This project started when I got in touch with the non-profit cultural centre Royal in Eskilstuna, a smallcity 1,5 hours west of Stockholm. They had just moved in to their new premises, an old cinema fromthe 30’s, and were gaining influence over the cultural scene of Eskilstuna. The first time I visited themwas late autumn last year and I quickly realised that I could develop a cooperation with them duringmy thesis project. I was interested in the phenomenon of cultural centres since I believed it to be a pro-grammatic padding or alibi used by politicians to gain popularity instead of more specifically payingculture a greater overall attention. Royal were keen to let me know that I had freedom in doing whate-ver I wanted as my project, but they also told me that they were interested in a renovation and concep-tualisation of their premises as well as a wardrobe, office space and conference room which were likelyto be built if it was what I decided to do.As I began the project with a scepticism towards cultural centres in general I wanted to look if it wasany need of Royal in Eskilstuna, and if it was – what kind of culture Royal should focus on. I realisedthat Eskilstuna had its cinema, theatre, concert stage etc., but there was no cultural centre that combi-ned the different activities, and there wasn’t anyone who did it non-profit. Further analysis of the cultu-ral atmosphere in Eskilstuna also showed that Royal were focusing on the right kind of culture attrac-ting a wide range of people from different ages, genders and backgrounds.I understood that part of the success of Royal was probably because it was driven by ambitious pe-ople residing in Eskilstuna and not by the municipality, and I realised that I had little to add in terms ofcultural content. I felt as Royal was doing something good and my motivation to contribute as an archi-tect grew. Solely to restore their premises and do what they told me to felt like taking the easy way out,hence I focused on what they needed but didn’t tell me.By then I had visited Eskilstuna a few times, and had become familiar with the location of Royal in rela-tion to the city. The area is called Nyfors and is located just south of the central station. Geographicallyin the centre of the city, however, the railway that separates Nyfors from the city centre works as a bar-rier reducing the importance of Nyfors to the point where it is regarded to be a suburb. Since there isn’tmuch activities in Nyfors there is no point for people not residing in that particular part of the city to gothere, which is one of the big reasons for why many people doesn’t even know that Royal exists. I figu-red that was one thing they didn’t tell me.While looking at the social aspects of Nyfors, I found that it had bad reputation due to high criminality.Much of the criminality was focused on a shady restaurant very close to Royal. I, however, discoveredthat the same restaurant had closed and reopened with new owners which had started a cooperationwith Royal. At that point I started to see Royal as a beginning of a development of the area alreadycausing a ripple effect.From that point my focus turned from being a restoration project of an old cinema to be more of a cityplanning project. Through my investigation I came to see that the square where Royal was locatedneeded a lift that in turn would draw attention to the cultural centre.I started to look at the existing square, with the intention to point out the things that worked and thethings that didn’t. Furthermore, I made an investigation of squares around the world that I had visitedaiming to figure out what characteristics defines a good square.My conclusion was that the square which I had decided to work with had almost everything against it.It was inclined to the north, it had no clear limits and it was too big for the surrounding city scale. Na-turally making it very hard to work with. At the same time I was motivated by the work of Royal, and Iwanted to give them something in form of a realistic proposal.By trying to understand the positives of the site and boost them while identifying the deficits and fixthem, I have meant to create a space for culture and recreation in the heart of Nyfors.I have intended to be perceptive towards my process and always see my customers’ needs in the big-ger picture, letting the project lead me towards the final intervention – a process successively runningto a result which came out to be a square far from my initial thoughts.

This thesis investigates the economic, and technical feasibility of using Additive Manufacturing (AM), specifically Laser Metal Deposition (LMD-p), for the repair of forging tool. Forging tools require frequent repairs and eventually replacement due to their severe wear and tear, especially from abrasive wear and plastic deformation. Although traditional machining is effective but the , traditional machining techniques wastes materials, wears tools more frequently, and use more energy. This study investigates AM as a more economical and environmentally friendly substitute. Furthermore, AM contributes to environmental goals by reducing energy use. Although AM infrastructure requires a capital investment up front, the long-term advantages it offers over conventional techniques make it a very compelling alternative for tool maintenance in the future. The study's concentration on a single particular tool used by a company, however, puts limits on it.

The Battery Regulation 2023/1542 necessitates recyclers to conduct extensive reporting on waste battery details, recycling efficiency, and material recovery. To ensure compliance, a robust traceability system must be implemented throughout the battery recycling supply chain. This research delves into the specific traceability requirements for recyclers, the suitability of information systems, and the effective translation of these requirements into a traceability system. To address these questions, a mixed-methods approach, combining a thorough literature review, regulation analysis, and a case study, was employed. The findings include defined achievable traceability levels for each recycling process step, an Entity-Relationship Diagram (ERD) for database design, a traceability system architecture, and Data Flow Diagrams (DFDs) for each recycling process step. These findings significantly contribute to research by focusing specifically on battery recycling and providing a detailed data collection process. The results empower recyclers to fully adhere to regulatory requirements while implementing a modern, scalable, accessible, and secure traceability system. Future research should explore the integration of advanced technologies into recycling traceability systems and adapt the proposed system to accommodate evolving regulations.

The aim of this study is to investigate the performances of semi-active suspension systems onto Scania AB terrain vehicles. This will be conducted by assessing the potential benefits for the road handling by implementing controllable dampers on 4x4 Scania AB chassis. The goal of the study is to conduct Multi-Body Simulation (MBS) on a virtual rugged terrain to analyze the impacts on parameters influencing the road-holding. The simulations will be performed through co-simulation between Adams and Matlab/Simulink. The truck will be modeled in Adams, and the shock absorbers will be controlled using a control algorithm developed in Matlab/Simulink. A comparison will be made between the results of the vehicle equipped with a passive damping system and those of the vehicle equipped with semi-active shock absorbers. Different scenarios will be explored based on the road profile, vehicle payload and driving speed. The ultimate goal is to identify benefits and/or challenges associated with integrating this system into the vehicle and to provide recommendations for future development. This study has demonstrated that, across all tested scenarios in the co-simulations, no significant improvements in road-holding or comfort were observed with the semiactive dampers. The main reasons identified include the inherent stiffness of the suspension system and the use of rigid axles, which limit suspension travel and the independent movement of the wheels. While the simplicity of the control algorithm may also contribute to the results, it is more likely that the mechanical and dynamic characteristics have a more substantial impact on performance. These findings suggest that, passive dampers may be as effective as semi-active dampers for off-road Scania AB vehicles.

This master’s thesis report set out to explore the user experience of mobile applications for public EV charging. The project was conducted in collaboration with VCG, a company that operates public EV charging infrastructure and provides a mobile application to manage charging sessions. The purpose of the project was to investigate the user needs related to mobile applications for EV charging and to identify the challenges and opportunities arising from this interaction. The project followed a double-diamond methodology, with a service design approach. The project began with a literature review exploring previous research on EV charging technology, as well as on consumer behavior and habits among EV drivers. This was followed by the main study, in which 25 users, EV retailers, and EV charger retailers were interviewed, and observations were conducted at 15 different locations. Various analysis methods were applied after the main study, resulting in seven major insights. These insights established the foundation for the subsequent concept development phase, which included ideation and prototyping. The concept development phase resulted in a main concept featuring a selection of key functions. These functions were categorized into five distinct solution areas. The project concluded with the development of a mobile application mockup that showcases the five different solution areas and functions identified in the study. In addition to the mockup, three recommendations are presented at a societal level to enhance the overall user experience.

