Chalmers Open Digital Repository

Welcome to Chalmers Open Digital Repository!

Here you can find:

Student theses and papers
Digital special collections, such as Chalmers modellkammare
Selected project reports

Research publications, reports and dissertations can be found in research.chalmers.se

Communities in Chalmers ODR

Select a community to browse its collections.

Now showing 1 - 2 of 2

Recent Submissions

Body Motion Estimation from Non-Dedicated Sensors
(2026) Nama, Kanishk
Accurate knowledge of vehicle body motion is essential for modern suspension control systems, such as continuously variable damping and air spring control. Traditionally, these systems rely on dedicated accelerometers mounted on the vehicle body. However, modern vehicles already carry inertial sensors within safety-critical subsystems such as the Supplemental Restraint System (SRS), alongside suspension deflection sensors at each wheel. This thesis investigates whether these existing onboard signals can be repurposed to estimate vehicle body accelerations, specifically heave, roll, and pitch, to a standard sufficient for suspension control, without introducing additional hardware. A Kalman filter-based sensor fusion algorithm is developed, combining SRS IMU measurements with vehicle speed from the CAN bus and suspension deflection signals. The algorithm first estimates vehicle attitude through a linear time-varying Kalman filter that propagates the gravity direction vector using gyroscope measurements and corrects it using compensated lateral and vertical accelerometer readings. The estimated attitude is then used to extract global vertical acceleration from the IMU, which is subsequently fused with suspension-derived body signals in a Kalman-based observer to produce heave, roll, and pitch acceleration estimates. The algorithm was evaluated against ground-truth signals derived from a dedicated accelerometer system currently used at Volvo Cars Corporation. The results demonstrate that heave and pitch accelerations can be reconstructed with strong accuracy across the primary ride frequency band, while roll estimation captures the dominant low-frequency dynamics with more moderate agreement. Overall, the proposed framework shows that SRS-based sensor fusion is a technically viable alternative to dedicated body accelerometers, offering a path toward reduced hardware cost and improved sensor redundancy in production vehicles.
Adapting Without Forgetting
(2026) Haraldsson, Max; Widengård, Anton
Automated cell segmentation in microscopy images is a critical task in pharmaceutical research and biological discovery. Existing supervised approaches require extensive annotated data and must be retrained for each new cell type or imaging condition. Foundation models offer a promising alternative through their ability to generalise across diverse domains without retraining. However, a systematic comparison of their generalisation across diverse microscopy conditions has been lacking. Moreover, adapting such models to specific datasets through fine-tuning risks catastrophic forgetting, potentially undermining the very generalisation that makes them valuable. This thesis benchmarks two state-of-the-art foundation models for cell segmentation, CellSAM and Cellpose-SAM, across several datasets spanning multiple imaging modalities. Cellpose-SAM consistently outperformed CellSAM, particularly at stricter evaluation thresholds, attributed to its flow-based instance separation, which handles touching and overlapping cells more effectively than CellSAM’s promptbased approach. Cellpose-SAM was subsequently fine-tuned using QLoRA, a parameter-efficient method that adapts only a small fraction of the model’s weights. Fine-tuning on a small curated dataset yielded substantial improvement on challenging microscopy images while retaining near-baseline performance on held-out general data, demonstrating that catastrophic forgetting can be avoided with careful, parameter-efficient finetuning. The entire training process required approximately 15 minutes on a single A100 GPU, illustrating that meaningful adaptation of large vision foundation models is achievable with modest computational resources.
Machine Learning for Diverse and Adaptive Waveform Generation
(2026) Gradin, Per-Ola; Melin, Gabriel
