APRÈN. Portal de la producció docent del professorat de la UPC

Artificial intelligence (AI) has become a transformative force in today's world, revolutionising industries by enabling smarter decision-making and automation. Its ability to analyse vast datasets, learn from patterns, and adapt to new challenges makes it essential for solving complex problems in areas like cybersecurity, healthcare, and communications. As networks and systems grow in complexity, AI plays a critical role in enhancing efficiency, improving security, and driving innovation across various fields. This project focuses on three key objectives: first, to explore artificial intelligence, with a deeper emphasis on machine learning techniques, and their applicability in the context of traffic generation. AI and ML models have shown great promise in analysing complex datasets, making them suitable for modelling the behaviour of real-world traffic and generating synthetic traffic that mimics real-world conditions. A thorough analysis of various ML algorithms will be conducted, examining how they can be applied to predict traffic patterns and generate synthetic data that aligns with the characteristics of real IoT traffic. The second component of the project involves capturing real-world traffic from two specific mobile games. Mobile gaming is a growing area where fast and stable communication is crucial, with constant data exchange between the game servers and players' devices. By analysing the traffic from these games, the project will gain insights into the patterns, volume, and nature of data transmitted in real-time, providing a solid foundation for modelling synthetic traffic. This real-world traffic captured will form the basis for developing a realistic dataset, which will be instrumental in generating synthetic traffic using AI-based approaches. Finally, the project will use the captured data to generate synthetic traffic and assess its accuracy and realism compared to the actual traffic data. The goal is to determine how closely the synthetic traffic matches the real traffic in terms of key characteristics, such as transmission frequency, data volume, and event-based spikes. A comprehensive evaluation will be performed to measure the quality of the generated traffic, identifying strengths and potential areas for improvement.

In today's digital landscape, the demand for cutting-edge networks has soared, particu- larly with the rise of eXtended Reality (XR) services. These services come with stringent requirements, demanding high bandwidth coupled with reliable and low latency commu- nications, positioning XR between the URLLC and eMBB paradigms of 5G, with extreme requirements on both sides. This thesis explores a method for addressing these demands leveraging the Network as a Service (NaaS) paradigm, utilizing the communication between the user application and the network to provide rate recommendations for XR applications based on network ca- pacity. The proposed method exploits the powerful capability of xApps in ORAN networks to tune the network in the absence to delve into vendor specific implementations. The solution has been tested in a simulated environment using NS-3 and compared to the commonly used over-the-top (OTT) approach in a scenario with severe network conges- tion. The results obtained show considerable improvements, with up to a 72% reduction in outage probability

Aquest treball analitza l'estàndard IEEE 802.11be, conegut com a Wi-Fi 7, la pròxima generació de tecnologia de comunicacions Wi-Fi. Els objectius principals d'aquest estudi són els d'explorar les millores tècniques que ofereix Wi-Fi 7 en comparació amb les generacions anteriors, avaluar el seu rendiment en diferents escenaris d'aplicació i proporcionar una visió general de les potencials innovacions que aportarà a les comunicacions sense fils. Per assolir aquests objectius, s'ha seguit una metodologia que inclou una revisió exhaustiva de la literatura existent sobre les diferents generacions de Wi-Fi, amb un enfocament particular en les característiques tècniques de Wi-Fi 7. Això inclou l'anàlisi de les millores en la capa física, com ara l'ús de 4096-QAM, l'augment en l'amplada de banda del canal i l'ús de la banda de 6 GHz. També s'han analitzat les novetats en la capa MAC, especialment l'operació multi-enllaç (MLO). A més, s'han descrit els dispositius de hardware i les eines software utilitzades en les proves, incloent-hi punts d'accés, targetes de xara i analitzadors d'espectre. Les proves s'han dut a terme en diferents escenaris, tant interiors com exteriors, per avaluar el rendiment del nou estàndard en situacions reals. Aquestes mesures també han servit per millorar les prediccions d'alguns models teòrics de propagació. Els resultats mostren que Wi-Fi 7 proporciona un augment significatiu en la velocitat de transmissió màxima i una millor gestió de l'espectre, molt útil sobretot en entorns amb alta densitat de dispositius. També es destaca la capacitat del nou estàndard per mantenir connexions més estables i amb menor latència, gràcies a les seves millores tecnològiques. En conclusió, aquest estudi demostra que l'estàndard IEEE 802.11be suposa una evolució important en el camp de les comunicacions sense fils, oferint avantatges clau que responen a les necessitats creixents de banda ampla i connectivitat en diferents àmbits d'aplicació.

