Search

This project consists of the development of an all-in-one system that gathers data from a live e-sports match and displays it, in real time, on a publicly accessible website. The idea was motivated by the fact that, compared to traditional sports, e-sports already reside digitally inside computers, making it much easier to get data from them. Moreover, the concept of watching a live stream while doing other activities on a secondary screen, especially among young, tech-savvy individuals, is very widespread. The objectives of the project aim to solve the problem of creating the previously mentioned system, which continuously uploads game data to a database in order to be analyzed and shown on a public website in real time. Additionally, another objective involves creating a private space on the website where the administrators can monitor the user traffic of the public site. The entire process has been carried out and documented through a very technical scope, mainly focusing on data management and transmission, and using Javascript as the predominant programming language throughout the whole system. The development has been structured using the Scrum methodology, dividing the work in five sprints, each of which with a specific deadline and objective. Although some scheduling modifications have had to be made during development due to unforeseen problems, the project has been completed in time thanks to the sprint-segmented workflow and careful initial time planning. Then, it has been tested in different ways to check its resilience to uncommon edge cases and validate that the objectives have been met. The result is a robust system comprised of three modules: a local application that connects to the game client and gathers the data, a server that communicates with the database and allows the other modules to interact with it, and a website containing both the public data visualizations and the private administrator dashboard. Even though they are independent pieces o

Purpose: Clinical needle insertion into tissue, commonly assisted by 2D ultrasound imaging for real-time navigation, faces the challenge of precise needle and probe alignment to reduce out-of-plane movement. Recent studies investigate 3D ultrasound imaging together with deep learning to overcome this problem, focusing on acquiring high-resolution images to create optimal conditions for needle tip detection. However, high-resolution also requires a lot of time for image acquisition and processing, which limits the real-time capability. Therefore, we aim to maximize the US volume rate with the trade-off of low image resolution. We propose a deep learning approach to directly extract the 3D needle tip position from sparsely sampled US volumes. Methods: We design an experimental setup with a robot inserting a needle into water and chicken liver tissue. In contrast to manual annotation, we assess the needle tip position from the known robot pose. During insertion, we acquire a large data set of low-resolution volumes using a 16 × 16 element matrix transducer with a volume rate of 4 Hz. We compare the performance of our deep learning approach with conventional needle segmentation. Results: Our experiments in water and liver show that deep learning outperforms the conventional approach while achieving sub-millimeter accuracy. We achieve mean position errors of 0.54 mm in water and 1.54 mm in liver for deep learning. Conclusion: Our study underlines the strengths of deep learning to predict the 3D needle positions from low-resolution ultrasound volumes. This is an important milestone for real-time needle navigation, simplifying the alignment of needle and ultrasound probe and enabling a 3D motion analysis.

This thesis examines the integration of 5G technology with edge computing to address the challenge of reducing latency in telecommunications, towards enhancing performance for real-time applications such as autonomous driving, telemedicine, online gaming, and smart city infrastructures. Central to this research is the concept that hosting applications, processing and routing data at the network's edge, closer to its source, can significantly decrease transmission delays, offering critical improvements in safety, decision-making speed, and interactive experiences. By deploying a practical 5G testbed with edge computing capabilities, the study not only demonstrates considerable latency reductions but also validates the feasibility of hosting and managing application data directly at the edge, thereby presenting a more efficient, cost-effective, and secure alternative to conventional cloud processing. This work underscores the transformative potential of combining 5G and edge computing in creating responsive, high-performance telecommunications systems.

This paper presents an algorithm for synthesizing a supervisor for timed automata (TA) using the conventional supervisory control theory. The algorithm is directly applicable to TA without explicit transformation into finite automata, and iteratively strengthens the guards of edges labeled by controllable events and invariants of locations where the progression of time can be preempted by forcible events. The synthesized supervisor, also a TA, is controllable, maximally permissive, and guarantees a non-blocking and safe supervised plant. The use of real-valued clocks in TA makes it a practical modeling framework, however, the infinite state space brings challenges. The proposed algorithm addresses these by providing a synthesis method that avoids the state-space explosion of finite automata and the loss of information that can result from abstraction of real-time values.

A practical CV-QKD system has been presented and characterized. Our real-time GPU- based receiver (Bob) implementation demonstrates effective calibration, digital signal processing, and post- processing techniques. The system achieves peak finite-size secret key rates up to 5.8 Mbps over 12.8 km and 0.5 Mbps over a 25.7 km fiber distance. Additionally, we conducted a numerical simulation analysis to enhance the system’s resilience to quantum channel loss. Various scenarios were explored, focusing on potential improvements in excess noise, reconciliation efficiency, and quantum efficiency, aiming to extend distance tolerances beyond 50 km

Emerging computing paradigms like mist, edge, and fog computing address challenges in the real-time processing of vast Internet of Things (IoT) applications. Alongside, cloud computing offers a suitable platform for executing services. Together, they form a multi-tier computing environment known as compute-continuum to efficiently enhance data management and task execution of real-time tasks. The primary considerations for compute-continuum include variations in resource configuration and network architecture, rental cost model, application security needs, energy consumption, transmission latency, and system reliability. To address these problems, we propose two scheduling algorithms (RCSECH and RSECH) for real-time multi-workflow scheduling frameworks. Both algorithms optimize for rental cost, energy consumption, and task reliability when scheduling real-time workflows while considering deadlines and security requirements as constraints. RCSECH also factors in reliability alongside these constraints. The environment under investigation consists of a compute-continuum architecture consisting of mist, edge, fog, and cloud layers, each potentially composed of heterogeneous resources. The framework undergoes evaluation via simulation experiments, revealing promising results. Specifically, the framework exhibits the capability to enhance reliability by up to 7%, reduce energy consumption by 8%, surpass reliability constraints by more than 25%, and generate cost savings by at least 15%.

Initiating communication and conveying critical information to the human operator is a key problem in human-automation collaboration. This problem is particularly pronounced in time-constrained safety critical domains such as in Air Traffic Management. A visual representation should aid operators understanding why the system initiates the communication, when the operator must act, and the consequences of not responding to the cue. Data glyphs can be used to present multidimensional data, including temporal data in a compact format to facilitate this type of communication. In this paper, we propose a glyph design for communication initialization for highly automated systems in Air Traffic Management, Vessel Traffic Service, and Train Traffic Management. The design was assessed by experts in these domains in three workshop sessions. The results showed that the number of glyphs to be presented simultaneously and the type of situation were domain-specific glyph design aspects that need to be adjusted for each work domain. The results also showed that the core of the glyph design could be reused between domains, and that the operators could successfully interpret the temporal data representations. We discuss similarities and differences in the applicability of the glyph design between the different domains, and finally, we provide some suggestions for future work based on the results from this study., This research was funded by the Swedish Transport Administration through project F AUTO (part I: TRV 2018/41347 and part II: TRV 2020/138317). Special thanks to Jimmy Johansson Westberg, who, during his time at Linköping University, participated in the concept development and early drafts of the paper., F AUTO part I TRV 2018/41347, F AUTO part II TRV 2020/138317

El treball de final de màster titulat "Smart City - Augmented Testbed" se centra en la creació d'una plataforma de realitat augmentada, basada en una infraestructura Edge-To-Cloud que possibilita la gestió de components i la visualització de dispositius reals, emulats i/o virtuals en un banc de proves enfocat en una Smart City. L'objectiu principal és facilitar a investigadors, empreses o clients particulars la possibilitat de crear, personalitzar i utilitzar testbeds. Gràcies a l'escalabilitat i dinamisme que ofereix el projecte, els usuaris poden observar el comportament dels dispositius de manera immersiva i realista, així com adaptar-se a les necessitats de cada producte o experiment. Des d'aquesta plataforma s'ofereixen tres funcionalitats complementàries perquè els usuaris puguin dur a terme les seves proves amb èxit. La primera consisteix en una interfície web que permet al gestor dels testbeds crear, modificar, visualitzar i eliminar els mateixos segons les necessitats dels usuaris. La segona funcionalitat implica la provisió de plantilles i codis d'exemple per a diversos tipus de dispositius presents en una ciutat intel·ligent, com vehicles, drons, semàfors, entre d'altres. Així mateix, també ofereix components que gestionen la capa d'intel·ligència d'una Smart City, com controladors capaços de conèixer la posició i estat dels dispositius i d'emetre ordres. Finalment, s'ofereix una aplicació de realitat augmentada que permet a l'usuari visualitzar l'escena, la qual mostrarà la interacció dels diferents dispositius en temps real, amb la possibilitat d'observar el seu comportament des de diferents angles. Per il·lustrar un exemple pràctic, considerem l'escenari de controlar el trànsit d'una ciutat mitjançant semàfors intel·ligents. En primer lloc, un gestor registra el testbed, després l'usuari executa una prova en un encreuament amb semàfors connectats al controlador que monitora la ubicació dels vehicles a la ciutat (generats mitjançant un algoritme de tràn