Multiple-input multiple-output (MIMO) radar systems offer high flexibility in the design of transmit waveforms, something that can be leveraged for several different radar objectives. This thesis focuses on using neural-network based approaches for designing waveforms that suppress unwanted signal returns (clutter), while maintaining strong target returns. The waveform design problem is formulated around shaping the range-angle ambiguity function, where the objective is to maximize signal-to-clutter-noise ratio (SCNR) under a soft constraint on the bandwidth of the phase-coded waveforms. Unlike conventional optimization-based methods of waveform design, the proposed approach aims to train neural networks capable of generalizing to arbitrary, previously unseen clutter environments at test-time. Several neural network architectures are explored throughout this thesis, namely Convolutional Autoencoders, Vision Transformers, Residual Networks, and Diffusionbased models. The models are trained and evaluated on synthetically generated clutter maps, with varying waveform dimensionalities and clutter distributions. The performance is evaluated using held-out test sets of clutter maps, with corresponding optimization-based waveforms as a benchmark. Additionally, methods for generating diverse sets of waveforms for the same clutter scenario are explored, motivated by the non-convexity of the waveform design objective as well as the potential mitigation against adversarial identification and counter-measure techniques. The results demonstrate that neural networks can achieve performance comparable to the provided optimization-based benchmark on previously unseen clutter scenarios, particularly for clutter environments with fewer and larger regions of clutter. Models trained directly using differentiable waveform performance objectives are shown to significantly outperform a supervised training approach. Moreover, the results show that diverse waveform generation is achievable, although there is an indication of a tradeoff between achieved diversity and average waveform performance, and the effective diversity of the generated waveforms is shown to be significantly dependent on the method of generating multiple outputs. Overall, the findings indicate that machine learning and neural networks offers a promising strategy for adaptive MIMO radar waveform synthesis, while also highlighting some important limitations and challenges, with room for future research.
A Method for Using Model-Based Systems Engineering to Drive Multidisciplinary Design Optimization Studies of Jet Engine Components
(2026) Karlsson, Edvin
The aerospace industry faces immense pressure to maintain efficient innovation due to increasing system complexity, strict environmental regulations and a very competitive market. To manage these conditions, organizations move increasingly towards exploring Model Based Systems Engineering (MBSE) as the potential new way of working to represent the decisions made during requirement breakdown for a novel system. Multidisciplinary Design Optimization (MDO) is a mature method commonly implemented when performing a Design Space Exploration (DSE) and currently MBSE and MDO are implemented in complete digital isolation from one another. This thesis explores the potential benefits and challenges with of adopting MBSE into the designated way of working at GKN Aerospace and develops an integration scheme with a corresponding method to evaluate the possibilities of utilizing MBSE to drive MDO studies. The developed method was evaluated based on three success criteria: Traceability, automation and results reuse ability. Implementation and evaluation of the developed method revealed that MBSE enables the possibility of having a digital thread from the Product Breakdown (PB) to the MDO environment, however, the current available tools and methods do not satisfy the needs to achieve an efficient integration and there is currently no established way of working with MBSE at GKN.
Modulbaserad pergola med träning och gemenskap i centrum
(2026) Bredberg, Nathanael; Ferm, Gustav
This thesis covers the concept development of a module-based pergola with training possibilities for the consumer market. The aim of this project was to create a solution that merged the boundaries between social spaces and fitness equipment through the use of a modular framework. Questions that guided the process were how simple assembly could be combined with strong structural integrity. Based on these questions, the process resulted in a concept based on beams of glued laminated timber, also known as “glulam”, and powder-coated sheet-metal. Key solutions include steel corner joints that ensure structural integrity for the beam assembly, steel rails that allow modules to be installed and steel pillar bases that allow the structure to be mounted on a wooden deck. User based studies confirm the need for ease of installment and an aesthetic that appeals to a garden environment. This resulted in a design language that combined a high level of technical functionality with an aesthetically pleasing appearance. An environmental analysis of the concept indicated that the majority of the climate impact came from the extracting and refining of raw materials. This impact could be reduced by offering spare parts and options for personal modification of the structure and therefore extending its lifetime. The final concept proves the possibilities of developing a stable modular construction that incorporates training equipment and social interactions in a garden.