The conference pays close attention to questions concerning interspecies relations, aiming to promote multidisciplinary research collaborations. 

The purpose is to gather researchers in a new Nordic network focusing on the study of human-animal relations in narrative forms, such as children’s literature and young adult fiction, film, art, games, fanfiction, fanart, and exhibitions.

Keynote speakers are Zoe Jaques (University of Cambridge), children’s literature scholar within posthumanism and ecocriticism, as well as Helena Pedersen (University of Gothenburg), researcher within critical animal studies and critical animal pedagogy.

When developing an agent-based model (ABM), an important decision that has to be made is whether it should be two- or three-dimensional. 3D is often more realistic, while 2D typically has better computational performance. Since requirements on simulators vary, there is no single correct answer as to whether one should use 2D or 3D. The Unity game engine is popular both among video game developers and simulation developers, and has support for both 2D and 3D. However, there is little research regarding the performance differences between the two. To this end, this sets out to determine the difference in performance between 2D and 3D simulations in Unity. We develop two functionally identical simulators in Unity, one in 2D and one in 3D, and compare their computational performances. The results of this study show that with a low number of agents (500 and less) there is no meaningful difference in performance except a slightly higher memory usage in the 3D simulator. However, larger agent counts (1000 and higher) cause the 2D simulator to use more CPU time than its 3D counterpart. We conclude that the use of 3D simulators is preferable to 2D as it is more realistic and has better CPU performance. However, 2D simulators can still be useful for low agent counts when memory is a concern.

Elections are politically significant events that generate uncertainty in financial markets. This uncertainty influences investor behavior, shapes expectations regarding future policy, and alters the perception of various types of risk. This thesis investigates how election-related uncertainty affects four dimensions of policy and market uncertainty: financial uncertainty, economic policy uncertainty, geopolitical policy uncertainty, and climate policy uncertainty. A particular emphasis is placed on comparing the effects of partisan election uncertainty, which captures the probability that the incumbent party remains in power, with nonpartisan election uncertainty, which reflects the closeness of the electoral race regardless of party identity. The empirical analysis focuses on United States presidential elections, utilizing daily data from multiple electoral cycles and applying a structural vector autoregression model with panel data. The impulse response analysis reveals that partisan uncertainty produces strong and statistically significant effects on both financial and economic policy uncertainty, while nonpartisan uncertainty has minimal impact. Although climate and geopolitical policy uncertainty do not exhibit pronounced short-term reactions, the forecast error variance decomposition shows that partisan uncertainty accounts for a substantial portion of their variation over longer horizons. These findings demonstrate that financial markets react more strongly to expected changes in political direction than to general uncertainty about electoral outcomes.

I uppsatsen analyseras huruvida utredarens föreslagna lagstiftning i promemorian Beskattning av viss avkastning från riskkapitalfonder för beskattning av inkomster som härrör från särskild vinstandel kan leda till en förutsebar beskattning av inkomsten. Analysen görs i ljuset av legalitetsprincipen, som syftar till att bidra till rättssäkerhetsmålet om beskattningens förutsebarhet.

Affärsidén i riskkapitalbranschen är att ett riskkapitalbolag etablerar en riskkapitalfond som har till syfte att investera i portföljbolag. Kapitalet i riskkapitalfonden tillskjuts av externa investerare och anställda i riskkapitalbolaget, vilka inom ramen för sin anställning i riskkapitalbolaget också arbetar med förvaltning av portföljbolagen. Förvaltningen utförs vanligtvis genom förvaltningsbolag som finns inom riskkapitalbolaget. Upp till en viss nivå –en nivå som bestäms när riskkapitalfonden inrättas genom avtal mellan de externa investerarna och de aktiva delägarna som är anställda i riskkapitalbolaget – fördelas avkastning och vinst vid försäljning av portföljbolaget i förhållande till satsat kapital i riskkapitalfonden. Avkastning och vinst inom ett intervall över den nivån fördelas asymmetriskt till förmån för de aktiva delägarna. Rättigheten till asymmetrisk avkastning är vad som kallas för särskild vinstandel.

Hur inkomst från särskild vinstandel ska beskattas har under mer än 15 år varit föremål för skatteprocesser mellan Skatteverket och mottagaren av inkomsten. I Sverige finns det ingen specifik lagstiftning som stipulerar hur inkomsten ska beskattas, utan generella skatteregler har tillämpats. Inkomsten har beskattats i såväl inkomstslaget tjänst som kapital och med tillämpning av 3:12-reglerna. Det har resulterat i en osäkerhet om hur inkomsten ska beskattas, vilket har medfört bristande förutsebarhet för den skattskyldige som mottagit inkomsten.

I promemorian föreslås det att inkomst från särskild vinstandel ska beskattas med tillämpning av 3:12-reglerna, men med vissa avvikande regler från det befintliga regelverket i 57 kap. IL. I uppsatsen analyseras de föreslagna bestämmelserna i ljuset av de aspekter som legalitetsprincipen ställer upp av förutsebarhetsskäl. Den föreslagna lagstiftningen kan i vissa delar anses leda till en förutsebar beskattning av inkomster som härrör från särskild vinstandel, men de föreslagna särregleringarna riskerar att medföra osäkerhet för de berörda skattskyldiga. Ett separat kapitel i IL, innehållande särskilda bestämmelser för beskattning av inkomst som härrör från särskild vinstandel, vore av förutsebarhetsskäl att föredra.

Although scientific research is commonly regarded as objective, prior studies have shown that it is prone to different cognitive biases. Adding to previous research in behavioral economics and meta-science, our study investigates whether students’ methodological orientation influence how they interpret empirical findings, and whether such differences hinder people from objective interpretation.

Using an experimental design, based on Kahan et al. (2017), we presented participants with numerical scenarios that were identical in data but framed differently to suggest either qualitative or quantitative as the correct method. Participants were recruited from Linköping University’s JEDI-lab, encompassing a diverse academic background. To capture methodological orientation, participants got to answer which program they study and whether they primarily engage with qualitative or quantitative methods in their studies. These two measures were then combined into a single variable representing overall methodological orientation. A numeracy test was also included to assess participants’ numerical ability. We examined motivated reasoning by analyzing with binomial logistic regression whether methodological orientation alone influences interpretation, and motivated numeracy with the same regression model by testing the interaction between methodological orientation and numeracy.

The results show no evidence of motivated reasoning or motivated numeracy; neither methodological orientation nor numerical ability systematically influenced participants' interpretations. These findings suggest that methodological polarization does not shape data interpretation among students in this context, and that higher numeracy promotes accuracy rather than reinforcing bias.

As AI continues to expand into new domains, it has become an increasingly researched field in the military sector. At the Swedish Defence Research Agency (FOI), MATLAB/Simulink models are currently used for conducting simulations of different combat scenarios, such as between an aircraft and a missile. However, this method is time-consuming and computationally expensive. To address this challenge, two deep learning models have been explored, aiming at speeding up the trajectory simulations and expanding the result dataset. Two neural network architectures are proposed and evaluated: a CNN-LSTM hybrid model and a Transformer model based on the iTransformer architecture. Using data from almost 45 000 simulated scenarios, the deep learning models are trained on static input parameters, as well as multivariate time series data representing the 3D positions of the aircraft and the missile. The models are trained to predict a continuation of trajectories from initial conditions, and then assessed based on standard regression metrics as well as computational efficiency. Results show that the proposed models capture most trajectory patterns well with a low positional error in meters, and also proves a significant reduction in simulation time compared to the MATLAB/Simulink models. However, the models are not directly comparable due to differences in input sequence lengths, as well as the degree of downsampling applied. In their current state, the proposed models demonstrate a strong potential but may not yet reliably serve as a complementary tool to the MATLAB/Simulink models. Further testing, parameter tuning and validation are needed to ensure both a consistent and reliable performance across a wider range of conditions.