Understanding which factors affect people's choices when traveling helps planners and policymakers build better infrastructure that fosters more sustainable practices. Gaining insights into how cyclists choose their routes is one major key to improving infrastructure and promoting cycling. Particularly commuters and e-bike users have been noted to have preferences unconstrained from shortest path logic. There are other more human centric factors can be of importance, such as closer to nature and away from busy intersections. To that end, this paper aims to uncover cyclists' preferences and the affecting nature-related attributes of routes for commuters to high school in Nijmegen, Netherlands. Based on a Dutch dataset, the study analyzed 1284 e-bike cycling trips, each with four route alternatives, including the chosen one. The primary objective is to identify the most influential parameters affecting e-bike commuters' route choices and understand their contributions. The approach employed both a simple path size Logit (PSL) and a Pairwise Combinatorial Logit (PCL) model, incorporating nature-related and interaction variables. Additionally, the research compared the predictive performance of Logit-based models with deep learning models. The findings shed light on the factors influencing e-bike commuters' route choices and demonstrate the superior predictive capabilities of deep learning techniques, with a validation accuracy of 80.16\%. By adopting a sensitivity analysis approach, we uncovered the key factors that influence our deep learning model in predicting cycling routes, giving a precise interpretation of our results. Our findings show that commuter e-bike cyclists prefer shorter routes with fewer traffic lights and favor routes that have more natural settings. However, their primary concern is efficiency in their commutes.

Dynamic Discrete Choice Models (DDCMs) allow us to analyze trip attribute choices within an activity-based framework where the sequence of activities matters. These models have been primarily developed for travel survey data. However, the advent of big data presents new opportunities for mobility analysis. This paper bridges traditional approaches with novel data sources by adapting an established activity-based DDCM to handle anonymized mobile network data. The proposed framework advances methods within transportation planning in two ways. First, dynamic models allow for detailed assessments of sequential travel decisions, for instance related to activity, departure time, and modes. Second, automatically collected data are advantageous as compared to travel surveys that are limited by respondents’ ability and willingness to disclose travel behavior. However, this approach poses methodological challenges, particularly because individuals’ exact locations over time are not directly observed. We address these in a latent-based framework in which the positions of the individuals are treated as latent variables and the location of cell towers as observations. We estimate the model using a stochastic expectation-maximization (StEM) algorithm. The performance is highlighted in a case study of daily travel demand in Stockholm, Sweden. Our paper demonstrates the agility of activity-based DDCMs in adapting to new data sources, by incorporating appropriate modifications within a well-established modeling paradigm.

This study proposes a novel strategy for the development of macro-sized bone tissue. To address the challenges presented by current jaw surgery techniques, this thesis proposes a novel approach, which is tested in vitro utilizing adipose-derived stem cells (ADSCs) and their ability to differentiate into e.g. osteoblasts. The implementation of Fibronectin (FN)-silk is intended to provide a 3D-structural support for the ADSCs. FN-silk is a biocompatible biomaterial engineered from the recombinant spider silk protein 4RepCT, proven to assemble into a strong, elastic, network which supports 3D cell cultivation similar to in vivo conditions. The aim was to establish a bone tissue 3D model and to investigate the potential of ADSCs in combination with FN-silk networks.

This study demonstrates that ADSCs were successfully integrated within the FN-silk network and - after the addition of osteogenic factors - did undergo osteogenic differentiation within the FN-silk network, as proven by increased Alkaline Phosphatase activity and signs of mineralization demonstrated by Alizarin Red S staining. It also reveals that Simulated Body Fluid (SBF) alone in cultivation media does not promote osteogenesis in vitro. Additionally, the establishment of a method to incorporate ADSCs within the FN-silk network into zirconium molds for clinical applications has been successfully conducted.

Developing crosslinked elastomers that are easily produced and easily recyclable is complex, usually requiring a compromise between performance and recyclability. However, combining reversible exchange reactions together with phase separation phenomena appears as a promising approach. Herein, a simple and up‐scalable extrusion process is proposed, involving commercial maleated ethylene propylene rubber (EPRgMA), maleated polypropylene (PPgMA), and a suitable crosslinker. It is shown that a crosslinker enabling disulfide exchange reactions can provide local and long‐range rearrangements required for extrusion, yielding a robust crosslinked blend (BlendSS) with strength of 15 MPa and an impressive elongation of 1000%. Moreover, the presence of the disulfide crosslinker provided the required fast exchanges for three repetitive recycling cycles by extrusion with close to 80% retention of initial properties. In comparison, the use of a crosslinker without the capability to establish reversible reactions (BlendCC), yielded crosslinked blends of marginal compatibility, strength of 4 MPa and only 40% elongation. The absence of reversible reactions restricted chain rearrangements and consecutive recycling is only possible by compression molding. The recycled blends presented even lower compatibility, elasticity and thermomechanical performance, demonstrating that the proper design of interfacial interactions between PPgMA and EPRgMA can build a bridge between processability, performance, and high‐throughput recyclability.

Motion planning in uncertain environments like complex urban areas is a key challenge for autonomous vehicles (AVs). The aim of our research is to investigate how AVs can navigate crowded, unpredictable scenarios with multiple pedestrians while maintaining a safe and efficient vehicle behavior. So far, most research has concentrated on static or deterministic traffic participant behavior. This paper introduces a novel algorithm for motion planning in crowded spaces by combining social force principles for simulating realistic pedestrian behavior with a risk-aware motion planner.

We evaluate this new algorithm in a 2D simulation environment to rigorously assess AV-pedestrian interactions, demonstrating that our algorithm enables safe, efficient, and adaptive motion planning, particularly in highly crowded urban environments—a first in achieving this level of performance. This study has not taken into consideration real-time constraints and has been shown only in simulation so far. Further studies are needed to investigate the novel algorithm in a complete software stack for AVs on real cars to investigate the entire perception, planning and control pipeline in crowded scenarios. We release the code developed in this research as an open-source resource for further studies and development. It can be accessed at the following link: https://github.com/TUM-AVS/PedestrianAwareMotionPlanning.

Självklart! Här är texten med de saknade mellanslagen tillagda, utan andra ändringar:

 The steel industry is a key sector for modern society. Due to the extreme versatility of steel, it finds application in very different sectors. However, the steel industry is also very polluting, accounting for roughly 7-9 % of the yearly global anthropogenic CO2 emissions. Steel is produced either from iron ore, with the Blast Furnace-Basic Oxygen Furnace (BF-BOF) process, or from steel scrap, following the Electric Arc Furnace (EAF) process. With the latter being a more sustainable alternative (roughly half of the energy required and 1.5 tonnes of CO2 emissions), the steel industry looks forward to shifting the production towards the EAF, which today accounts for 29 % of the global production. However, differently from the BF-BOF, the EAF has still not been optimized up to its full potential.

In the process of achieving EAF process’ maximum efficiency, the integration of artificial intelligence could represent a significant step forward. The power and computational ability of machine learning models may indeed be used to further analyse the data collected from industry. In this way, it could be possible to improve the understanding of the EAF process and help optimizing it. This thesis focuses on using data clustering algorithms (Gaussian Mixture Model and K-Means) to analyse real industrial data. The objective was to examine the influence of the steel scrap type and its preparation on the EAF energy consumption. The information gathered during the preliminary data analysis and the data clustering were then used to develop and support an optimizer tool used to predict the most energy-efficient configuration of the EAF charging bucket.

This paper explores the aesthetic value shifts required for sustainable design of so called ‘high-tech’ products, highlighting an increasing down-to-earth ethos within the field. Using a spaceship metaphor and drawing from principles of post-industrial design and visions for long-term sustainable transformation, a high-level analysis is presented of how grassroot activist cultures and alternative tastemaking practices are currently steering designs towards the systemic, earthy and organic. This direction is illustrated by diverse examples from within the TEI discourse that investigates new material approaches and that critically challenge conventional aesthetic orientations. While such approaches can be criticized for insufficiently addressing interactive and electronic components, this research underscores their essential role in reimagining technology and materials, to navigate complex cultural interdependencies and advance sustainable design futures.