El presente trabajo estudia la viabilidad como interfaz de comunicaciones inalámbricas de la nueva enmienda IEEE 802.11ax, certificada como Wi-Fi 6/6E. Para lograr este objetivo, se evalúan parámetros clave de esta tecnología, como el throughput máximo, el tiempo de escaneo y el tiempo de traspaso, mediante pruebas con dispositivos comerciales de Wi-Fi 6/6E. Además, se examinan las bandas y canales de frecuencia más congestionados, especialmente dentro de la banda de 6 GHz, con el fin de minimizar las interferencias en un enlace entre el dron y el control en tierra. Finalmente, se analiza la cobertura que esta tecnología proporciona en las misiones realizadas en el DroneLab del Parque Mediterráneo de la Tecnología. Estas metas se organizan en siete capítulos. El primero ofrece una visión general del estándar IEEE 802.11, repasando su historia, conceptos clave y enmiendas hasta la fecha. El segundo capítulo se centra en los drones, sus principios de vuelo y como Wi-Fi 6/6E puede mejorar sus operaciones. Los siguientes capítulos componen la parte experimental, que incluye pruebas de throughput, escaneo y traspaso en el laboratorio, así como estudios sobre la distribución de redes Wi-Fi en tres ciudades españolas. Finalmente, en el DroneLab, se elabora un mapa 3D de cobertura Wi-Fi, estudiando su comportamiento en cada banda. El trabajo concluye con recomendaciones para futuras investigaciones y un análisis sobre la sostenibilidad de esta tecnología. Los resultados muestran altos valores de throughput, alcanzando hasta 2 Gbps. También, se observan mejoras significativas en los tiempos de escaneo y traspaso; sin embargo, no son suficientes para ciertos casos de uso en drones. Aun así, la certificación Wi-Fi 6/6E aún no está ampliamente adoptada, dado que cuenta con una presencia del 0.1\% en las ciudades. En cuanto a la cobertura, las tres bandas analizadas son capaces de soportar aplicaciones como el streaming de video en la mayoría de los puntos del DroneLab. No obstante, la banda de 6 GHz presenta ciertas limitaciones en los extremos de este volumen. En resumen, considerando las limitaciones y potenciando las ventajas, esta tecnología podría suponer un avance significativo en el ámbito de los drones, mejorando el rendimiento de sus comunicaciones inalámbricas.

The main purpose of this thesis is to explore the benefits of combining two radio interfaces onboard a Unmanned Aerial Vehicle (UAV) to communicate with a Ground Control Station (GCS) and other UAVs inside a swarm. The goals are to use the IEEE 802.11ah standard (Wi-Fi HaLow) with other Wi-Fi technologies, such as IEEE 802.11ax specification (Wi-Fi 6/6E), to enable real-time video transmission from UAVs to the GCS. Several scenarios in which an UAV swarm performs infrastructure inspection are emulated. During the missions, UAVs have to send real-time video to the GCS through a multihop network when some damage in the infrastructure is found. The emulated scenarios are defined using different architectures and radio technologies. Once the emulations finish, the collected data is transferred to spreadsheets to be analysed. An expression to find the multi-hop network capacity from hops distances is found using the NS-3 simulator and this expression is used to determine the capacity of the multi-hop networks used in the emulations. Then, this expression is applied in the algorithms in charge of composing the multi-hop network to offer on demand quality video. This contribution among others related to new techniques and strategies to build the multi-hop network, are implemented in the emulation program. After performing the emulations with the proposed strategies, the resulting data is analysed and compared among the different scenarios.

