Chalmers Open Digital Repository

Välkommen till Chalmers öppna digitala arkiv!

Här hittar du:

Studentarbeten utgivna på lärosätet, såväl kandidatarbeten som examensarbeten på grund- och masternivå
Digitala specialsamlingar, som t ex Chalmers modellkammare
Utvalda projektrapporter

Forskningspublikationer, rapporter och avhandlingar hittar du i research.chalmers.se

Enheter i Chalmers ODR

Välj en enhet för att se alla samlingar.

Visar 1 - 2 av 2

Senast publicerade

Equivariant Inductive Biases for Weather Prediction with PEAR - Investigating the exploitation of rotational symmetries for accurate transformer-based weather forecasting over the HEALPix discretisation
(2026) Rosso, Pietro
Weather forecasting is a complex challenge due to its intrinsically complex physical dynamics that define the evolution of the system. In recent years, deep learning weather prediction has emerged as a promising alternative to classical numerical weather prediction, matching or outperforming it on several benchmarks at a fraction of the inference time. This thesis contributes to this direction by analysing the symmetries of this system in relation to the group SO(2): the rotation of the Earth around its own axis. The study builds on Pangu Equal ARea (PEAR), a transformer-based model operating on the Hierarchical Equal Area isoLatitude Pixelization (HEALPix) discretisation of the sphere, and examines whether the symmetry awareness of this architecture can be increased from two complementary perspectives: the data on which the model is trained, and the architecture itself. In the first part, starting from ERA5, the reanalysis dataset that provides global estimates of the surface and atmospheric variables, we introduce a new 2-hourly sampling, which allows a comparison of PEAR’s equivariance behaviour across three configurations of dataset and forecast horizon. The analysis shows that the equivariance error is dominated by the architecture and the forecast horizon rather than by the sampling. The second part introduces two architectural modifications, an iso-latitude interspersed windowing scheme and a set of HEALPix-aware convolutions, designed to better align the model with the rotational symmetry of the sphere. These modifications successfully reduce the equivariance error at the surface level, but fail to improve it at the upper atmospheric levels, and do not translate into a forecasting advantage over the baseline. This outcome highlights the difficulty of embedding inductive biases in the case of domains that involve using high-dimensional samples, specifically in relation to window-based attention mechanisms.
Testing the Semigroup Property of Generative Models for Dynamical Systems - Developing a test based on the Chapman–Kolmogorov equation
(2026) Green, Max; Wennberg, Hedvig
Surrogate models for molecular dynamics, particularly those based on generative artificial intelligence, offer an efficient way to model molecular systems across timescales that may be difficult to access through simulation. However, such models should remain consistent with the underlying physics. For Markovian dynamics, the Chapman Kolmogorov equation is a cornerstone of this consistency, describing how transition dynamics across different timescales should relate to each other. One such surrogate model, the Implicit Transfer Operator (ITO) framework, learns transition dynamics across multiple timescales, making it natural to question whether the learned dynamics remain consistent. Existing methods to assess this quantitatively use comparisons of distributions in the molecular space, while the test proposed in this work instead evaluates distributions in latent space, enabling metrics that were previously unavailable. In this thesis, we develop and evaluate a Chapman–Kolmogorov test for ITO models operating in the latent space of the model. The test is evaluated on both a one-dimensional model trained on the dynamics from a potential well and a three-dimensional transferable model trained on molecular dynamics data. The one dimensional model passes the test consistently, while the three-dimensional model gives more uncertain results, leading to a discussion about both the model and the multivariate version of the test. We further show that the CK-test’s performance improves alongside the learning of correct dynamics during training, suggesting that the semigroup property is learned rather than being inherent to the model architecture. However, passing the test does not guarantee that the model has learned the correct dynamics, as models with poor dynamical accuracy can still satisfy the CK-test.
Detection and Adaptation of Staggered Concept Drift in Federated Online Learning - A Dynamic Cluster-Based Framework for Machine Learning with Distributed Heterogenous Data Streams
(2026) Jakobson, Josef; Forsman, Fabian
Federated Learning (FL) enables collaborative model training across distributed edge devices without requiring the exchange of raw data. However, most FL approaches assume stationary data distributions, whereas real-world environments often exhibit concept drift, where the relationship between input data and target variables changes over time. In federated settings, such drifts may occur independently across clients, creating staggered concept drift that can lead to model poisoning, degraded predictive performance, and ineffective aggregation. This thesis investigates how staggered concept drift can be automatically detected and adapted to in a federated online learning setting under the computational constraints of edge devices. A drift-aware FL framework is proposed in which clients continuously train local online models, monitor prediction errors using lightweight drift detectors, and perform local adaptation when drift is detected. To prevent aggregation of models representing different concepts, a server-side clustering mechanism dynamically groups clients according to their current concepts and maintains separate cluster models for aggregation. Additionally, the framework is evaluated on the computational cost of the edge-level algorithms, the accuracy of the detectors and performance of the global models.