This thesis explored the use of a Gated Recurrent Unit (GRU) based neural network, used as a feedforward controller in the electro-hydraulic driven robotic arm of DeLavals Voluntary Milking System (VMS). The research focused on developing a GRU-based feedforward controller and comparing its performance to the existing controller of the VMS in terms of velocity tracking accuracy, as the non-linear dynamics of the electro-hydraulic actuators pose a challenge for traditional methods of system modelling. The network was trained on data collected from the robotic arm’s movements, enabling it to learn the complex relationship between desired velocities and the corresponding outputs. This trained network was implemented as a feedforward controller, directly generating actuator commands to achieve the desired velocity trajectory. The performance comparison utilises Mean Squared Error (MSE) to evaluate the tracking accuracy of both controllers. The results demonstrate that the GRU-based feedforward controller significantly outperforms the existing controller, achieving a 53% lower MSE value. These results further signify the potential of GRU neural networks for improved control of robotic manipulators with complex dynamics.

This report analyzes strategies to optimize the production of electric motors with a focus on stator winding and quality controls. Two winding methods, hairpin and continuous winding, are compared in terms of performance, cost-efficiency, and adaptation to different production volumes. Hairpin winding appears to be more suitable for high-volume production due to automation possibilities and improved electrical efficiency, while continuous winding is more flexible and cost-effective for low-volume production.

The study is based on the analysis of Scania's production data, interviews with suppliers, and literature studies to identify appropriate quality controls and improvement areas.

The report also highlights the importance of quality controls such as Automated Optical Inspection (AOI), electrical testing, and Non-Destructive Testing (NDT) to reduce defect levels and ensure product quality. The results show that automated controls are more cost-effective at large volumes, while manual inspection is advantageous for smaller volumes. For example, manual inspection can save approximately 45 625 EUR per year at a production volume of 1 500 stators compared to AOI.

In conclusion, the report emphasizes the importance of adapting both winding methods and quality control strategies to production volume and specific requirements to optimize efficiency and profitability in the manufacturing process.

Modern automotive systems critically depend on efficient software for their operation. To improve the software development efficiency, automatic code generation tools are used to generate the C/C++ codes for model-based software for Electronic Control Units (ECU). However, the generated code often has low execution efficiency when deployed on the ECU. Thus, it seriously impacts the functionalities of automobile-embedded systems. To address this problem, the degree project conducted at ZEEKR Technology Europe R&D Center (ZEEKR) investigates a method for improving ECU efficiency by decreasing data dependencies while deployed on the onboard embedded systems. The study covers algorithm design and code optimization, contributing a novel model-based code generation method. First, the Simulink model will be transferred into an Extensible Markup Language (XML) structure. Then, a methodology will be utilized to analyze the data flow of the model, which will also build a data graph in C++. Then, an algorithm will be implemented to optimize the original C code sequence to improve the efficiency of the embedded system in a designated automobile caused by data dependencies. The algorithm's core is the Breadth First Search (BFS) methodology which scans and assigns each node. Finally, an optimized code execution sequence is generated based on these hierarchical relationships, reducing data dependencies. Processor in the Loop (PiL) simulations were performed for system testing to examine the performance and correctness of the layer-based new algorithm. The benchmark of the experiment will be the clock cycle, which is the minor operation unit executed within the processor. By measuring the number of clock cycles executed, the performance of the optimized code compared with the original can be accurately evaluated. The results show that the optimized methodology could take 0.5% to 2.5% less time owing to reduced data dependencies in the model.

This thesis explores the design and implementation of modern data pipelines within the Google Cloud Platform, with a focus on identifying essential components, methodologies, and best practices to enhance their efficiency, reliability, and scalability. Conducted in collaboration with Astrafy, this research provides critical insights into the construction of automated data pipelines. The study specifically addresses the complexities of integrating various open source tools and microservices, aiming to simplify the development process by proposing a structured methodology for selecting and combining these technologies. A key contribution of this work is the creation of an automated one-click deployment system, which streamlines the setup and management of data pipelines for processing Google Cloud billing data. The system is evaluated through a series of experiments, comparing various technologies based on performance metrics such as cost, deployment time, and ease of use. The results offer valuable guidelines and replicable models for organizations seeking to implement scalable and reliable data pipelines on cloud platforms.

Free space optical communication (FSOC) is rapidly gaining traction in the space industry due to its advantages over traditional radiowave communication. It offers higher data rates, increased security, a license-free wavelength spectrum, and easy deployment in a commercial field with growing demands on transmission capacity between Earth and space due to an increasing number of satellites. The main limit of FSOC is the adversarial impact on the optical signal from atmospheric effects such as turbulence, absorption, and scattering. Major research efforts are underway to understand these effects further, model their impact on FSOC operations, and develop effective mitigation techniques. Characterization of local atmospheric conditions onsite is also critical when evaluating optimal locations for new optical ground stations. This thesis reviews the state-of-the-art in FSOC, the impact and characteristics of the atmospheric channel, and existing mitigation techniques. It analyzes instrumental data from a characterization station located at an upcoming optical ground station (OGS) to determine if local data can be used to predict the suitability of a potential location with regard to atmospheric conditions using traditional statistics models and machine learning. Results show a low correlation between weather parameters and the measured optical turbulence. Attempts to predict optical turbulence using nonlinear models such as random forest regressor and k-nearest neighbors show strong predictive capability with an R² of around 0.81 for both models. This indicates that optical turbulence can be modeled and predicted but that the underlying relationships are nonlinear and complex.

Transparent electronics is an emerging field concentrating on invisible circuitry. A nearly transparent complementary metal oxide semiconductor (CMOS) technology can be a novel solution for optoelectronic and lab-on-chip applications. Indium tin oxide (ITO) is known to be electrically conductive (ρ = 100 μΩcm) [1] and optically transparent (> 85%) [2]. This study is focused on using ITO as a metal stack for fully-depleted silicon-on-insulator (FDSOI) CMOS technology. ITO is shown to meet the compatibility requirements on fabrication for CMOS technology. Annealing ITO with 5% N₂/H₂ at 450 ^◦C for 30 seconds after deposition is shown to decrease local sheet resistance down to 5 Ω/□and improve the sheet resistance uniformity across the wafer up to 95%. Inductively-coupled plasma reactive ion etching (ICP-RIE) is selected to be the primary method to etch ITO with CH₄/Cl₂/H₂ chemistry. A wafer with a poly-Si gate and an ITO metal layer is successfully transferred onto a glass substrate, demonstrating near transparency to visible light. An FDSOI CMOS wafer with a single 200 nm ITO metal layer is fabricated with a 25 nm active silicon device layer and a 100 nm p-type poly-Si gate. ITO to ITO contacts and ITO to gate contacts are tested with a probe station, contacts were found to have 35 Ω and 5 kΩ of resistance, respectively. The lowest subthreshold swing (SS) value was found to be 130 mV/dec and V_T was higher than 1 V.

The construction industry is digitally underdeveloped compared to other sectors. A large amount of documents is used during processes such as quality inspection, maintenance planning, incident reporting, or issuing permits to work. All these documents are traditionally on sheets of paper that are filled in by hand and transferred between the different stakeholders. A few companies are starting to digitize their processes, making them both more efficient and more reliable. With their new platform, they now have access to past construction data, and a key issue is to be able to learn from them. In particular, during the quality inspection, many construction defects are reported, and then rectified. These steps are logged, and data gathered during the process can be leveraged to improve the planning of defect rectification. Two solutions are examined to pursue this goal: a visualization application is developed to easily explore defects data, dive into the relevant details, and identify possible improvements to prevent these defects or speed up their correction. In addition, a forecasting model is designed to predict the time to correct a defect and help to estimate the duration of the inspection stage. The visualisation tool is successfully used by several customers, the main remaining issue is to use it at a larger scale for around a hundred different companies. The forecasting attempt shows encouraging results but may require further improvements to be usable.