The present work aims to design and implement a parking assistance system for articulated vehicles called MASDAV (Manoeuvre Assistance System with Drones for Articulated Vehicles). The system is based on the use of a mini drone that deploys when the driver wishes to park the vehicle. This drone provides a bird's-eye view of the vehicle and its maneuvers, thus facilitating the parking process and reducing the risk of collisions. The methodology used includes SITL (Software In The Loop) configuration to simulate the drone's flight environment and integration with QGroundControl to plan and execute take-off, vehicle tracking and landing missions. A Python script has been developed using the dronekit library to control the drone autonomously, simulating parking scenarios and adjusting the height and position of the drone in real time. In addition, the basis is laid for a practical implementation of the proposed mechanisms. For possible implementation, the Holybro X500 V2 drone, compatible with the Ardupilot platform, which offers advanced flight control capabilities and precise navigation via GPS RTK, is recommended. Communication between the drone and the vehicle's navigation system is carried out via radio frequency, ensuring fast and reliable data transmission, even in adverse conditions. The on-board computer chosen is the NVIDIA Jetson Xavier NX, which provides the processing power needed to handle real-time image monitoring and analysis tasks. Ubuntu has been selected as the operating system because of its compatibility with drone control software. The results obtained show that the MASDAV system is capable of providing effective assistance in the parking of articulated vehicles. The drone can track the vehicle precisely, adjusting its position to maintain an optimal view. In the event of signal loss or low battery, the drone automatically returns to its bay on the roof of the vehicle, ensuring the safety and continuity of the system. In addition, it has been proven that the system can operate within Spanish regulations, classifying the drone in category A2, which allows its use in industrial environments with adequate safety measures.

In the modern world, the use of Internet based communications has become a daily activity for a big part of the population. In particular, mobile communications are experiencing a relentless growth in the number of users, which has become an issue as they must be allocated in the channels with the limited available time and frequency resources in the physical layer. This is why Non-Orthogonal Multiple Access (NOMA) schemes are being used in 5G mobile communications and are expected to be a key factor in the future 6G communications. Another factor to take into account is the evolution in technologies such as Artificial Intelligence (AI) and Machine Learning (ML) that, as proven in literature, can perform tasks in the telecommunications world. NOMA will play an important role in 6G communications, while AI/ML has proven to play a key role in many layers including the physical layer, to improve the performance. In particular, constellation design to optimize BER vs SNR performance has proven to be feasible via Deep Learning (DL). In this study, we focus on neural design of demappers for an uplink NOMA system, to demonstrate the performance advantages of these systems against classical demapping techniques. The traditional demapping techniques to evaluate are the Successive Interference Cancellation (SIC) and the Joint Demapper. On the other hand, the SICNet demapper , a ML version of the traditional SIC, has been adapted to the uplink NOMA system. The comparison of the performance of all the demappers has been done with BER-EbN0 curves. The results show that the Joint demapper is the most optimal technique in a two-user uplink system. Meanwhile, the SICNet demapper has proven to outperform the SIC demapper in most of the studied scenarios.

This final degree thesis is an analysis of the IEEE 802.11ax standard and the two subsequent certifications, Wi-Fi 6 and Wi-Fi 6E, with the aim of obtaining information to understand these new IEEE 802.11 standards. We will start by presenting the evolution of the IEEE 802.11 standard throughout its history, detailing the main and most important features such as frequencies, bandwidths, and maximum transmission rates. Then, we will discuss the IEEE 802.11ax specification, distinguishing the mechanisms defined at the physical layer (PHY) and the medium access control layer (MAC). Afterwards, we will proceed with the performance analysis, starting with the description of the scenarios, where we will explain the specifications of the elements used (access points, and adapters) to work in this new frequency band and the dispositions of each of the scenarios. Also, with the help of a spectrum analyser, a visualization of the spectrum found in the 3 bands will be made, which will help us to see how congested they are. The analysis will consist of tests carried out in a laboratory to measure, mainly, the maximum throughput, outdoor tests to measure throughput as a function of distance and, finally, an analysis of the association time will be carried out for the different frequency bands.

The following project is based on the study of different solutions to manage a drone swarm by implementing a Wi-Fi HaLow control interface. To do this, a case study is carried out where a number of drones are deployed on the runway of Josep Tarradellas Barcelona-El Prat airport using different strategies. Then, an analysis is made of the different parameters obtained in order to determine which of the proposed strategies best suits the different scenarios.