Laser Speckle Contrast Imaging (LSCI) is an imaging modality for real-time visualization of microcirculatory blood flow, commonly referred to as perfusion. The technology works by illuminating biological tissue with a laser and capturing the backscattered speckle pattern. The perfusion estimate is based on quantifying the image blurring that occurs due to the speckle pattern fluctuation. During captures, the imaged object has to be almost completely still, or there will be motion artifacts of overestimated perfusion data. PeriCam MultiFlow is a camera system based on LSCI technology that suffers from this limitation. In this thesis, we present an algorithm based on Shi-Tomasi corner detection and sparse Lucas-Kanade optical flow motion estimation, to predict in real-time whether motion artifacts are present in the perfusion images produced by MultiFlow. We evaluate the algorithm’s predictive performance as a binary classification problem using Matthew’s Correlation Coefficient (MCC) metric. We also assess the real-time capability by performing run-time measurements of the algorithm in isolation and when integrated into MultiFlow. We demonstrate that the algorithm performs better in certain cases, particularly on larger motions, while the algorithm struggles to differentiate between small motions and temporal noise. There are also performance differences between imaged objects. For instance, the algorithm is better at making predictions from small motions of a posterior hand than an anterior forearm or foot sole. However, the algorithm’s predictive performance can be consistently decent (MCC above 0.5) by choosing a balanced parameter configuration. Furthermore, we demonstrate that the algorithm is efficient enough for real-time processing in MultiFlow without impairing frame throughput from the camera., Laser-speckel-kontrastavbildning (LSCI) är en teknik för realtidsvisualisering av mikrocirkulatorisk blodflöde, ofta kallad perfusion. Tekniken fungerar genom att belysa vävnad med en laser och fånga det reflekterade dynamiska speckelmönstret. Perfusionsuppskattningen baseras på att kvantifiera bildoskärpan som uppstår på grund av variationer i speckelmönstret. Under en mätning måste det som avbildas vara nästan helt stilla annars kommer det att finnas rörelseartefakter av överskattade perfusionsvärden. PeriCam MultiFlow är en kameraapplikation baserad på LSCIteknik som har denna begränsning. I denna uppsats föreslår vi en algoritm baserad på Shi-Tomasi hörndetektion och Lucas-Kanade optiskt-flöde rörelseuppskattning för att förutsäga i realtid om rörelseartefakter är närvarande i de perfusionsbilder som produceras av MultiFlow. Vi utvärderar algoritmens prestanda som ett binärt klassificeringsproblem och använder Matthew’s Correlation Coefficient (MCC). Vi utvärderar också realtidskapaciteten genom att utföra körtidsmätningar av algoritmen både enskilt och integrerad i MultiFlow. Vi påvisar att algoritmens prestanda är hög i vissa fall, särskilt vid större rörelser. Algoritmen har däremot svårt att skilja brus och små rörelser åt. Det finns också variation mellan olika inspelningar. Algoritmen är bättre på att detektera små rörelser på inspelning av baksidan av handen än på inspelning av framsidan av underarmen eller fotsulan. Dock så är algoritmens allmänna prestanda hyfsad (MCC över 0.5) genom att välja en balanserad parameterkonfiguration. Vidare visade vi att algoritmen är tillräckligt effektiv för realtidsbehandling i MultiFlow utan att begränsa genomströmningen av bilder från kameran.

The rapid advancements in computer vision technology have paved the way for significant innovations across various fields, such as augmented reality, virtual reality and robotics. Part of these advancements is the generation of 3D scene graphs, which enable machines to semantically understand and structure 3D environments, to enhance perception of the physical world. This thesis explores the integration of real-time semantic segmentation into the Hydra 3D framework to construct dynamic and accurate 3D scene graphs. Real-time semantic segmentation is crucial for various applications, providing pixel-level scene understanding, essential for tasks such as obstacle avoidance in robotics, and enhancing augmented and virtual reality applications. The research investigates the challenges and potential of embedding state-of-the-art real-time semantic segmentation algorithms into the Hydra framework, which traditionally relies on preprocessed input data. By conducting this integration, the study aims to enhance the real-time capabilities of Hydra, enabling it to process live data and generate 3D scene graphs efficiently. Conducted in collaboration with the company “Ericsson”, this research compares the performance and suitability of various segmentation algorithms, seeking to optimize Hydra’s functionality for real-world applications. The study addresses three primary research questions, the integration methods of real-time segmentation within Hydra, the identification of suitable algorithms, and the performance comparison between the enhanced Hydra framework and its original configuration. The research provides a proof of concept for integrating real-time semantic segmentation into Hydra, demonstrating potential system improvements. The findings contribute to the broader field of computer vision, by proposing a robust framework for real-time 3D scene graph construction, with implications for advanced applications in the field of robotics. This work not only highlights the

The vertical jump test evaluates leg strength and explosiveness and has many applications, mainly in sports. The gold standard technologies for assessing jump test performance are force plates and optoelectronic motion capture systems. Their drawback is that they are expensive and non-portable. Previous research has found potential in inertial measurement units (IMU) as an affordable and portable alternative. However, applying this technology often involves manual data processing, which can be difficult without professional oversight. There is also a scarcity of studies regarding the optimal positioning of the IMU on the body to maximize the accuracy of measurements. This study therefore aimed to develop and validate a real-time, automated algorithm for vertical jump tests using IMU technology and investigating optimal sensor placement. The algorithm estimates and gives feedback on jump height, net impulse, and peak values of force, power, and velocity. For validation, 8 participants performed jump tests on force plates while wearing inertial sensors placed on the chest, hip, ankle, and heel. The algorithm identified jumps in real time with perfect or near-perfect accuracy across all sensor positions. The hip position provided the least systematic errors in its feedback values. However, the chest position exhibited the largest correlation to the gold standard force plate measures. Values measured from the chest also showed the highest within-session reliability, meaning that they were consistent across jumps performed by the same participant. The study shows potential for real-time usage of IMUs during jump tests when sensors are placed on the chest and correction equations are applied to deal with systematic bias., Vertikala hopptest mäter benstyrka och explosivitet och har många användningsområden, framförallt inom idrott. För att utvärdera ett hopptest används oftast kraftplattor och optoelektroniska motion capture system. Nackdelen med dessa är att de är dyra och icke-portabla. Forskning har visat potential för inertial measurement units (IMU) som ett prisvärt och portabelt alternativ. Denna metod kräver dock ofta manuell bearbetning av data, vilket kan vara svårt utan professionell översyn. Dessutom finns det få studier om var på kroppen en IMU bör placeras för bästa mätnoggrannhet. Denna studie syftade därför till att utveckla och validera en IMU-baserad automatiserad algoritm för vertikala hopptest, samt undersöka vilken den optimala sensorplaceringen är. Den utvecklade algoritmen beräknar och ger återkoppling på hopphöjd, impuls och toppvärden för kraft, effekt och hastighet. För att validera metoden genomfördes hopptest på kraftplattor med 8 deltagare som samtidigt bar IMU-sensorer på bröstet, höften, vristen och hälen. Algoritmen kunde identifiera hopp med perfekt eller nästintill perfekt precision från alla sensorpositioner. Höftpositionen gav minst systematiska fel i återkopplingsvärdena. Mätvärden från bröstpositionen uppvisade dock den starkaste korrelationen med kraftplattornas mätningar. Bröstpositionen hade också högst reliabilitet, vilket innebär att hopp genomförda av samma deltagare gav konsekventa värden. Studien visar potential för realtidsanvändning av IMU under hopptest, om sensorer placeras på bröstet och korrigeringsekvationer tillämpas för att hantera systematiska fel.