This thesis investigates gaze-based intent recognition within Pandemic: Hot Zone – Europe, a collaborative board game adapted into a Virtual Reality (VR) environment. Intent recognition, a critical area in human-computer interaction, uses observable behaviors such as gaze to infer human goals and actions. Building upon prior research in gaze-based intent modeling and heuristic planning, this study develops a model that combines gaze data with probabilistic and heuristic models to predict player actions in this complex and dynamic environment. A VR version of the game was developed to enable real-time data collection, including gaze fixations, which were used to infer player intents. The results demonstrate that the inclusion of gaze data notably improves the model performance, particularly after the players first action, where the large initial set of intents is drastically narrowed. For instance, broader metrics such as Hit@5 and Hit@10 showed marked improvements, with the gaze-based model reaching 55.8% and 65.1% accuracy before the second action respectively, compared to 34.9% and 51.2% without gaze data. However, the results also highlight limitations in early- turn predictions, where overlapping fixation mappings and the large number of intents reduced the model’s effectiveness. Evident in Hits@1, the strictest evaluation criteria, which produced a notably worse performance, 27.9% using gaze data compared to 18.6% without. Despite these challenges, the findings indicate that gaze behavior is a valuable input for reducing the number of intents and providing actionable predictions. Before the final action, the model utilizing gaze data demonstrated over 90% accuracy in broader metrics such as Hit@5 and Hit@10, and 69.8% for Hit@1, showcasing its ability to reliably capture the player’s plan. While predicting the exact sequence earlier in the turn is less consistent, the strong performance in broader metrics highlights the model’s potential to support decision-making by narrowing down plausible intents, even in a complex and dynamic game environment. Finally, while the small dataset limits the ability to achieve statistical significance, the observed trends provide valuable insights and highlight promising directions for future research and applications. This includes exploring larger datasets, dynamic fixation mapping strategies, and the many capabilities of the VR environment, such as capturing additional modalities like speech and hand gestures, to enhance the robustness and practical application of such models.

Over time, the amount of networking equipment has increased exponentially. As the networks grow, so does the complexity of managing them. Numerous errors and misconfigurations can appear in a network over its lifetime at any given time. Detecting and correcting those as soon as possible is a priority for many businesses. In this thesis, a new network automation system called NetHealth is presented. It can actively monitor failures and misconfigurations that can occur across multiple networks at any time. In the case of misconfigurations, it effectively corrects them according to the original intended state. Also, a network simulator called FakeNOS and a chaos engine that randomly creates failures and misconfigurations have been developed to test the system’s response under a controlled chaotic environment. The results prove that NetHealth can detect failures in under 5 minutes and resolve misconfigurations in less than 3 minutes. Although it only works in Passive Optical Network (PON) as of today, it has been designed so it can be easily extended to any other network. Moreover, a significant contribution of this thesis is all the statistics of the libraries used to connect to the devices and the developed tools to test the systems, such as the chaos engine and the FakeNOS network simulator.

In 4G/5G Radio Access Network (RAN), the scale of data traffic in the network becomes larger, which leads to increasing demand for ensuring the good connection of User Equipments (UEs). With the help of Carrier Aggregation (CA), UE can use expanded bandwidth by connecting to multiple cells. There are multiple carrier frequencies with overlapping cells, which can result in several potential cell sets for each UE. In this case, the traffic steering algorithm is applied to distribute UEs to their best cell sets. While the data traffic in RAN is complicated, highly dynamic, and challenging to predict, this project aims to use AI/ML algorithm to predict the network traffic and then optimize the current traffic steering in a simulator of Ericsson by finding a balance between the throughput and number of cell handover of UEs based on the prediction results. While frequent handover can help the UEs to be distributed to their current best cell, there could also be some unnecessary handovers which may be useless or even harmful to the network. Unnecessary handover wastes the network resources, and if the handover failure occurs, it may cause the degradation of user experience or cause disconnection. In this context, it is worth finding a trade-off between throughput and the number of handovers. In this thesis, a network traffic is generated by an Ericsson simulator, and the AI/ML algorithm is developed and implemented to predict two targets for the next 15 minutes which are median UE throughput and average number of handover per UE. The algorithm dynamically adjusts two controllable parameters according to a linear weight optimization function of the two targets which can indirectly control the frequency of cell handover of UE. Several models are trained with data from simulations and then processed. LSTM, fully connected neural network, transformer and random forest model have been compared, where random forest provided the best performance. To evaluate the algorithm, a simulation without the assistance of AI/ML algorithm is executed as a baseline, throughput and the number of handovers are recorded per 15 minutes in the time of the simulator. After that, simulations with the same configuration as the baseline but with AI/ML algorithm and different values of weight parameters in the optimization function are executed. Throughput and number of handovers are also recorded for every simulation and compared with the baseline to get the values of weight parameters which can provide an acceptable trade off. As a result, the number of handovers can be reduced by approximately 40% when the throughput is reduced by 2%.

Understanding dynamic driving scenarios poses significant challenges in the development of Automated Driving Systems (ADS). ADS need to perceive crucial information from dynamic driving scenarios to maintain functional safety. The information from the scenarios is perceived through sensors (e.g., cameras, IMU, LiDAR) and is usually unstructured data. Current approaches often rely on Convolutional Neural Networks (CNNs) to process this data. However, while CNNs are effective at handling unstructured sensor inputs, they struggle to capture the complex relational dynamics between traffic participants, which are crucial for a deeper understanding of the driving environment. To address this gap, Graph Neural Networks (GNNs), designed for relational data, offer a promising alternative. This thesis proposes a framework by adopting Relational Graph Convolutional Networks (RGCNs), a special-designed GNNs for relational data, to understand the relational data regarding dynamic driving scenarios. Specifically, by categorizing the types of traffic participants, a workflow is designed to convert unstructured data into spatial relational data. By analyzing the frames within a time series using the rule-based approach, the spatial relational data is converted into spatial-temporal relational data. Next, the RGCNs model infers the behaviours of the traffic participants and the relationships among them. As results, with the adoption of the RGCNs, the framework outperforms in classifying nodes and predicting their relationships in the open-source real-world dataset. Compared to classic Graph Convolutional Networks (GCNs), the proposed methods achieve an improvement of about 10% in prediction accuracy, increasing from 0.77 to 0.85 in node classification, and from 0.74 to 0.82 in link prediction. Furthermore, apart from the tasks on node classification and link prediction, this thesis supports the reasoning of unknown conditions by encoding known relationships with a top-1 accuracy of 0.78 and hits@2 of 0.91, which can help improve comprehension of the risk of the scenario.

Testing is an essential activity of software development, and despite the vast use of testing by the industry, little formal reasoning can be found in the scientific literature of the field. The project presents an extensively formal approach to glass box testing from the underlying language to the graph model of a program, its execution, and reasoning on preconditions for its paths. In this paper, programs are modelled as executable Control Flow Graphs and test requirements are paths of the graphs. We study a logic to infer the weakest precondition of a given path of the graph. Weakest preconditions let us reason on relations between test requirements, which in turn let us optimise the test requirements set. The model uses a minimal language with the integer data type and first-order logic. This language can be extended as defined in this work to suit the need of the reader. Several metatheorems are proved for any language extending the minimal language we provide, including soundness, completeness under the restrictions to obtain a weakest precondition formula, and undecidability.

Spiking neural networks (SNNs) are a kind of neural networks capable of processing event-based data in real-time, resource-constrained environments. This study focuses on evaluating the performance of SNNs for object tracking, specifically tracking cars. The primary goal of this research is to investigate and compare different SNN architectures in the context of traffic monitoring, then adjust the number of parameters on the best-performing architecture to investigate differences in performance. The results are analyzed to understand further how different SNN designs perform, and the implications of these findings are discussed. The network demonstrated the ability to learn key features of cars and disregard irrelevant events, validating the viability of using a spiking neural network for traffic tracking. The thesis highlights the importance of optimizing the trade-off between network size and performance, particularly when working with resource-constraint applications.