This thesis presents the design and development of a tablet-based touch interface for OpenSpace, an open-source astro-visualization software that can be used for live presenta- tions in planetariums and educational settings. The project was conducted in collaboration with the Scientific Computing and Imaging Institute (SCI) at the University of Utah and Linköping University (LiU) in Norrköping. The primary aim was to enhance the usability of OpenSpace for live presentations by creating a user-friendly interface that allows pre- senters to navigate complex visual environments directly from a tablet, reducing the need for secondary operators. The development process began with an analysis of the current limitations in using OpenSpace for live presentations, specifically the need for a separate pilot to manage the navigation, while the presenter focuses on speaking. The current desktop interface works well when controlled at a computer, but offers a poor experience on touch devices like tablets, making it less effective for presenters wishing to control OpenSpace directly. To address this, the interface was iteratively designed based on user feedback and functional requirements. The development process involved continuous user testing with presenters and technical staff, allowing for on-going adjustments and improvements to the design. Wireframing and prototyping were carried out in the design tool Figma, with early designs evaluated through user surveys to gather insights on the usability of the interface. Based on this feedback, the system integrates both gesture-based interactions and a frame-based interaction method to provide precise control over navigation and scene manipulation. Extra attention was given to balancing usability and flexibility, trying to ensure that the interface could handle complex navigation tasks smoothly. In the end, a evaluation was performed to test the interface in real-world scenarios, which identified areas that were lacking, particularly in refining the precision of touch movements for smaller adjustments. The resulting interface is a responsive touch-interface that allow the presenter themselves to control OpenSpace live during presentations. It is described as intuitive and natural by users and provides a solid foundation for touch control over OpenSpace. However, there are remaining challenges in achieving greater precision for finer movements. There is also room for refining touch sensitivity, improving interaction for complex navigation tasks, and optimizing the interface for different tablet sizes.

 Climate change induced hydrological changes affect the risk of extreme climate and weather events such as floods. To improve climate resilience there is a need for adaptation of the Swedish water management system; the foundation of which is the regulation and authorization of water operations in water bodies of all sizes. This study examines the legislation for water operations and a sample of water permits from the Motala Str¨ om catchment area with the intention of identifying institutional barriers and opportunities for adaptation. First, the thesis reveals a legislative tradition of reluctance to deviate from existing principles and requirements, and review or revoke existing authorizations, allowing thousands of water operations to continue operating using century-old permits, and perpetuating a principle of perpetuity that is incompatible with the principles of sustainability and the adaptive reasoning of the EU Water Framework Directive. Second, the age and lack of accessibility of permits creates a knowledge gap limiting the potential for effective coordination, necessitating coordinated archiving efforts. Third, the absence of environmental considerations and flexibilities for societal or hydrological changes necessitates the review or revocation of older permits to ensure compliance with modern regulation and create climate resilience. Finally, the lack of time limitations and mandatory general revision requirements, and the regulatory uncertainty and financial burden of forgoing permits limits the opportunities for adaptation. The thesis concludes that expanding the regulatory opportunities to forgo permits in emergency situations while conducting mandatory revisions for all water operations would create new opportunities for adaptation.

Dokumentet utgör en rapport över ett kandidatprojekt som genomfördes vid Linköpings universitet inom kursen TDDD96 – Kandidatprojekt i programvaruutveckling under vårterminen 2025. Projektet resulterade i en webbapplikation för att underlätta organisering och kommunikation. Kunden till projektet var studentföreningen FAD, en förening som hjälper till under mottagningsperioden för nya studenter vid Linköpings universitet. FAD efterfrågade en webbapplikation vid namn Fadderportalen. Fadderportalen syftar till att centralisera information under mottagningsperioden i syfte att underlätta kommunikationen mellan de olika parter som är med och realiserar mottagningsperioden. Rapporten beskriver projektgruppens arbetsprocess samt de metoder och teknologier som använts. Den beskriver även de utmaningar och lärdomar som identifierats under projektets gång. Arbetet genomfördes med en agil metodik där kontinuerliga utvärderingar och iterativa förbättringar användes för att säkerställa en fungerande slutprodukt. En central del av utvecklingen var att skapa en användarvänlig plattform eftersom det var ett av de huvudsakliga krav som ställdes från kunden. Fadderportalen samlar och presenterar information på ett tydligt sätt som möjliggör en smidig samordning. Vidare diskuteras i rapporten hur projektet skapade värde för kunden och användarna genom att minska spridningen av kommunikationskanaler och förbättra struktureringen av information. Utvärderingar och tester genomfördes för att säkerställa att Fadderportalen uppfyller användarnas behov och de krav som ställts på produkten. Slutligen reflekterar rapporten över projektgruppens erfarenheter.

Denna rapport detaljerar arbetet på projektet Spacetime – spatial realtids-presentation av drönarvideo för Remote Aero. Projektet utförs som kandidatarbete för civilingenjörsutbildningar inom data- och mjukvaruteknik på Linköpings universitet. Rapporten tar upp hur gruppen har planerat, utfört och reflekterat över arbetet och de erfarenheter som detta har lett till. Kanban användes som process i fokus och var en gynnsam arbetsmetod för ett grupparbete på åtta personer.

Produkten som har utvecklats för kunden innehåller den centrala funktionen: projicering av video från en drönare på en karta i en webbapplikation. Gränssnittet är enkelt designat, och användarvänligt även om produkten kräver en viss mängd teknisk kunskap att köra på grund av simuleringen av video från drönare. Projektgruppen har skapat värde för kunden, som kommer att kunna fortsätta utvecklingen av produkten som finns tillgänglig som öppen källkod. Värdet kommer från uppfyllandet av 76% av de planerade kraven. Projektgruppen har använt sig av simulerad data, men att använda produkten med en riktig drönare har tagits i hänsyn under utvecklingen.

Enhetstest, reliabilitetstest och integrationstest utfördes, samt utforskande tester. 100% av koden för projektionsalgoritmen är testad med hjälp av enhetstest och integrationstest.

This bachelor's thesis details the work process and development of a time reporting system commissioned by the Digitalization Division (DIGIT) at Linköping University. The report addresses questions regarding how to provide value for the client, which experiences may be valuable for future projects, the benefits of creating and following a system anatomy, and the consequences of separating a web application into frontend and backend servers with separate development teams. Key findings include that the use of a Scrum framework with clearly defined sprint tasks contributed to increased productivity. Furthermore, the division between frontend and backend imposed high demands on communication between the teams but also enabled efficient parallel work. An important technical lesson was that early planning of data types and database schema facilitated implementation for both frontend and backend. The final product's usability was measured according to the SUS method to an average score of 79, indicating a good user experience.

Aseptic loosening, the failure of prosthetic fixation, is a common complication following joint replacement surgery and often requires revision procedures. This condition is linked to cellular stress caused by mechanical loading, which alters cellular behavior. One consequence is a change in the release of extracellular vesicles: under normal conditions, cells predominantly release microvesicles, but under increased mechanical load, they instead secrete exosomes. Additional intracellular responses include elevated ATP release, altered expression of key proteins, and shifts in intracellular signaling. Key components of this cellular response include Ca2+ homeostasis, where SERCA2, an intracellular calcium pump, plays an important role, as well as the ATP release and the expression of ACP5/TRAP, a marker linked to bone remodeling. In this thesis, a mathematical model of a hematopoietic cell was developed using ordinary differential equations (ODEs) to simulate intracellular dynamics under mechanical loading. Due to the complexity of biological systems, the model was constructed part-wise, with iterative hypothesis testing aimed at aligning the model with experimental observations. The available data reflected two mechanical loads, high and low, which guided model development toward capturing how different levels of stress affect cellular output. Under the project, various models were developed, and several were rejected during the process. The hypothesis about how the system works was modified under the process, and additional biological states were added, which led to a more complex model. In the end, a model that could predict some of the key states well was developed. The model provides valuable insight into how the intracellular mechanisms of a hematopoietic progenitor cell react to mechanical load, but there is still a lot that can be improved or expanded upon. Further refinement of the model structure would be required to investigate whether the switch between the loads could be identified. As the model developed in this project is the first of its kind, there are endless possibilities for how it can be extended. Some of the main weaknesses of the final model are the high uncertainty in predicting certain states and its inability to predict the expression of active SERCA2, a particularly important state. Despite its limitations, the model represents a novel and promising foundation for future work, with broad potential for extension and application.

Säkerhet är en av de största utmaningarna inom modern mjukvaruutveckling, delvis till följd av industrins omfattande användning av tredjepartskomponenter. Dessa beroenden kan introducera sårbarheter om de inte hanteras korrekt.

På uppdrag av kunden Ida Infront vidareutvecklades Dependency-Track, ett verktyg för att spåra och hantera mjukvaruberoenden samt identifiera säkerhetsbrister.

Projektets syfte var att anpassa verktyget efter kundens behov och arbetsflöden för att stödja säkrare mjukvaruutveckling. Fokus låg på att förbättra notifikationssystemet samt införa automatiseringar i webbapplikationen.

