Chalmers Open Digital Repository
Not a member yet
26247 research outputs found
Sort by
Rethinking Code ReviewWorkflows with LLM Assistance
No full textCode reviews are a critical yet time-consuming aspect of modern software development, increasingly challenged by growing system complexity and the demand for faster delivery. This thesis presents a study conducted at WirelessCar Sweden AB, combining an exploratory field study of current code review practices with a field experiment involving two variations of an LLM-assisted code review tool. The field study identifies key challenges in traditional code reviews, including frequent context switching, insufficient contextual information, and highlights both opportunities (e.g., automatic summarization of complex pull requests) and concerns (e.g., false positives and trust issues) in using LLMs. In the field experiment, two prototype variations were developed: one offering LLM-generated reviews upfront and the other enabling on-demand interaction. Both utilize a semantic search pipeline based on retrieval-augmented generation to assemble relevant contextual information for the review, thereby tackling the uncovered challenges. Developers evaluated both variations in real-world settings: AI-led reviews are overall more preferred, while still being conditional on the reviewers’ familiarity with the code base, as well as on the severity of the pull request
Designing Passwordless Authentication with the Tillitis TKey for Secure Web Login - A user-friendly authentication method using cryptographic hardware
No full textThe growing frequency of cyber threats, coupled with increasing regulatory demands,
has intensified the need for secure user-friendly alternatives to traditional
password-based authentication systems. This thesis presents a proof of concept implementation
of a passwordless authentication solution using the Tillitis TKey, a
cryptographic USB device designed for secure identity verification. The proposed
solution uses a hardware based challenge-response authentication, implemented using
a modern web stack, using a Svelte frontend and utilizing the Web Serial API
for browser based hardware communication. To further increase security, and as
an option for future hardware development, optional biometric authentication using
facial recognition was introduced as a second factor. The project involved translating
TKey Go libraries written by Tillitis into TypeScript, to simplify browser based
execution, thereby enhancing usability for the development team. Although the
system has certain limitations, such as restricted comparability with some browsers
and operating systems, it demonstrates the feasibility and advantages of using of
using hardware-based passwordless authentication on the web. The thesis also discusses
technical trade-offs, ethical considerations that were made, outlines future
work, including OAuth integration, secure key life-cycle management and improved
usability. Another limitation of the project is lack of formal security audit, which
was deemed to be outside the scope of this project. The Tillitis TKey is treated as a
trusted component from the manufacturer, and the web application serves as a proof
of concept rather than a production ready authentication service that is equipped
to handle real world user data and account protection. Overall, the project provides
a foundation for developing secure and privacy-conscious authentication systems as
alternatives to traditional password-based authentication services
Voltage Stability Assessment of System with High share of Inverter Interfaced Generation - Long Term Voltage Stability Assessment using DIgSILENT Powerfactory simulations
No full textAbstract
A significant amount of renewable energy-based generators are being connected to the distribution grid and provide reactive power support to the transmission grid in a less effective
way compared to traditional centralized synchronous machine-based generation. Consequently, the voltage stability phenomenon is affected in the transmission buses. In this Master thesis, the impact of distributed generation (DG) on voltage stability at the transmission grid is studied. The analyses are done in a simplified two-bus system and in the IEEE/CIGRE Nordic32 benchmark model using DigSILENT PowerFactory simulations. The impact on the PV curves of the transmission buses has been evaluated considering different control modes of DGs at different control points. One of the synchronous generators in South Sweden is replaced by DG, and its impact on the maximum power transfer from North to Central Sweden has been studied. Furthermore, the loss of one transmission line between North and Central Sweden has also been studied. When one of the synchronous generators is replaced by DG in South Sweden, the control of the voltage at the DG terminal causes a reduction in transfer capacity compared to the current scenario of the synchronous generator. This is caused by an increase in impedance between the DG and the TSO bus. With voltage control at the 135 kV bus to 1 p.u., the transfer capacity is worse than the voltage control in the DG terminal scenario. DG tries to keep the voltage to 1 p.u. at the 135 kV bus by supplying Q appropriately, causing more reactive power demand compared to DG voltage control at its terminal. So, DG reaches its Q limit earlier. This results in reduced transfer capacity. In contrast, with Q control on the 135 kV bus, it is observed that the reactive power of the DG is not fully utilized,
as the transfer capacity reaches its limit even before the reactive power limit of the DG is reached
From Gap to Gain: Insourcing Inbound Logistics in the Heavy Truck Industry
No full textIn today’s increasingly complex and uncertain business environment, maintaining a cost-
efficiency, resilient, and environmentally sustainable inbound supply chain is crucial for
remaining competitive. This thesis investigates how heavy truck manufacturers can
improve their inbound supply chain by identifying key challenges and exploring potential
areas for improvement, with a particular focus on long-distance suppliers.