Analysis of High-Dimensional Time-Series Data
(2026) Naumanen, Jonathan; Leiditz Thorsson, Joel
Modern software systems generate large volumes of high-dimensional multivariate time-series data during testing and operation. In industrial settings such as continuous integration and continuous deployment (CI/CD) pipelines, manually diagnosing recurring failures across thousands of performance metrics is time-consuming and error-prone. This thesis investigates whether representation learning can produce stable, interpretable, and diagnostically useful groupings of anomalous behaviour in such data. This allows for recurring behavioural patterns to be identified and related to meaningful system characteristics, thereby supporting failure investigation at scale. We develop and compare several autoencoder architectures, including standard, variational, Long Short-Term Memory-based (LSTM), and Transformer-based variants, against traditional dimensionality reduction baselines and state-of-the-art anomaly detection models. Each architecture encodes multivariate time windows into a compact latent representation, from which reconstruction error is derived as an anomaly score and clustering is subsequently performed. We evaluate these methods across four dimensions: anomaly detection performance, clustering stability under perturbation, interpretability of the resulting cluster structure, and computational cost. Experiments are conducted on three established benchmark datasets: Server Machine Dataset (SMD), Secure Water Treatment System (SWaT), and Soil Moisture Active Passive (SMAP) and an unlabelled industrial dataset from an Ericsson CI/CD testing pipeline. No single model consistently outperforms across all datasets. LSTM-based and probabilistic architectures achieve the highest anomaly detection scores, yet traditional methods retain approximately 90% of peak performance at a fraction of the computational cost. Clustering stability is high on the benchmark datasets, with simpler models frequently matching more complex architectures under perturbation. This means that the models capture meaningful structure instead of noise which is important for downstream analysis. The LSTM encoder combined with Uniform Manifold Approximation and Projection (UMAP) with K-Means clustering produces the most interpretable groupings, with cluster prototypes showing clearly distinct feature-level patterns. In the industrial case study, the approach groups test runs by performance-metric dynamics rather than textual error similarity, offering a complementary perspective for fault diagnosis. Across all tasks, the relationship between model complexity and performance proves non-linear: moving beyond linear projections yields meaningful gains, but further architectural complexity delivers diminishing returns.
Human-Autonomy Teaming: Effects of Autonomy Level and UAV Integration on Operator Decision-Making
(2026) Andersson, Sofie; Othman, Amin
As autonomous systems become increasingly integrated into safety-critical operations, understanding how human operators collaborate with heterogeneous robot teams is essential for effective system design. This thesis investigates how the level of automation (LOA) of an unmanned ground vehicle (UGV) and the integration of unmanned aerial vehicle (UAV) information influence operator state, performance, and decision-making in a human-UGV-UAV team. A controlled experiment was designed, varying UGV automation level — Assisted-Autonomous (AA-LOA), where the operator receives take-over requests (TORs), and Autonomous (A-LOA), where the system operates without operator input — and UAV information type — Conflicting or Non-Conflicting with the UGV sensor data. Thirty-three participants completed video-based simulations, representing logistics transport scenarios in which the UAV flew ahead of the UGV acting as a forward-looking sensor. Quantitative (e.g., operator response time, take-over probability, and eye-tracking gaze) and qualitative (e.g., mental workload and observations) data was collected. Results show that AA-LOA significantly increases perceived responsibility compared to A-LOA, without a corresponding increase in mental workload. Operators integrated the UAV-view throughout the mission, and their subjective estimates of UAV reliance closely matched that of the objective eye-tracking data. Conflicting UAV information increased the likelihood of manual intervention but did not significantly extend decision time. Despite a 50% conflict rate, operators consistently valued the UAV-view for its contextual overview and sense of safety. Demographic factors showed meaningful effects: gamers reported lower mental workload and trended toward faster decisions, while high driving frequency and strong sense of direction were the strongest predictors of longer decision times and greater confidence in own judgment during conflict scenarios. Based on these findings, 15 design guidelines were developed addressing LOA selection, transparency, UAV integration, and temporal awareness — including UAV recap functionality and a time-delta indicator to help operators contextualize temporally misaligned information. A dual-screen operator interface concept design implementing these guidelines is presented. To conclude, the findings in this thesis suggests that assisted-autonomy combined with time-transparent UAV integration provides operators with the best conditions for effective situational awareness and decision-making in heterogeneous human-autonomy systems.