Projektet bedrevs enligt en Scrumban-metodik och organiserades i sprintar med Trello som stödverktyg. En systemanatomi togs fram och flera metoder för erfarenhetsfångst användes. Arbetet inkluderade även en hållbarhetsanalys samt ett nära samarbete med fri programvara.

Resultatet blev en förbättrad version av Dependency-Track med utökad funktionalitet och ökad användbarhet, vilket möjliggjorde en effektivare hantering av sårbara beroenden. Projektet gav även värdefulla erfarenheter av agila arbetssätt, nya verktyg och tekniker samt samarbete. Systemanatomin bidrog till ökad förståelse för systemets arkitektur och underlättade både implementation och kommunikation. Dokumentationen producerades med Typst, ett tidseffektivt alternativ till LaTeX.

Avslutningsvis gav arbetet med fri programvara viktiga insikter i öppen källkod och betonade betydelsen av dokumentation, transparens och samverkan med externa aktörer.

Denna kandidatrapport redogör för utvecklingen av en webbapplikation för interaktiv positionering av hjärtklaffar i 3D--renderade CT--bilder, med syfte att stödja klinisk personal vid planering av hjärtkirurgi. Projektet genomfördes vid Linköpings universitet inom kursen TDDD96 Kandidatarbete i programvaruutveckling och i samarbete med Cordicity AB. De frågeställningar rapporten syftar till att besvara är hur värde för kunden kunde skapas vid utvecklingen av applikationen, vilka erfarenheter som kan dokumenteras samt hur en systemanatomi kunde stödja utvecklingsprocessen. 

Arbetet inleddes med en behovsanalys genom möten med kunden och framtagande av en kravspecifikation, uppdelad i hög-- och lågprioriterade krav. En systemanatomi togs fram för att visuellt beskriva applikationens arkitektur, som omfattade klient-- och serverfunktioner för inläsning av DICOM--, NIFTI-- och VTK--filer, 3D--visualisering med VTK via Trame samt ett användargränssnitt baserat på Vuetify.

Utvecklingen organiserades med Kanban, parprogrammering och kodgranskning i GitLab. Resultatet blev en applikation som innehåller både 3D--volym-- och planbaserad navigering och låter användaren välja, rotera och translatera STL--modeller av hjärtklaffar. En SUS--utvärdering genomfördes som gav applikationen ett resultat på 87,1, vilket tyder på en användarvänlig applikation. 

Projektet skapade värde för kunden genom att utforma en kravspecifikation, sätta upp relevanta kvalitetskrav och överlämna lättförståelig vältestad kod. De erfarenheter som gruppen dokumenterade var de positiva effekterna av att använda Kanban och regelbundna möten samt tekniska erfarenheter gällande de verktyg som använts. Den systemanatomi som framställdes upplevde gruppen gav ett bra stöd i den initiala fasen av projektet. 

Interoperability within the healthcare sector means the ability to share medical data in an organized manner. To achieve interoperability there is a need for structures and standards on multiple levels, including the semantic level. openEHR is an open framework that works to improve this area by providing technical standards for data within an electronic health record, but these standards are not yet widely used in Sweden. There is, however, already a lot of medical data being produced. One method used for writing reports within the healthcare sector is Structured Reporting (SR). These reports get their structure from SR templates and the fields in these templates can in theory be matched to fields in an openEHR template to make the report openEHR compatible. The process of matching fields between SR templates and openEHR templates is known as mapping.

This thesis project investigates what the mapping process looks like today, what difficulties are associated with it, and how they can be alleviated. After interviewing informaticians it was found that the work they do is still largely manual in nature and could benefit from partial automatization. When two fields are being mapped, the person doing the mapping has to write out or copy the path to each field. This can be a time consuming task, especially for more complex templates where a deep structure creates long paths. From this it was decided the thesis would develop a tool with a graphical user interface which can produce and store the correct paths for mapped fields. Another issue with the state of mapping is the differences between SR and openEHR. A secondary focus of this thesis is thus looking at to what extent mapping is possible and why any unmapped fields could not be mapped.

The resulting tool is a web application that can produce a mapping configuration file based on user input. The application takes SR and openEHR templates as input and after pre-processing, displays them in two separate tree structures. The user can then click on a node in each tree and map these to each other. The main benefits of the web application are the simple graphical user interface, the automatic generation of paths, and the pre-processing of the SR templates. To further reduce the manual labor associated with the mapping process future efforts should be made on implementing artificial intelligence and natural language processing. However, it should be stated that the real goal is for medical data to be openEHR compatible to begin with, completely erasing the need for mapping in the first place. 

The mappings possible in the application did not achieve 100% coverage for any of the SR or openEHR templates used in this project. While some of the disconnect can be erased by implementing further functionalities in the application, the results still show that there is currently a disconnect between the two structures. Neither SR nor openEHR has a widespread use, and as more work is done on implementing both SR and openEHR in healthcare there should be an awareness of the other structure to reduce the disconnect at the source. The effect that SR and openEHR have on interoperability today is therefore hard to conclude.

Small fluorescent probes serve as the primary tools for studying protein aggregates that occur in neurodegenerative diseases including Alzheimer’s disease (AD). Research on amyloid pathology benefits from bi-thiophene-vinyl-benzothiazole (bTVBT) analogues because their conjugated backbones produce unique photophysical characteristics. The development of probes for biological targets continues as an active area of research. In this project, a library of bTVBT analogues was synthesized by modifying the R1 moiety of the bTVBT2 scaffold to study their optical properties and binding characteristics. Among this library, 18 new analogues with phenyl, pyridyl and fused heterocyclic R1 cores were successfully synthesized through a two-step synthetic method. The first Suzuki coupling reactions produced aldehyde intermediates at yields between 19% and 89% while the subsequent condensation reactions produced final products with yields between 5% and 57%. Then the research examined how R1 core modifications affected photophysical responses through spectral detection and differentiation of protein aggregates in AD. The staining of AD brain tissue showed that meta-substituted analogues together with some pyridinyl-based compounds failed to bind specifically but a major portion of synthesized probes including para-substituted phenyl analogues and one meta-substituted phenyl derivative (Ligand 9) successfully stained protein aggregates. The effective probes showed discrimination between Aβ plaques and neurofibrillary tangles (NFTs) while most of them displayed strong tau selectivity. This research describes the synthesis of bTVBT2 analogue libraries while showing their photophysical properties and their application in staining AD brain tissue to identify potential candidates for future development.

This thesis presents the first successful demonstration of atomic layer deposition (ALD) of gallium oxide (Ga₂O₃) thin films using a homoleptic tris(1,3-diisopropyltriazenide)gallium(III) precursor with water as the oxidant. Driven by the need for cleaner, safer, and more controllable gallium sources, the work extends triazenide chemistry which was previously validated for GaN, InN, and In₂O₃ to Ga₂O₃, filling a gap in Group 13 oxide research. 

The process is self-limiting, achieving a growth per cycle (GPC) of ~ 1.5 Å at precursor pulses ≥ 2 s. A narrow ALD temperature window was identified between 400 and 415 °C, with deposition behavior exhibiting three distinct regimes across 395 – 420 °C, including precursor desorption at the upper bound. Film thickness scaled linearly with number of cycles following an initial nucleation delay of ~ 24 cycles. 

Time-of-flight elastic recoil detection analysis (ToF-ERDA) confirmed near-stoichiometric Ga₂O₃ containing trace amounts of C, H, N, and Cl, all of which decreased after annealing. Grazing Incidence X-ray diffraction (GI-XRD) and scanning electron microscopy (SEM) revealed that annealing at 700 °C converted amorphous films into phase-pure, (111)-oriented β-Ga₂O₃ with enhanced grain structure. As-deposited films were highly transparent (> 96 %), and their optical bandgap increased modestly from ~ 3.74 eV to ~ 4.0 eV upon annealing. Electrical conductivity rose to 1.05 S/cm after annealing, driven by crystallization and improved stoichiometry; one sample showed anomalously high conductivity due to unintended Cu contamination. 

Overall, Ga(triaz)₃ emerges as an essentially carbon-free, thermally stable, and water-compatible precursor that enables high-purity Ga₂O₃ films at moderate temperatures. These results lay a foundation for future exploration of gallium triazenide ALD, including doped and ternary Ga₂O₃ systems and plasma-enhanced variants, with promising applications in high-power electronics and UV photonics. 