Analyzing the usage of material handling equipment (MHE) in a warehouse to increase operational efficiency and reduce costs was the main objective of this master thesis. In order to find inefficiencies and bottlenecks, the investigation examined how the present MHE was used. It also investigated the proper approach to apply for determining the size and type of future MHE. This thesis was conducted as a case study of a warehouse in Stockholm, Sweden, belonging to the Northern Europe branch of a Home Appliance Manufacturing Company (HAMC). The research was based on the analysis of the types, quantities, and use of MHE used by the HAMC, along with an analysis of problems that employees experienced. A literature review was conducted to get a deeper understanding of the different methods used for selecting MHE and how these problems are resolved to improve operational efficiency. During this literature review, four distinct approaches to MHE selection and material handling system design were identified: conventional, analytical, knowledge-based, and hybrid. The literature review also helped in determining which key performance indicators should be tracked and how to compensate for their absence. Modifications were made to the future MHE’s composition based on efficiency criteria to enhance their performance and overall efficiency, reduce operating, maintenance, and leasing costs, and streamline operations. Key modifications included equipment elimination, system upgrades, safety protocols, parameter measurement, utilization analysis, and an MHE allocation tool. Employee participation in the research and the design of the new MHE system was crucial in identifying the real operational problems faced daily in the warehouse. By involving employees in the design process, the new system is tailored to their specific tasks and expectations, ensuring that it will be more efficient. After confirming the research and methodology, the project sponsors were pleased with the outcomes for the new material handling system’s composition. In addition, they committed to track important activities over time and incorporate the data into a simulation tool they were developing so they could compare the two sets of findings.

Future mobile networks like 6G face key challenges, including ultra-reliable, low-latency communications, massive device connectivity, and high data rates. These become increasingly complex in environments with poor signal propagation, leading to inadequate coverage, reduced data rates, and unreliable communication. Reconfigurable Intelligent Surfaces (RIS) have emerged as a promising solution since they can redirect signals and create a virtual Line-of-Sight (LoS), effectively improving coverage and the Signal-to-Noise Ratio (SNR). While most studies focus on two-dimensional (2D) rectangular RIS surfaces, there is no theoretical evidence to suggest that this shape is optimal. An additional challenge arises when the user is in mobility, as the RIS configurations need to be updated to maintain optimal performance. How often these updates should occur remains an open problem. This thesis investigates the effectiveness of various RIS geometries using Sionna, a GPU-accelerated ray tracing simulator developed by Nvidia. The evaluation includes one-dimensional (1D) uniform linear arrays (ULA), 2D uniform planar arrays (UPA), and three-dimensional (3D) shapes. Simulations are conducted across various scenarios, including 3.5 GHz and 28 GHz frequency bands, single and dual-antenna transmitters, and differing mobility conditions. The results indicate that 3D geometries outperform UPA and ULA configurations as they provide higher SNR. Additionally, the study investigates how often the RIS control signaling should be transmitted to change the RIS configuration considering user mobility and demonstrates that transmitting it every two meters of receiver movement is sufficient to maintain high SNR levels. These values correspond to a transmission interval of approximately 1.42 seconds when the receiver walks at an average speed of 1.4 m/s and 0.24 seconds when moving in a car at 30 km/h.

In certain microwave systems where balanced signals are used, it can be crucial to minimize the phase and amplitude imbalance. To convert an unbalanced signal to a balanced signal a balun can be used. However, for extremely wideband applications, for example electronic warfare, it can be impractical and expensive to use baluns which exhibit satisfactory phase and amplitude imbalance across the whole working band. A potential solution, which is investigated in this thesis, is constructing a phase and amplitude compensator, which can be actively controlled in order to improve the phase and amplitude imbalance in a balanced transmission line. A full design cycle was completed to create the phase and amplitude compensator PCB circuit operating between 0.5 − 6 GHz, which improves the phase and amplitude imbalance at one frequency point at a time. The phase and amplitude compensator consisted of a voltage variable phase shifter followed by a voltage controlled attenuator. The phase and amplitude compensator was controlled by a DC voltage bias control network. The phase shifter was realized using a bandpass topology with voltage controlled varactor diodes for variable phase shift. The phase shifter was designed and simulated using the software "Cadence AWR Microwave Office". A COTS voltage controlled attenuator was used. Lastly, an optimization algorithm was devised, to find the optimal settings of the phase and amplitude compensator. The performance of the manufactured phase and amplitude compensator PCB and its subsystems was measured. The variable phase shifter achieved a phase shift of at least ±10◦ with S₁₁ < −15 dB and S₂₁ > −1.5 dB across the whole working frequency band. The phase and amplitude compensator achieved S₁₁ < −12 dB and S₂₁ > −3 dB across the whole frequency band. A real life test attempting to compensate the phase and amplitude imbalance of a balun was performed. The phase and amplitude imbalance was successfully improved at 3 GHz, with the phase imbalance improved from 6.8◦ to 1.7◦, and amplitude imbalance improved from 0.65 dB to 0.02 dB. The successful phase and amplitude imbalance improvement is a proof of concept that a phase and amplitude compensator circuit can be used to improve the phase and amplitude imbalance of a balun.

All receivers, especially wide band ones, are sensitive to strong interfering signals. Due to non-linearities, strong interference can cause unwanted spurious signals to appear in the received band. These spurs can be mistaken for, or obscure signals of interest, giving false information about the received spectrum and the surrounding electromagnetic environment. Additionally, if the signal is strong enough, the receiver may even be damaged. This vulnerability is especially problematic in defence applications, where misleading information can lead to bad decision-making and reliability is highly important. In order to address this problem, an early prototype of a Spectrum sensor has been developed at SAAB (patent pending). This sensor generates DC voltages that are dependent on the power and frequency of an incoming signal. These signals can be used to control a fix filter bank in order to selectively filter out unwanted interference before it reaches the receiver front end. The project at hand is to implement this sensor in such a system. The sensor prototype was first adjusted and characterised such that the frequency of an incoming signal can be determined from the output voltage. A filter bank was designed and manufactured consisting of three different notch filters and a bypass path. Having defined the spectrum sensor output signals and the filter bank input signals, a control circuit was designed and constructed. The control circuit takes the DC voltages from the spectrum sensor and controls the filter bank such that the most appropriate filter is applied to the signal path. The final system was assembled and tested, and achieved robust interference suppression with a response time below 100 ns. While some simplifications were made to the final system, it was able to demonstrate the function of the spectrum sensor, while offering valuable insight into it’s implementation.

School principals play a vital role in guiding schools and shaping teaching practices. While students and teachers use various technologies for teaching and learning practices, the challenge remains in turning raw data into actionable insights for decision-making. Although teachers receive support for their close work with students, less attention is given to principals, who make higher-level decisions and allocate resources. This thesis explores how data visualization tools can help principals gain insights to improve students' reading skills, which have been considerably declining during the last decade, and especially during the last few years. The research examines how data visualizations can support principals in identifying students needing reading assistance and tracking reading progress through a mixed-methods approach. Seven school principals with varying levels of computer experience participated in the study, which involved interacting with prototype visualizations across three rounds. Methods used include Think Aloud protocols, A/B testing, and follow-up interviews. The findings show that well-designed data visualization tools can help principals better identify students struggling with reading. The study highlights the importance of iterative design in improving these tools and emphasizes the need for clear, informative visualizations. Overall, effective data visualization enhances principals' decision-making, ultimately supporting students' reading development.

Generative Artificial Intelligence (Gen AI) models have become very popular, especially after the public release of ChatGPT in late 2022. The demonstrated capabilities of these models have led researchers to use them to build agentic workflows, including Multi-Agent Systems (MAS), and explore their potential. Previous studies that investigate these systems’ coding abilities have shown promising results. However, these studies have mainly used coding benchmarks consisting of simple coding problems, leaving a gap in understanding coding ability of agentic workflows when facing coding problems of greater difficulty. This study tries to fill this gap by developing a MAS and evaluating its coding ability against 120 selected coding problems taken from Kattis, with the difficulty levels ranging from easy to hard. In the experiment, both the developed MAS (configured with Llama 3-70b) and a single Large Language Model (LLM) (Llama 3-70b) were tested on the selected coding problems using Kattis’ assessment system. The experiment showed that the MAS was significantly better than the single LLM when zero- shot prompting was used, increasing acceptance rates by 6.7% and decreasing failure rates by 11.7%. Overall, the results of this study suggest that using a designed LLM-based MAS offers a significant performance boost compared to a single LLM.