To catch up with the growing trend of parallelism, this thesis work focuses on the adaption of embedded real-time systems to a multicore platform. We use the embedded system of Xilinx ZCU-102, a multicore board, as an example of an embedded system without getting deep into its architecture. First, we deal with the tasks required to be able to make an embedded system operational and discuss why they are different from those for normal computer systems. The processes it takes to make a custom operating system for the given Xilinx embedded system are examined and patching the custom operating system along with customizing it is studied. We then take a look at related work in the field of benchmarking real-time systems and embedded systems and with a good understanding of related work propose a design similar to the related work for benchmarking embedded systems. The benchmarks we use run on multiple cores and aim at challenging the Xilinx board’s capabilities of running real-time tasks when the other cores on the board are occupied with performing independent tasks. We test the designed benchmarks on different conditions under two different operating systems of RT-Linux and Embedded Linux to study the differences between them. We then note how the RT-Linux would be a real upgrade for real-time systems if multicore operations are considered. The final result we have obtained is that core idling might decrease the performance of real-time tasks and RT-Linux might experience more interrupts but it is also better at recovering from interrupts.

As the world is evolving faster and technology gets cheaper and more powerful, simulations are proving to be more and more beneficial. Simulations allow for faster development with less lead time between iterations which means that product versions can be released more often and at less cost. The automotive industry is heavily affected by emission legislation and environmental politics. The development of more environmentally friendly engines forces the rate of development to speed up. Simulations allow the industry to keep up with the increasing requests for more complex systems. For software development, it speeds up the process significantly. Although real-life testing in engine test cells and test cars still sets the foundation for and validates the results of the simulations, simulations can drastically decrease the amount needed. Less unnecessary real-life testing with bugs in the software causing wasted time and cost. In the meantime, more iterations can be tested in a smaller time frame, making the actual real-life testing more valuable and giving. The aim of this thesis is to develop a method to convert an accurate but relatively slow simulation model of the airflow through an engine, into a faster-running format preparing it to run in close to Real-Time and with a fixed timestep. Then exporting the converted model as a functional mock-up unit, FMU, a standardised entity, part of the Functional Mock-up Interface standard. The FMU was to be used in Software in Loop, SiL, simulations using it in cooperation with Matlab Simulink and Synopsys Silver. The SiL environment is then to be used to develop and test calibration software for the engine in question.

As the world is evolving faster and technology gets cheaper and more powerful, simulations are proving to be more and more beneficial. Simulations allow for faster development with less lead time between iterations which means that product versions can be released more often and at less cost. The automotive industry is heavily affected by emission legislation and environmental politics. The development of more environmentally friendly engines forces the rate of development to speed up. Simulations allow the industry to keep up with the increasing requests for more complex systems. For software development, it speeds up the process significantly. Although real-life testing in engine test cells and test cars still sets the foundation for and validates the results of the simulations, simulations can drastically decrease the amount needed. Less unnecessary real-life testing with bugs in the software causing wasted time and cost. In the meantime, more iterations can be tested in a smaller time frame, making the actual real-life testing more valuable and giving. The aim of this thesis is to develop a method to convert an accurate but relatively slow simulation model of the airflow through an engine, into a faster-running format preparing it to run in close to Real-Time and with a fixed timestep. Then exporting the converted model as a functional mock-up unit, FMU, a standardised entity, part of the Functional Mock-up Interface standard. The FMU was to be used in Software in Loop, SiL, simulations using it in cooperation with Matlab Simulink and Synopsys Silver. The SiL environment is then to be used to develop and test calibration software for the engine in question.

To catch up with the growing trend of parallelism, this thesis work focuses on the adaption of embedded real-time systems to a multicore platform. We use the embedded system of Xilinx ZCU-102, a multicore board, as an example of an embedded system without getting deep into its architecture. First, we deal with the tasks required to be able to make an embedded system operational and discuss why they are different from those for normal computer systems. The processes it takes to make a custom operating system for the given Xilinx embedded system are examined and patching the custom operating system along with customizing it is studied. We then take a look at related work in the field of benchmarking real-time systems and embedded systems and with a good understanding of related work propose a design similar to the related work for benchmarking embedded systems. The benchmarks we use run on multiple cores and aim at challenging the Xilinx board’s capabilities of running real-time tasks when the other cores on the board are occupied with performing independent tasks. We test the designed benchmarks on different conditions under two different operating systems of RT-Linux and Embedded Linux to study the differences between them. We then note how the RT-Linux would be a real upgrade for real-time systems if multicore operations are considered. The final result we have obtained is that core idling might decrease the performance of real-time tasks and RT-Linux might experience more interrupts but it is also better at recovering from interrupts.

Estimating human posture from an image or video is an essential task in computer vision. This task has detected body key points from a camera for body posture and gesture recognition technology, which enables the following applications: assisted living in the case of fall detection, yoga pose identification, character animation, and an autonomous drone control system. The rapid development of AI-based posture estimation algorithms for picture recognition has resulted in the availability of quick and dependable solutions for recognizing the human body joint in collected films. One major issue in human posture assessment is the system’s capacity to perform with high accuracy in real-time under shifting ambient conditions. The ultimate goal of the proposed transfer learning-based posture estimation assignment is to achieve real-time speed with virtually no drop accuracy. In this research paper, assisted living program (ALP) is implemented by using a single-shot deep estimation network and a pose key points angular feature. Experimental results show that transfer learning-based pose identifies and estimates posture with a frame rate of about 30 frames per second and a detection accuracy rate of 96.81%.

In the landscape of robotics applications, the pivotal role of occupancy grids in preserving environmental data cannot be overstated. This significance becomes particularly pronounced in the realm of Simultaneous Localization and Mapping (SLAM), where mobile robots dynamically navigate and scan their surroundings. Addressing the challenges posed by loop closure detection, FRAGG-Map—an innovative 3D grid expressly crafted to optimize efficiency during sustained, long-term exploration. Diverging from conventional methods, FRAGG-Map minimizes the necessity for map reconstruction at loop closures, thereby enhancing overall operational efficiency., Peer Reviewed

The detection of small targets remains a critical challenge within the field of image processing. Traditional techniques, such as image subtraction with frame-to-frame registration, suffer from high false alarm rates. Even state-of-the-art deep learning architectures, like YOLO and Masked R-CNN, exhibit limitations in this domain. In overextended distances, the inherent feature quality of small targets degrades significantly, leading to a scarcity of informative data for conventional detection algorithms. Consequently, accurate visual recognition becomes a particularly hard task.This work presents a novel detection approach that draws inspiration from the human visual attention mechanism. By leveraging dense optical flow, the model prioritizes moving objects within the scene, facilitating effective target detection. Furthermore, the proposed method employs K-Means clustering to achieve robust foreground-background separation based on color intensity characteristics. To address the limitations of dense optical flow with stationary targets, a dedicated tracking algorithm is also introduced. Our approach demonstrated a high level of accuracy (98%) when evaluated on unseen test data. Additionally, the algorithm functioned in real-time, enabling immediate processing.

Estimating human posture from an image or video is an essential task in computer vision. This task has detected body key points from a camera for body posture and gesture recognition technology, which enables the following applications: assisted living in the case of fall detection, yoga pose identification, character animation, and an autonomous drone control system. The rapid development of AI-based posture estimation algorithms for picture recognition has resulted in the availability of quick and dependable solutions for recognizing the human body joint in collected films. One major issue in human posture assessment is the system’s capacity to perform with high accuracy in real-time under shifting ambient conditions. The ultimate goal of the proposed transfer learning-based posture estimation assignment is to achieve real-time speed with virtually no drop accuracy. In this research paper, assisted living program (ALP) is implemented by using a single-shot deep estimation network and a pose key points angular feature. Experimental results show that transfer learning-based pose identifies and estimates posture with a frame rate of about 30 frames per second and a detection accuracy rate of 96.81%.