Sweden's upper secondary school curriculum does not explicitly state that existential questions should be discussed in a classroom setting. However, studies have shown that such questions are important for young adults and their well-being, as they help students become aware of existential issues and provide them with tools to reflect on their existence. The purpose of this paper is to explore the potential of using fiction to illustrate existential questions in upper secondary education. Furthermore, the paper aims to investigate this by using James Joyce's short story The Dead. As there are no previous studies of this exact kind, a literary analysis of The Dead is conducted based on Jean-Paul Sartre's existential philosophy. This will be followed by a didactic analysis grounded in the literary findings and supported by the didactic theories of Louise Rosenblatt and Judith Langer. The literary analysis shows that The Dead can be interpreted as containing existential themes, and that Sartre's concepts of freedom, anguish, nothingness, and abandonment are evident in the story. The didactic analysis indicates that these concepts can be used to discuss existential questions in the classroom. In the paper's discussion, the book discussion is presented as a possible method to use in the classroom to discuss the short story and its existential themes. The conclusion suggests that working with The Dead and existential questions can contribute to the students' well being.

This thesis examines how the feminist neologism ``lao deng" functions as a counter-discursive tool in Chinese digital culture. Based on 199 posts from Douban, Xiaohongshu, and Weibo, and analyzed using thematic analysis and Feminist Critical Discourse Analysis (FCDA), the study shows how feminist users deploy emotionally charged strategies to critique patriarchal authority and construct resistant subjectivities. Situating the findings within counter-discourse and affect theory, the research argues that ``lao deng" reflects an emergent feminist practice that uses everyday language to perform critique, build solidarity, and normalize resistance. The thesis highlights how feminist discourse in China adapts to censorship through affective creativity and offers pathways for future research on digital feminist expression.

Detta dokument är en rapport som beskriver arbetet som genomförts i kursen TDDD96

- Kandidatprojekt i programvaruutveckling vid Linköpings universitet. I projektet framställdes en modulär webbapplikation som kan användas för att skapa webbapplikationer där man kan träna på quiz genererade av en AI. Generering av quiz sker via artificiell intelligens och agerar som ännu en modul för att skapa webbapplikationen. I rapporten resoneras det kring hur projektet skapats så att det i framtiden kan vidareutvecklas och förbättras samthur projektgruppen skapat värde för kunden. I rapporten diskuteras även hur skapandet av en systemanatomi ger värde för arbetsprocessen. Slutligen tar rapporten upp hur erfarenheterna projektgruppen fått från projektet är intressanta för framtida projekt.

This bachelor thesis presents the EagleEye project, a real-time visualization tool designed for multi-camera systems, developed for Axis Communications at Linköping University. The aim of the project was to create a system demonstrating the analytical capabilities of Axis cameras, specifically leveraging their new features for geo-coordinate metadata. Through low-latency video streaming, movement analysis via heatmaps, and alarm functionalities, EagleEye effectively showcases the commercial potential of these cameras.

The project was executed using agile methodologies, combining iterative sprints with Kanban, pair programming, and workshops to ensure efficiency and quality. The EagleEye system comprises a Python-based backend and a TypeScript/React frontend, integrating third-party tools like Mosquitto for MQTT communication and RTSPtoWebRTC for converting camera RTSP streams to WebRTC, achieving latency below one second. An initial system anatomy laid the technical foundation, while regular meetings with Axis communications representatives ensured alignment with their requirements.

EagleEye facilitates efficient product demonstrations of Axis communications camera systems through metadata visualization, highlighting analytical capabilities and commercial applications. Project experiences underline the importance of flexible planning and clear role distribution.

Road traffic injuries present the leading cause of death among young adults, with efforts to improve road safety and educate safe drivers constantly ongoing. This study examines how road commentary in a simulator training environment could contribute to facilitating safe driving behaviors, by investigating the effects of road commentary training in a simulator environment on the hazard perception, risk assessment and self-evaluated risk awareness of Swedish learner drivers. Previous research has found road commentary training and driving simulators effective for improving novice drivers’ abilities, but the combination of expert-produced road commentary training while performing a secondary task such as simulator drives remains largely unexplored. A mixed methods comparative test was conducted, with 48 participants completing a driving simulator course consisting of a training route and a test route. Half of participants received expert-produced road commentary training during the training route, whereas remaining participants did not. Each participant also self-assessed their risk awareness prior to the training route and after the test route, as well as asked to assess risks in the environments. Driving performance during the test route was compared in regards to hazard detection. No differences were detected between the groups on hazard perception and self-evaluated risk awareness. However, the test group was more prone to link different risks assessed to each other, indicating a greater understanding of traffic situations. Future research is needed to support the findings of this study and to further investigate how expert-produced road commentary training in a simulator environment can foster safe drivers.

Detta kandidatarbete, genomfört i samarbete med Karolinska Universitetssjukhuset, syftar till att utveckla ett webbaserat användargränssnitt som ett konceptbevis för interaktiv visualisering av patientjournaler inför Region Stockholms planerade övergång till openEHR-standarden för bl.a. långtidslagring och åtkomst av journalanteckningar från det befintliga journalsystemet TakeCare innan detta avvecklas. Projektet kombinerar teknisk utveckling med ett starkt fokus på process och arbetssätt. Med hjälp av ramverken Svelte och Tailwind-CSS skapades ett gränssnitt som möjliggör effektiv navigering, filtrering och visuell analys av journalanteckningar via bland annat listvyer, en dynamisk tidslinje och ett canvasläge. Parallellt utvecklades en välstrukturerad arbetsprocess baserad på de agila metoderna Scrum och Kanban, där dagliga standups, sprintplaneringar, retrospektiva utvärderingar och kontinuerlig kundkontakt bidrog till ett systematiskt och lärorikt arbetssätt.

Resultatet blev en fungerande prototyp som uppfyller kundens krav på prestanda, responsivitet och funktionalitet, och som tydligt demonstrerar potentialen i ett modernt gränssnitt baserat på OpenEHR. Utöver den tekniska leveransen resulterade arbetet även i en fördjupad förståelse för hur agila metoder kan tillämpas och anpassas i ett verkligt utvecklingsprojekt. Genom löpande utvärderingar och mätningar kunde gruppen förbättra sin interna samverkan, planering och kvalitetssäkring.

Arbetet visar hur en väldefinierad utvecklingsprocess i kombination med tekniskt fokus kan leda till både ett konkret och användarcentrerat system samt värdefulla insikter för framtida mjukvaruprojekt. Rapporten beskriver inte bara det tekniska resultatet, utan även den arbetsprocess och de lärdomar som utvecklingsarbetet genererade.

Monitoring wildlife is essential for biodiversity research and for managing threats to agriculture and forestry. However, manual analysis of large volumes of camera trap images is both time-consuming and inefficient. This thesis presents the development and evaluation of a machine learning model designed to detect Swedish wild animals in natural environments. The aim is to support farmers, foresters, and cattle owners in protecting their livelihood by enabling real-time animal classification. The model focuses on recognizing key species such as predators (foxes, wolves, lynxes, bears, and wolverines), ungulates (roedeer and moose), large birds, wild boars, and hares. The work began with a detailed analysis of available data sources, including imagery from Qulinda, Artdatabanken, eBird, GBIF, and MS COCO, to construct a robust and diverse training dataset. Multiple YOLO versions, ranging from YOLOv8 to YOLO12, and their variants were evaluated based on their detection accuracy, inference speed, and class generalization. YOLO11s was ultimately selected for its balance of performance and efficiency. Iterative experiments with class structures and training configurations refined the model further. Evaluation results show reliable detection performance across various environments, including scenarios with out of distribution objects such as humans, vehicles and other animals. This work offers a structured approach to wildlife detection model development and provides considerations for deploying AI-based support tools in agriculture, forestry, and wildlife monitoring.

Post-quantum (PQ) cryptography bases its security on mathematical problems believed to be hard even with respect to quantum computers. This thesis evaluates the performance of the Transport Layer Security (TLS) handshake when using hybrid PQ digital signatures that combine two or three PQ schemes. A signature combiner supporting 17 standardized or NIST-proposed schemes is implemented, enabling both concatenation and strong nesting for an arbitrary number of signatures. The Mbed TLS library is extended to integrate this combiner into its handshake process.