This work consists of finding a software simulator that allows the study of WLAN (Wireless Local Area Network) networks based on the 802.11 standard. To this end, it was decided to analyse the suitability of the Cisco Packet Tracer, GNS3 and OMNeT++ programs as wireless network simulation environments. The analysis consists of an introduction to the software, an installation tutorial, the setup of one or two simple scenarios and the conclusions after working with each software. More specifically, the introduction includes a brief explanation of the programme, that is to say, what language it is programmed in, whether it is free software or not, what types of users are used to working with it, what it is usually used for, etc. In the installation tutorial, all the necessary steps to have the software available on the computer are detailed, from the web page where it is recommended to download the executable, to the demonstration of some useful advice. In the assembly of the different scenarios, both the way in which the assembly is carried out and the programme itself are explained, i.e., the processes that must be followed to create each environment are specified and, at the same time, the most important parts of the programme are defined. Finally, in the conclusions, the most positive and negative points of each programme are expressed in the form of an opinion, considering the objective of this work. Once the three programs have been examined, it is concluded that the most suitable software for studying wireless networks through simulated environments is OMNeT++, as it is the program that allows the most detailed specifications related to the configuration of this type of networks. Therefore, a fourth chapter describes the results obtained from the simulation programmed with OMNeT++, which consists of two computers connected through an ad-hoc wireless network. Finally, the last section presents the general conclusions obtained after working with and studying the Cisco Packet Tracer, GNS3 and OMNeT++ programs. It also details the reasons for choosing OMNeT++ as the most suitable program for dealing with the simulation of wireless networks and concludes with a brief explanation of how this work could have been continued.

The fifth generation (5G) mobile networks are enabling operators and stakeholders to enhance and innovate new services in response to an increasing market demand. 5G architecture provides scalability and flexibility for adapting its infrastructure to a customizable communication system by means of Cloudification. Softwarization and virtualization are key terms for upcoming industries that will require ultra-low latency, only possible if the infrastructure equipment that traditionally was centralized in the communication network core is physically moved closer to the user, at the network edge. The main objective of this master thesis was to implement a Reinforcement Learning algorithm (Q-Learning Temporal Difference) aimed at next generation networks to optimally allocate Virtualized Network Functions (VNF) to 5G network Edge Computing (EC) centers. In order to evaluate the algorithm performance and compare it, two more algorithms have been developed to achieve a solution under the same network circumstances. The first one, Best Fit, was inspired by a classical network load balancing algorithm (Weighted Round Robin), whereas the second, MDP, was approached through dynamic programming (Policy Iteration), having posed the network dynamics as a finite Markov Decision Process. The several tests that have been carried out indicate that Q-Learning performs better than the Best Fit and almost as close as the MDP. It shows that the Q-Learning algorithm is able to allocate optimally the incoming VNF demands when EC centers' available resources are somehow restricted.

The increasing need for faster, reliable and latency bounded solutions, especially in theindustrial, medical and audiovisual fields, alongside the demand for mobility and easeof configuration, make the fusion of IEEE 802.11 (“Wi-Fi”) and Time-Sensitive Networking(TSN) more necessary than ever. The evolution of high-quality video, the implementation oftelesurgery in health care and haptic technology in industries brings a wide variety of newchallenges in time synchronization, latency, reliability and management of resources. Toachieve such specifications, formerly all the transmissions were wired, as the wireless so-lutions could not compete especially in terms of reliability and bounded delay. Nowadays,considering the current most notorious iteration of the wireless communications standard,IEEE 802.11ac, with a slight look into the newest improvements in TSN for the upcomingversion IEEE 802.11be and the up-to-date publications for the TSN related amendmentsunder IEEE 802.1, it is obvious that there is a remarkable interest in enhancing the Wi-Fitechnology for these kinds of scenarios. In this work, the concepts of the two technologiesmentioned above are studied, specifically at the physical and link layers, to then develop asimulator with which to study the behaviour of a wireless system in response to a set of in-put parameters and its operation is validated according to different theoretical calculations.In addition, the tool implements a behaviour similar to IEEE 802.1CB (Frame Duplication),in which the packets to be transmitted are sent over all parallel links to increase reliability.Finally, different combinations of parameters are tested to see how they affect the link con-ditions, the number of links, and the possible retransmissions to the system's behaviour.Moreover, use cases are exposed in which the tool could be useful to find the best con-figuration parameters to meet the requirements in terms of reliability and latency betweenpackets.