The detection of small targets remains a critical challenge within the field of image processing. Traditional techniques, such as image subtraction with frame-to-frame registration, suffer from high false alarm rates. Even state-of-the-art deep learning architectures, like YOLO and Masked R-CNN, exhibit limitations in this domain. In overextended distances, the inherent feature quality of small targets degrades significantly, leading to a scarcity of informative data for conventional detection algorithms. Consequently, accurate visual recognition becomes a particularly hard task.This work presents a novel detection approach that draws inspiration from the human visual attention mechanism. By leveraging dense optical flow, the model prioritizes moving objects within the scene, facilitating effective target detection. Furthermore, the proposed method employs K-Means clustering to achieve robust foreground-background separation based on color intensity characteristics. To address the limitations of dense optical flow with stationary targets, a dedicated tracking algorithm is also introduced. Our approach demonstrated a high level of accuracy (98%) when evaluated on unseen test data. Additionally, the algorithm functioned in real-time, enabling immediate processing.

Monitoring crop phenology is essential for managing field disasters, protecting the environment, and making decisions about agricultural productivity. Because of its high timeliness, high resolution, great penetration, and sensitivity to specific structural elements, synthetic aperture radar (SAR) is a valuable technique for crop phenology estimation. Particle filtering (PF) belongs to the family of dynamical approach and has the ability to predict crop phenology with SAR data in real time. The observation equation is a key factor affecting the accuracy of particle filtering estimation and depends on fitting. Compared to the common polynomial fitting (POLY), machine learning methods can automatically learn features and handle complex data structures, offering greater flexibility and generalization capabilities. Therefore, incorporating two ensemble learning algorithms consisting of support vector machine regression (SVR), random forest regression (RFR), respectively, we proposed two machine learning-aided particle filtering approaches (PF-SVR, PF-RFR) to estimate crop phenology. One year of time-series Sentinel-1 GRD SAR data in 2017 covering rice fields in Sevilla region in Spain was used for establishing the observation and prediction equations, and the other year of data in 2018 was used for validating the prediction accuracy of PF methods. Four polarization features (VV, VH, VH/VV and Radar Vegetation Index (RVI)) were exploited as the observations in modeling. Experimental results reveals that the machine learning-aided methods are superior than the PF-POLY method. The PF-SVR exhibited better performance than the PF-RFR and PF-POLY methods. The optimal outcome from PF-SVR yielded a root-mean-square error (RMSE) of 7.79, compared to 7.94 for PF-RFR and 9.1 for PF-POLY. Moreover, the results suggest that the RVI is generally more sensitive than other features to crop phenology and the performance of polarization features presented consistent among all methods, i.

Flow sensing exhibits significant potential for monitoring, controlling, and optimizing processes in industries, resource management, and environmental protection. However, achieving wireless real-time and omnidirectional sensing of gas/liquid flow on a simple, self-contained device without external power support has remained a formidable challenge. In this study, a compact-sized, fully self-powered wireless sensing flowmeter (CSWF) is introduced with a small size diameter of down to less than 50 mm, which can transmit real-time and omnidirectional wireless signals, as driven by a rotating triboelectric nanogenerator (R-TENG). The R-TENG triggers the breakdown discharge of a gas discharge tube (GDT), which enables flow rate wireless sensing through emitted electromagnetic waves. Importantly, the performance of the CSWF is not affected by the R-TENG's varied output, while the transmission distance is greater than 10 m. Real-time wireless remote monitoring of wind speed and water flow rate is successfully demonstrated. This research introduces an approach to achieve a wireless, self-powered environmental monitoring system with a diverse range of potential applications, including prolonged meteorological observations, marine environment monitoring, early warning systems for natural disasters, and remote ecosystem monitoring. © 2024 Wiley-VCH GmbH.

The introduction of modern connected vehicles has led to increased safety and comfort. It also introduces new attack surfaces that could be exploited by malicious actors during an attack targeting the electronic control units (ECU) in the vehicle. The ECUs control safety-critical components in the vehicle, and a successful attack could lead to catastrophic consequences. The most common network type for inter-ECU communication is CAN-bus, which is a reliable and lightweight protocol, but it was never intended to withstand cyber security threats.This master thesis explores different approaches to CAN-bus intrusion detection systems (IDS) and evaluates them in terms of capability to detect previously unknown attacks (zero-day), real-time performance and implementation feasibility on a typical embedded device. To the best of our knowledge, these requirements have not been evaluated together in previous research. High-level features are created and evaluated in order to include as many aspects of the data as possible. The different approaches are evaluated by exposing them to attacks commonly used in previous research, as well as a data-altering attack introduced in the thesis.The experiments show that the best-performing approach is to model the behaviour of every single ECU on the network with a separate support vector machine (SVM) and a set of high-level features that capture the timing and data payload aspects of CAN-bus traffic. This approach achieves a detection rate of more than 99% and a false positive rate during normal operation below 0.01% in the majority of cases. More long-term features are also explored, but they do not conform to the real-time requirements.

Denna studie utforskar möjligheterna och utmaningarna som är förknippade med implementeringen av AI-drivet ansiktsigenkänningssystem i realtid inom svensk brottsbekämpning, i kontexten av den nya EU-lagstiftningen. Viktiga möjligheter inkluderar förbättrad effektivitet i brottsbekämpande verksamheter genom mer effektiva övervaknings- och identifieringsprocesser. Utmaningar såsom övertro till teknologi, black box-effekter och etiska dilemman utgör dock kritiska problem. Metodiken för denna studie innefattar kvalitativa semistrukturerade intervjuer med både polisofficerare och AI-forskare, vilket ger en mer heltäckande bild av användningen av ansiktsigenkänningssystem. Resultaten indikerar en komplexitet där tekniska framsteg kan avsevärt förbättra brottsförebyggande och allmän säkerhet, men även medföra betydande risker för integritet och rättvisa. Studien avslutas med en diskussion om behovet av en balans mellan att förbättra brottsbekämpningens kapacitet och skydda grundläggande mänskliga rättigheter, samt rekommenderar starka riktlinjer och transparens för att säkerställa etisk användning av ansiktsigenkänningsteknologier., This study explores the opportunities and challenges associated with the implementation of AI-driven real-time facial recognition technology within Swedish law enforcement, in the context of the new EU legislation. Key opportunities include enhanced efficiency in law enforcement practices through more effective surveillance and identification processes. However, challenges such as overreliance on technology, black box-effects, and ethical dilemmas are critical concerns. The methodology for this study involves qualitative semi-structured interviews with both police officers and AI researchers, providing a comprehensive view of the deployment of facial recognition systems. Findings indicate a complex landscape where technological advances can significantly boost crime prevention and public security, yet pose substantial risks to privacy and fairness. The thesis concludes by discussing the need for a delicate balance between enhancing law enforcement capabilities and safeguarding fundamental human rights, recommending strong guidelines and continuous oversight to ensure ethical usage of facial recognition technologies.

With an ever worsening geopolitical situation, not least due to Russia’s offensive war in Ukraine, the need to detect enemy movement and disrupt their detection capabilities has increased in defensive value. Using real equipment to test different strategies and placements of radar systems is both costly and time consuming, and this thesis therefore investigates methods of simulating large-scale scenarios of electronic warfare in real-time. The proposed methods include using approximations, multithreading, simplified signal representations and fast convolutions. The results show that if high efficiency is required, the developed simulation structure with a simplified signal representation is able to process one million signals every 50 ms which indicates that realistic and large-scale simulations of this kind is possible. If a more realistic approach is required, a representation using sampled signals is proposed which also utilizes the GPU, though results show that the efficiency drops to around 185.1162 s per one million signals.

Networks which use real-time communication have high requirements on latency and packet loss. Improving one aspect may results in worse performance for another, and it can be difficult to prioritize one over the other as all the requirements need to be met in order for the network tooperate as expected. Many studies have investigated reducing the size of packet buffers to improve the latency. However, they have mainly focused on studying TCP traffic which may not be optimal for real-time traffic, where it instead could be more suitable to use UDP. We have performed an experiment where we compared the performance of real-time traffic over multiple different buffer sizes. We generated traffic using synchronized bursts of packets which were either sample value (SV) or IP packets, as defined by IEC 61850. We measured the packet loss and latency for situations where the traffic was either entirely composed of SV packets, or when it had mixed SV and IP traffic. For the mixed traffic, we also experimented with using different VLAN priorities for the two types of packets. We have determined deadline thresholds that show what size of packet buffer will start causing packets to miss their deadline, and what size will lead every packet in bursts oftraffic to miss their deadlines. We also found that increasing the priority of SV packets in mixed traffic can have either a positive or a negative impact on their performance, depending on how highly they are prioritized.