Benchmarking a wide range of hybrid configurations reveals that public key and signature sizes are major bottlenecks. Practical size limits for handshake messages in TLS implementations and imbalances in scheme performances severely restrict the set of usable hybrids. However, a small number of two- and three-scheme hybrids appear to be relatively viable. While there are two-scheme hybrids with only marginal performance overheads relatively suitable for TLS, combining three or more schemes leads to more significant performance overheads. While these overheads may be acceptable depending on the application, larger hybrids are deemed to be more immediately suited for use cases with infrequent, long-duration connections, such as IKEv3 or SSH.

This project explores how skeletal tracking can be used to create a cross-country skiingexergame that gets people moving in an enjoyable way. The game is created in Godot, runsdirectly in a web browser, and uses only a standard webcam as input for the skeletal track-ing neural network. The player’s real-life movements are used to control the game andgenerate forward motion. Players switch techniques based on the terrain, avoid obstaclesby balancing, and aim to ski as far as possible within two minutes.The goal was to design an experience that feels like a game while still providing aworkout at a moderate- to vigorous-intensity level. Another goal was to place the playerin a peaceful, natural environment that doesn’t feel like a typical screen-based activity. Wetested the game with 14 participants and measured perceived exertion (BORG), emotionalresponse (PANAS), and enjoyment (PACES). Most players reached the target intensity leveland reported that they had fun. The results show that it is possible to build a skiing ex-ergame using skeletal tracking that provides a workout at moderate- to vigorous-intensitywhile also being enjoyable to play.

Applications of forensic interest are frequently analyzed to understand functionalities that may impact the outcomes of digital forensic investigations. However, when new versions of an application are released, prior assumptions about its functionality may no longer be valid. Performing a full manual analysis of every version of all relevant applications would entail an unreasonable amount of work. To counter this, an automated analysis would improve the quality of investigations performed while maintaining a reasonable workload.

This thesis presents a novel approach for categorizing changes identified between two versions of an Android binary. Conducted on behalf of NFC (The Swedish National Forensic Centre)—a division of the Swedish Police Authority—the thesis presents a tool that streamlines the reverse engineering process by automatically categorizing changes found, using only the two Android binaries as input.

The tool is built upon the Ghidriff diffing engine, which leverages Ghidra to identify changes between the two binaries. For categorization, regular expressions are used to map detected operations to a set of predefined categories, based on common symbols and function names found in Android source code. Each change is marked as critical or non-critical based on its forensic relevance, and changes that fall outside the predefined categories are set to uncategorized.

The performance and accuracy of the developed tool were evaluated by comparing its output against a manual analysis of source code changes from Signal's Android GitHub repository across three release diffs. The tool detected 91.4% of the changes, with a categorization accuracy of 74.3%. Across the three release comparisons used as test data, only 3.3% of unique changes were marked as uncategorized, indicating they did not match any predefined category. As a result, the thesis presents convincing evidence that the proposed tool is effective both in filtering out insignificant changes and in highlighting meaningful ones.

The report describes the project Medarbetarpuls och Arbetsmiljöverktyg which is conducted as a part of the course TDDD96 - Bachelor's Project in Software Development. The project was commissioned by the company Personalkollen AB and aimed to develop a user-friendly survey tool to efficiently collect and analyze data from employees. The report examines how a survey tool is implemented to create value for the customer, what experiences may be relevant for future projects, and what support can be gained by creating and following up on a system anatomy. Furthermore, it explores how valuable trend analyses can be conducted and how the response development of anonymous individuals can be tracked. The goal of the project was to create a survey tool that encourages employees to participate in surveys and intuitively visualizes the collected data.

By gathering and defining customer requirements alongside creating a user-friendly design for anonymous data collection and analysis, the survey tool delivers tangible value. Usability tests indicated a positive reception, with an average SUS score of 85.94, reflecting its simplicity and user-friendliness.

Key experiences emphasize the importance of selecting tools based on functionality over prior experience, with Django chosen for the backend to enhance security and scalability. The advantages of structured methodologies, such as Scrum and pair programming, are highlighted, along with the necessity of a clear system anatomy to support requirements analysis and design.

A well-defined system anatomy aids in identifying central components and fostering a shared understanding within the team. To facilitate valuable trend analyses, standardized questions are implemented for comparability and measurability of responses. By filtering data and applying statistical measures, insights into temporal and group-related changes are achievable. The system effectively tracks changes in individual responses through anonymous IDs and flags anomalous response patterns, ensuring anonymity.

While anonymous responses can encourage employees to voice their opinions, the report also discusses the challenges anonymity poses, such as difficulties for managers in following up on specific issues and the increased risk of unresolved conflicts.

In recent years, the banking sector has experienced significant turbulence, with several banks collapsing due to financial weakness or loss of confidence for the banks, triggering bank runs. This has revived interest in understanding the factors that influence bank stability. This study investigates the impact of bank size and revenue diversification on insolvency risk within the Swedish banking sector.

The Swedish market is uniquely characterized by a few dominant banks, high household indebtedness, and the presence of "savings banks" with unique business models. As far as we are aware, the extensive existing literature on bank insolvency risk has not focused exclusively on a Swedish context and captured the unique features of the Swedish banking sector. To investigate this relationship between bank size and revenue diversification and insolvency risk, we use the Z-score, a widely used measure to capture the insolvency risk in banks and Correlated Random Effects (CRE) regression is used to analyse this relationship. The CRE model used in this study accounts for both within and between bank effects. The sample is an unbalanced panel consisting of 58 banks active in Sweden between the years 2005-2023. We find no robust evidence that revenue diversification or bank size directly affect insolvency risk for banks operating in Sweden. However, we find savings banks having a lower insolvency risk than other types of banks. These findings are fully robust when adding macroeconomic control variables. 

This thesis investigates the relationship between institutional ownership and stock price volatility among companies listed on the Swedish OMXS30 index during the period 2010 to 2024. Using GARCH models, daily volatility is estimated and then converted into monthly volatility. Through panel data regressions, the analysis explores how different categories of institutional ownership affect volatility. The categories examined are total, active, neutral, and passive institutional ownership. The study is conducted while controlling for firm-specific factors, such as market value and leverage ratio, as well as time-fixed effects. The results show that total institutional ownership has no significant effect on volatility, except in the case of financial firms. A higher proportion of active institutional ownership is associated with lower volatility, particularly for financial companies, while passive institutional ownership tends to increase volatility in firms that are neither in the financial nor industrial sectors. This suggests that the impact on volatility differs across ownership categories. The findings deviate from much of the previous research and can be explained by structural characteristics specific to the Swedish market. The study contributes to a deeper understanding of how different forms of institutional ownership influence market dynamics and offers valuable insights for both investors and policymakers.

Denna studie undersöker eventuella samband mellan ESG-betyg (Environmental, Social, Governance) och finansiell prestanda och volatilitet i portföljer optimerade enligt Markowitz teori. Syftet är att analysera hur ESG påverkar den riskjusterade avkastningen och volatiliteten inom fyra sektorer på den svenska aktiemarknaden. Dessa sektorer är industri, konsumentvaror, sjukvård & läkemedel samt finans. Den analyserade tidsperioden är mellan januari 2020 till december 2024. Studien bidrar till ett forskningsgap genom att koppla samman ESG-analys med portföljteori och volatilitet på sektorsnivå. Kombinationen av GARCH och Markowitz portföljoptimering möjliggör en djupare analys av både tidsvarierande volatilitet och effektiv portföljsammansättning, eftersom investerare i högre grad söker strategier som kombinerar hållbarhet och finansiell prestation.

Det empiriska tillvägagångssättet utgörs av kvantitativa metoder, där portföljer konstrueras med hjälp av ett "Best in Class"-urval baserat på ESG-betyg från LSEG Data & Analytics. Portföljerna optimeras enligt Markowitz portföljoptimeringsteori och volatilitetsperspektivet analyseras med Generalized Autoregressive Conditional Heteroskedacity (GARCH).

Analysen visar att ESG påverkar den riskjusterade avkastningen i tre av fyra sektorer, där portföljer med höga ESG-betyg uppvisar en högre sharpekvot. I tre sektorer visade sig även den förväntade avkastningen vara högre i de portföljer med ett högt ESG-betyg. Samtidigt framkommer det att effekterna varierar stort mellan sektorer och att ESG inte kan isoleras som en kausal förklaringsfaktor. Volatilitetsanalyserna visar inte på något konsekvent mönster mellan ESG-betyg och risknivå. I sektorerna sjukvård och läkemedel samt industri är volatiliteten lägre i portföljerna med ett högt ESG-betyg, men i sektorerna finans och konsumentvaror är förhållandet tvärtom. GARCH-analyserna indikerar att portföljernas känslighet för marknadsschocker och uthållighet i volatilitet påverkas av investerarpsykologi, såsom flockbeteende och förlustaversion kopplat till benägenheten att investera hållbart.3

The integration of Large Language Models with symbolic planners has emerged as apromising approach for automated task planning. This thesis investigates how multimodalLLM-based systems, incorporating both textual and visual input, can be used to generateaccurate Planning Domain Definition Language (PDDL) problem descriptions.