Los algoritmos de machine learning han permitido desarrollar, durante los últimos años una gran cantidad de aplicaciones que son utilizadas por muchas personas a diario, estas proporcionan funcionalidades en campos tan diversos como la radiología o los asistentes virtuales controlados por voz. Muchas de estas aplicaciones, hubiesen sido prácticamente imposibles de desarrollar sin los avances realizados en esta materia durante las últimas décadas, pero con algoritmos de machine learning son posibles. El principal objetivo de este trabajo es encontrar y explorar aplicaciones potenciales de machine learning en la gestión de un aeropuerto. Como prueba de concepto, se han estudiado en profundidad dos aplicaciones: el reconocimiento facial y la predicción del tiempo de espera encolas en aeropuertos. En lo que respecta al reconocimiento facial, se desarrolla, implementa y evalúa un modelo de reconocimiento facial edge, y también se implementa y evalúa un modelo de reconocimiento facial cloud utilizando la Face API de Azure. Ambas implementaciones se realizan en Raspberry Pi, con la intención de comparar el tiempo y consumo de energía que se requiere para cada uno de los dos modelos.Los resultados de reconocimiento facial obtenidos son buenos en ambos casos, pero son superiores en el modelo cloud, tanto en calidad del reconocimiento, como en tiempo y consumo de energía, siendo la mejor opción utilizar el modelo cloud con una conexión a internet mediante cable.Respecto a la herramienta para predecir el tiempo de espera en cola en un aeropuerto, se ha requerido un procesado de datos públicos sobre tiempos de esperas en aeropuertos. A continuación, se ha hecho una comparación entre varios algoritmos para ver cuál daba mejores resultados y, finalmente, se ha optimizado el algoritmo que ha dado mejores resultados.Los resultados obtenidos con esta herramienta son prometedores, y permiten ampliar el modelo desarrollado en este trabajo para futuras aplicaciones relacionadas con las colas en aeropuertos.

The evolution of wireless technologies has shifted the expectations on how sensors anddata collection is performed. By reducing the complexity to deploy and connect sensors, anincrease in the variety and quantity of data collected is leading to the concept of ubiquitoussensing.However, there are measurements that still rely on manual work. In particular, the auto-motive industry still performs manually the rain isolation quality tests. These tests requirenowadays for an employee to manually place a handheld moisture sensor device in con-tact with the area to analyse, taking long time, low scalability and prone to human errors.Additionally, workers are exposed to these humid conditions which is hazardous for theirhealth.In order to tackle this problem, low-cost wireless identification tags are proposed in thiswork to be used instead. These tags, known as Radio Frequency Identification (RFID),are bundled in pairs, one waterproofed and one exposed to work as moisture detectionsensor. By interacting with them with a RFID telecommunications system, low level datais collected to perform predictions about their status. This way, it is possible to measuresignal differences when subjected to different quantities of water exposure, both in staticand in-movement scenarios, which lead to predictive methods of water leaks.In this work, different methodologies and setups will be evaluated on their performanceas predictors of wet RFID tags. The project involves researching the state of the art onthis topic, empirical experimentation of the different proposed methods, development of anautomated software platform and potential contribution to the academia with the resultsfound, by means of paper publication in conferences and/or journals. In particular, twodifferent experimental setups were tested, a single antenna setup and dual antenna setup.These setups were implemented in a controlled scenario, where 48 different experimen-tal iterations were performed. Later, the setup was implemented on a in-vehicle scenarioto emulate the real industrial operating scenario. In this scenario up to 70 different ex-perimental iterations where performed. When analysing the data samples, four differentmethods are proposed and implemented. Finally, the project also includes the evaluationthrough machine learning algorithms of the best method which achieved higher than80%accuracy test on predicting water leakages in real life scenario such as automotive “raintests”, through the developed sensors and system.With this research, the viability of this methodology for wireless detection of leaks hasbeen shown. Next steps are to explore different Tags manufacturers or complex scenariosinside production lines.