Container-based virtualization is a promising deployment model in fog and edge computing applications, because it allows a seamless co-existence of virtualized applications in a heterogeneous environment without introducing significant overhead. Certain application domains (e.g., industrial automation, automotive, or aerospace) mandate that applications exhibit a certain degree of temporal predictability. Container-based virtualization cannot be easily used for such applications, since the technology is not designed to support real-time properties and handle temporal disturbances. This article proposes a framework consisting of a static offline and a dynamic online phase for resource allocation and adaptive re-dimensioning of real-time containers. In the offline phase, the optimal initial deployment and dimensioning of containers are decided based on ideal system models. Additionally, to adapt to dynamic variations caused by changing workloads or interferences, the online phase adapts the CPU usage and limits of real-time containers at runtime to improve the real-time behavior of the real-time containerized applications while optimizing resource usage. We implement the framework in a real Linux-based system and showthrough a series of experiments that the proposed framework is able to adjust and re-distribute computing resources between containers to improve the real-time behavior of containerized applications in the presence of temporal disturbances while optimizing resource usage.

All necessary Data to recreate the published plots and images in the publication: "Absolute energy-dependent scintillating screen calibration for real-time detection of laser-accelerated proton bunches". Included are the raw scintillating screen images, the plotting data and Python Scripts used for calculations and plotting.

Robust real-time moving object segmentation (MOS) in LiDAR point clouds is crucial for generating high-quality 3D maps and gaining vital information for path-planning and object avoidance. This thesis focuses on MOS within forest environments, particularly for drone applications, where the available computa- tional resources are limited. The study evaluates state-of-the-art methods with a focus on mean Intersection over Union (mIoU), the real-time capabilities when run on an Nvidia AGX Orin, and their generalization across different environ- ments. It also investigates the effectiveness of synthetic data generation and trans- fer learning for adaptation to forest scenarios, where labeled data is scarce. A variety of methods for MOS in LiDAR point clouds are compared, encom- passing the most common data representations, including range-view (LMNet, MotionSeg3D), bird’s-eye-view (MotionBEV, M-Detector), and voxel-based (Map- MOS, 4DMOS, Dynablox) representations. Both learning and non-learning meth- ods are considered. The non-learning-based methods, Dynablox and M-Detector, meet the real-time threshold of 100 ms. Furthermore, the learning-based meth- ods, LMNet and MotionBEV, achieve real-time performance using bird’s-eye and range-view approaches, respectively. In contrast, the learning-based volumetric methods, 4DMOS and MapMOS, are an order of magnitude away from the real- time threshold, while MotionSeg3D is marginally close. The results indicate that range-view and bird’s-eye-based methods exhibit poor generalization capabilities from urban to forest environments, while volumetric and non-learning-based methods adapt significantly better. The non-learning- based method Dynablox shows the best generalization performance, with an mIoU of 72.4% over all the test sequences from a forest environment. The study underscores the significant improvements obtained through synthetic data generation and fine-tuning. These techniques substantially enhance the per- formanc

The introduction of modern connected vehicles has led to increased safety and comfort. It also introduces new attack surfaces that could be exploited by malicious actors during an attack targeting the electronic control units (ECU) in the vehicle. The ECUs control safety-critical components in the vehicle, and a successful attack could lead to catastrophic consequences. The most common network type for inter-ECU communication is CAN-bus, which is a reliable and lightweight protocol, but it was never intended to withstand cyber security threats.This master thesis explores different approaches to CAN-bus intrusion detection systems (IDS) and evaluates them in terms of capability to detect previously unknown attacks (zero-day), real-time performance and implementation feasibility on a typical embedded device. To the best of our knowledge, these requirements have not been evaluated together in previous research. High-level features are created and evaluated in order to include as many aspects of the data as possible. The different approaches are evaluated by exposing them to attacks commonly used in previous research, as well as a data-altering attack introduced in the thesis.The experiments show that the best-performing approach is to model the behaviour of every single ECU on the network with a separate support vector machine (SVM) and a set of high-level features that capture the timing and data payload aspects of CAN-bus traffic. This approach achieves a detection rate of more than 99% and a false positive rate during normal operation below 0.01% in the majority of cases. More long-term features are also explored, but they do not conform to the real-time requirements.

With an ever worsening geopolitical situation, not least due to Russia’s offensive war in Ukraine, the need to detect enemy movement and disrupt their detection capabilities has increased in defensive value. Using real equipment to test different strategies and placements of radar systems is both costly and time consuming, and this thesis therefore investigates methods of simulating large-scale scenarios of electronic warfare in real-time. The proposed methods include using approximations, multithreading, simplified signal representations and fast convolutions. The results show that if high efficiency is required, the developed simulation structure with a simplified signal representation is able to process one million signals every 50 ms which indicates that realistic and large-scale simulations of this kind is possible. If a more realistic approach is required, a representation using sampled signals is proposed which also utilizes the GPU, though results show that the efficiency drops to around 185.1162 s per one million signals.

Denna studie utforskar möjligheterna och utmaningarna som är förknippade med implementeringen av AI-drivet ansiktsigenkänningssystem i realtid inom svensk brottsbekämpning, i kontexten av den nya EU-lagstiftningen. Viktiga möjligheter inkluderar förbättrad effektivitet i brottsbekämpande verksamheter genom mer effektiva övervaknings- och identifieringsprocesser. Utmaningar såsom övertro till teknologi, black box-effekter och etiska dilemman utgör dock kritiska problem. Metodiken för denna studie innefattar kvalitativa semistrukturerade intervjuer med både polisofficerare och AI-forskare, vilket ger en mer heltäckande bild av användningen av ansiktsigenkänningssystem. Resultaten indikerar en komplexitet där tekniska framsteg kan avsevärt förbättra brottsförebyggande och allmän säkerhet, men även medföra betydande risker för integritet och rättvisa. Studien avslutas med en diskussion om behovet av en balans mellan att förbättra brottsbekämpningens kapacitet och skydda grundläggande mänskliga rättigheter, samt rekommenderar starka riktlinjer och transparens för att säkerställa etisk användning av ansiktsigenkänningsteknologier., This study explores the opportunities and challenges associated with the implementation of AI-driven real-time facial recognition technology within Swedish law enforcement, in the context of the new EU legislation. Key opportunities include enhanced efficiency in law enforcement practices through more effective surveillance and identification processes. However, challenges such as overreliance on technology, black box-effects, and ethical dilemmas are critical concerns. The methodology for this study involves qualitative semi-structured interviews with both police officers and AI researchers, providing a comprehensive view of the deployment of facial recognition systems. Findings indicate a complex landscape where technological advances can significantly boost crime prevention and public security, yet pose substantial risks to privacy and fairness. The thesis concludes by discussing the need for a delicate balance between enhancing law enforcement capabilities and safeguarding fundamental human rights, recommending strong guidelines and continuous oversight to ensure ethical usage of facial recognition technologies.

This thesis presents an empirical analysis of the performance of WebTransport and WebSocket using the Socket.IO framework under varying network conditions. The thesis investigates two primary experiments, message round trip time and multiplexing performance, both conducted under three scenarios—perfect network (0% packet loss), moderate network (5% packet loss), and poor network (15% packet loss). The experiments were conducted on a system running Fedora 40 with a 12th Gen Intel Core i5-12600K processor and 16 GB of memory. The findings demonstrate that WebTransport mitigates Head-of-Line (HOL) blocking more effectively than WebSocket, especially in environments with degraded networks. WebTransport (Reliable) provided a significant performance advantage under suboptimal network conditions when measuring message round trip time, while WebTransport (Unreliable) consistently demonstrated slightly lower latency but with some data loss. When multiplexing, WebTransport showed significant performance benefits compared to using WebSocket.

