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

QuadriSparse: RISC-V Sparse Matrix Accelerator and ISA Extension - A Tightly-Coupled Sparse-Dense Matrix Multiplication Accelerator for RISC-V
(2026) Erlandsson, Nik; Swärd, Oskar
Sparse dense matrix multiplication (SpMM) is an important operation in many applications such as inference and training of pruned large language models, graph analytics and scientific computing. These applications often operate on data that is inherently sparse. Applying dense matrix multiplication (GEMM) to sparse data wastes computations on zero-valued elements, which SpMM avoids by skipping them. However, accelerators designed for dense matrix multiplication do not necessarily support sparse matrix multiplication efficiently. This motivates extensions that can exploit sparsity while retaining computation capabilities for dense workloads. This thesis explores the prospect of extending a small dense matrix multiplication accelerator with additional hardware for SpMM, evaluating the the performance benefits against the hardware overhead. We introduce QuadriSparse, a SpMM extension for unstructured sparsity that adds a partly new datapath to the small and efficient RISC-V accelerator Quadrilatero. The accelerator includes three new instructions to load a tile of the sparse matrix (SPLD_W), load a tile of the dense matrix (DLD_W), and multiply the two using Gustavson’s algorithm (SPMAC_W). We evaluate the resulting accelerator in terms of execution time across different sparsity levels and matrix sizes and measure the FPGA resource utilization through synthesis. QuadriSparse achieves up to 6.6x lower execution time than dense execution on Quadrilatero at 99%sparsity and first outperforms the dense baseline at 95% sparsity. FPGA synthesis shows an increase in resource utilization of 4.6% in LUTs and 23% in DSP-blocks relative to the baseline design.
Cybersecurity in Decentralized Machine Learning for Battery Management Systems: Threats, Detection, and Defense
(2026) Afrem, Johny; Haj Ibrahim, Zaid
Federated Learning (FL) enables distributed devices to collaboratively train machine learning models without sharing raw data, making it suitable for Battery Management Systems (BMSs) that estimate battery State of Health (SOH). However, FL remains vulnerable to attacks such as poisoning attacks in which malicious participants manipulate local training or model updates to influence the learned model. This thesis investigates the security of decentralized battery health prediction systems implemented using the FEDn framework and an Adaptive Iterative Clustered Federated Learning (AICFL) architecture. A controlled experimental environment was developed to evaluate the impact of poisoning attacks on both traditional FL and clustered FL. Three attack categories were implemented and analyzed: model poisoning, stealth-oriented poisoning, and targeted backdoor attacks. To support attack analysis, a server-side suspicious-client risk scoring mechanism was developed to identify anomalous client behavior based on model update characteristics collected during training. Experimental results compare the effectiveness of attacks in FL and AICFL environments and examine how clustering influences attack propagation and model robustness. Experimental results show that poisoning attacks can significantly affect model behavior in both FL and AICFL environments. The impact varies across attack types, while the clustered architecture influences how malicious updates propagate through the federation. The proposed risk-scoring mechanism was able to identify clients exhibiting suspicious update patterns during training. The findings provide insights into the security challenges of clustered federated learning systems for battery management applications and contribute practical methods for analyzing malicious behavior in decentralized AI systems.
Scenario Classification via Imagery and Video Data for ADAS Trigger Analysis
(2026) Zhao, Weirui
Advanced Driver Assistance Systems (ADAS) generate large volumes of trigger events during development and validation. For low-speed reversing functions such as Rear Auto Brake (RAB), these events must be reviewed to distinguish recurring traffic situations and valid interventions from ambiguous or undesired activations. Manual event-by-event inspection is difficult to scale, while extensive ground-truth labels are generally unavailable. This thesis therefore investigates whether pretrained visual representations and unsupervised clustering can structure industrial RAB trigger data for exploratory analysis. An offline, configuration-driven pipeline was developed to represent each event by one key frame, extract frozen embeddings using DINOv2 ViT-S/14 or ResNet50, optionally reduce their dimensionality to 128 principal components, and cluster them using HDBSCAN or K-Means. The evaluation comprises 18 controlled experiments on the same 8,017 RAB trigger images. Cluster count, noise ratio, Silhouette score, and Davies–Bouldin index are considered together with PCA and UMAP visualizations. DINOv2 combined with HDBSCAN produced the strongest internal cluster structure among the evaluated configurations, and PCA improved every paired DINOv2– HDBSCAN experiment. The best internal result reached a Silhouette score of 0.2709 and a Davies–Bouldin index of 1.4676, while assigning 46.8% of the samples to noise. DINOv2 also retained more events and separated them more strongly than ResNet50 in the two matched comparisons, whereas K-Means covered every event but produced substantially weaker separation. The high HDBSCAN noise ratios indicate considerable visual diversity and show that many events did not form sufficiently dense recurring groups. A manual annotation of 820 events drawn from one reference run examined whether the discovered groups carry consistent semantics. Weighted purity reached 0.931 for trigger validity, 0.947 for scenario cause, and 0.943 for fine object type, while individual clusters still combined opposing validity judgments and the sampled noise group consisted mostly of clear true-positive observations of ordinary static obstacles. Self-supervised embeddings and density-based clustering therefore expose measurable and partly interpretable structure in unlabeled RAB trigger data and can support grouped inspection. The approach remains an exploratory structuring tool rather than an automated scenario classifier, because the semantic evidence came from one run and one reviewer, single key frames omit motion and sensor context, and the effect on review effort was not measured.