This thesis researches if convolutional neural network models can be used to detect and classify multipath signal propagation in radar signals. Specifically, a convolutional neural network model is used to detect multipath propagation in a simulated frequency modulated pulsed radar signal. A combined determin- istic and stochastic signal propagation software model is used to generate a labeled data set that emulates pulses detected by radar warning system located on a vessel in a marine environment. The simulated data set is created and quantified with MATLAB and is represented in the form of in- phase and quadrature signals. The model is trained and evaluated on data that represent no multipath signal propagation, multipath signal propagation, and non-multipath interference (two separate emitters that interfere without multipath) in two separate scenarios for a wide range of signal-to-noise ratios. The machine learning model has an average accuracy of 91.78% on the test set and produces robust results for high signal-to-noise environments for the multipath classes.

In fifth generation (5G) cellular networks, millimeter wave (mmW) bands were introduced. Beamforming is widely used in mmW communication due to its very high path-loss. This paper investigates if beamforming Received Power (RP) measurements of regular Sounding Reference Signal (SRS) can be useful for User Equipment (UE) positioning. So far, precise mmW UE positioning techniques have focused on specialized methods such as dedicated Positioning Reference Signals (PRS) or sending specific transmission pilots. While these techniques are very accurate, they come with their own drawbacks. They require more resources, add complexity, and increase the overall cost. Regular SRS is a fundamental part of 5G communication and is readily available. For the work, a simulation set up of one Base station (BS) and many UE’s give the necessary data. Several methods are evaluated for their positioning accuracy: Neural Networks (NN), Mixture Density Networks (MDN), Kalman Filters, K-nearest- neighnbors (KNN). Results show that a combination of Kalman Filters and MDN works best, having a median euclidean positioning prediction error of 3-4 meters. This is slightly worse then the more complex mmW positioning techniques, but comparable in accuracy to GPS.

This thesis focuses on two simultaneous brute-force password cracking and DDoS attack scenarios: In scenario A, the flows of the two attackers can be analyzed separately based on their IP addresses. However, in scenario B, due to IP spoofing, IP addresses become unreliable, so the mixed attack sequences should be analyzed directly. Conventional intrusion detection systems (IDS) are inadequate for these sophisticated attacks, prompting the need for advanced methods. This thesis studies a novel approach that integrates statistical models into IDS. The method contains two main steps: data source selection (feature selection) and model learning. Four feature sets are selected by different feature selection methods. These selected features are then utilized in statistical models including Hidden Markov Model (HMM) and Long Short- Term Memory (LSTM) model, and can speed up the model learning. This combination aims to predict the start time of the attack and confirm the attack type in both scenarios, enabling a timely response to simultaneous multi- step attacks. Experimental evaluations demonstrate that the models achieve high accuracy in predicting attack start times and steps, with a very low misclassification rate for attack types on the investigated dataset. This method significantly enhances the capability of IDS to handle complex, multi-step, and simultaneous attacks, ensuring better network security.

This study evaluates generative AI’s effectiveness in generating full-stack application code based on detailed requirements. The research question is: “How effectively can generative AI, specifically ChatGPT-4, generate full-stack application code that aligns with detailed requirements specifications?” Despite AI's potential, research in this area is limited. This study explores its strengths and limitations, contributing to both academia and industry. The methodology consists of two main phases. In the first, a case study was conducted to test different prompting techniques in order to identify the most effective one. Four participants with experience in full-stack development and generative AI evaluated the techniques by creating a simple application and assessing them based on ease of use, accuracy, functionality, and code quality. The goal was to select the most effective technique for use in the second phase. The techniques tested included Zero-shot learning, One-shot learning, Few-shot learning, Prompt Chaining, and Multimodal Prompting. Of these, Prompt Chaining proved to be the most effective, increasing detail and accuracy through a two-step process. The second phase, which is the main phase of the study, involved the creation of five different applications in full-stack development. These applications were designed to test AI's ability to generate functional and correct code in more complex development scenarios. They were evaluated based on the criterias: accuracy, responsiveness, functionality and code quality. The results show that AI has the potential to generate code that meets many of the stated requirements, but that there are still limitations. AI showed strong performance in generating fullstack code, accurately capturing requirements and functional elements. However, challenges arose with complex layouts, responsive design, and connecting front-end to back-end. While code quality was generally qualitative, issues with maintainability and CSS practices were noted. This study highlights AI's potential to accelerate development but underscores the need for precise input and additional refinement.

Cooperative Intelligent Transport Systems (C-ITS) aim to create safer, more efficient, and more comfortable transportation systems. A cornerstone of these systems is the reliable exchange of information in Vehicular Ad Hoc Networks (VANETs), facilitated through communication modes such as Vehicle-to-Vehicle (V2V), Vehicle-to-Infrastructure (V2I), and Vehicle-to-Pedestrian (V2P). Wireless technologies like ITS-G5 enable this communication by transmitting messages such as Basic Safety Messages (BSMs) or Cooperative Awareness Messages (CAMs) in Europe, which carry critical data, including speed and position. To ensure secure communication in VANETs, Misbehavior Detection Systems (MDSs) are employed to validate received messages. However, real-world vehicular maneuvers, such as lane changes and turns, introduce significant dynamics into the network, potentially impacting the reliability and robustness of MDSs. Understanding these impacts is crucial to building a secure vehicular communication systems. The problem of evaluating the impact of vehicular maneuvers on MDSs in VANET is both significant and under-explored. Relevant researches often focus on the performances or fusion of algorithms of MDSs in random or idealized scenarios, neglecting the complexities introduced by dynamic maneuvers. This provides us an opportunity to explore MDSs under more dynamic conditions, advancing their reliability for real-world deployment. In addition, deployment of novel attacks gives us a new prospective of evaluating MDSs performance on these attacks. This research employs SUMO for mobility and map simulations, OMNeT++ for communication protocols and discrete event simulations, and the F2MD framework for deploying attacks, simulating BSMs, and implementing MDSs. By combining controlled simulations with statistical analyses, we evaluate the effectiveness of MDSs under varying conditions. Key findings indicate that while maneuvers such as lane changes and turns introduce minor false positives in non-attack scenarios, they have limited impact on the detection of already well-identified attacks. The Behavioral Application, with its time-based punishment mechanism, outperformed other MDSs, demonstrating enhanced detection for gradual attacks. However, attacks designed specifically for maneuvers, such as the lane change mirror attack, proved challenging to detect, revealing vulnerabilities in threshold-based MDSs. Based on our findings, researchers can consider implementing time-based punishment mechanism to their MDSs for performance improvements. These advancements would bring MDSs closer to real-world applications, ensuring robust performance in dynamic vehicular environments.

Nuwiq® is a recombinant Factor VIII (FVIII) product used to treat Hemophilia A, a recessive X-linked disorder characterized by severe FVIII deficiency, primarily affecting men. The product is sold as a lyophilized powder and undergoes extensive quality control before release, including determination of water content. Currently, this is done using the coulometric Karl-Fischer (KF) titration method, a highly accurate and well-established method. However, KF titration requires hazardous chemicals, is time-consuming, labor-intensive, and is sample-destructive. 

An alternative method for water content determination could be Near Infrared (NIR) spectroscopy, a non-destructive, simple and faster method. The goal of this project was to evaluate whether NIR spectroscopy could replace KF titration for water content determination in Nuwiq®. To do this, 30 different batches of Nuwiq® were spiked with varying amounts of moisture in a humidity chamber. Spectra of the spiked samples were collected using a FT-NIR Antaris II in diffuse reflectance mode and the Result Operation software. KF titration was then performed on the sample samples to obtain reference values. 