This thesis explores the design and implementation of a simulation framework created for C2 training environments, focusing on simulating the movement of aircraft and other moving targets between waypoints. The core objective was to develop an extensible, reliable, and realistic simulation system that can adapt to simulate a variety of entities, each characterized by unique movement patterns defined by their type. The simulated input scenarios involve a collection of targets, where each target traverses through a defined set of waypoints according to their specified movement behavior at initialization. The research involved a evaluation of established theories and ideologies of both aircraft and computer simulation, leading to the creation of a framework that leverages Java for its scalability and performance. The framework was designed from the ground up to promote extensibility of various types of moving targets beyond of just that of aircraft, such as helicopters, boats, missiles, and surface vehicles. The accuracy of the simulation was tested by comparing its output to real-world flight data, demonstrating its capability to mimic realistic aircraft movement within predefined training scenarios. Additionally, the system's extensibility is showcased by simulating two different categories of aircraft—a commercial airliner and a small fighter jet—and illustrating the differences in their generated trajectories. Another outcome is the framework's ability to support real-time simulation, crucial for training scenarios in a C2 context. The system integrates and utilizes concurrent design techniques and dependable design patterns to ensure high throughput and performance, even when scaling to large numbers of targets. The framework's architecture allows for future extension and integration of new target types, without requiring modifications to the existing codebase, by the utilization of Javas SPI.

Networks which use real-time communication have high requirements on latency and packet loss. Improving one aspect may results in worse performance for another, and it can be difficult to prioritize one over the other as all the requirements need to be met in order for the network tooperate as expected. Many studies have investigated reducing the size of packet buffers to improve the latency. However, they have mainly focused on studying TCP traffic which may not be optimal for real-time traffic, where it instead could be more suitable to use UDP. We have performed an experiment where we compared the performance of real-time traffic over multiple different buffer sizes. We generated traffic using synchronized bursts of packets which were either sample value (SV) or IP packets, as defined by IEC 61850. We measured the packet loss and latency for situations where the traffic was either entirely composed of SV packets, or when it had mixed SV and IP traffic. For the mixed traffic, we also experimented with using different VLAN priorities for the two types of packets. We have determined deadline thresholds that show what size of packet buffer will start causing packets to miss their deadline, and what size will lead every packet in bursts oftraffic to miss their deadlines. We also found that increasing the priority of SV packets in mixed traffic can have either a positive or a negative impact on their performance, depending on how highly they are prioritized.

Light Detection and Ranging (Lidar) is a kind of active sensor that emits a laser pulse and primarily measures the time of flight of the returning pulse and uses it to construct a 3D point cloud of the scene around the lidar sensor. The constructed point cloud is an essential asset for the control of autonomous vehicles, and especially today, an essential basis for the training of autonomous vehicle control models. However, it remains time-consuming, high-risk and expensive to acquire the amounts of data necessary to train the rather complex modern control models. As such, generating the point cloud through simulations becomes a natural solution. Yet, many lidar simulations today produce ideal point clouds, corrected only by random noise, without considering the physical reasons behind the imperfections visible in real lidar point clouds. The aim of this study was to investigate real-time simulation models for disturbances that may cause imperfections in lidar data. From a base investigation of lidar, disturbances were found, models were investigated and finally a real-time implementation of Atmospheric Effects and attenuation from Beam Divergence was evaluated. It was found that the implemented models could produce physically accurate lidar point placement while keeping the computational time low enough for real-time evaluation. However, to achieve correct separation of target hit rates under Atmospheric Effects, as high as 34% of the points had to be dropped. Additionally, the intensity of the return points could not be properly verified. From these results it can be concluded that, with additional verification and adjustment, the presented models can achieve good results for evaluation in real-time. The results of this study thus serve as a support for future developments of realistic real-time lidar simulations, for use in development of autonomous vehicle control models and implementation of digital twins.

Replicating a real-world environment is crucial for creating simulations, computer vision, global and local path planning, and localization. While computer-aided design software is a standard tool for such a task, it may not always be practical or effective. An alternative approach is mapping, which uses sensory input and computer vision technologies to reconstruct the environment. However, developing such software requires knowledge of various fields, making it a challenging task. This thesis deep-dives into a state-of-the-art mapping framework and explores potential improvements, providing a foundation for an open-source project. The resulting software can replicate a real-world environment while storing panoptic classification data on a voxel level. Through 3D object matching and probability theory, the mapping software is resilient to object misclassifications and retains consistency in the different instances of observed objects. The final software is designed to make it easy to use in a different project by substituting the simulation data with a semantic, instance, or panoptic segmentation model. Additionally, the software integrates certain functionalities that facilitate the visualization of diverse classes or a particular class instance., Att replikera en verklig miljö är avgörande för att skapa simuleringar, datorseende, global och lokal vägplanering samt lokalisering. Trots att ett datorstött designprogram är ett standardverktyg för sådana uppgifter kanske det inte alltid är praktiskt eller effektivt. Ett alternativt tillvägagångssätt är kartläggning, som använder sensorisk input och datorseendeteknik för att uppnå reskonstruering av omgivningar. Att utveckla sådan programvara kräver dock kunskap inom olika områden, vilket gör det till en utmanande uppgift. Den här avhandlingen fördjupar sig i ett toppmodernt kartläggningsramverk och utforskar potentiella förbättringar, vilket ger en grund för ett öppet källkodsprojekt. Resultatet av denna avhandling är en programvara som kan replikera en verklig miljö samtidigt som den lagrar panoptisk klassificeringsdata på en voxelnivå. Genom 3D-objektmatchning och sannolikhetsteori är kartläggningsprogramvaran motståndskraftig mot felaktiga objektklassificeringar och är koncekvent avseende förekomsten av olika observerade objekt. Den slutliga programvaran är utformad med fokus på att göra den enkel att använda i andra projekt genom att ersätta simuleringsdata med en semantisk, instans eller panoptisk segmenteringsmodell. Dessutom integrerar programvaran funktioner som underlättar visualiseringen av antingen olika klasser eller en specifik instans av en klass.

Robust real-time moving object segmentation (MOS) in LiDAR point clouds is crucial for generating high-quality 3D maps and gaining vital information for path-planning and object avoidance. This thesis focuses on MOS within forest environments, particularly for drone applications, where the available computa- tional resources are limited. The study evaluates state-of-the-art methods with a focus on mean Intersection over Union (mIoU), the real-time capabilities when run on an Nvidia AGX Orin, and their generalization across different environ- ments. It also investigates the effectiveness of synthetic data generation and trans- fer learning for adaptation to forest scenarios, where labeled data is scarce. A variety of methods for MOS in LiDAR point clouds are compared, encom- passing the most common data representations, including range-view (LMNet, MotionSeg3D), bird’s-eye-view (MotionBEV, M-Detector), and voxel-based (Map- MOS, 4DMOS, Dynablox) representations. Both learning and non-learning meth- ods are considered. The non-learning-based methods, Dynablox and M-Detector, meet the real-time threshold of 100 ms. Furthermore, the learning-based meth- ods, LMNet and MotionBEV, achieve real-time performance using bird’s-eye and range-view approaches, respectively. In contrast, the learning-based volumetric methods, 4DMOS and MapMOS, are an order of magnitude away from the real- time threshold, while MotionSeg3D is marginally close. The results indicate that range-view and bird’s-eye-based methods exhibit poor generalization capabilities from urban to forest environments, while volumetric and non-learning-based methods adapt significantly better. The non-learning- based method Dynablox shows the best generalization performance, with an mIoU of 72.4% over all the test sequences from a forest environment. The study underscores the significant improvements obtained through synthetic data generation and fine-tuning. These techniques substantially enhance the per- formanc

