected to two radiating microstrip patches. Two of the pairs of coupled rings excite the radiation of the two patches in horizontal linear polarization at two different bands, while the other two pairs excite the vertical linear polarization at those bands. One of the patches radiates at a center frequency of 5.5 GHz, while the other patch radiates at a center frequency of 9.0 GHz. The “diplectenna” shows a fractional bandwidth of roughly 6.5% in both bands for both polarizations with sharp frequency selectivity, gain values larger than 7 dBi and cross-polar radiation levels below -20 dB.

The use of asymmetric coupled waveguides enables efficient management of losses associated with curvature. This type of geometry is particularly interesting in platforms such as silicon nitride, where radiation loss limits the bend radii that can be used and, consequently, the attainable density of integration. This document presents a review of recent results related to this configuration and its applications in racetrack-type microresonators, the design of ultra-compact polarizers, and biosensors.

In this work, we present the characterization of an integrated Distributed Bragg Reflector (DBR) laser designed for space applications. The laser delivers a high output power of approximately 2 mW and a continuous tuning range exceeding 100 GHz, with a side-mode suppression ratio (SMSR) of up to 49 dB, ensuring stable single-mode operation. These results highlight the DBR laser's potential as a critical component in advanced photonic systems, particularly for space-based and remote sensing applications that require stable, high-power, and tunable light sources.

La combinación de arrays con lentes dieléctricas es una solución atractiva para mejorar el rendimiento de la antena en aplicaciones de ondas milimétricas. El diseño y la optimización de estas lentes presentan retos importantes. Esto se debe principalmente a los altos requisitos computacionales de su simulación con paquetes de software de onda completa, debido principalmente a la simulación extra que se requiere para simular la distribución de fase no lineal del array. Una posible solución para acelerar el proceso de diseño es la aplicación de un método de trazado de rayos ("ray tracing'" RT en su denominación en inglés). Para hacer frente a las limitaciones de nuestro anterior modelo de RT en 2D, actualmente lo estamos ampliando para poder analizar escenarios 3D. En este trabajo describimos los pasos necesarios para la creación de la geometría y la implementación de la óptica geométrica dentro del modelo. La parte de óptica geométrica incluye tanto el RT inverso para la obtención de la distribución de fase como el RT directo para el cálculo del diagrama de radiación. Los resultados de nuestro modelo simplificado muestran una buena concordancia con modelos de onda completa.

This work highlights the influence of sideband impedance on the stability of RF power amplifiers. A reconfigurable measurement setup is developed, allowing independent control of baseband and sideband impedances to experimentally assess their impact on stability. The proposed methodology employs the large-signal stability parameter, K</sub>LS</sub>, to predict instability under various baseband impedance conditions. Experimental results demonstrate that, even when the amplifier is matched at its fundamental frequency, certain sideband impedance combinations can trigger oscillations, as observed in a fabricated GaAs based microstrip amplifier. By correlating measured scattering parameters with the predicted K</sub>LS</sub> values, stable operating conditions are identified. The results emphasize the necessity of considering sideband impedance in power amplifier stability analysis to prevent undesired oscillations.

Esta comunicación describe el diseño de un amplificador de potencia (PA) híbrido, adecuado para el uso de modulación de carga (load modulation). El diseño está optimizado para aplicaciones a una frecuencia de 3.5 GHz y específicamente pensado para comunicaciones con vehículos aéreos no tripulados (UAVs). El diseño utiliza un transistor de alta movilidad de electrones de nitruro de galio (GaN-HEMT), seleccionado por su superior potencia y eficiencia, un factor crucial para aplicaciones embarcadas. Se basa en una configuración de amplificador balanceado (BA), que combina dos amplificadores de potencia individuales mediante acopladores. Los acopladores se diseñan utilizando "curvas de llenado de espacio" para mejorar la potencia mientras se minimizan las reflexiones de señal. El circuito puede operar como un BA convencional o como un amplificador balanceado modulado por carga (LMBA). El circuito diseñado fue fabricado y medido. En el punto de compresión P2dB, se midieron: una potencia de salida de 34.5 dBm, una ganancia de potencia de 15 dB y una eficiencia en el drenador cercana al 40%.