Building upon an existing planning framework, this work evaluates the performanceof various model types in generating accurate PDDL problem descriptions. The evaluatedmodel types include open- and closed-source versions of Large Language Models, VisionLanguage Models, and Reasoning Language Models. Additionally, two prompting strategies are compared: direct prompting and Chain-of-Thought prompting. The study alsoexamines the effect of using image captions and bounding boxes as input in the selectedframework. Experiments are conducted across four domains, including a Forklift domainspecifically created for this study.

Results show that closed-source models such as GPT-4o significantly outperform opensource models, and that Vision Language Models with image input can outperform LargeLanguage Models and Reasoning Language Models that rely solely on textual input. Removing image captions had minimal effect on performance, likely because the object descriptions from the caption model added minimal additional information. In contrast, removing both captions and bounding boxes led to a drop in performance, highlighting theimportance of accurate object detection. Additionally, the results showed that CoT prompting does not consistently improve performance.

This thesis presents the design of a system capable of handling trading orders within a cloud-hosted trading platform. The motivation stems from the nature of cloud environments, where network latencies are unpredictable and connections vary between clients. In such an environment, a real-time trading system requires an additional layer to prevent incoming orders from unfairly jumping ahead or falling behind due to latency differences. The proposed solution centers on a batch processor that temporarily stores and sorts orders using a selection of data structures, including Skip Lists and various balanced trees, aiming to optimize both fairness and execution speed. The processor delays order execution until all expected orders from a given time window have arrived. This delay is padded to ensure sufficient processing time before dispatching orders to the trading server. The system was tested in a simulated environment using high-frequency mock orders under controlled gateway latency scenarios. Results show that the system significantly improves fairness, particularly when batch durations are not perfectly divisible by gateway latencies. Among the tested data structures, Skip List consistently provided strong performance with minimal processing overhead. The work was conducted at Vermiculus Financial Technology in Stockholm, with the broader objective of demonstrating that fair and efficient cloud-based trading is feasible. The findings provide a proof of concept that, with appropriate latency compensation mechanisms, trading can be reliably hosted in the cloud.

AI chatbots are getting incorporated into every aspect of life and the bond between humans and AI is becoming increasingly psychologically loaded. This thesis explores how interactions with a chatbot may impact users' real-life regarding life satisfaction and emotions, sociability, and personality. A mixed-methods design was applied on an online survey with the quantitative data coming from the Oxford Happiness Questionnaire (OHQ) and the Ten Item Personality Measure (TIPI), and the qualitative data coming from a thematic analysis on the open-ended questions. Results showed that frequent chatbot users reported a significantly lower overall life satisfaction than infrequent users (t(41) = -3.34, p = .001.), despite perceiving the chatbot as emotionally supportive. Users who frequently replaced real-life interactions with chatbot interactions also displayed a significantly lower score on the OHQ (t(41) = 2.86, p = .007). Participants displayed themes of emotional comfort, perspective taking, emotional trade-offs and ethical concerns surrounding chatbot usage in their open-ended responses. These findings suggest that there is a potential for psychological consequences when frequently using chatbots, emphasizing the need for responsible chatbot integration into society. 

Characterizing the stoichiometry and surface roughness of tantalum carbide (TaC) thin films is typically expensive, time-consuming, and dependent on specialized equipment. This thesis investigates whether optical methods, specifically light scattering and reflection ellipsometry, can reliably distinguish between TaC samples with varying structural and chemical properties. Six samples were analyzed, including both commercially obtained films and CVD-grown specimens. The light scattering measurements revealed consistent differences in angular spread and reflectance behavior that likely correlate with variations in surface roughness. Similarly, the ellipsometric measurements showed clear differences in the refractive index among the samples, which are interpreted as indicative of stoichiometric variation. These findings suggest that it may be possible to construct a reference database that connects optical responses to physical properties, enabling fast and low-cost screening of TaC films in both research and industrial settings.

There have been significant advances in high resolution 3D reconstruction with the use of Gaussian Splatting. With this technique multi-view high resolution images can be reconstructed into high resolution 3D scenes by first building up a point cloud and optimizing 3D Gaussians projected on the point cloud, minimizing the difference between images and the rendered scene. Recent research has delved into using generative models to modify reconstructions created with Gaussian Splatting. Previous work is either limited in the scale of which scenes can be generated or focused on editing parts of existing reconstructions. To address the limitations, we propose a framework of iterative scene generation with the purpose of extending 3D reconstructions created using Gaussian Splatting. The framework extends a baseline reconstruction by adding new information created through inpainting and estimating point clouds in the missing areas of the reconstruction. In efforts to automate this process, functions for identifying poor reconstruction and a dynamic weighting loss are presented. The results show that it is possible to extend existing reconstructions without a significant loss in 3D representation.

High-speed aerial vehicles present a challenging control problem due to rapidly changing dynamics, nonlinear behaviour and environmental uncertainties. Traditional controllers rely on pre-computed parameters which lack the ability to adapt if the vehicle does not perform as desired. In contrast, adaptive control methods update parameters in real time during flight, enabling the vehicle to adjust to changing conditions and maintain performance despite limited knowledge of its dynamics. 

This thesis examines the performance of a high-speed aerial vehicle using one type of adaptive controller, Model Reference Adaptive Control (MRAC). The MRAC is compared to a Linear Quadratic Regulator (LQR), which uses pre-computed parameters. Two MRAC methods are investigated, a direct method based on direct updates of control parameters, and an indirect method which derives control parameters based on estimated plant dynamics. Three different uncertainty models are evaluated, based on State Space Functions, Radial Basis Functions, and Neural Networks. Concurrent Learning (CL) is also implemented to examine its performance implications. Two parameter update methods are investigated, one based on Gradient Descent (GD), and one based on a Recursive Least Squares with forgetting factor (RLS-β) algorithm. These methods are tested on two different paths, examining the performance for low and high altitudes, as well as different levels of system excitation. 

The results from simulations show that the direct MRAC handles the test cases well, with improved reference tracking compared to the LQR. The use of an uncertainty model further improves the performance of the system, although none clearly outperforms the others in all metrics. The direct MRAC manages to compensate for the uncertain dynamics rapidly, with minimal tracking error during aggressive manoeuvres without prior excitation. The direct MRAC is shown to also be able to adapt to different guidance parameters, whereas the LQR lacks this capability. The implemented indirect MRAC did not result in a working controller, likely due to difficulties in fitting the nonlinear vehicle to the chosen, highly simplified, reference model. 

The thesis concludes that a direct MRAC can provide several performance benefits to high-speed aerial vehicles. It improves performance compared to the LQR over the range of test cases, showing the capability of the direct MRAC to adapt to changing circumstances without changes to the controller. The direct MRAC method performs well despite having very little prior information about the dynamics of the plant, which proves the viability of the direct MRAC for controlling uncertain systems. 

This thesis investigates the feasibility of utilizing commercially available large language models (LLMs) for consultant–client matching. By utilizing embeddings to represent and compare text, the matching can be performed in a semantic manner, capturing even implicit details in text. Two embedding-based ranking strategies were compared: a full-profile embedding approach and a weighted attribute-based embedding approach. In the absence of ground truth labels, statistical evaluation metrics indicate that the weighted attribute-based method outperforms the full-profile method, showing greater discriminative power, clearer score dispersion among top candidates, and stronger correlation between ranking and consultant experience. In terms of cost and scalability, the pipeline demonstrated that integration into existing staffing workflows is feasible, even at scale. However, achieving real-time responsiveness would require additional optimization.

Advancements in sensor and missile technologies have transformed modern air combat, shifting engagements to long-range, data-driven scenarios known as beyond visual range (BVR) air combat. In such settings, real-time decision-making systems must process sequential, high-dimensional, non-stationary data while retaining essential information over time. This thesis explored the effectiveness of the recently introduced neural network architectures — Mamba, minGRU, and minLSTM — for processing such data. While these architectures have shown promising performance in supervised learning (SL) and offline reinforcement learning (RL), their performance in online RL, partially observable environments, and multi-agent settings were previously unexplored.