The collected spectra were used to build a calibration model in TQ Analyst, using Partial Least Squares (PLS) regression and various chemometric techniques to reduce noise, smooth the data and optimize results. The KF reference values were paired with their corresponding NIR spectra to create the calibration model. The model’s performance was evaluated by analyzing 12 different Nuwiq® batches not included in the calibration set. 

The final calibration model revealed strong predictive ability, with a correlation coefficient of 0.99899, a Root Mean Square Error of Calibration (RMSEC) of 0.0362, and a Standard Error of Prediction (SEP) of 0.0346 – suggesting that the model was not overfitted predicted the reference values with high precision. However, in some cases, the model did not predict the water content accurately, indicating a need for further evaluation. 

In conclusion, the NIR calibration model shows potential as replacement for KF titration in routine analysis of water content determination in Nuwiq®. However, the calibration sample set should be expanded to 50-100 to improve robustness, accuracy and precision. 

It is not a matter of if, but rather when, the next viral pandemic emerges. In order to be prepared, methods need to be developed to quickly engineer solutions for diagnostics and treatment of these, and antiviral antibodies are a promising option. If a system can be built to avoid the previously time-consuming and costly process of antibody development, they may have even more potential. As emerging viruses are frequently related to already existing ones, there may be structural similarities in their epitopes that can be taken advantage of. By introducing mutations in antibodies capable of binding these related viruses, one might be able to engineer the specificity towards the novel viruses, leading to more efficient antibody development.

This project was based on BreviA, a recently developed high-throughput interaction analysis system for antibody libraries, enabling the user to collect a large amount of interaction data and antibody sequences in a short amount of time. Two antigens, SARS-CoV-1 and SARS-CoV-2, were investigated, along with an antibody library based on a model antibody called Fab C1. Using BreviA, a set of 1152 antibody sequences and kinetic parameters were collected, resulting in 528 unique variants. The affinities towards both SARS-CoV-1 and SARS-CoV-2 were investigated, in hopes of obtaining a CoV-1 non-binder with high affinity towards CoV-2, which could potentially be used for diagnostic purposes. However, most obtained CoV-1 non-binders were also CoV-2 non-binders, and the few CoV-1 non-binders that retained their CoV-2 affinity showed severely reduced KD values. This may be due to the sequence similarities of the antigens and should be considered in future efforts.

In general, this study evaluated the potential of using the BreviA system to quickly collect enough data to enable machine learning, which in the future could be used to predict or design a CoV-2-specific antibody. Although no actual machine learning was performed, BreviA did enable a uniquely quick data collection process which, in combination with machine learning, could hopefully result in an accelerated antibody engineering process.

The purpose of the study is to develop a scalable innovation process for a case company in the Swedish grocery retail market. In the larger retail market, companies operate with small profit margins in a competitive and regulated market. This increases the need for established innovation processes to capture value long-term. The study offers insights on challenges associated with developing and utilizing innovation processes in a grocery retail company. An iterative literature study is conducted, alongside a semi-structured interview study, which is thematically analyzed to identify key challenges in the case company. The results showcase eight key challenges related to innovation work in retail, including ad-hoc initiatives, organizational silos as well as lack of collaboration and processes for storing ideas. A six step scalable innovation process is developed to address these challenges, supported by an Innovation Platform. To address the key challenges, the innovation process is required to be flexible and robust enough for implementation at full-scale across the case company’s business units to structure the way innovations are developed and implemented. The study's contribution to academia and the practical work in grocery retail includes industry insights and discussions on the development of scalable innovation processes for grocery retail.

As the global energy landscape shifts towards renewable sources, integrating flexible energy systems with existing nuclear power plants offers a promising solution to the challenges of grid stability, renewable energy integration, and decarbonization. This thesis explores the techno-economic feasibility of coupling flexible energy systems with the Oskarshamn Nuclear Power Plant (NPP) in Sweden. The systems analyzed include Pumped Hydro Storage (PHS), Hydrogen-Methanol production, Battery Energy Storage Systems (BESS), and synchronous compensators, all of which play critical roles in balancing energy supply and demand while optimizing economic returns. The study evaluates how these technologies can participate in various energy markets, including day-ahead (DA) electricity trading, Frequency Containment Reserves (FCR), and other ancillary services. The integration of PHS and hydrogen-methanol production is particularly advantageous, as both systems provide substantial revenue streams by responding strategically to electricity prices and system needs. BESS, while less financially viable when solely participating in Fast Frequency Reserves (FFR), contributes significantly to grid stability during peak periods. The synchronous compensator enhances system reliability through voltage and inertia support, generating additional revenues through grid services. In addition to the technical and operational analysis, the thesis examines the financial support available for these systems. Sweden’s incentives, including ERA-Net SES funding for PHS, EU and national grants for hydrogen production, and tax exemptions for renewable energy projects, significantly reduce capital costs and improve the financial feasibility of these technologies. The study underscores the importance of optimizing the operation of each system to maximize profitability while contributing to Sweden’s goal of becoming carbonneutral by 2045. Overall, the findings demonstrate that integrating flexible energy systems with nuclear infrastructure is not only feasible but essential for accelerating the transition to a sustainable, resilient, and decarbonized energy future. This research provides valuable insights for policymakers and industry stakeholders, showing that the combined use of PHS, hydrogenmethanol production, BESS, and synchronous compensators can help address both technical and economic challenges in the evolving energy market.

The hybrid renewable energy microgrid system designed in this thesis addresses the critical energy and water challenges faced by Assumber, a remote village in the Gilgit Baltistan region of Pakistan. The village, home to approximately 900 people, suffers from unreliable electricity, limited heating during harsh winters, and inadequate access to clean drinking water. These issues, compounded by its remote location and lack of infrastructure, necessitate a sustainable solution tailored to the village's unique geographic and climatic conditions. The significance of this project lies in its focus on integrating locally available renewable energy resources, such as micro-hydropower and biomass, into a scalable and selfsufficient microgrid. Despite the global shift towards renewable energy, remote regions like Assumber are often overlooked, and no one has yet implemented a comprehensive energy solution for such isolated, resource-constrained environments. This project was designed to bridge this gap, offering a practical and replicable model that addresses the urgent need for reliable energy in off-grid communities. The problem was tackled through a detailed assessment of the village's energy and water needs, followed by the design of a hybrid system that combines micro-hydropower, biomass gensets, and battery energy storage. The methodology involved simulations using HOMER PRO to evaluate different configurations, ultimately selecting a system that offers continuous energy supply year-round. Additionally, the project includes a reverse osmosis (RO) plant powered by the microgrid, providing clean drinking water sourced from nearby glacial streams. The key result of this project is the successful design of a cost-effective and sustainable hybrid microgrid capable of meeting 100% of the village’s energy needs through renewable sources. This system improves quality of life by providing consistent electricity, heating, and clean water, while minimizing environmental impact. Future projects can now adopt this model to address similar challenges in other remote regions, making sustainable energy and water access a reality for isolated communities.

This chapter focuses on a landmark change in ski wax production that took place in the 1940s. The production of synthetic waxes, developed as a result of scientific testing and engineering, replaced the previously small-scale and experience-based ski wax production. The chapter explores the historical circumstances and processes – such as industrialization and scientification – that affected the technological development of skiing and paved the way for this change.

This article is a study of technological change in cross-country (XC) skiing in Sweden from the late nineteenth century to the 1930s. While technological development in sport is usually seen as a linear and predetermined process, it is instead treated in this context as an arena where wills and intentions – grouped under the concepts of nostalgia and intensification – are negotiated. Nostalgia, in this sense, reflects a scepticism of innovation and change based on contemporary civilizational and rural-romantic concerns, whereas intensification represents the total mobilization of resources to improve sporting performance. The article shows that in its beginnings at the end of the nineteenth century, XC skiing expressed a romantic vision of a distinctive Swedish national landscape where skis were to be made using traditional and old-fashioned craftsmanship. In the early twentieth century, however, XC skiing changed and became progressive. The traditional Swedish skiing landscape was abandoned and the sport adapted to international conditions to enable Swedish athletes to compete successfully abroad. In other words, skiing was intensified because of the mobilization of resources by the Swedish Ski Association in close cooperation with the Swedish ski industry, which was then in an expansion phase.