Od ledna 2023 je pro registrované uživatele k dispozici služba Galileo High Accuracy Service (HAS). V rámci HAS mohou uživatelé získat korekce hodin a oběžné dráhy, stejně jako zkreslení kódu pro konstelace Galileo a GPS. Galileo HAS pokrývá téměř celou zeměkouli a je dalším systémem po japonském CLAS (Japanese Centimeter Level Augmentation Service) a čínském PPP-B2b, který poskytuje korekce prostřednictvím satelitního signálu. Tyto korekce Galileo jsou přenášeny v signálu Galileo E6, ale také prostřednictvím internetu. Oficiální dokument definice služby HAS uvádí, že přesnost polohování je lepší než 25 cm a lepší než 30 cm pro horizontální a vertikální komponenty. Korekce HAS lze použít v různých aplikacích v reálném čase bez připojení k internetu. Předložená práce ukazuje využití korekcí Galileo HAS v různých geovědních aplikacích jako je určování polohy, časování, monitorování troposféry nebo detekce koseismických vibrací. Kromě toho je kvalita přenášené korekce HAS také ověřována pomocí parametru SISRE. Výpočty se provádějí v in-house vyvinutém softwaru GNSS-WARP s korekcemi Internet HAS uloženými v souborech ASCII pomocí softwaru BKG Ntrip Client. Kvalita Galileo HAS korekcí pomocí parametru SISRE činí 8,6 cm pro systém GPS a 13,0 cm pro Galileo. Experiment s PPP (Precise Point Positioning) odhalil sub-decimetrovou a centimetrovou přesnost v kinematickém a statickém přístupu. Pokud jde o přesnost časování, Galileo HAS zajišťuje přesnost jedné nanosekundy. Odhad Zenith Tropospheric Delay s využitím vyhodnocených korekcí splnil požadavek monitorování troposféry a byla možná i detekce koseismických vibrací., Since January 2023 the Galileo High Accuracy Service (HAS) is available to registered users. Within the HAS users can retrieve clock and orbit corrections, as well as code biases for Galileo and GPS constellations. Galileo HAS covers almost the entire globe and is the next system, after the Japanese CLAS (Japanese Centimeter Level Augmentation Service) and the Chinese PPP-B2b, that provides corrections via a satellite signal. These Galileo corrections are transmitted in the Galileo E6 signal but also via the Internet stream. The official HAS Service Definition Document indicates that the positioning accuracy is better than 25 cm and better than 30 cm for horizontal and vertical components, respectively. HAS corrections can be used in difference real-time applications without Internet connection. The presented work shows the use of Galileo HAS corrections in various geoscience applications such as positioning, timing, troposphere monitoring or coseismic vibrations detection. Additionally, the quality of the transmitted HAS correction is also verified using the SISRE parameter. The computations are carried out in the in-house developed GNSS-WARP software with the Internet HAS corrections saved in the ASCII files by the BKG Ntrip Client software. The Galileo HAS corrections quality using SISRE parameter equal to 8.6 cm and 13.0 cm for GPS and Galileo respectively. The experiment with Precise Point Positioning reveal sub-decimeter and centimeters accuracy in kinematic and static approach, respectively. In terms of timing precision Galileo HAS ensure precision of single nanosecond. Zenith Tropospheric Delay estimation with the use of evaluated corrections meet the requirement of troposphere monitoring and coseismic vibrations detection was also possible.

Rozvoj družicové techniky v oblasti přesného určování polohy a dostupnost mobilních zařízení s vestavěným přijímačem multi-GNSS (Global Navigation Satellite System) nám umožňuje určovat polohy s větší přesností než před několika lety. Zároveň ze strany uživatelů roste poptávka po stále přesnějším určování polohy při snižování nákladů na finální zařízení (např. navigační systém pro autonomní sekačky na trávu). Tato práce představuje koncept řešení, které integruje různé techniky určování polohy: gyroskopy, počítadla kilometrů a GNSS pomocí volně vázaného Kalmanova filtru. Model integrace senzoru byl vyvinut pro horizontální komponenty (2D) se současným určením azimutu a přesné polohy. Integrační filtr navíc využívá dynamickou matici vah, jejíž hodnoty se volí v závislosti na typu řešení GNSS (FIX, FLOAT atd.). Zkoušky měřící plošiny byly prováděny za ideálních podmínek a v místech zastřeného horizontu (průjezdy podél vysoké zdi, pod prolamovanými ocelovými vazníky apod.). Přesnosti získané během testů pomocí EKF a dynamické hmotnostní matice jsou RMS 0,019 m pro severní a východní komponenty. Přesnost určení azimutu však byla 0,59°. Prezentace navíc představí koncept platformy, která integruje levné senzory v reálném čase. Tuto práci podpořila Wroclaw University of Environmental and Life Sciences (Polsko) v rámci projektu „POMOST“ (grant č. N110/0002/22)., The development of satellite techniques in the field of precise positioning and the availability of mobile devices with a built-in multi-GNSS (Global Navigation Satellite System) receiver allow us to determine positions with better accuracy than a dozen or so years ago. At the same time, there is a growing demand from users for more and more accurate positioning while reducing the costs of the final device (e.g., a navigation system for autonomous lawn mowers). This work presents the concept of a solution that integrates different positioning techniques: gyroscopes, odometers, and GNSS using a loosely coupled Kalman filter. The sensor integration model was developed for horizontal components (2D) with simultaneous determination of azimuth and precise position. Additionally, the integration filter uses a dynamic weight matrix, the values of which are selected depending on the type of GNSS solution (FIX, FLOAT, etc.). Tests of the measurement platform were carried out under ideal conditions and in places where the horizon is obscured (passes along a high wall, under openwork steel trusses, etc.). The accuracies obtained during tests using EKF and the dynamic weight matrix are RMS 0.019 m for the North and East components. However, the accuracy of the azimuth determination was 0.59°. In addition, the presentation will present a platform concept that integrates low-cost sensors in real time. This work was supported by Wroclaw University of Environmental and Life Sciences (Poland) under the project “POMOST” (grant no. N110/0002/22).

Indoor Positioning Systems (IPSs) have multiple applications. For example, they can be used to guide people, to locate items in a warehouse and to support the navigation of Automated Guided Vehicles (AGV). Currently most AGVs use local pre-defined navigation systems, but they lack a global localisation system. Integrating both systems is uncommon due to the inherent challenge in balancing accuracy with coverage. Visible Light Position (VLP) offers accurate and fast localisation, but it encounters scalability limitations. To overcome this, this paper presents a novel Image Sensor-based VLP (IS-VLP) identification method that harnesses existing Light Emitting Diode (LED) lighting infrastructure to substitute both navigation and localisation systems effectively in the whole area. We developed an IPS that achieves six-axis positioning at 90 Hz refresh rate using OpenCV’s solvePnP algorithm and embedded computing. This IPS has been validated in a laboratory environment and successfully deployed in a real factory to position an operative AGV. The system has resulted in accuracies better than 12 cm for 95% of the measurements. This work advances towards positioning VLP as an appealing choice for IPS in industrial environments, offering an inexpensive, scalable, accurate and robust solution., This research received funding from the Departament de Recerca de Universitats de la Generalitat de Catalunya under grant 2021 SGR 01524, from the European Union’s Horizon 2020 research and innovation programme under Grant Agreement No. 957246 (project IoT-NGIN) and from the Spanish MCIN/AEI/10.13039/501100011033 through project PID2019-106808RAI00., Peer Reviewed, Postprint (published version)

Parameter estimation is crucial for polymer electrolyte membrane fuel cell monitoring and control. Nonetheless, most parameter estimation algorithms rely on a persistence of excitation condition, which is rarely satisfied and not convenient in fuel cell systems. For this reason, this work presents and compares three algorithms to estimate in real-time some critical PEMFC parameters in the voltage equation: the ohmic resistance, the charge transfer coefficient and the oxygen activity of a proton exchange fuel cell. The first algorithm is a standard gradient descent, while the other two are based on a set of pre-preprocessing dynamics. It is shown that, while the gradient descent requires the persistence of excitation condition, the addition of the pre-processing dynamics ensures reliable estimation under significantly weaker excitation assumptions. Moreover, it is shown that the pre-processing dynamics improves the transient behaviour and noise performance of the estimators. The results are validated through a set of numerical simulations and in an experimental prototype, where sensor noise and unmodelled disturbances are considered., This work has been supported by the Spanish Ministry of Universities funded by the European Union - NextGenerationEU (2022UPC-MSC-93823). This work is part of the Project MAFALDA (PID2021- 126001OB-C31) funded by MCIN/ AEI /10.13039/501100011033 and by "ERDF A way of making Europe" This work is part of the project MASHED (TED2021-129927BI00), funded by MCIN/ AEI/10.13039/501100011033 and by the European Union Next GenerationEU/PRTR. This research has been developed within the CSIC Interdisciplinary Thematic Platform (PTI+) Transición Energética Sostenible+ (PTI-TRANSENER+) as part of the CSIC program for the Spanish Recovery, Transformation and Resilience Plan funded by the Recovery and Resilience Facility of the European Union, established by the Regulation (EU) 2020/2094., Peer Reviewed, Preprint