Earlier work suggests that stationary SL benchmarks may be poor proxies for RL performance. To investigate this further, this thesis evaluated whether SL performance on non-stationary data provides a better indication of performance in RL. Three non-stationary SL tests were designed to mimic types of non-stationarity found in RL, followed by experiments in four partially observable RL environments, with one being a high-dimensional multi-agent air combat scenario. Results showed minimal differences in SL performance across models when comparing stationary and non-stationary conditions. In RL the performance diverged more, where minGRU demonstrated superior learning progress. Moreover, while minLSTM displayed stable SL results, it struggled significantly in RL. The findings suggest that both stationary and non-stationary SL performance has limited predictive value for RL performance.

This thesis examines the digital signal processing (DSP) chain in a modulation transmitter, focusing on the interpolation, resampling, and frequency mixing stages. There is a working system for 256-QAM in place. The goal is to model it and reduce the usage of FPGA DSP blocks without degrading performance. A DSP block is a hardware unit designed to efficiently perform arithmetic operations such as multiplying, adding, and accumulating digital signals.

For interpolation, the system employs multistage structures combining half-band Type I linear phase FIR filters (designed using the Parks–McClellan algorithm) and polyphase decomposition to reduce computational load. Fractional delay resampling is implemented using a Farrow structure filter, optimized through minimax FIR filter design and a two-rate-based architecture. Frequency mixing is performed using Gauss's method for complex multiplication, which reduces the number of required real multipliers from four to three. A Python-based model simulates the entire signal chain and provides a platform to analyze resource usage under various selections of design parameters. Additionally, VHDL implementations of the interpolator and the complex multiplier were developed to evaluate their performance in real hardware, with special attention to finite word-length effects such as quantization noise, coefficient truncation, and overflow behavior. The results show that the optimized system achieves significant savings in DSP blocks (about 50\%) compared to the original implementation, while maintaining the required signal quality for 256-QAM transmission.

Radiology reports are the main form of communication between radiologists and referring physicians, and their quality is important for patient safety. Errors and inconsistencies within these reports do occur and can lead to confusion and misunderstandings, potentially delaying the patient's treatment. This thesis explores methods for reducing errors in radiology reports by evaluating the effectiveness of a generic large language model (LLM) in identifying inconsistencies and errors. Specifically, this study analyzes the text content of radiology reports together with associated metadata in order to mark sex, laterality and age mismatches. Automatic prompt optimization using two algorithms, genetic evolution and Prompt Optimization with Textual Gradients (ProTeGi), have been applied for optimizing the prompts used for error detection.

The datasets used for development and evaluation were created from de-identified radiology reports, with errors manually inserted and annotated to create a ground truth with known error positions. In the final evaluation, 3000 reports were used, of which 5% contained errors.

The LLM based error detection system developed in the thesis achieved a consistent recall of at least 0.9 across all error types, with age errors achieving a perfect recall of 1 for both the initial and optimized prompts. Larger variance was reported for the precision, with the highest scores being 0.85 for sex, 0.67 for laterality and  0.90 for age errors.

Both optimization algorithms produced prompts that achieved a higher F₁-score compared to the manually written baseline prompt. The prompt optimized using the genetic evolution algorithm achieved an average F₁-score of 0.81 and the ProTeGi optimized prompt 0.86, compared to the baseline prompt reporting an F₁-score of 0.76.

Event cameras offer unique advantages over traditional frame-based cameras, including high temporal resolution, high dynamic range, absence of motion blur, and much lower bandwidth. However, their asynchronous and sparse output introduces new challenges for tasks like depth estimation. This thesis investigates whether a machine learning model can predict per-pixel probability distributions over depth from a single event camera stream in indoor environments. The study evaluates six different event representations: spatio-temporal voxel grids (with 1, 5, 10, and 20 temporal bins), Compact Spatio-Temporal Representation (CSTR), and Time-Ordered Recent Event (TORE) volumes. Training initially utilizes synthetic data from the BlinkVision dataset, followed by retraining on real-world sequences from the MVSEC dataset.

Quantitative analysis reveals that first training on synthetic data followed by retraining with real-world data reduces absolute relative error and root mean squared error compared to using either dataset alone, highlighting the value of synthetic data, especially when real-world data collection is challenging. Among the tested representations, TORE, CSTR, and the 20-bin voxel grid exhibit better uncertainty calibration, indicating that detailed temporal information is important for improved uncertainty estimation. However, the quality of uncertainty predictions remains modest overall, as reflected by relatively high Area Under Sparsification Error.

För närvarande saknas ett IT-stöd inom hälso- och sjukvården för att underlätta, dokumentera och sprida förbättringsarbeten inom Region Östergötland. Denna brist medför att vårdpersonal har begränsad tillgång till tidigare genomförda förbättringsprojekt, vilket försvårar arbetet och utvecklingen. Syftet med denna studie är att identifiera och analysera behovet av ett IT-stöd som kan stödja förbättringsarbetet inom hälso- och sjukvården, samt att ta fram behov av förbättringsarbeten i form av funktionella krav. Med utgångspunkt i dessa krav presenteras ett förslag på hur ett IT-stöd, med fokus på interaktionsdesign, kan utformas för att skapa en effektivare hantering av förbättringsarbeten. Studien har genomförts med hjälp av forskning genom design, tematisk analys, där centrala teman identifierats och legat till grund för specifikationer. De funktioner som tagits fram är ett digitalt bibliotek för tidigare förbättringsarbeten, möjlighet att registrera nya förbättringsarbeten, funktion för att skicka in förbättringsförslag samt tillgång till tips på arbetssätt. Utifrån de framtagna specifikationerna har en prototyp utvecklats. Prototypens syfte är att uppfylla behoven och vara ett stöd i förbättringsarbetet samt att prototypen är utformad med avseende på tydlighet, användarvänlighet, samt anpassad till designriktlinjer. Det bör dock noteras att prototypen inte har genomgått några avslutande användartester. Studien är fortsättningsvis menad att bidra med insikter och kunskap kring vilka behov som finns i samband med förbättringsarbete, vilken typ av IT-stöd som kan vara till hjälp i det dagliga arbetet samt att ge underlag för vidare forskning inom området. 

In the automotive industry, the development of advanced safety systems is crucial to improve customer safety and maintain market competitiveness. With the rapid advancements in the automotive industry, more safety systems are introduced. With more technology being integrated into cars, there is an increased focus on both software and hardware. Therefore, these components need to be thoroughly tested.\\

This thesis aims to develop an automated software testing framework to improve the reliability and efficiency of a software testing tool that is responsible for testing newly developed hardware. The research explores two different software testing methods. The first method utilizes the unit testing method using the built-in testing tool in winIDEA, which showed promising results for testing simple functions but failed to handle more complex cases. The second method involves the development of a software testing framework using the system test method together with the Pytest package in Python. A lot of time was allocated to developing a robust testing framework, so additional tests would be easy to implement at a later stage. In addition to this framework, two test modules were successfully developed to evaluate the temperature and Ethernet communication on the Micro Controller Unit (MCU). An additional module was planned, but due to time constraints, it was not finished in time.

