Chalmers Open Digital Repository
Not a member yet
26247 research outputs found
Sort by
Overstretched Parameters in Lattice-Based Cryptography
No full textLattice-based cryptosystems such as Learning With Errors (LWE) and NTRU have gained popularity as one of the main methods for constructing cryptosystems secure against a quantum attacker. The process of instantiating these involves fixing multiple parameters based on cost estimates for performing lattice reduction attacks to find the secret key. Instantiations of NTRU using large moduli q, known as overstretched NTRU, are known to suffer from an attack that allows discovery of a dense sublattice of the q-ary NTRU lattice, breaking the cryptosystem at a significantly smaller cost than the traditional secret key recovery attack.
The question of whether ring-LWE, module-LWE and NTWE suffer from the dense sublattice attack remains unresolved. We apply the attack model developed by Ducas and van Woerden to these problems to produce asymptotic cost estimates for the dense sublattice attack. We find module-LWE and ring-LWE are not expected to suffer from this attack, while NTWE is. This is further corroborated by experiments performed on NTWE. Additionally, the analysis performed on NTRU by Ducas and van Woerden is extended to larger polynomial coefficients.
Taken together, we show that lattice-based cryptographic problems that have a corresponding problem on a q-ary lattice with a dense sublattice could be vulnerable to the dense sublattice attack and that the attack model can be used to predict such a vulnerability
Quantum Optimization of Physician Scheduling For Maximal Healthcare Capacity
No full textPhysician scheduling is a critical challenge in healthcare systems, demanding a balance between operational efficiency, fairness, and individual preferences. This thesis investigates the use of quantum computing, specifically the Quantum Approximate Optimization Algorithm (QAOA), as a novel approach to solving the Physician Scheduling Problem (PSP), a known combinatorial optimization task. Classical methods, Mixed Integer Linear Programming (MILP) with Gurobi and Satisfiability Modulo Theories (SMT) with Z3, are implemented as benchmarks and used to establish feasible, constraint-satisfying solutions. The PSP is formulated as a Quadratic Unconstrained Binary Optimization (QUBO) problem, which serves as input to the QAOA algorithm. This is then executed on both quantum simulators and IBM’s real quantum hardware. A modular scheduling framework is developed to encode fairness, availability, preferences, and contractual work extent into the objective functions, enabling both short- and long-term optimization scenarios. Comparative evaluations reveal that while classical solvers consistently yield feasible schedules, QAOA demonstrates potential for competitive solution quality despite current hardware limitations
Developing a Robotic Lawn Mower for Climbing
No full textThis Bachelor’s thesis presents the development of a prototype robotic lawn mower capable of climbing vertical obstacles such as curbs and small garden walls. Traditional robotic lawn mowers are limited by their inability to traverse such elevations, requiring manual repositioning. The project addressed this limitation by designing and integrating a height-adjustable wheel system allowing step-by-step elevation of the lawn mower’s body. Utilizing components from Husqvarna’s 450x and 550 EPOS models, a three-pair wheel configuration was developed where the middle and rear wheels can independently lift themselves and lower via a mechanical elevator system powered by motors. Emphasis was placed on maintaining stability during climbing through center of gravity management and the use of specialized middle wheels. Testing confirmed that the system functioned as intended and was effective in concept. However, manual assistance was occasionally required due to excessive friction in the elevators, which prevented the
mower from completing the climbing process independently
Evaluation of emulsion gels and bigels as animal fat substitutes inmulti-material food 3D printing
No full textTo successfully develop plant-based meat analogs, it is essential not only to mimic
the protein content of conventional meat, but to also replicate the structural and
functional roles of animal fat. This thesis investigates two approaches for structuring
liquid oils, emulsion gels and bigels, for use as fat substitutes in food 3D printing.
The study was conducted in two phases. In the first, the rheological properties,
printability in single material printing and microstructure of both gels were evaluated.
The second investigated their performance in multi-material 3D printing
when combined with a pea protein isolate-based ink, using dual and coaxial extrusion
techniques.
Rheological analyses, including amplitude sweep, shear-viscosity test, frequency
sweep, three interval thixotropy test and temperature sweep, revealed distinct differences
in behaviours for each gel. The bigel showed higher initial viscosity and greater
thermal sensitivity, but also stronger shear-thinning and self-supporting properties.
The emulsion gel was softer and less structurally stable. Confocal laser scanning
microscopy provided further insights into the gels’ microstructure and phase distribution,
supporting the rheological findings.
In 3D printing, both gels were printable using the same G-code with the same
printing parameters. However, the bigel retained better definition and buildability,
and could withstand increased layer heights, where emulsion gel collapsed. In
multi-material 3D printing, the bigel maintained structural integrity when printed
together with PPI30, in contrast to emulsion gel, which exhibited poor material
distribution and inconsistent extrusion behaviour. The bigel also showed superior
storage stability, maintaining their form over extended periods at room temperature.
The results demonstrate that while both gel systems are potential options for structured
fat replacement, bigel offer greater mechanical stability and compatibility for
use in food 3D printing. These findings contribute to the development of more realistic
plant-based meat analogs and highlight the importance of optimizing both
material formulation and printing techniques
Evaluation of wheel torque coordination strategies for heavy battery electric vehicles
No full textWith the rapid evolution of technology and growing environmental concerns, the demand for electric
vehicles has increased significantly. The main challenge for a heavy battery electric vehicle is to combat the efficiency and load carrying capacity on different kinds of roads (country roads and highway)
ranging from a tarmac road with high coefficient of friction to a low friction road.