Objective.Artificial neural networks (ANNs) are state-of-the-art tools for modeling and decoding neural activity, but deploying them in closed-loop experiments with tight timing constraints is challenging due to their limited support in existing real-time frameworks. Researchers need a platform that fully supports high-level languages for running ANNs (e.g. Python and Julia) while maintaining support for languages that are critical for low-latency data acquisition and processing (e.g. C and C++).Approach.To address these needs, we introduce the Backend for Realtime Asynchronous Neural Decoding (BRAND). BRAND comprises Linux processes, termednodes, which communicate with each other in agraphvia streams of data. Its asynchronous design allows for acquisition, control, and analysis to be executed in parallel on streams of data that may operate at different timescales. BRAND uses Redis, an in-memory database, to send data between nodes, which enables fast inter-process communication and supports 54 different programming languages. Thus, developers can easily deploy existing ANN models in BRAND with minimal implementation changes.Main results.In our tests, BRAND achieved <600 microsecond latency between processes when sending large quantities of data (1024 channels of 30 kHz neural data in 1 ms chunks). BRAND runs a brain-computer interface with a recurrent neural network (RNN) decoder with less than 8 ms of latency from neural data input to decoder prediction. In a real-world demonstration of the system, participant T11 in the BrainGate2 clinical trial (ClinicalTrials.gov Identifier: NCT00912041) performed a standard cursor control task, in which 30 kHz signal processing, RNN decoding, task control, and graphics were all executed in BRAND. This system also supports real-time inference with complex latent variable models like Latent Factor Analysis via Dynamical Systems.Significance.By providing a framework that is fast, modular, and language-agnostic, BRAND lowers th

This thesis explores the design and implementation of a simulation framework created for C2 training environments, focusing on simulating the movement of aircraft and other moving targets between waypoints. The core objective was to develop an extensible, reliable, and realistic simulation system that can adapt to simulate a variety of entities, each characterized by unique movement patterns defined by their type. The simulated input scenarios involve a collection of targets, where each target traverses through a defined set of waypoints according to their specified movement behavior at initialization. The research involved a evaluation of established theories and ideologies of both aircraft and computer simulation, leading to the creation of a framework that leverages Java for its scalability and performance. The framework was designed from the ground up to promote extensibility of various types of moving targets beyond of just that of aircraft, such as helicopters, boats, missiles, and surface vehicles. The accuracy of the simulation was tested by comparing its output to real-world flight data, demonstrating its capability to mimic realistic aircraft movement within predefined training scenarios. Additionally, the system's extensibility is showcased by simulating two different categories of aircraft—a commercial airliner and a small fighter jet—and illustrating the differences in their generated trajectories. Another outcome is the framework's ability to support real-time simulation, crucial for training scenarios in a C2 context. The system integrates and utilizes concurrent design techniques and dependable design patterns to ensure high throughput and performance, even when scaling to large numbers of targets. The framework's architecture allows for future extension and integration of new target types, without requiring modifications to the existing codebase, by the utilization of Javas SPI.

This thesis presents an empirical analysis of the performance of WebTransport and WebSocket using the Socket.IO framework under varying network conditions. The thesis investigates two primary experiments, message round trip time and multiplexing performance, both conducted under three scenarios—perfect network (0% packet loss), moderate network (5% packet loss), and poor network (15% packet loss). The experiments were conducted on a system running Fedora 40 with a 12th Gen Intel Core i5-12600K processor and 16 GB of memory. The findings demonstrate that WebTransport mitigates Head-of-Line (HOL) blocking more effectively than WebSocket, especially in environments with degraded networks. WebTransport (Reliable) provided a significant performance advantage under suboptimal network conditions when measuring message round trip time, while WebTransport (Unreliable) consistently demonstrated slightly lower latency but with some data loss. When multiplexing, WebTransport showed significant performance benefits compared to using WebSocket.

Light Detection and Ranging (Lidar) is a kind of active sensor that emits a laser pulse and primarily measures the time of flight of the returning pulse and uses it to construct a 3D point cloud of the scene around the lidar sensor. The constructed point cloud is an essential asset for the control of autonomous vehicles, and especially today, an essential basis for the training of autonomous vehicle control models. However, it remains time-consuming, high-risk and expensive to acquire the amounts of data necessary to train the rather complex modern control models. As such, generating the point cloud through simulations becomes a natural solution. Yet, many lidar simulations today produce ideal point clouds, corrected only by random noise, without considering the physical reasons behind the imperfections visible in real lidar point clouds. The aim of this study was to investigate real-time simulation models for disturbances that may cause imperfections in lidar data. From a base investigation of lidar, disturbances were found, models were investigated and finally a real-time implementation of Atmospheric Effects and attenuation from Beam Divergence was evaluated. It was found that the implemented models could produce physically accurate lidar point placement while keeping the computational time low enough for real-time evaluation. However, to achieve correct separation of target hit rates under Atmospheric Effects, as high as 34% of the points had to be dropped. Additionally, the intensity of the return points could not be properly verified. From these results it can be concluded that, with additional verification and adjustment, the presented models can achieve good results for evaluation in real-time. The results of this study thus serve as a support for future developments of realistic real-time lidar simulations, for use in development of autonomous vehicle control models and implementation of digital twins.

Replicating a real-world environment is crucial for creating simulations, computer vision, global and local path planning, and localization. While computer-aided design software is a standard tool for such a task, it may not always be practical or effective. An alternative approach is mapping, which uses sensory input and computer vision technologies to reconstruct the environment. However, developing such software requires knowledge of various fields, making it a challenging task. This thesis deep-dives into a state-of-the-art mapping framework and explores potential improvements, providing a foundation for an open-source project. The resulting software can replicate a real-world environment while storing panoptic classification data on a voxel level. Through 3D object matching and probability theory, the mapping software is resilient to object misclassifications and retains consistency in the different instances of observed objects. The final software is designed to make it easy to use in a different project by substituting the simulation data with a semantic, instance, or panoptic segmentation model. Additionally, the software integrates certain functionalities that facilitate the visualization of diverse classes or a particular class instance., Att replikera en verklig miljö är avgörande för att skapa simuleringar, datorseende, global och lokal vägplanering samt lokalisering. Trots att ett datorstött designprogram är ett standardverktyg för sådana uppgifter kanske det inte alltid är praktiskt eller effektivt. Ett alternativt tillvägagångssätt är kartläggning, som använder sensorisk input och datorseendeteknik för att uppnå reskonstruering av omgivningar. Att utveckla sådan programvara kräver dock kunskap inom olika områden, vilket gör det till en utmanande uppgift. Den här avhandlingen fördjupar sig i ett toppmodernt kartläggningsramverk och utforskar potentiella förbättringar, vilket ger en grund för ett öppet källkodsprojekt. Resultatet av denna avhandling är en programvara som kan replikera en verklig miljö samtidigt som den lagrar panoptisk klassificeringsdata på en voxelnivå. Genom 3D-objektmatchning och sannolikhetsteori är kartläggningsprogramvaran motståndskraftig mot felaktiga objektklassificeringar och är koncekvent avseende förekomsten av olika observerade objekt. Den slutliga programvaran är utformad med fokus på att göra den enkel att använda i andra projekt genom att ersätta simuleringsdata med en semantisk, instans eller panoptisk segmenteringsmodell. Dessutom integrerar programvaran funktioner som underlättar visualiseringen av antingen olika klasser eller en specifik instans av en klass.

Parameter estimation is crucial for polymer electrolyte membrane fuel cell monitoring and control. Nonetheless, most parameter estimation algorithms rely on a persistence of excitation condition, which is rarely satisfied and not convenient in fuel cell systems. For this reason, this work presents and compares three algorithms to estimate in real-time some critical PEMFC parameters in the voltage equation: the ohmic resistance, the charge transfer coefficient and the oxygen activity of a proton exchange fuel cell. The first algorithm is a standard gradient descent, while the other two are based on a set of pre-preprocessing dynamics. It is shown that, while the gradient descent requires the persistence of excitation condition, the addition of the pre-processing dynamics ensures reliable estimation under significantly weaker excitation assumptions. Moreover, it is shown that the pre-processing dynamics improves the transient behaviour and noise performance of the estimators. The results are validated through a set of numerical simulations and in an experimental prototype, where sensor noise and unmodelled disturbances are considered.