In this paper, we demonstrate the improvement in circuit performance when time-varying circuits are modulated by non-sinusoidal waveforms. Specially we focus on triangular and rectangular modulations. By this, bandpass filters are designed and compared to conventional cases in order to exhibit the enhancement. Specially, the bandpass filters are desined to invoke non-reciprocity, an exotic and very interesting feature in time-modulated systems. A method of analysis based on the spectral domain is employed, establishing a general approach that is valid and applicable to other types of signals, as long as it is possible to obtain its decomposition in Fourier series. The results obtained from our model match those simulated using the Keysight ADS, thus validating the present approach.

Este trabajo presenta la integración de un sistema receptor de satélite en banda Ka basado en componentes comerciales disponibles en el mercado (COTS). El receptor consta de un amplificador de bajo ruido, un filtro pasa banda, un mezclador subarmónico, un sintetizador de frecuencia y un amplificador de frecuencia intermedia (FI). La banda de frecuencias del receptor es de 27 - 30 GHz, siendo una de las bandas asignadas a las comunicaciones Satcom para enlace de usuario. La banda de FI es DC - 2.5 GHz, que se establece mediante el rango de operación del mezclador. Todos los COTS seleccionados son circuitos integrados monolíticos de microondas (MMIC), excepto el sintetizador de frecuencia, que está compuesto por un MMIC encapsulado. Se ha diseñado un chasis a medida para integrar el receptor, que sirve como precursor de la integración de estos componentes en un módulo System-in-Package (SiP) para su evaluación en aplicaciones espaciales. El primer objetivo del sistema desarrollado en formato convencional (chasis) es caracterizar y validar la funcionalidad de los componentes COTS como receptor de banda Ka. Este enfoque va más allá de la simple conexión de estos componentes; implica planificar cuidadosamente el diseño estructural del chasis en el menor espacio posible.

The current responsivity of zero-bias microwave detectors based on FETs is assessed for the case of a GaN HEMT by means of Monte Carlo (MC) simulations and a small-equivalent circuit (SSEC). A MC analysis is presented at very intrinsic level, including RF power reflection and transmission effects. Optimization of the detector is proposed using the SSEC and a circuit simulator in order to study the impact of parasitic elements. An analytical expression for the responsivity allows us to incorporate the effect of adding external Csub<gs> and Csub<gd> capacitors. The inclusion of these elements in the layout may improve the bandwidth in fabricated detectors.

This work presents a monolithic transmit/receive (T/R) frontend for the 6–18-GHz band, implemented in 150-nm gallium-nitride (GaN) technology. Single-die integration is leveraged by co-designing all subcircuits: the T/R asymmetric switch simplifies the transmit (Tx) path to save for Tx output power and efficiency, while its receive (Rx) branch merges into the low-noise amplifier input matching network for compactness and reduced noise. On-wafer measurements show an average Tx output power of 9.5 W (minimum of 5.6 W) with 19.4 % efficiency (minimum of 13.5 %) across the band. In Rx mode, the prototypes exhibit over 21 dB gain with an average noise figure of 3 dB (maximum of 3.3 dB). These metrics compete with the state of the art of ultrawideband GaN-based single-chip T/R frontends.

The majority of advancements and research conducted in the field of non-contact vital sign monitoring using radar technology are not suitable for implementation in low-latency, real-time scenarios, such as in apnea monitoring. This paper presents a real time tool implemented in Simulink, which is independent of the radar device, for multiple patient vital sign monitoring, including an apnea detector module. Its performance has been proven using a phase array radar, the AWR21243BOOST, in which the breathing activity in multiple patient scenarios and the detection of apnea episodes have been successfully evaluated.