Modular E-axles such as cruise and startability axles have been introduced in this research with different types of electric machines and gear ratios to make it a reliable, cost efficient and effective setup
for achieving a higher driving range along with less power losses. The different types of power losses
that have been considered in this study are drivetrain losses, longitudinal tyre slip losses, rolling resistance losses, friction brake losses. However, only the drivetrain losses have been minimised in this work.
Three different kinds of wheel torque coordination strategies have been discussed in this thesis for
allocating force/torque requests to the actuators (electric machine, brakes) in order to evaluate the
energy savings for different types of trucks. Out of the three strategies, two of them are based on
power loss minimisation and is compared to the third strategy where equal friction is achieved at the
wheels. The performance of these strategies were evaluated using real world driving cycles.
However, the primary challenge lies in determining the optimal balance between energy efficiency and
the vehicle’s safety factor. Different methods like finding the lateral margins, friction circles of the
tyres at the axle level have been formulated and implemented in order to find a safety metric. This
thesis aimed to identify an optimal energy-efficient strategy while also defining a suitable safety metric
Determining the Optimal Treatment for Contaminated Dredged Sediment using MCDA and LCA
No full textContaminated sediments that have been dredged are impacted by industrial, agricultural, and urban activities, and present health and environmental risks due to pollutants like Potentially Toxic Elements (PTEs) and organic pollutants. Treatment methods are crucial, as after the dredging process, one end of the life point is the disposal of the sediment back into the sea. If the sediment is not treated for TBT, the current regulation dictates that the sediment must be disposed of in landfills, of which availability is more scarce compared to the sea. The study provides guidance on selecting sustainable sediment treatment methods that balance environmental impacts and organic pollutant removal. This study evaluated four sediment treatment methods: Fenton + Photoelectrocatalysis (PEC), Fenton at pH 3, Methanol + Ultrapure Water + PEC, and Density Separation + PEC. The evaluation uses a Comparative Life Cycle Assessment (CLCA) and Multi-Criteria Decision Analysis (MCDA). The goal of the study is to determine the most sustainable treatment method to remove tributyltin (TBT) from 1 ton of dredged sediment. The results indicate that the treatment method including the Fenton process, particularly Fenton + PEC, achieves high TBT removal, but generates environmental impacts due to energy and chemical use, with hydrogen peroxide production being a major contributor. Methanol + Ultrapure Water + PEC is impactful due to the production of methanol; this contributes to Eutrophication. Density Separation + PEC shows the lowest environmental impact, driven mainly by salt usage and resulted in the most sustainable option according to the MCDA results. Improvement points include using renewable energy, process integration, and closed-loop systems for better recourse efficiency and process development
Enabling Energy Efficient Training for AI Algorithms by Controlling Resource Allocation
No full textTraining deep learning models typically involves large-scale computations that require significant energy resources, making the process both costly and environmentally unsustainable. One reason for this that the default strategy of using high frequencies during deep neural network training. However, the various layers in a deep learning network have varying computational and memory access patterns, leading to potential mismatches and bottlenecks. The purpose of this thesis was to address this challenge by exploring resource allocation strategies that can reduce the energy consumption on a fine-grained level when training CNNs on GPUs. The research focuses on predicting the computational and memory demands of different deep network layers, and creating appropriate execution strategies to reduce energy consumption by reducing idle times of compute and memory units. These resource allocation strategies are based on both analysis of arithmetic intensity as well as exhaustive searches, allocating the appropriate resources by adjusting the compute and memory clock frequency combinations of each layer. This thesis demonstrates that resource allocation strategies can potentially reduce energy consumption during deep learning training. This was analysed for two deep learning models, ResNet50 and VGG16, on two different GPUs, NVIDIA RTX A4000 and NVIDIA RTX 2000 Mobile. For full training executions using our execution strategies, there were no significant improvements to the energy efficiency that did not increase the execution time. With a slight increase in execution time, one strategy achieved moderate energy savings. Focusing on the forward propagation phase there was improved results. The same strategy yielded execution times comparable to the default, in some cases even better, with moderate energy savings. If users are willing to sacrifice some performance, another execution strategy achieves a significant reduction in energy consumption with only a slight increase in execution time
Development of an in vitro biological effect assay to assess delivery and efficacy of targeted siRNA to the lung
No full textCFD-simuleringar av urbana värmeöar
No full textI denna studie undersöktes hur atmosfärisk stabilitet påverkar luftföde och temperaturfördelning i urbana miljöer, med fokus på den urbana värmeöeffekten.
Det genomfördes Computational Fluid Dynamics (CFD) simuleringar i programmet STAR-CCM+. En förenklad tvådimensionell modell av en stad gjordes och användes under tre olika stabilitetsförhållanden, definierade av Richardson-talet, Rb, det stabila fallet Rb = 0.79, det neutrala fallet Rb = 0 och det instabila fallet Rb = −0.21.
Resultatet visade att det stabila fallet begränsade den vertikala omblandningen, vilket resulterade i högre temperaturer vid gatunivå men lägre temperaturer över byggnaderna. Detta kan i sin tur bidra till den urbana väörmeöeffekten.
I det neutrala fallet noterades inte några termiskt drivna flöden. Luftcirkulationen styrdes främst av den mekaniska vinden. I det instabila fallet visade på vertikala luftströmmar och en effektiv värmetransport, vilket orsakades av uppvärmda ytor.
Denna studie uppmärksammar vikten av att inkludera betydelsen av bärkraftseffekter i simuleringarna för att kunna framställa luftflödesmönster och temperaturfördelningar, särskilt under de instabila atmosfäriska förhållandena.
Resultaten bidrar till en ökad förståelse för urbana miljöer och hur de negativa effekterna kan minskas genom en medveten stadsplanering