The study begins by examining logistical challenges faced in the inbound supply chain of
a heavy truck manufacturer, when sourcing from distant suppliers. The starting point for
identifying the challenges is based on cost-efficiency, resilience, and environmental
sustainability. It then explores how B2B companies can draw insights from the practices
of high-performing B2C companies to manage these challenges. Finally, drawing
inspiration from B2C strategies, this thesis assesses the potential of insourcing to manage
the logistical challenges in long-distance inbound logistics.
The findings reveal that the heavy truck manufacturer has logistical challenges with limited
supply chain overview, lack of control in the supply chain, supplier dependence, and weak
incentives for supplier resilience and environmental sustainability. These interrelated
issues are rooted in strategic decisions. While some B2C practices are not directly
transferable to a B2B context, the study finds that they nonetheless offer valuable insights,
particularly the potential benefits of a more insourced approach for long-distance inbound
logistics. The study finds that while insourcing can improve supply chain overview and
control in the case, it does not substantially reduce supplier dependency and may even lead
to new challenges.
Based on the findings, the study recommends that the heavy truck manufacturer should
retain its current partial outsourcing strategy while developing a more dynamic and flexible
model. This allows for continuous adaptation to changing conditions. Insourcing should
be considered as a strategic measure rather than a standard solution, applied selectively
based on context-specific factors such as internal capacity, risk, and supplier criticality.
Furthermore, by expanded benchmarking efforts and drawing inspiration from successful
B2C companies, the heavy truck manufacturer can strengthen its ability to manage long-
distance inbound logistics and address external challenges more effectively
FE-based simulation and validation of a pedestrian impact detection system
No full textAccording to the European Commission, pedestrians accounted for 18.4% of road
traffic fatalities in the EU in 2023 [5]. Modern vehicles integrate systems such as
the Active Hood Lift System (AHLS), which lifts the bonnet to reduce head trauma
during pedestrian collisions[3]. A key component enabling AHLS is the pressuresensitive
sensor tube, which detects frontal impacts and triggers hood deployment
within milliseconds. To support the deployment of AHLS, this project aimed to
develop and validate a Finite Element (FE) model of an air-filled pressure sensor
tube for use in pedestrian impact detection. The model was implemented in LSDYNA
and evaluated against physical impact tests conducted at three velocities.
Waveform correlation between simulated and measured pressure signals was assessed
using the Waveform Index Factor (WIFac) to quantify model accuracy.
The model was assessed under varying configurations to evaluate its predictive accuracy.
Foam material properties and fixation methods were systematically tested.
Among them, the calibrated foam material model generally yielded the highest waveform
correlation with physical test results, achieving WIFac values typically above
70% across the tested velocities. By contrast, plastic strap fixation and alternative
foam models resulted in lower WIFac values, often below 60%. A nonlinear
inverse relationship was observed between the tube’s Young’s modulus and the resulting
peak pressure, with 0.001 GPa providing the closest match for the advanced
dynamic test. This is physically consistent, as lower stiffness allows for greater deformation
and a more gradual pressure increase. Sensitivity analysis was performed
by introducing small perturbations in initial velocity and impactor position to evaluate
the model’s robustness. Results showed that the waveform structure remained
stable under moderate variations, supporting the model’s reliability within expected
test tolerances. Therefore, the calibrated foam model is recommended as the reference
configuration for simulation-based evaluations of AHLS pressure tube systems.
To explore the model’s applicability in more complex scenarios, additional simulations
were conducted using the PDI-2 full-leg impactor. These simulations exhibited
increased signal deviations, indicating that further refinement is necessary for integration
into biomechanical assemblies. Future work should focus on experimental
validation of material parameters and extension to full-vehicle implementations
Learning Chern Numbers of Multiband Topological Insulators with Gauge Equivariant Neural Networks
No full textEquivariant network architectures are a well-established tool for predicting invariant
or equivariant quantities. However, almost all learning problems considered in
this context feature a global symmetry, i.e. each point of the underlying space is
transformed with the same group element, as opposed to a local “gauge” symmetry,
where each point is transformed with a different group element, exponentially
enlarging the size of the symmetry group. Gauge equivariant networks have so far
mainly been applied to problems in quantum chromodynamics. Here, we introduce a
novel application domain for gauge-equivariant networks in the theory of topological
condensed matter physics. We use gauge equivariant networks to predict topological
invariants (Chern numbers) of multiband topological insulators. The gauge symmetry
of the network guarantees that the predicted quantity is a topological invariant.
We introduce a novel gauge equivariant normalization layer to stabilize the training
and prove a universal approximation theorem for our setup. We train on samples
with trivial Chern number only but show that our models generalize to samples with
non-trivial Chern number. We provide various ablations of our setup. Our code is
available at https://github.com/sitronsea/GENet/tree/main
3D visualisering av roterande magnetfält: Användning av Unity och Arduino för att mäta och visa ett roterande magnetfält i realtid
No full textDetta examensarbete genomfördes under våren 2025. Syftet med projektet var att skapa en applikation som visualiserar magnetflödet i en asynkronmotor i realtid. Idén grundade sig i behovet av att bättre kunna illustrera hur flödesrörelsen ser ut under drift, något som trots omfattande teoretiska studier kan vara svårt att visa på ett pedagogiskt sätt.