La monitorización de la marcha puede ayudar a diagnosticar ciertas enfermedades, como el párkinson. Para que sea efectiva, esta debe realizarse en entornos domésticos. Los radares representan una solución prometedora para este propósito, ya que ofrecen fiabilidad sin comprometer la privacidad. La instalación de radares en hogares puede beneficiarse de la extracción inalámbrica de datos, lo que reduce la complejidad de la instalación. Este trabajo presenta tres algoritmos que permiten la extracción de datos de radar FMCW y CW mediante Bluetooth Low Energy (BLE) hacia una unidad de procesamiento externa. Estos algoritmos son extremadamente eficientes desde el punto de vista computacional, ya que se implementan en microcontroladores comerciales de bajo costo. Se evalúan tanto en configuraciones de un solo nodo como en configuraciones multinodo, analizando su rendimiento en términos de relación señal-ruido (SNR) y precisión en los parámetros de la marcha. Este trabajo es relevante para futuras implementaciones domésticas de radar, ya que propone soluciones efectivas y eficientes que facilitan su adopción.

La detección temprana de la enfermedad de Parkinson (EP) es fundamental para mejorar la calidad de vida de los pacientes. La tecnología radar puede ayudar a anticipar su detección mediante la monitorización de la marcha. Sin embargo, ningún estudio ha identificado cuáles son los parámetros de la marcha más relevantes para detectar la EP utilizando radares. En este artículo, se compara la marcha de dos individuos sanos y tres pacientes con EP utilizando una red de radares para determinar qué parámetros de la marcha son más determinantes para detectar la enfermedad. Los resultados muestran que el tiempo medio y máximo de zancada cuando los sujetos caminan a su ritmo habitual son los más determinantes para la detección de la EP. Esta contribución es relevante para futuras implementaciones de redes de radares, ya que señala qué parámetros deben analizarse durante las evaluaciones clínicas.

It is becoming increasingly important for winegrowers to have sensor-generated data to make decisions about vineyard management in a way that improves the quality of the final product. Currently, there are no techniques to measure the volume of grapes in each vineyard and no non-destructive techniques to detect trunk, such as those caused by Esca. This work presents the studies carried out to test the application of SAR and ISAR techniques in the millimeter frequency range in the field of viticulture for the detection of the volume of grapes present on vines as well as for the detection of dry trunk sectors due to the presence of Esca. The measurement setups used and the images obtained by applying the SAR and ISAR algorithms are shown, demonstrating the feasibility of their use in the sector.

This paper presents an experimental comparison of multiple Reconfigurable Intelligent Surface (RIS)-enabled self-localization techniques using Frequency-Modulated Continuous Wave (FMCW) radar. The localization techniques leverage Angle-of-Departure (AoD) and distance estimates from each RIS to enable precise localization of a User Equipment (UE). A dedicated millimeter-wave band demonstrator, incorporating a TI FMCW radar as the UE and a Sivers multiantenna transceiver to emulate RIS functionality, is used to validate the proposed techniques. The results demonstrate the feasibility of self-localization using a limited number of RIS in a linear deployment. Additionally, the robustness of AoD-based and distance difference-based methods against unknown bias errors, introduced due to the physical limitations of the demonstrator, is highlighted. This work identifies the potential of RIS-enabled self-localization in future 6G wireless communication systems, offering insights into the trade-offs between system complexity and localization accuracy in scenarios where no base station is utilized.

En los últimos años, se ha procedido al despliegue de escáneres de seguridad basados en radar MIMO en banda milimétrica para detectar amenazas ocultas en equipajes y bajo la ropa. Sin embargo, enfrentan desafíos tecnológicos debido a la distribución de señales de alta frecuencia, aumentando su complejidad y coste. Este trabajo presenta un radar MIMO en banda D, donde dichas señales se generan localmente en cada nodo. El diseño, basado en MMICs (Monolithic Microwave Integrated Circuits) comerciales con antenas integradas, es compacto y escalable. El sistema combina un radar MIMO horizontal con un barrido mecánico vertical, logrando imágenes 3D a 2-3 metros con resolución inferior al centímetro.