Energy consumption is increasing, and with it comes an increase in Greenhouse Gas emissions, which threaten humanity. It is pressing to decrease consumption through an increase in the efficiency of all energy-consuming systems. Tighter regulations on heavy-duty vehicle (HDV) emissions in the European Union for a not-so-close future imply that all the systems in this means of transport need to be improved. One of these systems is wheel hub bearings. In previous studies, it has been shown that by simply changing the grease choice, either by varying the thickener type or oil blend, energy losses of bearings can be reduced, thus improving the efficiency of HDV as a system. Finding a methodology capable of such evaluation is part of the process of optimising this line contact tribological system. This thesis aimed at mechanically validate and improve an already existing Cylindrical Roller Thrust Bearing test rig and fully develop the motor control and safety algorithms, sensor integration and respective data acquisition, and data analysis algorithms. Tests were performed as part of the machine's continuous evaluation and improvement process. Concurrently, a methodology was developed to evaluate the energy losses of Cylindrical Roller Thrust Bearings under different loads and drive cycle conditions with different lubricating grease formulations. The imposed test conditions were chosen to align with the conditions experienced by wheel hub bearings of heavy-duty vehicles. The results indicate that grease choice formulations greatly influence energy consumption and that friction energy can vary by 10.1% under drive cycle conditions depending on the grease choice. Recommendations for future work are given to improve the machine and the methodology and pursue more tests to better understand the tribological phenomena of line contact bearings and their impact on HDV wheel hub bearings.  

This thesis investigates non-reciprocity in a microwave frequency comb generated by multifrequency pumping of a superconducting parametric circuit.Non-reciprocal behavior is essential to microwave quantum technologies as it helps to isolate the readout of quantum signals from the back-action of noisy amplifiers.We conduct measurements of linearly independent, orthonormal modes emitted from a Josephson Parametric Amplifier.Mode-to-mode interaction in a frequency comb is achieved by modulating the natural frequency of the Josephson Parametric Amplifier with a suitable pump signal.We provide a detailed derivation of a scattering theory that encapsulates the input-output relations for the modes. We implement the scattering theory numerically to predict the outcome of experiments and guide the measurement configuration.We discuss two different kinds of modulation, high-frequency and low-frequency pumps, using simple setups to illuminate the interplay between the different types of pumps.Experiments with a specific combination of two high-frequency pumps with integer frequency offsets $n_{1,2} > 0$ and one low-frequency pump with a frequency spacing of $n_1 - n_2$ times the spacing of the frequency comb demonstrate non-reciprocity.We show that varying the relative amplitude and phase of the low-frequency pump impacts the strength of the non-reciprocal effect. These experiments demonstrate precise control over, and the ability to externally reconfigure, non-reciprocity.

Software testing is essential to deliver quality code in software development. Code coverage is a metric that measures how much of the code is tested. Code coverage can be calculated in several ways, for example, lines of code activated by the test or percentage of branches explored. Mutation testing is a method that tests the test suites by modifying the code and then running the test suites to see if the modification (called mutants) can be detected. A higher number of mutants detected leads to a higher mutation score. Currently, mutation testing is primarily used in research since it can be costly and complex to integrate into real-life industry software development. Previous research is divided on whether mutation score correlates with real errors in the code but most agree that is a good metric for determining test suite quality. This thesis uses mutation score to assess the correlation with different code coverage metrics for unit tests. Mutation score is compared and analyzed with branch and line coverage through regression analysis to find the correlation, R², and t-test statistic. The mutations are also used to find if there are any recurring deficiencies in the test suites. The implementation of mutation testing in a small-scale industry development is also examined and evaluated. The experiments and evaluation are done in collaboration with a software development team at Trafikförvaltningen Region Stockholm. The team works primarily in the Java programming language in a microservice environment using cloud architecture. The results indicate that line and branch coverage are useful and correlate well with the team’s mutation score. The project’s size and structure worked well with mutation testing, and both the results and the team found value in using it. The existing metrics became more reliable, and some common errors in the test suites were found. Further research could examine integration and system tests and larger-scale code bases.

Transition metal oxides (TMOs) exhibit a wide range of electronic and magnetic properties, making them essential in condensed matter physics. In magnetic TMOs, the ability to tune the magnetic properties offers valuable insights into correlated electron systems and potential functionalities in next-generation materials.

This Licentiate thesis investigates how antiferromagnetic (AFM) ordering can be tuned in powder AReO₄ (A = Mg, Zn), and LiFePO₄ using large-scale facility techniques. We explore how the application of hydrostatic pressure or the substitution of the non-magnetic ion affects the magnetic structure and ordering temperatures.

The work utilises muon spin spectroscopy and resonance ($\mu^+$SR) and neutron powder diffraction (NPD) to probe the magnetic properties of these materials. For LiFePO₄, high-pressure μ⁺SR experiments reveal that compressive strain enhances AFM ordering, contrary to theoretical predictions. For AReO₄, NPD and μ⁺SR suggest two possible AFM spin structures. Our measurements show a remarkably low ordered magnetic moment for both MgReO₄ and ZnReO₄. Bond valence sum (BVS) analysis supports a Re⁶⁺ oxidation state in both compounds, and we attribute the low magnetic moment to strong spin-orbit coupling (SOC).

This thesis demonstrates how NPD and μ⁺SR serve as complementary techniques for investigating complex magnetic systems and how a local probe, the muon, sensing only its immediate environment, can provide insight into macroscopic magnetic properties. The findings contribute to a deeper understanding of the magnetic phase of LiFePO₄ and Re⁶⁺ magnetism in octahedral coordination with oxygen.

The thesis "Business Case Analysis of Mobile and Modular Battery Charging Technology and Service" focuses on evaluating the viability and impact of off-grid, mobile, and modular charging solutions for electric vehicles (EVs). These innovative technologies address traditional grid-connected infrastructure's limitations, such as accessibility, scalability, and grid dependency, especially in urban and remote areas. The research examines technical feasibility, economic viability, environmental sustainability, and market acceptance of such solutions. It highlights how mobile and modular charging systems can provide on-demand, flexible, and eco-friendly alternatives using advanced battery technologies, predictive algorithms, and renewable energy sources. The proposed "Mobile Charging as a Service" (M-CaaS) model enhances EV accessibility, reduces greenhouse gas emissions, and promotes decentralized energy systems. The study employs a comprehensive methodology, integrating technical, economic, and environmental analyses, alongside stakeholder input and market assessments. Key considerations include regulatory frameworks, business risks, and operational strategies. The work ultimately advances sustainable energy infrastructure and informs policy and industry developments in EV charging technologies.

This thesis aims to explore the capabilities of Large Language Models (LLM) for the extractionof PERSON-PLACE relations from biographic text. The use of these models would reducethe required knowledge and training data from established methods and ease the generation ofspatial data from those text sources. An easier extraction method could advance the generationof spatial datasets based on biographies in many areas.With a newly compiled PERSON-PLACE relation dataset based on Wikipedia biographyarticles, the capabilities of an example LLM are compared with the performance of a baselinemethod based on the spaCy Python package. The baseline works by looping through extractedsentences and named entities, while the LLM extraction is based on few-shot prompting. Thecomparison is done by evaluating a sentence-based F1 score and a secondary document-basedFβ. The data set gives insight into the distributions of the types of relations and spatial entities.The results also show that the baseline outperforms the Large Language Model in almost allexamined metrics with regard to the F1 scores. The secondary analysis Fβ reveals more equalperformances. The under performance of the Large Language Model is likely based on thecomplexity of the extraction task, strictness of the comparison, and the rudimentary interactionswith it. However, the novel dataset and developed methods lay the groundwork for furtherresearch in this area. The specific few-shot method and comparison do not show promise forthe future.