För att mäta det magnetiska fältet användes tre Hall-effektsensorer, strategiskt placerade med 120 graders mellanrum runt motorn. Sensorerna kopplades till en Arduino Uno som samlade in mätvärden från varje sensor. Dessa data skickades trådlöst via en Bluetooth-modul till en visualiseringsapplikation utvecklad i Unity. I applikationen representerades varje sensors mätvärde av en färgad sfär vars nyans skiftade beroende på styrkan i det uppmätta magnetfältet. Dessutom beräknades och visades en riktningspil i 3D som visade den totala fältvektorns riktning och storlek i realtid. Visualiseringen gav en tydlig bild av hur det magnetiska flödet roterar i motorn.
Arbetet resulterade i en fungerande applikation med låg fördröjning och stabil överföring. Det finns potential för vidareutveckling, men den nuvarande lösningen uppfyller målet att ge en lättförståelig bild av motorflödets beteende. Systemet testades stegvis: först hårdvara och kommunikation, sedan helheten genom att rotera magneten och jämföra sensordata med visualisering. Genomförandet följde en systemkonstruktion som omfattade hårdvarudesign, kommunikationslösning och mjukvaruutveckling för ett komplett visualiseringssystem
En AI-chattbot som förenklar arbetet för studievägledare
No full textAs the student population continues to grow, so does the demand for accessible
and effective academic guidance. This bachelor’s thesis investigates how AI-based
chatbots, powered by large language models (LLMs), can be utilized to support the
work of academic advisors. Through interviews with study guidance counselors at
the University of Gothenburg, it was found that responses to most student inquiries
are already available on the university’s official websites. Accordingly, the project
commenced with the collection and structuring of relevant information from these
sources through automated web scraping.
To develop a domain-specific chatbot based on this data, a Retrieval-Augmented
Generation (RAG) system was implemented. Given the focus on addressing frequently asked questions from students, the potential of fine-tuning a large language
model was also explored. This was carried out through an automated pipeline that
generated question–answer pairs for training purposes.
The attempt at using fine-tuning did not yield significant results; however, it holds
potential for future experiments given more time and resources. In contrast, the
RAG-based system showed promising results, although it requires further development to be practically implemented
VOMB
No full textEver since our first observation of the universe, outer space has been something that many of us have thought about. When refering to space, it has usually been called something underlining its depth or unknown character, e.g. “nothingness” or “way beyond”. Today, our fascination for space has spread to multiple sectors, one of them being architecture.
Space exploration has brought valuable knowledge and inventions, however, not without consequences. As of today, all resources for space projects have been extracted from Earth, both financially and in terms of material. Space travel has also brought many health conerns to the human body and mind. Given this situation, the aim of this thesis was to design a visionary, functional and conceptual architectural proposal, focusing on the following questions:
What can space architecture do for longterm resilience in space,
both for human well-being and the human being as a species?
If we want to expand in space, where would we settle first and why, from a health perspective?
In order to continue with space research, we need to construct better material cycles for future space projects to come. In order to carry out longer manned space missions, more knowledge on human health needs to be applied
Clustering and Sentiment Analysis of Customer NVH Feedback in the Automotive Domain A Machine Learning Pipeline to Facilitate Extraction of Relevant Information from Large-Scale Textual Data
No full textNoise, Vibration, and Harshness (NVH) attributes play a significant role in shaping
overall customer satisfaction with a vehicle. Automotive manufacturers often
collect large volumes of textual customer feedback through surveys, offering valuable
insights into how various vehicle attributes are perceived. This information is
intended to help engineering teams in making informed decisions about where to
focus vehicle improvement efforts. However, its unstructured nature and scale make
it difficult for individual teams to extract the feedback relevant to them.
This thesis investigated the feasibility of a clustering and sentiment analysis pipeline
to support NVH teams in making better use of customer feedback. The proposed
pipeline combined sentence embeddings, dimensionality reduction, and clustering to
group semantically similar feedback. Cluster labels were automatically generated
using a large language model and manually refined when necessary. Both sentencebased
and aspect-based sentiment analysis were applied to quantify sentiment and
extract relevant subtopics for each cluster.
The final configuration produced 15 semantically coherent clusters from approximately
36,000 customer feedback sentences. These clusters captured distinct themes,
ranging from high-level impressions of driving and ownership to specific issues regarding
individual components. Sentence-level sentiment analysis successfully distinguished
between positive and negative feedback showing its potential to guide
improvement efforts. In contrast, aspect-based sentiment analysis was less reliable:
although per-cluster aspect distributions often aligned with cluster themes,
individual aspect terms were too frequently inaccurate. Nonetheless, the method
shows potential, and its effectiveness could likely be substantially enhanced through
domain-specific fine-tuning. Overall, the pipeline effectively facilitated the identification
of relevant feedback and could aid future data-driven design and product
improvement efforts