För att möta den framtida efterfrågan på el är en viktig aspekt utbyggnaden av förnybar elproduktion. Troligtvis kommer en stor del av den tillkommande elproduktionen i Sverige att genereras från storskalig vindkraft men även solcellsmarknaden har expanderat kraftigt och blivit ett alltmer relevant alternativ, där tillväxten förväntas fortsätta framöver. Dessutom har användningen av batterilagringssystem ökat exponentiellt i Sverige, både som fristående lösningar för stabilisering av elnätet och som integrerade system tillsammans med sol- och vindkraftsparker. En hög efterfrågan av etablering av ny elproduktion har resulterat i många kundförfrågningar gällande elnätsanslutningar till elnätsföretagen, vilket skapar långa köer. Även fast det visar på hög efterfrågan att ansluta ny elproduktion driftsätts aldrig majoriteten av projekten. Anledningen är att projektutvecklarna ofta avbryter sina anslutningsförfrågningar, vilket innebär att den efterfrågade effekten inte realiseras. För att främja tillväxten av förnybar elproduktion är det avgörande att fler anslutningar sker till elnätet. Därför är det väsentligt att undersöka orsakerna till att projektutvecklare avbryter sina anslutningsförfrågningar. Studiens syfte är därmed att identifiera orsakerna till att anslutningar av solkraft, vindkraft och batterilagring avbryts under anslutningsprocessen till regionnätet, samt att skapa en djupare förståelse för varför efterfrågad effekt inte kan realiseras. Även om orsakerna skiljer sig mellan aktörer inom solkraft, vindkraft och batterilagring undersöks. Studien är genomförd i samarbete med E.ON Energidistibution där avbrutna anslutningsärenden från cirka 40 projektutvecklare har studerats tillsammans med muntlig information från tre intervjuer. En av de främsta orsakerna till att anslutningsprocesser avbryts är för höga anslutningskostnader, vilket delvis kan härledas till ökade transformatorkostnader. Utvecklare inom solkraft och batterilagring påverkas mest av höga kostnader då de påvisats ha lägre betalningsvilja. Andra orsaker inkluderar problem med att erhålla tillstånd och mark, där kommunalt veto och avslag från länsstyrelsen ofta hindrar projektutvecklare från att fortsätta, främst vindkraftsutvecklare. Långa ledtider, orsakade av kapacitetsbrist, samt organisatoriska faktorer, där utvecklare prioriterar mer lönsamma projekt, bidrar också till avbrutna processer.  En del av orsakerna till att anslutningsförfrågningarna avbryts kunde inte identifieras i studien. För en fullständig kartläggning krävs mer tillgängliga data och att elnätsföretag tar del av utvecklarnas orsaker till avbruten anslutning och dokumenterar de på ett systematiskt sätt. Det är påtagligt att det behövs vidare undersökning kring hur anslutningsprocessen kan effektiviseras för att öka antalet anslutningar. Fler studier för att öka kunskapen, implementering av nya regleringar eller alternativa arbetssätt för att förbättra kommunikationen mellan elnätsföretagen och projektutvecklarna skulle kunna bidra till att utveckla anslutningsprocessen till samhällsekonomiskt effektiv. 

För samtliga livsmedelsgrossister i Sverige är emballage en stor del av verksamheten, och en avde ledande typerna av emballage är produkter från Svenska Retursystem. Dessa emballage,benämnda SRS-lådor och SRS-pallar, är en del i ett cirkulärt flöde tillhandahållet av SvenskaRetursystem. Detta medför ett flöde av SRS-lådor och SRS-pallar som cirkulerar mellan SvenskaRetursystem, grossister samt grossisternas kunder. En av grossisterna som använder SRSprodukter flitigt är Martin & Servera som är en av Sveriges största livsmedel- ochrestauranggrossister. I sitt norrköpingslager identifierade Martin & Servera att mängden varor debehandlar medförde att en stor del av lagerytan ockuperades för sortering och hantering av dessaSRS-produkter, speciellt av SRS-lådorna. För att frigöra lageryta och personal överväger Martin& Servera att outsourca hanteringen av SRS-lådor, vilket resulterade i denna studie vars syfte äratt skapa ett underlag för Martin & Serveras outsourcingbeslut.

Skapandet av beslutsunderlaget utfördes i tre olika delar som följer studiens huvudfrågor. Denförsta delen relaterar till att identifiera på vilka sätt outsourcingen skulle kunna lösas, då olikaalternativ skulle kunna medföra olika resultat och konsekvenser. För detta ändamål intervjuadesrelevanta anställda hos Martin & Servera samt potentiella leverantörer för outsourcingen. När demöjliga outsourcingalternativen identifierats inleddes den andre delen i skapandet avbeslutsunderlaget vilket var utredning av kostnader genom en totalkostnadsanalys. Relevantakostnadsposter att inkludera i beräkningarna identifierades både genom litteratursökning ochgenom intervjuer med Martin & Servera. I syfte av att göra beslutsunderlaget så heltäckande sommöjligt undersöktes även hur övriga, mer kvalitativa aspekter kunde påverkas vid en outsourcingvilket var den tredje och slutgiltiga delen i beslutsunderlaget. Övriga aspekter samlades primärtgenom litteratursökning och kompletterades sedan med svar från respondenterna hos Martin &Servera. Därefter analyserades hur dessa aspekter förändrades vid en potentiell outsourcing ochhur de påverkade de olika alternativen.

Resultatet visade att det fanns två potentiella leverantörer i närhet till Martin & Serverasnorrköpingslager som tjänsten skulle kunna outsourcas till, leverantör A och leverantör B. Meddessa leverantörer identifierades därefter tre olika outsourcingalternativ med leverantör A och ettoutsourcingalternativ med leverantör B som togs med till vidare analys. Två av alternativen medleverantör A innefattade att tomma SRS-lådor från kunder skickas direkt till outsourcingleverantör.Det tredje alternativet studerades både för leverantör A & B och medförde att ett tomt släp står viden av Martin & Serveras lastportar och lastas med osorterade SRS-lådor. När släpet väl är fyllthämtar leverantören släpet och ersätter det med ett nytt tomt släp. 

Beräkningar visade att outsourcingalternativet med ett stående släp med leverantör B var det minstkostsamma, och var dessutom det enda alternativet som medförde lägre kostnader än dagslägetsSRS-hantering. Från analysen av övriga aspekter framgick det att aspekterna motiveraroutsourcingalternativen med leverantör A då de har bättre möjlighet att hantera kapaciteten idagsamt större framtida volymer även om leverantör B hade starkare finansiella förutsättningar.iiUtöver detta medförde samtliga outsourcingalternativ en förbättrad arbetsmiljö hos Martin &Servera, men även en ökad miljöpåverkan. Studiens rekommendation utifrån beslutsunderlaget äratt uppdragsgivaren bör besluta om de vill utföra alternativet – stående släp med leverantör B föroutsourcingen av SRS-lådor. 

Sweden’s rental market is currently experiencing a significant housing deficit, resulting inextended queue times, particularly in greater metropolitan regions. To mitigate the severehousing deficit, the Swedish government appointed an official inquiry SOU 2021:50 toinvestigate the implementation of market-rents on newly constructed rental apartments. Thisstudy contributes to the policy debate by estimating individuals’ willingness to pay for reducedwaiting times, offering insights into how tenants value faster access to housing under currentregulatory framework. Using a binary choice contingent valuation method, we evaluatewillingness to pay in two key scenarios; Moving out from parental home and relocating for workopportunities. The analysis was conducted using logit regression, as well as non-parametricmethods to ensure robust results. Findings from logit regression reveal a significantly higherwillingness to pay among respondents relocating for work opportunities at approximately 60,156SEK for one-year faster access to a rental unit, compared to those moving out from the parentalhome where the willingness to pay was estimated to 13,910 SEK. Adding on to the estimatedwillingness to pay, this study has also identified economic effects to facilitate a future costbenefit analysis on implementing market-based rents. By providing a framework for measuringindividual preferences and utility, this study demonstrates how stated preference methods cansupport a more comprehensive and fair policy evaluation. In doing so, it complements theexisting inquiry SOU 2021:50 by capturing stated behavioral responses that are critical forunderstanding the full impact of market-based rent reforms, offering a valuable decision-supporttool for assessing future housing policies.

Physical crash testing is expensive and time consuming. Being able to simulate crashes virtually would significantly reduce costs and enable more testing focused on injury prevention, with the potential to improve vehicle safety. The discrepancy that may arise between physical tests response and virtual simulation response remains an obstacle to start using simulations in practice. This since it requires validated methods that ensure that the simulation recreates the physical crash. The aim of this project was to develop and apply a method for setting up virtual simulations and to evaluate them by comparing outputs to physical crash tests. The study focuses on lowspeed rear impact crashes using existing seat evaluation tools as dummies. This study was conducted primarily as a first step to integrating physical test results into virtual crash simulations for future development in injury prevention.  Two primary models were chosen for the method development, a dummy representing an average woman, SET 50F, and the laboratory car seat developed for high replicability. A twophase simulation setup was developed, consisting of a positioning phase and a crash pulse phase. When the simulation was executed, output data essential for future neck injury research was extracted and analyzed against physical test data to see how well the simulation recreated the physical crash.  The developed method was successfully applied to both the primary model SET 50F and the male counterpart SET 50M. Overall, the simulation response followed the trends of the physical test data well with just minor differences in, for example, peak acceleration and timing. A difference in general neck motion was identified, which could arise from discrepancies in the neck and/or torso region. The simulated Neck Injury Criteria closely followed the physical test trend throughout the whole simulation which is a great indication of well captured behavior.