Design of HV to LV DC/DC converter using resonant topology
(2026) Gouse Vaddinakatti, Sareena
The increasing demand for high-voltage electric vehicle architectures requires efficient power converters. Additionally, the increasing adoption of 800 V battery architectures alongside conventional 400 V systems has increased the requirements placed on the high-voltage to low-voltage (HV-LV) DC-DC converter. Hence, a wider variation in battery voltage requires converters with a wide input-voltage range while maintaining high efficiency, compact size, and low voltage stress on the components. This thesis presents the design and evaluation of a 9 kW isolated HV-LV DC-DC converter based on a Stacked Half-Bridge (SHB) LLC resonant topology, designed to operate over an input-voltage range of 330–950 V while providing a regulated low-voltage output. The proposed converter uses the LLC resonant tank together with the SHB structure to achieve the required wide voltage-gain range. The large variation in input voltage would otherwise require a conventional LLC converter to operate over a wide gain range, potentially increasing the required operating frequency range or requiring additional conversion stages. The SHB structure provides an additional voltage gain, allowing the required conversion ratio to be achieved over the 330–950 V input range. At the same time, the LLC resonant tank enables soft-switching operation, while the SHB structure reduces the voltage stress across the primary-side switches. The converter design includes the development of the LLC resonant tank, highfrequency transformer, resonant inductor, modulation strategy, and control structure. First Harmonic Approximation (FHA) was employed for resonant tank design, and detailed loss models were implemented in MATLAB/Simulink to evaluate converter performance under practical operating conditions. Several capacitor-voltage balancing techniques were investigated, and an interleaved PWM modulation strategy was selected due to its improved balancing performance. Simulation results demonstrate that the proposed converter achieves the required voltage conversion over the specified operating range while maintaining capacitorvoltage balance and high efficiency. A maximum efficiency of 97.2% was obtained under nominal operating conditions. From the results, it can be concluded that the SHB LLC converter is a promising solution for wide-input-voltage automotive DC-DC conversion applications. It offers high efficiency, reduced component stress and increased voltage gain range of the converter.
Sputtered Titanium Nitride Catalysts on Sulfur Hosts and Current Collectors for Lithium–Sulfur Batteries
(2026) Rosvall, Hugo
Lithium–sulfur batteries (LSBs) offer high theoretical capacities (1675 mAh g−1) and are a strong prospect for next-generation energy storage, but their performance is heavily limited by the polysulfide shuttle effect and slow redox kinetics. In this study, ultrathin layers of titanium nitride (TiN) were deposited by physical vapor deposition (PVD) onto multi-walled carbon nanotubes (MWCNTs) and aluminum current collectors to evaluate their effectiveness as a catalytic coating. The cells fab ricated with TiN-coated current collectors delivered superior discharge capacities of 1261.8 and 920.3 mAh g−1 at 0.1 and 0.5 C, respectively, significantly outperforming the reference cells (1010.1 and 770.8 mAh g−1), indicating enhanced interfacial con ductivity and sulfur utilization. Modification of the MWCNT sulfur host material with TiN coatings also improved electrochemical performance, with TiN@MWCNT cathodes showing increased initial discharge capacities of 1088.9 mAh g−1 for a ∼3 nm TiN layer and 1183.1 mAh g−1 for a ∼6 nm layer at 0.1 C, compared to the uncoated MWCNT reference (1010.1 mAh g−1 ), and consistently delivering higher capacities throughout extended cycling. An initial voltage decay was observed only in the TiN-modified MWCNT electrodes, after which the voltage profiles gradually stabilized over subsequent cycles, indicating an activation period associated with the TiN@MWCNT architecture. While PVD proved effective for both the planar aluminum current collector and the nanoscale MWCNT host, the fabrication of uniform TiN coatings was considerably more straightforward on the flat current col lector surface, suggesting that nanotube–based architectures warrant further study using PVD-optimized or alternative coating strategies. Overall, the results highlight TiN as a promising interfacial modifier for LSBs.