La densificación de la red de acceso radioeléctrico es crucial para el despliegue rápido y eficiente de redes celulares que vayan más allá de la actual red 5G. Este hecho implica integrar celdas pequeñas en zonas donde las macro celdas tradicionales experimentan problemas de cobertura o capacidad. La coordinación y gestión eficaces de estas celdas recién desplegadas son esenciales para garantizar un rendimiento adecuado de la red. Con los recientes avances en inteligencia artificial, esta tarea puede automatizarse mediante técnicas de aprendizaje automático. En este estudio, se propone un método automático para generar listas de celdas vecinas en una red celular basado en redes neuronales de grafos. El modelo propuesto se valida con un conjunto de datos extraídos de una red "Long Term Evolution" (LTE) real. Los resultados demuestran que la metodología propuesta se anticipa con éxito a las listas de vecinas generadas por la función Automatic Neighbor Relations (ANR) y ajustadas por el operador de red.

In beyond 5G mobile networks, the arise of low-power end-devices and data-hungry services with high UpLink (UL) traffic demands makes interference minimization crucial for network operators. Identifying the root cause of interference experienced in a cell is key for effective decision-making. With recent advances in artificial intelligence, such a task can be tackled through supervised learning provided that a labeled training dataset is available. Consequently, automatic data labeling is essential to enable fully automated interference hunting in zero-touch networks. This work presents an algorithm for clustering interference patterns experienced in the UL of a cellular radio access network by root cause through unsupervised learning. The proposed scheme has isolated 13 interference types on a dataset collected in 4 live LTE networks, comprising samples from cells with different bandwidths and radio channel degradation.

This paper presents the design of an automated alarm-based incident prediction system developed in collaboration with Cellnex, the Spanish digital television broadcasting network operator. Due to confidentiality agreements restricting the disclosure of results derived from proprietary data, the system was evaluated using synthetically generated scenarios, Baseline, Degraded, and Alarm-storm, each simulating different network conditions. For each scenario, alarm events are preprocessed, normalized, and encoded into numerical features across four contextual settings: no additional information, pattern sequences, time-based statistics, and their combination. Experimental results show that incorporating temporal context significantly improves incident prediction, particularly in highly imbalanced scenarios. Among the classifiers evaluated, LightGBM consistently achieves the highest balanced accuracy when using time-based statistical features. These results highlight the importance of temporal modeling and imbalance handling in improving alarm-based incident prediction.

Este trabajo presenta un sistema integrado que utiliza drones para identificar y analizar estaciones base en redes de comunicación, combinando tecnologías de posicionamiento GPS, sensores de orientación y aprendizaje automático mediante el modelo YOLOv8 para procesar imágenes capturadas durante el vuelo. El sistema ajusta dinámicamente la altitud del dron y recalcula distancias y direcciones hacia las antenas más cercanas, asegurando una cobertura óptima y una detección precisa incluso en entornos complejos, mientras recopila datos georreferenciados que se exportan a archivos CSV y Excel para su posterior análisis, destacando su potencial para optimizar redes de comunicación asistidas por drones.

For cellular networks, the knowledge of user position is key for improving its services; such as millimetre waves, user-based service or self-healing. The main researches in this area focus on improving the localisation systems, with a special interest in machine learning (ML)-based fingerprinting solutions. Nevertheless, the system performance is measured by prediction RMSE, or other metrics, ignoring any methodology of sensitivity of the system to non-ideal inputs. In this work, statistical-based methods are proposed for characterising the data used for training, the ML model sensitivity and prediction error margins. These aspects are evaluated with a dataset from a commercial mobile network from a European city.

Las redes móviles se enfrentan cada vez más a problemas de congestión en escenarios muy poblados. En este contexto, el despliegue de estaciones base aéreas (ABS) en drones ofrece una solución eficaz y flexible para ampliar la cobertura y aumentar la capacidad de la red sin necesidad de infraestructura fija adicional. Este artículo propone un algoritmo de navegación basado en el aprendizaje profundo por refuerzo (DRL) que optimiza la posición de una ABS para maximizar el cumplimiento del acuerdo de nivel de servicio global. Se comparan cuatro versiones del algoritmo, que se diferencian en la función de recompensa y en los datos utilizados para entrenar al agente DDQN. Los resultados de la simulación muestran que un agente con una función de recompensa basada en mediciones de potencia y entrenado con experiencias recogidas en múltiples entornos logra el mejor rendimiento global, aumentando el cumplimiento del acuerdo de nivel de servicio en un 16,27% en escenarios en los que los usuarios se concentran en una pequeña zona del escenario.